fs/nfs/nfs4proc.c

   1 /*
   2  *  fs/nfs/nfs4proc.c
   3  *
   4  *  Client-side procedure declarations for NFSv4.
   5  *
   6  *  Copyright (c) 2002 The Regents of the University of Michigan.
   7  *  All rights reserved.
   8  *
   9  *  Kendrick Smith <kmsmith@umich.edu>
  10  *  Andy Adamson   <andros@umich.edu>
  11  *
  12  *  Redistribution and use in source and binary forms, with or without
  13  *  modification, are permitted provided that the following conditions
  14  *  are met:
  15  *
  16  *  1. Redistributions of source code must retain the above copyright
  17  *     notice, this list of conditions and the following disclaimer.
  18  *  2. Redistributions in binary form must reproduce the above copyright
  19  *     notice, this list of conditions and the following disclaimer in the
  20  *     documentation and/or other materials provided with the distribution.
  21  *  3. Neither the name of the University nor the names of its
  22  *     contributors may be used to endorse or promote products derived
  23  *     from this software without specific prior written permission.
  24  *
  25  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  27  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  29  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  32  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  33  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  34  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  35  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 #include <linux/mm.h>
  39 #include <linux/delay.h>
  40 #include <linux/errno.h>
  41 #include <linux/string.h>
  42 #include <linux/ratelimit.h>
  43 #include <linux/printk.h>
  44 #include <linux/slab.h>
  45 #include <linux/sunrpc/clnt.h>
  46 #include <linux/nfs.h>
  47 #include <linux/nfs4.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/nfs_mount.h>
  51 #include <linux/namei.h>
  52 #include <linux/mount.h>
  53 #include <linux/module.h>
  54 #include <linux/xattr.h>
  55 #include <linux/utsname.h>
  56 #include <linux/freezer.h>
  57 #include <linux/iversion.h>
  58
  59 #include "nfs4_fs.h"
  60 #include "delegation.h"
  61 #include "internal.h"
  62 #include "iostat.h"
  63 #include "callback.h"
  64 #include "pnfs.h"
  65 #include "netns.h"
  66 #include "sysfs.h"
  67 #include "nfs4idmap.h"
  68 #include "nfs4session.h"
  69 #include "fscache.h"
  70 #include "nfs42.h"
  71
  72 #include "nfs4trace.h"
  73
  74 #define NFSDBG_FACILITY         NFSDBG_PROC
  75
  76 #define NFS4_BITMASK_SZ         3
  77
  78 #define NFS4_POLL_RETRY_MIN     (HZ/10)
  79 #define NFS4_POLL_RETRY_MAX     (15*HZ)
  80
  81 /* file attributes which can be mapped to nfs attributes */
  82 #define NFS4_VALID_ATTRS (ATTR_MODE \
  83         | ATTR_UID \
  84         | ATTR_GID \
  85         | ATTR_SIZE \
  86         | ATTR_ATIME \
  87         | ATTR_MTIME \
  88         | ATTR_CTIME \
  89         | ATTR_ATIME_SET \
  90         | ATTR_MTIME_SET)
  91
  92 struct nfs4_opendata;
  93 static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
  94 static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
  95 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
  96 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
  97                               struct nfs_fattr *fattr, struct inode *inode);
  98 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
  99                             struct nfs_fattr *fattr, struct iattr *sattr,
 100                             struct nfs_open_context *ctx, struct nfs4_label *ilabel);
 101 #ifdef CONFIG_NFS_V4_1
 102 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
 103                 const struct cred *cred,
 104                 struct nfs4_slot *slot,
 105                 bool is_privileged);
 106 static int nfs41_test_stateid(struct nfs_server *, const nfs4_stateid *,
 107                               const struct cred *);
 108 static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
 109                               const struct cred *, bool);
 110 #endif
 111
 112 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 113 static inline struct nfs4_label *
 114 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 115         struct iattr *sattr, struct nfs4_label *label)
 116 {
 117         int err;
 118
 119         if (label == NULL)
 120                 return NULL;
 121
 122         if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
 123                 return NULL;
 124
 125         label->lfs = 0;
 126         label->pi = 0;
 127         label->len = 0;
 128         label->label = NULL;
 129
 130         err = security_dentry_init_security(dentry, sattr->ia_mode,
 131                                 &dentry->d_name, NULL,
 132                                 (void **)&label->label, &label->len);
 133         if (err == 0)
 134                 return label;
 135
 136         return NULL;
 137 }
 138 static inline void
 139 nfs4_label_release_security(struct nfs4_label *label)
 140 {
 141         if (label)
 142                 security_release_secctx(label->label, label->len);
 143 }
 144 static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 145 {
 146         if (label)
 147                 return server->attr_bitmask;
 148
 149         return server->attr_bitmask_nl;
 150 }
 151 #else
 152 static inline struct nfs4_label *
 153 nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
 154         struct iattr *sattr, struct nfs4_label *l)
 155 { return NULL; }
 156 static inline void
 157 nfs4_label_release_security(struct nfs4_label *label)
 158 { return; }
 159 static inline u32 *
 160 nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
 161 { return server->attr_bitmask; }
 162 #endif
 163
 164 /* Prevent leaks of NFSv4 errors into userland */
 165 static int nfs4_map_errors(int err)
 166 {
 167         if (err >= -1000)
 168                 return err;
 169         switch (err) {
 170         case -NFS4ERR_RESOURCE:
 171         case -NFS4ERR_LAYOUTTRYLATER:
 172         case -NFS4ERR_RECALLCONFLICT:
 173         case -NFS4ERR_RETURNCONFLICT:
 174                 return -EREMOTEIO;
 175         case -NFS4ERR_WRONGSEC:
 176         case -NFS4ERR_WRONG_CRED:
 177                 return -EPERM;
 178         case -NFS4ERR_BADOWNER:
 179         case -NFS4ERR_BADNAME:
 180                 return -EINVAL;
 181         case -NFS4ERR_SHARE_DENIED:
 182                 return -EACCES;
 183         case -NFS4ERR_MINOR_VERS_MISMATCH:
 184                 return -EPROTONOSUPPORT;
 185         case -NFS4ERR_FILE_OPEN:
 186                 return -EBUSY;
 187         case -NFS4ERR_NOT_SAME:
 188                 return -ENOTSYNC;
 189         default:
 190                 dprintk("%s could not handle NFSv4 error %d\n",
 191                                 __func__, -err);
 192                 break;
 193         }
 194         return -EIO;
 195 }
 196
 197 /*
 198  * This is our standard bitmap for GETATTR requests.
 199  */
 200 const u32 nfs4_fattr_bitmap[3] = {
 201         FATTR4_WORD0_TYPE
 202         | FATTR4_WORD0_CHANGE
 203         | FATTR4_WORD0_SIZE
 204         | FATTR4_WORD0_FSID
 205         | FATTR4_WORD0_FILEID,
 206         FATTR4_WORD1_MODE
 207         | FATTR4_WORD1_NUMLINKS
 208         | FATTR4_WORD1_OWNER
 209         | FATTR4_WORD1_OWNER_GROUP
 210         | FATTR4_WORD1_RAWDEV
 211         | FATTR4_WORD1_SPACE_USED
 212         | FATTR4_WORD1_TIME_ACCESS
 213         | FATTR4_WORD1_TIME_METADATA
 214         | FATTR4_WORD1_TIME_MODIFY
 215         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 216 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 217         FATTR4_WORD2_SECURITY_LABEL
 218 #endif
 219 };
 220
 221 static const u32 nfs4_pnfs_open_bitmap[3] = {
 222         FATTR4_WORD0_TYPE
 223         | FATTR4_WORD0_CHANGE
 224         | FATTR4_WORD0_SIZE
 225         | FATTR4_WORD0_FSID
 226         | FATTR4_WORD0_FILEID,
 227         FATTR4_WORD1_MODE
 228         | FATTR4_WORD1_NUMLINKS
 229         | FATTR4_WORD1_OWNER
 230         | FATTR4_WORD1_OWNER_GROUP
 231         | FATTR4_WORD1_RAWDEV
 232         | FATTR4_WORD1_SPACE_USED
 233         | FATTR4_WORD1_TIME_ACCESS
 234         | FATTR4_WORD1_TIME_METADATA
 235         | FATTR4_WORD1_TIME_MODIFY,
 236         FATTR4_WORD2_MDSTHRESHOLD
 237 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
 238         | FATTR4_WORD2_SECURITY_LABEL
 239 #endif
 240 };
 241
 242 static const u32 nfs4_open_noattr_bitmap[3] = {
 243         FATTR4_WORD0_TYPE
 244         | FATTR4_WORD0_FILEID,
 245 };
 246
 247 const u32 nfs4_statfs_bitmap[3] = {
 248         FATTR4_WORD0_FILES_AVAIL
 249         | FATTR4_WORD0_FILES_FREE
 250         | FATTR4_WORD0_FILES_TOTAL,
 251         FATTR4_WORD1_SPACE_AVAIL
 252         | FATTR4_WORD1_SPACE_FREE
 253         | FATTR4_WORD1_SPACE_TOTAL
 254 };
 255
 256 const u32 nfs4_pathconf_bitmap[3] = {
 257         FATTR4_WORD0_MAXLINK
 258         | FATTR4_WORD0_MAXNAME,
 259         0
 260 };
 261
 262 const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 263                         | FATTR4_WORD0_MAXREAD
 264                         | FATTR4_WORD0_MAXWRITE
 265                         | FATTR4_WORD0_LEASE_TIME,
 266                         FATTR4_WORD1_TIME_DELTA
 267                         | FATTR4_WORD1_FS_LAYOUT_TYPES,
 268                         FATTR4_WORD2_LAYOUT_BLKSIZE
 269                         | FATTR4_WORD2_CLONE_BLKSIZE
 270                         | FATTR4_WORD2_CHANGE_ATTR_TYPE
 271                         | FATTR4_WORD2_XATTR_SUPPORT
 272 };
 273
 274 const u32 nfs4_fs_locations_bitmap[3] = {
 275         FATTR4_WORD0_CHANGE
 276         | FATTR4_WORD0_SIZE
 277         | FATTR4_WORD0_FSID
 278         | FATTR4_WORD0_FILEID
 279         | FATTR4_WORD0_FS_LOCATIONS,
 280         FATTR4_WORD1_OWNER
 281         | FATTR4_WORD1_OWNER_GROUP
 282         | FATTR4_WORD1_RAWDEV
 283         | FATTR4_WORD1_SPACE_USED
 284         | FATTR4_WORD1_TIME_ACCESS
 285         | FATTR4_WORD1_TIME_METADATA
 286         | FATTR4_WORD1_TIME_MODIFY
 287         | FATTR4_WORD1_MOUNTED_ON_FILEID,
 288 };
 289
 290 static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
 291                                     struct inode *inode, unsigned long flags)
 292 {
 293         unsigned long cache_validity;
 294
 295         memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
 296         if (!inode || !nfs_have_read_or_write_delegation(inode))
 297                 return;
 298
 299         cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
 300
 301         /* Remove the attributes over which we have full control */
 302         dst[1] &= ~FATTR4_WORD1_RAWDEV;
 303         if (!(cache_validity & NFS_INO_INVALID_SIZE))
 304                 dst[0] &= ~FATTR4_WORD0_SIZE;
 305
 306         if (!(cache_validity & NFS_INO_INVALID_CHANGE))
 307                 dst[0] &= ~FATTR4_WORD0_CHANGE;
 308
 309         if (!(cache_validity & NFS_INO_INVALID_MODE))
 310                 dst[1] &= ~FATTR4_WORD1_MODE;
 311         if (!(cache_validity & NFS_INO_INVALID_OTHER))
 312                 dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
 313
 314         if (nfs_have_delegated_mtime(inode)) {
 315                 if (!(cache_validity & NFS_INO_INVALID_ATIME))
 316                         dst[1] &= ~FATTR4_WORD1_TIME_ACCESS;
 317                 if (!(cache_validity & NFS_INO_INVALID_MTIME))
 318                         dst[1] &= ~FATTR4_WORD1_TIME_MODIFY;
 319                 if (!(cache_validity & NFS_INO_INVALID_CTIME))
 320                         dst[1] &= ~FATTR4_WORD1_TIME_METADATA;
 321         } else if (nfs_have_delegated_atime(inode)) {
 322                 if (!(cache_validity & NFS_INO_INVALID_ATIME))
 323                         dst[1] &= ~FATTR4_WORD1_TIME_ACCESS;
 324         }
 325 }
 326
 327 static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
 328                 struct nfs4_readdir_arg *readdir)
 329 {
 330         unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
 331         __be32 *start, *p;
 332
 333         if (cookie > 2) {
 334                 readdir->cookie = cookie;
 335                 memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
 336                 return;
 337         }
 338
 339         readdir->cookie = 0;
 340         memset(&readdir->verifier, 0, sizeof(readdir->verifier));
 341         if (cookie == 2)
 342                 return;
 343
 344         /*
 345          * NFSv4 servers do not return entries for '.' and '..'
 346          * Therefore, we fake these entries here.  We let '.'
 347          * have cookie 0 and '..' have cookie 1.  Note that
 348          * when talking to the server, we always send cookie 0
 349          * instead of 1 or 2.
 350          */
 351         start = p = kmap_atomic(*readdir->pages);
 352
 353         if (cookie == 0) {
 354                 *p++ = xdr_one;                                  /* next */
 355                 *p++ = xdr_zero;                   /* cookie, first word */
 356                 *p++ = xdr_one;                   /* cookie, second word */
 357                 *p++ = xdr_one;                             /* entry len */
 358                 memcpy(p, ".\0\0\0", 4);                        /* entry */
 359                 p++;
 360                 *p++ = xdr_one;                         /* bitmap length */
 361                 *p++ = htonl(attrs);                           /* bitmap */
 362                 *p++ = htonl(12);             /* attribute buffer length */
 363                 *p++ = htonl(NF4DIR);
 364                 p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
 365         }
 366
 367         *p++ = xdr_one;                                  /* next */
 368         *p++ = xdr_zero;                   /* cookie, first word */
 369         *p++ = xdr_two;                   /* cookie, second word */
 370         *p++ = xdr_two;                             /* entry len */
 371         memcpy(p, "..\0\0", 4);                         /* entry */
 372         p++;
 373         *p++ = xdr_one;                         /* bitmap length */
 374         *p++ = htonl(attrs);                           /* bitmap */
 375         *p++ = htonl(12);             /* attribute buffer length */
 376         *p++ = htonl(NF4DIR);
 377         p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
 378
 379         readdir->pgbase = (char *)p - (char *)start;
 380         readdir->count -= readdir->pgbase;
 381         kunmap_atomic(start);
 382 }
 383
 384 static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
 385 {
 386         if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
 387                 fattr->pre_change_attr = version;
 388                 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
 389         }
 390 }
 391
 392 static void nfs4_test_and_free_stateid(struct nfs_server *server,
 393                 nfs4_stateid *stateid,
 394                 const struct cred *cred)
 395 {
 396         const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
 397
 398         ops->test_and_free_expired(server, stateid, cred);
 399 }
 400
 401 static void __nfs4_free_revoked_stateid(struct nfs_server *server,
 402                 nfs4_stateid *stateid,
 403                 const struct cred *cred)
 404 {
 405         stateid->type = NFS4_REVOKED_STATEID_TYPE;
 406         nfs4_test_and_free_stateid(server, stateid, cred);
 407 }
 408
 409 static void nfs4_free_revoked_stateid(struct nfs_server *server,
 410                 const nfs4_stateid *stateid,
 411                 const struct cred *cred)
 412 {
 413         nfs4_stateid tmp;
 414
 415         nfs4_stateid_copy(&tmp, stateid);
 416         __nfs4_free_revoked_stateid(server, &tmp, cred);
 417 }
 418
 419 static long nfs4_update_delay(long *timeout)
 420 {
 421         long ret;
 422         if (!timeout)
 423                 return NFS4_POLL_RETRY_MAX;
 424         if (*timeout <= 0)
 425                 *timeout = NFS4_POLL_RETRY_MIN;
 426         if (*timeout > NFS4_POLL_RETRY_MAX)
 427                 *timeout = NFS4_POLL_RETRY_MAX;
 428         ret = *timeout;
 429         *timeout <<= 1;
 430         return ret;
 431 }
 432
 433 static int nfs4_delay_killable(long *timeout)
 434 {
 435         might_sleep();
 436
 437         __set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
 438         schedule_timeout(nfs4_update_delay(timeout));
 439         if (!__fatal_signal_pending(current))
 440                 return 0;
 441         return -EINTR;
 442 }
 443
 444 static int nfs4_delay_interruptible(long *timeout)
 445 {
 446         might_sleep();
 447
 448         __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
 449         schedule_timeout(nfs4_update_delay(timeout));
 450         if (!signal_pending(current))
 451                 return 0;
 452         return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
 453 }
 454
 455 static int nfs4_delay(long *timeout, bool interruptible)
 456 {
 457         if (interruptible)
 458                 return nfs4_delay_interruptible(timeout);
 459         return nfs4_delay_killable(timeout);
 460 }
 461
 462 static const nfs4_stateid *
 463 nfs4_recoverable_stateid(const nfs4_stateid *stateid)
 464 {
 465         if (!stateid)
 466                 return NULL;
 467         switch (stateid->type) {
 468         case NFS4_OPEN_STATEID_TYPE:
 469         case NFS4_LOCK_STATEID_TYPE:
 470         case NFS4_DELEGATION_STATEID_TYPE:
 471                 return stateid;
 472         default:
 473                 break;
 474         }
 475         return NULL;
 476 }
 477
 478 /* This is the error handling routine for processes that are allowed
 479  * to sleep.
 480  */
 481 static int nfs4_do_handle_exception(struct nfs_server *server,
 482                 int errorcode, struct nfs4_exception *exception)
 483 {
 484         struct nfs_client *clp = server->nfs_client;
 485         struct nfs4_state *state = exception->state;
 486         const nfs4_stateid *stateid;
 487         struct inode *inode = exception->inode;
 488         int ret = errorcode;
 489
 490         exception->delay = 0;
 491         exception->recovering = 0;
 492         exception->retry = 0;
 493
 494         stateid = nfs4_recoverable_stateid(exception->stateid);
 495         if (stateid == NULL && state != NULL)
 496                 stateid = nfs4_recoverable_stateid(&state->stateid);
 497
 498         switch(errorcode) {
 499                 case 0:
 500                         return 0;
 501                 case -NFS4ERR_BADHANDLE:
 502                 case -ESTALE:
 503                         if (inode != NULL && S_ISREG(inode->i_mode))
 504                                 pnfs_destroy_layout(NFS_I(inode));
 505                         break;
 506                 case -NFS4ERR_DELEG_REVOKED:
 507                 case -NFS4ERR_ADMIN_REVOKED:
 508                 case -NFS4ERR_EXPIRED:
 509                 case -NFS4ERR_BAD_STATEID:
 510                 case -NFS4ERR_PARTNER_NO_AUTH:
 511                         if (inode != NULL && stateid != NULL) {
 512                                 nfs_inode_find_state_and_recover(inode,
 513                                                 stateid);
 514                                 goto wait_on_recovery;
 515                         }
 516                         fallthrough;
 517                 case -NFS4ERR_OPENMODE:
 518                         if (inode) {
 519                                 int err;
 520
 521                                 err = nfs_async_inode_return_delegation(inode,
 522                                                 stateid);
 523                                 if (err == 0)
 524                                         goto wait_on_recovery;
 525                                 if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
 526                                         exception->retry = 1;
 527                                         break;
 528                                 }
 529                         }
 530                         if (state == NULL)
 531                                 break;
 532                         ret = nfs4_schedule_stateid_recovery(server, state);
 533                         if (ret < 0)
 534                                 break;
 535                         goto wait_on_recovery;
 536                 case -NFS4ERR_STALE_STATEID:
 537                 case -NFS4ERR_STALE_CLIENTID:
 538                         nfs4_schedule_lease_recovery(clp);
 539                         goto wait_on_recovery;
 540                 case -NFS4ERR_MOVED:
 541                         ret = nfs4_schedule_migration_recovery(server);
 542                         if (ret < 0)
 543                                 break;
 544                         goto wait_on_recovery;
 545                 case -NFS4ERR_LEASE_MOVED:
 546                         nfs4_schedule_lease_moved_recovery(clp);
 547                         goto wait_on_recovery;
 548 #if defined(CONFIG_NFS_V4_1)
 549                 case -NFS4ERR_BADSESSION:
 550                 case -NFS4ERR_BADSLOT:
 551                 case -NFS4ERR_BAD_HIGH_SLOT:
 552                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 553                 case -NFS4ERR_DEADSESSION:
 554                 case -NFS4ERR_SEQ_FALSE_RETRY:
 555                 case -NFS4ERR_SEQ_MISORDERED:
 556                         /* Handled in nfs41_sequence_process() */
 557                         goto wait_on_recovery;
 558 #endif /* defined(CONFIG_NFS_V4_1) */
 559                 case -NFS4ERR_FILE_OPEN:
 560                         if (exception->timeout > HZ) {
 561                                 /* We have retried a decent amount, time to
 562                                  * fail
 563                                  */
 564                                 ret = -EBUSY;
 565                                 break;
 566                         }
 567                         fallthrough;
 568                 case -NFS4ERR_DELAY:
 569                         nfs_inc_server_stats(server, NFSIOS_DELAY);
 570                         fallthrough;
 571                 case -NFS4ERR_GRACE:
 572                 case -NFS4ERR_LAYOUTTRYLATER:
 573                 case -NFS4ERR_RECALLCONFLICT:
 574                 case -NFS4ERR_RETURNCONFLICT:
 575                         exception->delay = 1;
 576                         return 0;
 577
 578                 case -NFS4ERR_RETRY_UNCACHED_REP:
 579                 case -NFS4ERR_OLD_STATEID:
 580                         exception->retry = 1;
 581                         break;
 582                 case -NFS4ERR_BADOWNER:
 583                         /* The following works around a Linux server bug! */
 584                 case -NFS4ERR_BADNAME:
 585                         if (server->caps & NFS_CAP_UIDGID_NOMAP) {
 586                                 server->caps &= ~NFS_CAP_UIDGID_NOMAP;
 587                                 exception->retry = 1;
 588                                 printk(KERN_WARNING "NFS: v4 server %s "
 589                                                 "does not accept raw "
 590                                                 "uid/gids. "
 591                                                 "Reenabling the idmapper.\n",
 592                                                 server->nfs_client->cl_hostname);
 593                         }
 594         }
 595         /* We failed to handle the error */
 596         return nfs4_map_errors(ret);
 597 wait_on_recovery:
 598         exception->recovering = 1;
 599         return 0;
 600 }
 601
 602 /*
 603  * Track the number of NFS4ERR_DELAY related retransmissions and return
 604  * EAGAIN if the 'softerr' mount option is set, and we've exceeded the limit
 605  * set by 'nfs_delay_retrans'.
 606  */
 607 static int nfs4_exception_should_retrans(const struct nfs_server *server,
 608                                          struct nfs4_exception *exception)
 609 {
 610         if (server->flags & NFS_MOUNT_SOFTERR && nfs_delay_retrans >= 0) {
 611                 if (exception->retrans++ >= (unsigned short)nfs_delay_retrans)
 612                         return -EAGAIN;
 613         }
 614         return 0;
 615 }
 616
 617 /* This is the error handling routine for processes that are allowed
 618  * to sleep.
 619  */
 620 int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
 621 {
 622         struct nfs_client *clp = server->nfs_client;
 623         int ret;
 624
 625         ret = nfs4_do_handle_exception(server, errorcode, exception);
 626         if (exception->delay) {
 627                 int ret2 = nfs4_exception_should_retrans(server, exception);
 628                 if (ret2 < 0) {
 629                         exception->retry = 0;
 630                         return ret2;
 631                 }
 632                 ret = nfs4_delay(&exception->timeout,
 633                                 exception->interruptible);
 634                 goto out_retry;
 635         }
 636         if (exception->recovering) {
 637                 if (exception->task_is_privileged)
 638                         return -EDEADLOCK;
 639                 ret = nfs4_wait_clnt_recover(clp);
 640                 if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 641                         return -EIO;
 642                 goto out_retry;
 643         }
 644         return ret;
 645 out_retry:
 646         if (ret == 0)
 647                 exception->retry = 1;
 648         return ret;
 649 }
 650
 651 static int
 652 nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
 653                 int errorcode, struct nfs4_exception *exception)
 654 {
 655         struct nfs_client *clp = server->nfs_client;
 656         int ret;
 657
 658         ret = nfs4_do_handle_exception(server, errorcode, exception);
 659         if (exception->delay) {
 660                 int ret2 = nfs4_exception_should_retrans(server, exception);
 661                 if (ret2 < 0) {
 662                         exception->retry = 0;
 663                         return ret2;
 664                 }
 665                 rpc_delay(task, nfs4_update_delay(&exception->timeout));
 666                 goto out_retry;
 667         }
 668         if (exception->recovering) {
 669                 if (exception->task_is_privileged)
 670                         return -EDEADLOCK;
 671                 rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
 672                 if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
 673                         rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
 674                 goto out_retry;
 675         }
 676         if (test_bit(NFS_MIG_FAILED, &server->mig_status))
 677                 ret = -EIO;
 678         return ret;
 679 out_retry:
 680         if (ret == 0) {
 681                 exception->retry = 1;
 682                 /*
 683                  * For NFS4ERR_MOVED, the client transport will need to
 684                  * be recomputed after migration recovery has completed.
 685                  */
 686                 if (errorcode == -NFS4ERR_MOVED)
 687                         rpc_task_release_transport(task);
 688         }
 689         return ret;
 690 }
 691
 692 int
 693 nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
 694                         struct nfs4_state *state, long *timeout)
 695 {
 696         struct nfs4_exception exception = {
 697                 .state = state,
 698         };
 699
 700         if (task->tk_status >= 0)
 701                 return 0;
 702         if (timeout)
 703                 exception.timeout = *timeout;
 704         task->tk_status = nfs4_async_handle_exception(task, server,
 705                         task->tk_status,
 706                         &exception);
 707         if (exception.delay && timeout)
 708                 *timeout = exception.timeout;
 709         if (exception.retry)
 710                 return -EAGAIN;
 711         return 0;
 712 }
 713
 714 /*
 715  * Return 'true' if 'clp' is using an rpc_client that is integrity protected
 716  * or 'false' otherwise.
 717  */
 718 static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
 719 {
 720         rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
 721         return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
 722 }
 723
 724 static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
 725 {
 726         spin_lock(&clp->cl_lock);
 727         if (time_before(clp->cl_last_renewal,timestamp))
 728                 clp->cl_last_renewal = timestamp;
 729         spin_unlock(&clp->cl_lock);
 730 }
 731
 732 static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
 733 {
 734         struct nfs_client *clp = server->nfs_client;
 735
 736         if (!nfs4_has_session(clp))
 737                 do_renew_lease(clp, timestamp);
 738 }
 739
 740 struct nfs4_call_sync_data {
 741         const struct nfs_server *seq_server;
 742         struct nfs4_sequence_args *seq_args;
 743         struct nfs4_sequence_res *seq_res;
 744 };
 745
 746 void nfs4_init_sequence(struct nfs4_sequence_args *args,
 747                         struct nfs4_sequence_res *res, int cache_reply,
 748                         int privileged)
 749 {
 750         args->sa_slot = NULL;
 751         args->sa_cache_this = cache_reply;
 752         args->sa_privileged = privileged;
 753
 754         res->sr_slot = NULL;
 755 }
 756
 757 static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
 758 {
 759         struct nfs4_slot *slot = res->sr_slot;
 760         struct nfs4_slot_table *tbl;
 761
 762         tbl = slot->table;
 763         spin_lock(&tbl->slot_tbl_lock);
 764         if (!nfs41_wake_and_assign_slot(tbl, slot))
 765                 nfs4_free_slot(tbl, slot);
 766         spin_unlock(&tbl->slot_tbl_lock);
 767
 768         res->sr_slot = NULL;
 769 }
 770
 771 static int nfs40_sequence_done(struct rpc_task *task,
 772                                struct nfs4_sequence_res *res)
 773 {
 774         if (res->sr_slot != NULL)
 775                 nfs40_sequence_free_slot(res);
 776         return 1;
 777 }
 778
 779 #if defined(CONFIG_NFS_V4_1)
 780
 781 static void nfs41_release_slot(struct nfs4_slot *slot)
 782 {
 783         struct nfs4_session *session;
 784         struct nfs4_slot_table *tbl;
 785         bool send_new_highest_used_slotid = false;
 786
 787         if (!slot)
 788                 return;
 789         tbl = slot->table;
 790         session = tbl->session;
 791
 792         /* Bump the slot sequence number */
 793         if (slot->seq_done)
 794                 slot->seq_nr++;
 795         slot->seq_done = 0;
 796
 797         spin_lock(&tbl->slot_tbl_lock);
 798         /* Be nice to the server: try to ensure that the last transmitted
 799          * value for highest_user_slotid <= target_highest_slotid
 800          */
 801         if (tbl->highest_used_slotid > tbl->target_highest_slotid)
 802                 send_new_highest_used_slotid = true;
 803
 804         if (nfs41_wake_and_assign_slot(tbl, slot)) {
 805                 send_new_highest_used_slotid = false;
 806                 goto out_unlock;
 807         }
 808         nfs4_free_slot(tbl, slot);
 809
 810         if (tbl->highest_used_slotid != NFS4_NO_SLOT)
 811                 send_new_highest_used_slotid = false;
 812 out_unlock:
 813         spin_unlock(&tbl->slot_tbl_lock);
 814         if (send_new_highest_used_slotid)
 815                 nfs41_notify_server(session->clp);
 816         if (waitqueue_active(&tbl->slot_waitq))
 817                 wake_up_all(&tbl->slot_waitq);
 818 }
 819
 820 static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
 821 {
 822         nfs41_release_slot(res->sr_slot);
 823         res->sr_slot = NULL;
 824 }
 825
 826 static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
 827                 u32 seqnr)
 828 {
 829         if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
 830                 slot->seq_nr_highest_sent = seqnr;
 831 }
 832 static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
 833 {
 834         nfs4_slot_sequence_record_sent(slot, seqnr);
 835         slot->seq_nr_last_acked = seqnr;
 836 }
 837
 838 static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
 839                                 struct nfs4_slot *slot)
 840 {
 841         struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
 842         if (!IS_ERR(task))
 843                 rpc_put_task_async(task);
 844 }
 845
 846 static int nfs41_sequence_process(struct rpc_task *task,
 847                 struct nfs4_sequence_res *res)
 848 {
 849         struct nfs4_session *session;
 850         struct nfs4_slot *slot = res->sr_slot;
 851         struct nfs_client *clp;
 852         int status;
 853         int ret = 1;
 854
 855         if (slot == NULL)
 856                 goto out_noaction;
 857         /* don't increment the sequence number if the task wasn't sent */
 858         if (!RPC_WAS_SENT(task) || slot->seq_done)
 859                 goto out;
 860
 861         session = slot->table->session;
 862         clp = session->clp;
 863
 864         trace_nfs4_sequence_done(session, res);
 865
 866         status = res->sr_status;
 867         if (task->tk_status == -NFS4ERR_DEADSESSION)
 868                 status = -NFS4ERR_DEADSESSION;
 869
 870         /* Check the SEQUENCE operation status */
 871         switch (status) {
 872         case 0:
 873                 /* Mark this sequence number as having been acked */
 874                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 875                 /* Update the slot's sequence and clientid lease timer */
 876                 slot->seq_done = 1;
 877                 do_renew_lease(clp, res->sr_timestamp);
 878                 /* Check sequence flags */
 879                 nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
 880                                 !!slot->privileged);
 881                 nfs41_update_target_slotid(slot->table, slot, res);
 882                 break;
 883         case 1:
 884                 /*
 885                  * sr_status remains 1 if an RPC level error occurred.
 886                  * The server may or may not have processed the sequence
 887                  * operation..
 888                  */
 889                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 890                 slot->seq_done = 1;
 891                 goto out;
 892         case -NFS4ERR_DELAY:
 893                 /* The server detected a resend of the RPC call and
 894                  * returned NFS4ERR_DELAY as per Section 2.10.6.2
 895                  * of RFC5661.
 896                  */
 897                 dprintk("%s: slot=%u seq=%u: Operation in progress\n",
 898                         __func__,
 899                         slot->slot_nr,
 900                         slot->seq_nr);
 901                 goto out_retry;
 902         case -NFS4ERR_RETRY_UNCACHED_REP:
 903         case -NFS4ERR_SEQ_FALSE_RETRY:
 904                 /*
 905                  * The server thinks we tried to replay a request.
 906                  * Retry the call after bumping the sequence ID.
 907                  */
 908                 nfs4_slot_sequence_acked(slot, slot->seq_nr);
 909                 goto retry_new_seq;
 910         case -NFS4ERR_BADSLOT:
 911                 /*
 912                  * The slot id we used was probably retired. Try again
 913                  * using a different slot id.
 914                  */
 915                 if (slot->slot_nr < slot->table->target_highest_slotid)
 916                         goto session_recover;
 917                 goto retry_nowait;
 918         case -NFS4ERR_SEQ_MISORDERED:
 919                 nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
 920                 /*
 921                  * Were one or more calls using this slot interrupted?
 922                  * If the server never received the request, then our
 923                  * transmitted slot sequence number may be too high. However,
 924                  * if the server did receive the request then it might
 925                  * accidentally give us a reply with a mismatched operation.
 926                  * We can sort this out by sending a lone sequence operation
 927                  * to the server on the same slot.
 928                  */
 929                 if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
 930                         slot->seq_nr--;
 931                         if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
 932                                 nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
 933                                 res->sr_slot = NULL;
 934                         }
 935                         goto retry_nowait;
 936                 }
 937                 /*
 938                  * RFC5661:
 939                  * A retry might be sent while the original request is
 940                  * still in progress on the replier. The replier SHOULD
 941                  * deal with the issue by returning NFS4ERR_DELAY as the
 942                  * reply to SEQUENCE or CB_SEQUENCE operation, but
 943                  * implementations MAY return NFS4ERR_SEQ_MISORDERED.
 944                  *
 945                  * Restart the search after a delay.
 946                  */
 947                 slot->seq_nr = slot->seq_nr_highest_sent;
 948                 goto out_retry;
 949         case -NFS4ERR_BADSESSION:
 950         case -NFS4ERR_DEADSESSION:
 951         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
 952                 goto session_recover;
 953         default:
 954                 /* Just update the slot sequence no. */
 955                 slot->seq_done = 1;
 956         }
 957 out:
 958         /* The session may be reset by one of the error handlers. */
 959         dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
 960 out_noaction:
 961         return ret;
 962 session_recover:
 963         set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
 964         nfs4_schedule_session_recovery(session, status);
 965         dprintk("%s ERROR: %d Reset session\n", __func__, status);
 966         nfs41_sequence_free_slot(res);
 967         goto out;
 968 retry_new_seq:
 969         ++slot->seq_nr;
 970 retry_nowait:
 971         if (rpc_restart_call_prepare(task)) {
 972                 nfs41_sequence_free_slot(res);
 973                 task->tk_status = 0;
 974                 ret = 0;
 975         }
 976         goto out;
 977 out_retry:
 978         if (!rpc_restart_call(task))
 979                 goto out;
 980         rpc_delay(task, NFS4_POLL_RETRY_MAX);
 981         return 0;
 982 }
 983
 984 int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
 985 {
 986         if (!nfs41_sequence_process(task, res))
 987                 return 0;
 988         if (res->sr_slot != NULL)
 989                 nfs41_sequence_free_slot(res);
 990         return 1;
 991
 992 }
 993 EXPORT_SYMBOL_GPL(nfs41_sequence_done);
 994
 995 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
 996 {
 997         if (res->sr_slot == NULL)
 998                 return 1;
 999         if (res->sr_slot->table->session != NULL)
1000                 return nfs41_sequence_process(task, res);
1001         return nfs40_sequence_done(task, res);
1002 }
1003
1004 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1005 {
1006         if (res->sr_slot != NULL) {
1007                 if (res->sr_slot->table->session != NULL)
1008                         nfs41_sequence_free_slot(res);
1009                 else
1010                         nfs40_sequence_free_slot(res);
1011         }
1012 }
1013
1014 int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
1015 {
1016         if (res->sr_slot == NULL)
1017                 return 1;
1018         if (!res->sr_slot->table->session)
1019                 return nfs40_sequence_done(task, res);
1020         return nfs41_sequence_done(task, res);
1021 }
1022 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1023
1024 static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
1025 {
1026         struct nfs4_call_sync_data *data = calldata;
1027
1028         dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
1029
1030         nfs4_setup_sequence(data->seq_server->nfs_client,
1031                             data->seq_args, data->seq_res, task);
1032 }
1033
1034 static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
1035 {
1036         struct nfs4_call_sync_data *data = calldata;
1037
1038         nfs41_sequence_done(task, data->seq_res);
1039 }
1040
1041 static const struct rpc_call_ops nfs41_call_sync_ops = {
1042         .rpc_call_prepare = nfs41_call_sync_prepare,
1043         .rpc_call_done = nfs41_call_sync_done,
1044 };
1045
1046 #else   /* !CONFIG_NFS_V4_1 */
1047
1048 static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1049 {
1050         return nfs40_sequence_done(task, res);
1051 }
1052
1053 static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1054 {
1055         if (res->sr_slot != NULL)
1056                 nfs40_sequence_free_slot(res);
1057 }
1058
1059 int nfs4_sequence_done(struct rpc_task *task,
1060                        struct nfs4_sequence_res *res)
1061 {
1062         return nfs40_sequence_done(task, res);
1063 }
1064 EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1065
1066 #endif  /* !CONFIG_NFS_V4_1 */
1067
1068 static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1069 {
1070         res->sr_timestamp = jiffies;
1071         res->sr_status_flags = 0;
1072         res->sr_status = 1;
1073 }
1074
1075 static
1076 void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1077                 struct nfs4_sequence_res *res,
1078                 struct nfs4_slot *slot)
1079 {
1080         if (!slot)
1081                 return;
1082         slot->privileged = args->sa_privileged ? 1 : 0;
1083         args->sa_slot = slot;
1084
1085         res->sr_slot = slot;
1086 }
1087
1088 int nfs4_setup_sequence(struct nfs_client *client,
1089                         struct nfs4_sequence_args *args,
1090                         struct nfs4_sequence_res *res,
1091                         struct rpc_task *task)
1092 {
1093         struct nfs4_session *session = nfs4_get_session(client);
1094         struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1095         struct nfs4_slot *slot;
1096
1097         /* slot already allocated? */
1098         if (res->sr_slot != NULL)
1099                 goto out_start;
1100
1101         if (session)
1102                 tbl = &session->fc_slot_table;
1103
1104         spin_lock(&tbl->slot_tbl_lock);
1105         /* The state manager will wait until the slot table is empty */
1106         if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1107                 goto out_sleep;
1108
1109         slot = nfs4_alloc_slot(tbl);
1110         if (IS_ERR(slot)) {
1111                 if (slot == ERR_PTR(-ENOMEM))
1112                         goto out_sleep_timeout;
1113                 goto out_sleep;
1114         }
1115         spin_unlock(&tbl->slot_tbl_lock);
1116
1117         nfs4_sequence_attach_slot(args, res, slot);
1118
1119         trace_nfs4_setup_sequence(session, args);
1120 out_start:
1121         nfs41_sequence_res_init(res);
1122         rpc_call_start(task);
1123         return 0;
1124 out_sleep_timeout:
1125         /* Try again in 1/4 second */
1126         if (args->sa_privileged)
1127                 rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1128                                 jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1129         else
1130                 rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1131                                 NULL, jiffies + (HZ >> 2));
1132         spin_unlock(&tbl->slot_tbl_lock);
1133         return -EAGAIN;
1134 out_sleep:
1135         if (args->sa_privileged)
1136                 rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1137                                 RPC_PRIORITY_PRIVILEGED);
1138         else
1139                 rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1140         spin_unlock(&tbl->slot_tbl_lock);
1141         return -EAGAIN;
1142 }
1143 EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1144
1145 static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1146 {
1147         struct nfs4_call_sync_data *data = calldata;
1148         nfs4_setup_sequence(data->seq_server->nfs_client,
1149                                 data->seq_args, data->seq_res, task);
1150 }
1151
1152 static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1153 {
1154         struct nfs4_call_sync_data *data = calldata;
1155         nfs4_sequence_done(task, data->seq_res);
1156 }
1157
1158 static const struct rpc_call_ops nfs40_call_sync_ops = {
1159         .rpc_call_prepare = nfs40_call_sync_prepare,
1160         .rpc_call_done = nfs40_call_sync_done,
1161 };
1162
1163 static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1164 {
1165         int ret;
1166         struct rpc_task *task;
1167
1168         task = rpc_run_task(task_setup);
1169         if (IS_ERR(task))
1170                 return PTR_ERR(task);
1171
1172         ret = task->tk_status;
1173         rpc_put_task(task);
1174         return ret;
1175 }
1176
1177 static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1178                              struct nfs_server *server,
1179                              struct rpc_message *msg,
1180                              struct nfs4_sequence_args *args,
1181                              struct nfs4_sequence_res *res,
1182                              unsigned short task_flags)
1183 {
1184         struct nfs_client *clp = server->nfs_client;
1185         struct nfs4_call_sync_data data = {
1186                 .seq_server = server,
1187                 .seq_args = args,
1188                 .seq_res = res,
1189         };
1190         struct rpc_task_setup task_setup = {
1191                 .rpc_client = clnt,
1192                 .rpc_message = msg,
1193                 .callback_ops = clp->cl_mvops->call_sync_ops,
1194                 .callback_data = &data,
1195                 .flags = task_flags,
1196         };
1197
1198         return nfs4_call_sync_custom(&task_setup);
1199 }
1200
1201 static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1202                                    struct nfs_server *server,
1203                                    struct rpc_message *msg,
1204                                    struct nfs4_sequence_args *args,
1205                                    struct nfs4_sequence_res *res)
1206 {
1207         unsigned short task_flags = 0;
1208
1209         if (server->caps & NFS_CAP_MOVEABLE)
1210                 task_flags = RPC_TASK_MOVEABLE;
1211         return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1212 }
1213
1214
1215 int nfs4_call_sync(struct rpc_clnt *clnt,
1216                    struct nfs_server *server,
1217                    struct rpc_message *msg,
1218                    struct nfs4_sequence_args *args,
1219                    struct nfs4_sequence_res *res,
1220                    int cache_reply)
1221 {
1222         nfs4_init_sequence(args, res, cache_reply, 0);
1223         return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1224 }
1225
1226 static void
1227 nfs4_inc_nlink_locked(struct inode *inode)
1228 {
1229         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1230                                              NFS_INO_INVALID_CTIME |
1231                                              NFS_INO_INVALID_NLINK);
1232         inc_nlink(inode);
1233 }
1234
1235 static void
1236 nfs4_inc_nlink(struct inode *inode)
1237 {
1238         spin_lock(&inode->i_lock);
1239         nfs4_inc_nlink_locked(inode);
1240         spin_unlock(&inode->i_lock);
1241 }
1242
1243 static void
1244 nfs4_dec_nlink_locked(struct inode *inode)
1245 {
1246         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1247                                              NFS_INO_INVALID_CTIME |
1248                                              NFS_INO_INVALID_NLINK);
1249         drop_nlink(inode);
1250 }
1251
1252 static void
1253 nfs4_update_changeattr_locked(struct inode *inode,
1254                 struct nfs4_change_info *cinfo,
1255                 unsigned long timestamp, unsigned long cache_validity)
1256 {
1257         struct nfs_inode *nfsi = NFS_I(inode);
1258         u64 change_attr = inode_peek_iversion_raw(inode);
1259
1260         if (!nfs_have_delegated_mtime(inode))
1261                 cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1262         if (S_ISDIR(inode->i_mode))
1263                 cache_validity |= NFS_INO_INVALID_DATA;
1264
1265         switch (NFS_SERVER(inode)->change_attr_type) {
1266         case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1267                 if (cinfo->after == change_attr)
1268                         goto out;
1269                 break;
1270         default:
1271                 if ((s64)(change_attr - cinfo->after) >= 0)
1272                         goto out;
1273         }
1274
1275         inode_set_iversion_raw(inode, cinfo->after);
1276         if (!cinfo->atomic || cinfo->before != change_attr) {
1277                 if (S_ISDIR(inode->i_mode))
1278                         nfs_force_lookup_revalidate(inode);
1279
1280                 if (!nfs_have_delegated_attributes(inode))
1281                         cache_validity |=
1282                                 NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1283                                 NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1284                                 NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1285                                 NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1286                 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1287         }
1288         nfsi->attrtimeo_timestamp = jiffies;
1289         nfsi->read_cache_jiffies = timestamp;
1290         nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1291         nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1292 out:
1293         nfs_set_cache_invalid(inode, cache_validity);
1294 }
1295
1296 void
1297 nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1298                 unsigned long timestamp, unsigned long cache_validity)
1299 {
1300         spin_lock(&dir->i_lock);
1301         nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1302         spin_unlock(&dir->i_lock);
1303 }
1304
1305 struct nfs4_open_createattrs {
1306         struct nfs4_label *label;
1307         struct iattr *sattr;
1308         const __u32 verf[2];
1309 };
1310
1311 static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1312                 int err, struct nfs4_exception *exception)
1313 {
1314         if (err != -EINVAL)
1315                 return false;
1316         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1317                 return false;
1318         server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1319         exception->retry = 1;
1320         return true;
1321 }
1322
1323 static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1324 {
1325          return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1326 }
1327
1328 static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1329 {
1330         fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1331
1332         return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1333 }
1334
1335 static u32
1336 nfs4_fmode_to_share_access(fmode_t fmode)
1337 {
1338         u32 res = 0;
1339
1340         switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1341         case FMODE_READ:
1342                 res = NFS4_SHARE_ACCESS_READ;
1343                 break;
1344         case FMODE_WRITE:
1345                 res = NFS4_SHARE_ACCESS_WRITE;
1346                 break;
1347         case FMODE_READ|FMODE_WRITE:
1348                 res = NFS4_SHARE_ACCESS_BOTH;
1349         }
1350         return res;
1351 }
1352
1353 static u32
1354 nfs4_map_atomic_open_share(struct nfs_server *server,
1355                 fmode_t fmode, int openflags)
1356 {
1357         u32 res = nfs4_fmode_to_share_access(fmode);
1358
1359         if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1360                 goto out;
1361         /* Want no delegation if we're using O_DIRECT */
1362         if (openflags & O_DIRECT) {
1363                 res |= NFS4_SHARE_WANT_NO_DELEG;
1364                 goto out;
1365         }
1366         /* res |= NFS4_SHARE_WANT_NO_PREFERENCE; */
1367         if (server->caps & NFS_CAP_DELEGTIME)
1368                 res |= NFS4_SHARE_WANT_DELEG_TIMESTAMPS;
1369         if (server->caps & NFS_CAP_OPEN_XOR)
1370                 res |= NFS4_SHARE_WANT_OPEN_XOR_DELEGATION;
1371 out:
1372         return res;
1373 }
1374
1375 static enum open_claim_type4
1376 nfs4_map_atomic_open_claim(struct nfs_server *server,
1377                 enum open_claim_type4 claim)
1378 {
1379         if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1380                 return claim;
1381         switch (claim) {
1382         default:
1383                 return claim;
1384         case NFS4_OPEN_CLAIM_FH:
1385                 return NFS4_OPEN_CLAIM_NULL;
1386         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1387                 return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1388         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1389                 return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1390         }
1391 }
1392
1393 static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1394 {
1395         p->o_res.f_attr = &p->f_attr;
1396         p->o_res.seqid = p->o_arg.seqid;
1397         p->c_res.seqid = p->c_arg.seqid;
1398         p->o_res.server = p->o_arg.server;
1399         p->o_res.access_request = p->o_arg.access;
1400         nfs_fattr_init(&p->f_attr);
1401         nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1402 }
1403
1404 static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1405                 struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1406                 const struct nfs4_open_createattrs *c,
1407                 enum open_claim_type4 claim,
1408                 gfp_t gfp_mask)
1409 {
1410         struct dentry *parent = dget_parent(dentry);
1411         struct inode *dir = d_inode(parent);
1412         struct nfs_server *server = NFS_SERVER(dir);
1413         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1414         struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1415         struct nfs4_opendata *p;
1416
1417         p = kzalloc(sizeof(*p), gfp_mask);
1418         if (p == NULL)
1419                 goto err;
1420
1421         p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1422         if (IS_ERR(p->f_attr.label))
1423                 goto err_free_p;
1424
1425         p->a_label = nfs4_label_alloc(server, gfp_mask);
1426         if (IS_ERR(p->a_label))
1427                 goto err_free_f;
1428
1429         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1430         p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1431         if (IS_ERR(p->o_arg.seqid))
1432                 goto err_free_label;
1433         nfs_sb_active(dentry->d_sb);
1434         p->dentry = dget(dentry);
1435         p->dir = parent;
1436         p->owner = sp;
1437         atomic_inc(&sp->so_count);
1438         p->o_arg.open_flags = flags;
1439         p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1440         p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1441         p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1442                         fmode, flags);
1443         if (flags & O_CREAT) {
1444                 p->o_arg.umask = current_umask();
1445                 p->o_arg.label = nfs4_label_copy(p->a_label, label);
1446                 if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1447                         p->o_arg.u.attrs = &p->attrs;
1448                         memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1449
1450                         memcpy(p->o_arg.u.verifier.data, c->verf,
1451                                         sizeof(p->o_arg.u.verifier.data));
1452                 }
1453         }
1454         /* ask server to check for all possible rights as results
1455          * are cached */
1456         switch (p->o_arg.claim) {
1457         default:
1458                 break;
1459         case NFS4_OPEN_CLAIM_NULL:
1460         case NFS4_OPEN_CLAIM_FH:
1461                 p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1462                                   NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1463                                   NFS4_ACCESS_EXECUTE |
1464                                   nfs_access_xattr_mask(server);
1465         }
1466         p->o_arg.clientid = server->nfs_client->cl_clientid;
1467         p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1468         p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1469         p->o_arg.name = &dentry->d_name;
1470         p->o_arg.server = server;
1471         p->o_arg.bitmask = nfs4_bitmask(server, label);
1472         p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1473         switch (p->o_arg.claim) {
1474         case NFS4_OPEN_CLAIM_NULL:
1475         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1476         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1477                 p->o_arg.fh = NFS_FH(dir);
1478                 break;
1479         case NFS4_OPEN_CLAIM_PREVIOUS:
1480         case NFS4_OPEN_CLAIM_FH:
1481         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1482         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1483                 p->o_arg.fh = NFS_FH(d_inode(dentry));
1484         }
1485         p->c_arg.fh = &p->o_res.fh;
1486         p->c_arg.stateid = &p->o_res.stateid;
1487         p->c_arg.seqid = p->o_arg.seqid;
1488         nfs4_init_opendata_res(p);
1489         kref_init(&p->kref);
1490         return p;
1491
1492 err_free_label:
1493         nfs4_label_free(p->a_label);
1494 err_free_f:
1495         nfs4_label_free(p->f_attr.label);
1496 err_free_p:
1497         kfree(p);
1498 err:
1499         dput(parent);
1500         return NULL;
1501 }
1502
1503 static void nfs4_opendata_free(struct kref *kref)
1504 {
1505         struct nfs4_opendata *p = container_of(kref,
1506                         struct nfs4_opendata, kref);
1507         struct super_block *sb = p->dentry->d_sb;
1508
1509         nfs4_lgopen_release(p->lgp);
1510         nfs_free_seqid(p->o_arg.seqid);
1511         nfs4_sequence_free_slot(&p->o_res.seq_res);
1512         if (p->state != NULL)
1513                 nfs4_put_open_state(p->state);
1514         nfs4_put_state_owner(p->owner);
1515
1516         nfs4_label_free(p->a_label);
1517         nfs4_label_free(p->f_attr.label);
1518
1519         dput(p->dir);
1520         dput(p->dentry);
1521         nfs_sb_deactive(sb);
1522         nfs_fattr_free_names(&p->f_attr);
1523         kfree(p->f_attr.mdsthreshold);
1524         kfree(p);
1525 }
1526
1527 static void nfs4_opendata_put(struct nfs4_opendata *p)
1528 {
1529         if (p != NULL)
1530                 kref_put(&p->kref, nfs4_opendata_free);
1531 }
1532
1533 static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1534                 fmode_t fmode)
1535 {
1536         switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1537         case FMODE_READ|FMODE_WRITE:
1538                 return state->n_rdwr != 0;
1539         case FMODE_WRITE:
1540                 return state->n_wronly != 0;
1541         case FMODE_READ:
1542                 return state->n_rdonly != 0;
1543         }
1544         WARN_ON_ONCE(1);
1545         return false;
1546 }
1547
1548 static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1549                 int open_mode, enum open_claim_type4 claim)
1550 {
1551         int ret = 0;
1552
1553         if (open_mode & (O_EXCL|O_TRUNC))
1554                 goto out;
1555         switch (claim) {
1556         case NFS4_OPEN_CLAIM_NULL:
1557         case NFS4_OPEN_CLAIM_FH:
1558                 goto out;
1559         default:
1560                 break;
1561         }
1562         switch (mode & (FMODE_READ|FMODE_WRITE)) {
1563                 case FMODE_READ:
1564                         ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1565                                 && state->n_rdonly != 0;
1566                         break;
1567                 case FMODE_WRITE:
1568                         ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1569                                 && state->n_wronly != 0;
1570                         break;
1571                 case FMODE_READ|FMODE_WRITE:
1572                         ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1573                                 && state->n_rdwr != 0;
1574         }
1575 out:
1576         return ret;
1577 }
1578
1579 static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1580                 enum open_claim_type4 claim)
1581 {
1582         if (delegation == NULL)
1583                 return 0;
1584         if ((delegation->type & fmode) != fmode)
1585                 return 0;
1586         switch (claim) {
1587         case NFS4_OPEN_CLAIM_NULL:
1588         case NFS4_OPEN_CLAIM_FH:
1589                 break;
1590         case NFS4_OPEN_CLAIM_PREVIOUS:
1591                 if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1592                         break;
1593                 fallthrough;
1594         default:
1595                 return 0;
1596         }
1597         nfs_mark_delegation_referenced(delegation);
1598         return 1;
1599 }
1600
1601 static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1602 {
1603         switch (fmode) {
1604                 case FMODE_WRITE:
1605                         state->n_wronly++;
1606                         break;
1607                 case FMODE_READ:
1608                         state->n_rdonly++;
1609                         break;
1610                 case FMODE_READ|FMODE_WRITE:
1611                         state->n_rdwr++;
1612         }
1613         nfs4_state_set_mode_locked(state, state->state | fmode);
1614 }
1615
1616 #ifdef CONFIG_NFS_V4_1
1617 static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1618 {
1619         if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1620                 return true;
1621         if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1622                 return true;
1623         if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1624                 return true;
1625         return false;
1626 }
1627 #endif /* CONFIG_NFS_V4_1 */
1628
1629 static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1630 {
1631         if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1632                 wake_up_all(&state->waitq);
1633 }
1634
1635 static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1636 {
1637         struct nfs_client *clp = state->owner->so_server->nfs_client;
1638         bool need_recover = false;
1639
1640         if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1641                 need_recover = true;
1642         if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1643                 need_recover = true;
1644         if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1645                 need_recover = true;
1646         if (need_recover)
1647                 nfs4_state_mark_reclaim_nograce(clp, state);
1648 }
1649
1650 /*
1651  * Check for whether or not the caller may update the open stateid
1652  * to the value passed in by stateid.
1653  *
1654  * Note: This function relies heavily on the server implementing
1655  * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1656  * correctly.
1657  * i.e. The stateid seqids have to be initialised to 1, and
1658  * are then incremented on every state transition.
1659  */
1660 static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1661                 const nfs4_stateid *stateid)
1662 {
1663         if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1664                 /* The common case - we're updating to a new sequence number */
1665                 if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1666                         if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1667                                 return true;
1668                         return false;
1669                 }
1670                 /* The server returned a new stateid */
1671         }
1672         /* This is the first OPEN in this generation */
1673         if (stateid->seqid == cpu_to_be32(1))
1674                 return true;
1675         return false;
1676 }
1677
1678 static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1679 {
1680         if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1681                 return;
1682         if (state->n_wronly)
1683                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1684         if (state->n_rdonly)
1685                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1686         if (state->n_rdwr)
1687                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1688         set_bit(NFS_OPEN_STATE, &state->flags);
1689 }
1690
1691 static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1692                 nfs4_stateid *stateid, fmode_t fmode)
1693 {
1694         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1695         switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1696         case FMODE_WRITE:
1697                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1698                 break;
1699         case FMODE_READ:
1700                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1701                 break;
1702         case 0:
1703                 clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1704                 clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1705                 clear_bit(NFS_OPEN_STATE, &state->flags);
1706         }
1707         if (stateid == NULL)
1708                 return;
1709         /* Handle OPEN+OPEN_DOWNGRADE races */
1710         if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1711             !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1712                 nfs_resync_open_stateid_locked(state);
1713                 goto out;
1714         }
1715         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1716                 nfs4_stateid_copy(&state->stateid, stateid);
1717         nfs4_stateid_copy(&state->open_stateid, stateid);
1718         trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1719 out:
1720         nfs_state_log_update_open_stateid(state);
1721 }
1722
1723 static void nfs_clear_open_stateid(struct nfs4_state *state,
1724         nfs4_stateid *arg_stateid,
1725         nfs4_stateid *stateid, fmode_t fmode)
1726 {
1727         write_seqlock(&state->seqlock);
1728         /* Ignore, if the CLOSE argment doesn't match the current stateid */
1729         if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1730                 nfs_clear_open_stateid_locked(state, stateid, fmode);
1731         write_sequnlock(&state->seqlock);
1732         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1733                 nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1734 }
1735
1736 static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1737                 const nfs4_stateid *stateid, nfs4_stateid *freeme)
1738         __must_hold(&state->owner->so_lock)
1739         __must_hold(&state->seqlock)
1740         __must_hold(RCU)
1741
1742 {
1743         DEFINE_WAIT(wait);
1744         int status = 0;
1745         for (;;) {
1746
1747                 if (nfs_stateid_is_sequential(state, stateid))
1748                         break;
1749
1750                 if (status)
1751                         break;
1752                 /* Rely on seqids for serialisation with NFSv4.0 */
1753                 if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1754                         break;
1755
1756                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1757                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1758                 /*
1759                  * Ensure we process the state changes in the same order
1760                  * in which the server processed them by delaying the
1761                  * update of the stateid until we are in sequence.
1762                  */
1763                 write_sequnlock(&state->seqlock);
1764                 spin_unlock(&state->owner->so_lock);
1765                 rcu_read_unlock();
1766                 trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1767
1768                 if (!fatal_signal_pending(current)) {
1769                         if (schedule_timeout(5*HZ) == 0)
1770                                 status = -EAGAIN;
1771                         else
1772                                 status = 0;
1773                 } else
1774                         status = -EINTR;
1775                 finish_wait(&state->waitq, &wait);
1776                 rcu_read_lock();
1777                 spin_lock(&state->owner->so_lock);
1778                 write_seqlock(&state->seqlock);
1779         }
1780
1781         if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1782             !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1783                 nfs4_stateid_copy(freeme, &state->open_stateid);
1784                 nfs_test_and_clear_all_open_stateid(state);
1785         }
1786
1787         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1788                 nfs4_stateid_copy(&state->stateid, stateid);
1789         nfs4_stateid_copy(&state->open_stateid, stateid);
1790         trace_nfs4_open_stateid_update(state->inode, stateid, status);
1791         nfs_state_log_update_open_stateid(state);
1792 }
1793
1794 static void nfs_state_set_open_stateid(struct nfs4_state *state,
1795                 const nfs4_stateid *open_stateid,
1796                 fmode_t fmode,
1797                 nfs4_stateid *freeme)
1798 {
1799         /*
1800          * Protect the call to nfs4_state_set_mode_locked and
1801          * serialise the stateid update
1802          */
1803         write_seqlock(&state->seqlock);
1804         nfs_set_open_stateid_locked(state, open_stateid, freeme);
1805         switch (fmode) {
1806         case FMODE_READ:
1807                 set_bit(NFS_O_RDONLY_STATE, &state->flags);
1808                 break;
1809         case FMODE_WRITE:
1810                 set_bit(NFS_O_WRONLY_STATE, &state->flags);
1811                 break;
1812         case FMODE_READ|FMODE_WRITE:
1813                 set_bit(NFS_O_RDWR_STATE, &state->flags);
1814         }
1815         set_bit(NFS_OPEN_STATE, &state->flags);
1816         write_sequnlock(&state->seqlock);
1817 }
1818
1819 static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1820 {
1821         clear_bit(NFS_O_RDWR_STATE, &state->flags);
1822         clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1823         clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1824         clear_bit(NFS_OPEN_STATE, &state->flags);
1825 }
1826
1827 static void nfs_state_set_delegation(struct nfs4_state *state,
1828                 const nfs4_stateid *deleg_stateid,
1829                 fmode_t fmode)
1830 {
1831         /*
1832          * Protect the call to nfs4_state_set_mode_locked and
1833          * serialise the stateid update
1834          */
1835         write_seqlock(&state->seqlock);
1836         nfs4_stateid_copy(&state->stateid, deleg_stateid);
1837         set_bit(NFS_DELEGATED_STATE, &state->flags);
1838         write_sequnlock(&state->seqlock);
1839 }
1840
1841 static void nfs_state_clear_delegation(struct nfs4_state *state)
1842 {
1843         write_seqlock(&state->seqlock);
1844         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1845         clear_bit(NFS_DELEGATED_STATE, &state->flags);
1846         write_sequnlock(&state->seqlock);
1847 }
1848
1849 int update_open_stateid(struct nfs4_state *state,
1850                 const nfs4_stateid *open_stateid,
1851                 const nfs4_stateid *delegation,
1852                 fmode_t fmode)
1853 {
1854         struct nfs_server *server = NFS_SERVER(state->inode);
1855         struct nfs_client *clp = server->nfs_client;
1856         struct nfs_inode *nfsi = NFS_I(state->inode);
1857         struct nfs_delegation *deleg_cur;
1858         nfs4_stateid freeme = { };
1859         int ret = 0;
1860
1861         fmode &= (FMODE_READ|FMODE_WRITE);
1862
1863         rcu_read_lock();
1864         spin_lock(&state->owner->so_lock);
1865         if (open_stateid != NULL) {
1866                 nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1867                 ret = 1;
1868         }
1869
1870         deleg_cur = nfs4_get_valid_delegation(state->inode);
1871         if (deleg_cur == NULL)
1872                 goto no_delegation;
1873
1874         spin_lock(&deleg_cur->lock);
1875         if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1876            test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1877             (deleg_cur->type & fmode) != fmode)
1878                 goto no_delegation_unlock;
1879
1880         if (delegation == NULL)
1881                 delegation = &deleg_cur->stateid;
1882         else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1883                 goto no_delegation_unlock;
1884
1885         nfs_mark_delegation_referenced(deleg_cur);
1886         nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1887         ret = 1;
1888 no_delegation_unlock:
1889         spin_unlock(&deleg_cur->lock);
1890 no_delegation:
1891         if (ret)
1892                 update_open_stateflags(state, fmode);
1893         spin_unlock(&state->owner->so_lock);
1894         rcu_read_unlock();
1895
1896         if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1897                 nfs4_schedule_state_manager(clp);
1898         if (freeme.type != 0)
1899                 nfs4_test_and_free_stateid(server, &freeme,
1900                                 state->owner->so_cred);
1901
1902         return ret;
1903 }
1904
1905 static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1906                 const nfs4_stateid *stateid)
1907 {
1908         struct nfs4_state *state = lsp->ls_state;
1909         bool ret = false;
1910
1911         spin_lock(&state->state_lock);
1912         if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1913                 goto out_noupdate;
1914         if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1915                 goto out_noupdate;
1916         nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1917         ret = true;
1918 out_noupdate:
1919         spin_unlock(&state->state_lock);
1920         return ret;
1921 }
1922
1923 static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1924 {
1925         struct nfs_delegation *delegation;
1926
1927         fmode &= FMODE_READ|FMODE_WRITE;
1928         rcu_read_lock();
1929         delegation = nfs4_get_valid_delegation(inode);
1930         if (delegation == NULL || (delegation->type & fmode) == fmode) {
1931                 rcu_read_unlock();
1932                 return;
1933         }
1934         rcu_read_unlock();
1935         nfs4_inode_return_delegation(inode);
1936 }
1937
1938 static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1939 {
1940         struct nfs4_state *state = opendata->state;
1941         struct nfs_delegation *delegation;
1942         int open_mode = opendata->o_arg.open_flags;
1943         fmode_t fmode = opendata->o_arg.fmode;
1944         enum open_claim_type4 claim = opendata->o_arg.claim;
1945         nfs4_stateid stateid;
1946         int ret = -EAGAIN;
1947
1948         for (;;) {
1949                 spin_lock(&state->owner->so_lock);
1950                 if (can_open_cached(state, fmode, open_mode, claim)) {
1951                         update_open_stateflags(state, fmode);
1952                         spin_unlock(&state->owner->so_lock);
1953                         goto out_return_state;
1954                 }
1955                 spin_unlock(&state->owner->so_lock);
1956                 rcu_read_lock();
1957                 delegation = nfs4_get_valid_delegation(state->inode);
1958                 if (!can_open_delegated(delegation, fmode, claim)) {
1959                         rcu_read_unlock();
1960                         break;
1961                 }
1962                 /* Save the delegation */
1963                 nfs4_stateid_copy(&stateid, &delegation->stateid);
1964                 rcu_read_unlock();
1965                 nfs_release_seqid(opendata->o_arg.seqid);
1966                 if (!opendata->is_recover) {
1967                         ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1968                         if (ret != 0)
1969                                 goto out;
1970                 }
1971                 ret = -EAGAIN;
1972
1973                 /* Try to update the stateid using the delegation */
1974                 if (update_open_stateid(state, NULL, &stateid, fmode))
1975                         goto out_return_state;
1976         }
1977 out:
1978         return ERR_PTR(ret);
1979 out_return_state:
1980         refcount_inc(&state->count);
1981         return state;
1982 }
1983
1984 static void
1985 nfs4_process_delegation(struct inode *inode, const struct cred *cred,
1986                         enum open_claim_type4 claim,
1987                         const struct nfs4_open_delegation *delegation)
1988 {
1989         switch (delegation->open_delegation_type) {
1990         case NFS4_OPEN_DELEGATE_READ:
1991         case NFS4_OPEN_DELEGATE_WRITE:
1992         case NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG:
1993         case NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG:
1994                 break;
1995         default:
1996                 return;
1997         }
1998         switch (claim) {
1999         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2000         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2001                 pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
2002                                    "returning a delegation for "
2003                                    "OPEN(CLAIM_DELEGATE_CUR)\n",
2004                                    NFS_SERVER(inode)->nfs_client->cl_hostname);
2005                 break;
2006         case NFS4_OPEN_CLAIM_PREVIOUS:
2007                 nfs_inode_reclaim_delegation(inode, cred, delegation->type,
2008                                              &delegation->stateid,
2009                                              delegation->pagemod_limit,
2010                                              delegation->open_delegation_type);
2011                 break;
2012         default:
2013                 nfs_inode_set_delegation(inode, cred, delegation->type,
2014                                          &delegation->stateid,
2015                                          delegation->pagemod_limit,
2016                                          delegation->open_delegation_type);
2017         }
2018         if (delegation->do_recall)
2019                 nfs_async_inode_return_delegation(inode, &delegation->stateid);
2020 }
2021
2022 /*
2023  * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
2024  * and update the nfs4_state.
2025  */
2026 static struct nfs4_state *
2027 _nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
2028 {
2029         struct inode *inode = data->state->inode;
2030         struct nfs4_state *state = data->state;
2031         int ret;
2032
2033         if (!data->rpc_done) {
2034                 if (data->rpc_status)
2035                         return ERR_PTR(data->rpc_status);
2036                 return nfs4_try_open_cached(data);
2037         }
2038
2039         ret = nfs_refresh_inode(inode, &data->f_attr);
2040         if (ret)
2041                 return ERR_PTR(ret);
2042
2043         nfs4_process_delegation(state->inode,
2044                                 data->owner->so_cred,
2045                                 data->o_arg.claim,
2046                                 &data->o_res.delegation);
2047
2048         if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2049                 if (!update_open_stateid(state, &data->o_res.stateid,
2050                                          NULL, data->o_arg.fmode))
2051                         return ERR_PTR(-EAGAIN);
2052         } else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode))
2053                 return ERR_PTR(-EAGAIN);
2054         refcount_inc(&state->count);
2055
2056         return state;
2057 }
2058
2059 static struct inode *
2060 nfs4_opendata_get_inode(struct nfs4_opendata *data)
2061 {
2062         struct inode *inode;
2063
2064         switch (data->o_arg.claim) {
2065         case NFS4_OPEN_CLAIM_NULL:
2066         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2067         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2068                 if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2069                         return ERR_PTR(-EAGAIN);
2070                 inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2071                                 &data->f_attr);
2072                 break;
2073         default:
2074                 inode = d_inode(data->dentry);
2075                 ihold(inode);
2076                 nfs_refresh_inode(inode, &data->f_attr);
2077         }
2078         return inode;
2079 }
2080
2081 static struct nfs4_state *
2082 nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2083 {
2084         struct nfs4_state *state;
2085         struct inode *inode;
2086
2087         inode = nfs4_opendata_get_inode(data);
2088         if (IS_ERR(inode))
2089                 return ERR_CAST(inode);
2090         if (data->state != NULL && data->state->inode == inode) {
2091                 state = data->state;
2092                 refcount_inc(&state->count);
2093         } else
2094                 state = nfs4_get_open_state(inode, data->owner);
2095         iput(inode);
2096         if (state == NULL)
2097                 state = ERR_PTR(-ENOMEM);
2098         return state;
2099 }
2100
2101 static struct nfs4_state *
2102 _nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2103 {
2104         struct nfs4_state *state;
2105
2106         if (!data->rpc_done) {
2107                 state = nfs4_try_open_cached(data);
2108                 trace_nfs4_cached_open(data->state);
2109                 goto out;
2110         }
2111
2112         state = nfs4_opendata_find_nfs4_state(data);
2113         if (IS_ERR(state))
2114                 goto out;
2115
2116         nfs4_process_delegation(state->inode,
2117                                 data->owner->so_cred,
2118                                 data->o_arg.claim,
2119                                 &data->o_res.delegation);
2120
2121         if (!(data->o_res.rflags & NFS4_OPEN_RESULT_NO_OPEN_STATEID)) {
2122                 if (!update_open_stateid(state, &data->o_res.stateid,
2123                                          NULL, data->o_arg.fmode)) {
2124                         nfs4_put_open_state(state);
2125                         state = ERR_PTR(-EAGAIN);
2126                 }
2127         } else if (!update_open_stateid(state, NULL, NULL, data->o_arg.fmode)) {
2128                 nfs4_put_open_state(state);
2129                 state = ERR_PTR(-EAGAIN);
2130         }
2131 out:
2132         nfs_release_seqid(data->o_arg.seqid);
2133         return state;
2134 }
2135
2136 static struct nfs4_state *
2137 nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2138 {
2139         struct nfs4_state *ret;
2140
2141         if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2142                 ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2143         else
2144                 ret = _nfs4_opendata_to_nfs4_state(data);
2145         nfs4_sequence_free_slot(&data->o_res.seq_res);
2146         return ret;
2147 }
2148
2149 static struct nfs_open_context *
2150 nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2151 {
2152         struct nfs_inode *nfsi = NFS_I(state->inode);
2153         struct nfs_open_context *ctx;
2154
2155         rcu_read_lock();
2156         list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2157                 if (ctx->state != state)
2158                         continue;
2159                 if ((ctx->mode & mode) != mode)
2160                         continue;
2161                 if (!get_nfs_open_context(ctx))
2162                         continue;
2163                 rcu_read_unlock();
2164                 return ctx;
2165         }
2166         rcu_read_unlock();
2167         return ERR_PTR(-ENOENT);
2168 }
2169
2170 static struct nfs_open_context *
2171 nfs4_state_find_open_context(struct nfs4_state *state)
2172 {
2173         struct nfs_open_context *ctx;
2174
2175         ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2176         if (!IS_ERR(ctx))
2177                 return ctx;
2178         ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2179         if (!IS_ERR(ctx))
2180                 return ctx;
2181         return nfs4_state_find_open_context_mode(state, FMODE_READ);
2182 }
2183
2184 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2185                 struct nfs4_state *state, enum open_claim_type4 claim)
2186 {
2187         struct nfs4_opendata *opendata;
2188
2189         opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2190                         NULL, claim, GFP_NOFS);
2191         if (opendata == NULL)
2192                 return ERR_PTR(-ENOMEM);
2193         opendata->state = state;
2194         refcount_inc(&state->count);
2195         return opendata;
2196 }
2197
2198 static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2199                                     fmode_t fmode)
2200 {
2201         struct nfs4_state *newstate;
2202         struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2203         int openflags = opendata->o_arg.open_flags;
2204         int ret;
2205
2206         if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2207                 return 0;
2208         opendata->o_arg.fmode = fmode;
2209         opendata->o_arg.share_access =
2210                 nfs4_map_atomic_open_share(server, fmode, openflags);
2211         memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2212         memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2213         nfs4_init_opendata_res(opendata);
2214         ret = _nfs4_recover_proc_open(opendata);
2215         if (ret != 0)
2216                 return ret;
2217         newstate = nfs4_opendata_to_nfs4_state(opendata);
2218         if (IS_ERR(newstate))
2219                 return PTR_ERR(newstate);
2220         if (newstate != opendata->state)
2221                 ret = -ESTALE;
2222         nfs4_close_state(newstate, fmode);
2223         return ret;
2224 }
2225
2226 static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2227 {
2228         int ret;
2229
2230         /* memory barrier prior to reading state->n_* */
2231         smp_rmb();
2232         ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2233         if (ret != 0)
2234                 return ret;
2235         ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2236         if (ret != 0)
2237                 return ret;
2238         ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2239         if (ret != 0)
2240                 return ret;
2241         /*
2242          * We may have performed cached opens for all three recoveries.
2243          * Check if we need to update the current stateid.
2244          */
2245         if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2246             !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2247                 write_seqlock(&state->seqlock);
2248                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2249                         nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2250                 write_sequnlock(&state->seqlock);
2251         }
2252         return 0;
2253 }
2254
2255 /*
2256  * OPEN_RECLAIM:
2257  *      reclaim state on the server after a reboot.
2258  */
2259 static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2260 {
2261         struct nfs_delegation *delegation;
2262         struct nfs4_opendata *opendata;
2263         u32 delegation_type = NFS4_OPEN_DELEGATE_NONE;
2264         int status;
2265
2266         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2267                         NFS4_OPEN_CLAIM_PREVIOUS);
2268         if (IS_ERR(opendata))
2269                 return PTR_ERR(opendata);
2270         rcu_read_lock();
2271         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2272         if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0) {
2273                 switch(delegation->type) {
2274                 case FMODE_READ:
2275                         delegation_type = NFS4_OPEN_DELEGATE_READ;
2276                         if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2277                                 delegation_type = NFS4_OPEN_DELEGATE_READ_ATTRS_DELEG;
2278                         break;
2279                 case FMODE_WRITE:
2280                 case FMODE_READ|FMODE_WRITE:
2281                         delegation_type = NFS4_OPEN_DELEGATE_WRITE;
2282                         if (test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags))
2283                                 delegation_type = NFS4_OPEN_DELEGATE_WRITE_ATTRS_DELEG;
2284                 }
2285         }
2286         rcu_read_unlock();
2287         opendata->o_arg.u.delegation_type = delegation_type;
2288         status = nfs4_open_recover(opendata, state);
2289         nfs4_opendata_put(opendata);
2290         return status;
2291 }
2292
2293 static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2294 {
2295         struct nfs_server *server = NFS_SERVER(state->inode);
2296         struct nfs4_exception exception = { };
2297         int err;
2298         do {
2299                 err = _nfs4_do_open_reclaim(ctx, state);
2300                 trace_nfs4_open_reclaim(ctx, 0, err);
2301                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2302                         continue;
2303                 if (err != -NFS4ERR_DELAY)
2304                         break;
2305                 nfs4_handle_exception(server, err, &exception);
2306         } while (exception.retry);
2307         return err;
2308 }
2309
2310 static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2311 {
2312         struct nfs_open_context *ctx;
2313         int ret;
2314
2315         ctx = nfs4_state_find_open_context(state);
2316         if (IS_ERR(ctx))
2317                 return -EAGAIN;
2318         clear_bit(NFS_DELEGATED_STATE, &state->flags);
2319         nfs_state_clear_open_state_flags(state);
2320         ret = nfs4_do_open_reclaim(ctx, state);
2321         put_nfs_open_context(ctx);
2322         return ret;
2323 }
2324
2325 static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2326 {
2327         switch (err) {
2328                 default:
2329                         printk(KERN_ERR "NFS: %s: unhandled error "
2330                                         "%d.\n", __func__, err);
2331                         fallthrough;
2332                 case 0:
2333                 case -ENOENT:
2334                 case -EAGAIN:
2335                 case -ESTALE:
2336                 case -ETIMEDOUT:
2337                         break;
2338                 case -NFS4ERR_BADSESSION:
2339                 case -NFS4ERR_BADSLOT:
2340                 case -NFS4ERR_BAD_HIGH_SLOT:
2341                 case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2342                 case -NFS4ERR_DEADSESSION:
2343                         return -EAGAIN;
2344                 case -NFS4ERR_STALE_CLIENTID:
2345                 case -NFS4ERR_STALE_STATEID:
2346                         /* Don't recall a delegation if it was lost */
2347                         nfs4_schedule_lease_recovery(server->nfs_client);
2348                         return -EAGAIN;
2349                 case -NFS4ERR_MOVED:
2350                         nfs4_schedule_migration_recovery(server);
2351                         return -EAGAIN;
2352                 case -NFS4ERR_LEASE_MOVED:
2353                         nfs4_schedule_lease_moved_recovery(server->nfs_client);
2354                         return -EAGAIN;
2355                 case -NFS4ERR_DELEG_REVOKED:
2356                 case -NFS4ERR_ADMIN_REVOKED:
2357                 case -NFS4ERR_EXPIRED:
2358                 case -NFS4ERR_BAD_STATEID:
2359                 case -NFS4ERR_OPENMODE:
2360                         nfs_inode_find_state_and_recover(state->inode,
2361                                         stateid);
2362                         nfs4_schedule_stateid_recovery(server, state);
2363                         return -EAGAIN;
2364                 case -NFS4ERR_DELAY:
2365                 case -NFS4ERR_GRACE:
2366                         ssleep(1);
2367                         return -EAGAIN;
2368                 case -ENOMEM:
2369                 case -NFS4ERR_DENIED:
2370                         if (fl) {
2371                                 struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2372                                 if (lsp)
2373                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2374                         }
2375                         return 0;
2376         }
2377         return err;
2378 }
2379
2380 int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2381                 struct nfs4_state *state, const nfs4_stateid *stateid)
2382 {
2383         struct nfs_server *server = NFS_SERVER(state->inode);
2384         struct nfs4_opendata *opendata;
2385         int err = 0;
2386
2387         opendata = nfs4_open_recoverdata_alloc(ctx, state,
2388                         NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2389         if (IS_ERR(opendata))
2390                 return PTR_ERR(opendata);
2391         nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2392         if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2393                 err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2394                 if (err)
2395                         goto out;
2396         }
2397         if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2398                 err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2399                 if (err)
2400                         goto out;
2401         }
2402         if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2403                 err = nfs4_open_recover_helper(opendata, FMODE_READ);
2404                 if (err)
2405                         goto out;
2406         }
2407         nfs_state_clear_delegation(state);
2408 out:
2409         nfs4_opendata_put(opendata);
2410         return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2411 }
2412
2413 static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2414 {
2415         struct nfs4_opendata *data = calldata;
2416
2417         nfs4_setup_sequence(data->o_arg.server->nfs_client,
2418                            &data->c_arg.seq_args, &data->c_res.seq_res, task);
2419 }
2420
2421 static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2422 {
2423         struct nfs4_opendata *data = calldata;
2424
2425         nfs40_sequence_done(task, &data->c_res.seq_res);
2426
2427         data->rpc_status = task->tk_status;
2428         if (data->rpc_status == 0) {
2429                 nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2430                 nfs_confirm_seqid(&data->owner->so_seqid, 0);
2431                 renew_lease(data->o_res.server, data->timestamp);
2432                 data->rpc_done = true;
2433         }
2434 }
2435
2436 static void nfs4_open_confirm_release(void *calldata)
2437 {
2438         struct nfs4_opendata *data = calldata;
2439         struct nfs4_state *state = NULL;
2440
2441         /* If this request hasn't been cancelled, do nothing */
2442         if (!data->cancelled)
2443                 goto out_free;
2444         /* In case of error, no cleanup! */
2445         if (!data->rpc_done)
2446                 goto out_free;
2447         state = nfs4_opendata_to_nfs4_state(data);
2448         if (!IS_ERR(state))
2449                 nfs4_close_state(state, data->o_arg.fmode);
2450 out_free:
2451         nfs4_opendata_put(data);
2452 }
2453
2454 static const struct rpc_call_ops nfs4_open_confirm_ops = {
2455         .rpc_call_prepare = nfs4_open_confirm_prepare,
2456         .rpc_call_done = nfs4_open_confirm_done,
2457         .rpc_release = nfs4_open_confirm_release,
2458 };
2459
2460 /*
2461  * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2462  */
2463 static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2464 {
2465         struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2466         struct rpc_task *task;
2467         struct  rpc_message msg = {
2468                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2469                 .rpc_argp = &data->c_arg,
2470                 .rpc_resp = &data->c_res,
2471                 .rpc_cred = data->owner->so_cred,
2472         };
2473         struct rpc_task_setup task_setup_data = {
2474                 .rpc_client = server->client,
2475                 .rpc_message = &msg,
2476                 .callback_ops = &nfs4_open_confirm_ops,
2477                 .callback_data = data,
2478                 .workqueue = nfsiod_workqueue,
2479                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2480         };
2481         int status;
2482
2483         nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2484                                 data->is_recover);
2485         kref_get(&data->kref);
2486         data->rpc_done = false;
2487         data->rpc_status = 0;
2488         data->timestamp = jiffies;
2489         task = rpc_run_task(&task_setup_data);
2490         if (IS_ERR(task))
2491                 return PTR_ERR(task);
2492         status = rpc_wait_for_completion_task(task);
2493         if (status != 0) {
2494                 data->cancelled = true;
2495                 smp_wmb();
2496         } else
2497                 status = data->rpc_status;
2498         rpc_put_task(task);
2499         return status;
2500 }
2501
2502 static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2503 {
2504         struct nfs4_opendata *data = calldata;
2505         struct nfs4_state_owner *sp = data->owner;
2506         struct nfs_client *clp = sp->so_server->nfs_client;
2507         enum open_claim_type4 claim = data->o_arg.claim;
2508
2509         if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2510                 goto out_wait;
2511         /*
2512          * Check if we still need to send an OPEN call, or if we can use
2513          * a delegation instead.
2514          */
2515         if (data->state != NULL) {
2516                 struct nfs_delegation *delegation;
2517
2518                 if (can_open_cached(data->state, data->o_arg.fmode,
2519                                         data->o_arg.open_flags, claim))
2520                         goto out_no_action;
2521                 rcu_read_lock();
2522                 delegation = nfs4_get_valid_delegation(data->state->inode);
2523                 if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2524                         goto unlock_no_action;
2525                 rcu_read_unlock();
2526         }
2527         /* Update client id. */
2528         data->o_arg.clientid = clp->cl_clientid;
2529         switch (claim) {
2530         default:
2531                 break;
2532         case NFS4_OPEN_CLAIM_PREVIOUS:
2533         case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2534         case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2535                 data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2536                 fallthrough;
2537         case NFS4_OPEN_CLAIM_FH:
2538                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2539         }
2540         data->timestamp = jiffies;
2541         if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2542                                 &data->o_arg.seq_args,
2543                                 &data->o_res.seq_res,
2544                                 task) != 0)
2545                 nfs_release_seqid(data->o_arg.seqid);
2546
2547         /* Set the create mode (note dependency on the session type) */
2548         data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2549         if (data->o_arg.open_flags & O_EXCL) {
2550                 data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2551                 if (clp->cl_mvops->minor_version == 0) {
2552                         data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2553                         /* don't put an ACCESS op in OPEN compound if O_EXCL,
2554                          * because ACCESS will return permission denied for
2555                          * all bits until close */
2556                         data->o_res.access_request = data->o_arg.access = 0;
2557                 } else if (nfs4_has_persistent_session(clp))
2558                         data->o_arg.createmode = NFS4_CREATE_GUARDED;
2559         }
2560         return;
2561 unlock_no_action:
2562         trace_nfs4_cached_open(data->state);
2563         rcu_read_unlock();
2564 out_no_action:
2565         task->tk_action = NULL;
2566 out_wait:
2567         nfs4_sequence_done(task, &data->o_res.seq_res);
2568 }
2569
2570 static void nfs4_open_done(struct rpc_task *task, void *calldata)
2571 {
2572         struct nfs4_opendata *data = calldata;
2573
2574         data->rpc_status = task->tk_status;
2575
2576         if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2577                 return;
2578
2579         if (task->tk_status == 0) {
2580                 if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2581                         switch (data->o_res.f_attr->mode & S_IFMT) {
2582                         case S_IFREG:
2583                                 break;
2584                         case S_IFLNK:
2585                                 data->rpc_status = -ELOOP;
2586                                 break;
2587                         case S_IFDIR:
2588                                 data->rpc_status = -EISDIR;
2589                                 break;
2590                         default:
2591                                 data->rpc_status = -ENOTDIR;
2592                         }
2593                 }
2594                 renew_lease(data->o_res.server, data->timestamp);
2595                 if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2596                         nfs_confirm_seqid(&data->owner->so_seqid, 0);
2597         }
2598         data->rpc_done = true;
2599 }
2600
2601 static void nfs4_open_release(void *calldata)
2602 {
2603         struct nfs4_opendata *data = calldata;
2604         struct nfs4_state *state = NULL;
2605
2606         /* If this request hasn't been cancelled, do nothing */
2607         if (!data->cancelled)
2608                 goto out_free;
2609         /* In case of error, no cleanup! */
2610         if (data->rpc_status != 0 || !data->rpc_done)
2611                 goto out_free;
2612         /* In case we need an open_confirm, no cleanup! */
2613         if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2614                 goto out_free;
2615         state = nfs4_opendata_to_nfs4_state(data);
2616         if (!IS_ERR(state))
2617                 nfs4_close_state(state, data->o_arg.fmode);
2618 out_free:
2619         nfs4_opendata_put(data);
2620 }
2621
2622 static const struct rpc_call_ops nfs4_open_ops = {
2623         .rpc_call_prepare = nfs4_open_prepare,
2624         .rpc_call_done = nfs4_open_done,
2625         .rpc_release = nfs4_open_release,
2626 };
2627
2628 static int nfs4_run_open_task(struct nfs4_opendata *data,
2629                               struct nfs_open_context *ctx)
2630 {
2631         struct inode *dir = d_inode(data->dir);
2632         struct nfs_server *server = NFS_SERVER(dir);
2633         struct nfs_openargs *o_arg = &data->o_arg;
2634         struct nfs_openres *o_res = &data->o_res;
2635         struct rpc_task *task;
2636         struct rpc_message msg = {
2637                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2638                 .rpc_argp = o_arg,
2639                 .rpc_resp = o_res,
2640                 .rpc_cred = data->owner->so_cred,
2641         };
2642         struct rpc_task_setup task_setup_data = {
2643                 .rpc_client = server->client,
2644                 .rpc_message = &msg,
2645                 .callback_ops = &nfs4_open_ops,
2646                 .callback_data = data,
2647                 .workqueue = nfsiod_workqueue,
2648                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2649         };
2650         int status;
2651
2652         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2653                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
2654
2655         kref_get(&data->kref);
2656         data->rpc_done = false;
2657         data->rpc_status = 0;
2658         data->cancelled = false;
2659         data->is_recover = false;
2660         if (!ctx) {
2661                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2662                 data->is_recover = true;
2663                 task_setup_data.flags |= RPC_TASK_TIMEOUT;
2664         } else {
2665                 nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2666                 pnfs_lgopen_prepare(data, ctx);
2667         }
2668         task = rpc_run_task(&task_setup_data);
2669         if (IS_ERR(task))
2670                 return PTR_ERR(task);
2671         status = rpc_wait_for_completion_task(task);
2672         if (status != 0) {
2673                 data->cancelled = true;
2674                 smp_wmb();
2675         } else
2676                 status = data->rpc_status;
2677         rpc_put_task(task);
2678
2679         return status;
2680 }
2681
2682 static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2683 {
2684         struct inode *dir = d_inode(data->dir);
2685         struct nfs_openres *o_res = &data->o_res;
2686         int status;
2687
2688         status = nfs4_run_open_task(data, NULL);
2689         if (status != 0 || !data->rpc_done)
2690                 return status;
2691
2692         nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2693
2694         if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2695                 status = _nfs4_proc_open_confirm(data);
2696
2697         return status;
2698 }
2699
2700 /*
2701  * Additional permission checks in order to distinguish between an
2702  * open for read, and an open for execute. This works around the
2703  * fact that NFSv4 OPEN treats read and execute permissions as being
2704  * the same.
2705  * Note that in the non-execute case, we want to turn off permission
2706  * checking if we just created a new file (POSIX open() semantics).
2707  */
2708 static int nfs4_opendata_access(const struct cred *cred,
2709                                 struct nfs4_opendata *opendata,
2710                                 struct nfs4_state *state, fmode_t fmode)
2711 {
2712         struct nfs_access_entry cache;
2713         u32 mask, flags;
2714
2715         /* access call failed or for some reason the server doesn't
2716          * support any access modes -- defer access call until later */
2717         if (opendata->o_res.access_supported == 0)
2718                 return 0;
2719
2720         mask = 0;
2721         if (fmode & FMODE_EXEC) {
2722                 /* ONLY check for exec rights */
2723                 if (S_ISDIR(state->inode->i_mode))
2724                         mask = NFS4_ACCESS_LOOKUP;
2725                 else
2726                         mask = NFS4_ACCESS_EXECUTE;
2727         } else if ((fmode & FMODE_READ) && !opendata->file_created)
2728                 mask = NFS4_ACCESS_READ;
2729
2730         nfs_access_set_mask(&cache, opendata->o_res.access_result);
2731         nfs_access_add_cache(state->inode, &cache, cred);
2732
2733         flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2734         if ((mask & ~cache.mask & flags) == 0)
2735                 return 0;
2736
2737         return -EACCES;
2738 }
2739
2740 /*
2741  * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2742  */
2743 static int _nfs4_proc_open(struct nfs4_opendata *data,
2744                            struct nfs_open_context *ctx)
2745 {
2746         struct inode *dir = d_inode(data->dir);
2747         struct nfs_server *server = NFS_SERVER(dir);
2748         struct nfs_openargs *o_arg = &data->o_arg;
2749         struct nfs_openres *o_res = &data->o_res;
2750         int status;
2751
2752         status = nfs4_run_open_task(data, ctx);
2753         if (!data->rpc_done)
2754                 return status;
2755         if (status != 0) {
2756                 if (status == -NFS4ERR_BADNAME &&
2757                                 !(o_arg->open_flags & O_CREAT))
2758                         return -ENOENT;
2759                 return status;
2760         }
2761
2762         nfs_fattr_map_and_free_names(server, &data->f_attr);
2763
2764         if (o_arg->open_flags & O_CREAT) {
2765                 if (o_arg->open_flags & O_EXCL)
2766                         data->file_created = true;
2767                 else if (o_res->cinfo.before != o_res->cinfo.after)
2768                         data->file_created = true;
2769                 if (data->file_created ||
2770                     inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2771                         nfs4_update_changeattr(dir, &o_res->cinfo,
2772                                         o_res->f_attr->time_start,
2773                                         NFS_INO_INVALID_DATA);
2774         }
2775         if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2776                 server->caps &= ~NFS_CAP_POSIX_LOCK;
2777         if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2778                 status = _nfs4_proc_open_confirm(data);
2779                 if (status != 0)
2780                         return status;
2781         }
2782         if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2783                 struct nfs_fh *fh = &o_res->fh;
2784
2785                 nfs4_sequence_free_slot(&o_res->seq_res);
2786                 if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2787                         fh = NFS_FH(d_inode(data->dentry));
2788                 nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2789         }
2790         return 0;
2791 }
2792
2793 /*
2794  * OPEN_EXPIRED:
2795  *      reclaim state on the server after a network partition.
2796  *      Assumes caller holds the appropriate lock
2797  */
2798 static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2799 {
2800         struct nfs4_opendata *opendata;
2801         int ret;
2802
2803         opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2804         if (IS_ERR(opendata))
2805                 return PTR_ERR(opendata);
2806         /*
2807          * We're not recovering a delegation, so ask for no delegation.
2808          * Otherwise the recovery thread could deadlock with an outstanding
2809          * delegation return.
2810          */
2811         opendata->o_arg.open_flags = O_DIRECT;
2812         ret = nfs4_open_recover(opendata, state);
2813         if (ret == -ESTALE)
2814                 d_drop(ctx->dentry);
2815         nfs4_opendata_put(opendata);
2816         return ret;
2817 }
2818
2819 static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2820 {
2821         struct nfs_server *server = NFS_SERVER(state->inode);
2822         struct nfs4_exception exception = { };
2823         int err;
2824
2825         do {
2826                 err = _nfs4_open_expired(ctx, state);
2827                 trace_nfs4_open_expired(ctx, 0, err);
2828                 if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2829                         continue;
2830                 switch (err) {
2831                 default:
2832                         goto out;
2833                 case -NFS4ERR_GRACE:
2834                 case -NFS4ERR_DELAY:
2835                         nfs4_handle_exception(server, err, &exception);
2836                         err = 0;
2837                 }
2838         } while (exception.retry);
2839 out:
2840         return err;
2841 }
2842
2843 static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2844 {
2845         struct nfs_open_context *ctx;
2846         int ret;
2847
2848         ctx = nfs4_state_find_open_context(state);
2849         if (IS_ERR(ctx))
2850                 return -EAGAIN;
2851         ret = nfs4_do_open_expired(ctx, state);
2852         put_nfs_open_context(ctx);
2853         return ret;
2854 }
2855
2856 static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2857                 const nfs4_stateid *stateid)
2858 {
2859         nfs_remove_bad_delegation(state->inode, stateid);
2860         nfs_state_clear_delegation(state);
2861 }
2862
2863 static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2864 {
2865         if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2866                 nfs_finish_clear_delegation_stateid(state, NULL);
2867 }
2868
2869 static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2870 {
2871         /* NFSv4.0 doesn't allow for delegation recovery on open expire */
2872         nfs40_clear_delegation_stateid(state);
2873         nfs_state_clear_open_state_flags(state);
2874         return nfs4_open_expired(sp, state);
2875 }
2876
2877 static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2878                                                const nfs4_stateid *stateid,
2879                                                const struct cred *cred)
2880 {
2881         return -NFS4ERR_BAD_STATEID;
2882 }
2883
2884 #if defined(CONFIG_NFS_V4_1)
2885 static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2886                                                const nfs4_stateid *stateid,
2887                                                const struct cred *cred)
2888 {
2889         int status;
2890
2891         switch (stateid->type) {
2892         default:
2893                 break;
2894         case NFS4_INVALID_STATEID_TYPE:
2895         case NFS4_SPECIAL_STATEID_TYPE:
2896                 return -NFS4ERR_BAD_STATEID;
2897         case NFS4_REVOKED_STATEID_TYPE:
2898                 goto out_free;
2899         }
2900
2901         status = nfs41_test_stateid(server, stateid, cred);
2902         switch (status) {
2903         case -NFS4ERR_EXPIRED:
2904         case -NFS4ERR_ADMIN_REVOKED:
2905         case -NFS4ERR_DELEG_REVOKED:
2906                 break;
2907         default:
2908                 return status;
2909         }
2910 out_free:
2911         /* Ack the revoked state to the server */
2912         nfs41_free_stateid(server, stateid, cred, true);
2913         return -NFS4ERR_EXPIRED;
2914 }
2915
2916 static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2917 {
2918         struct nfs_server *server = NFS_SERVER(state->inode);
2919         nfs4_stateid stateid;
2920         struct nfs_delegation *delegation;
2921         const struct cred *cred = NULL;
2922         int status, ret = NFS_OK;
2923
2924         /* Get the delegation credential for use by test/free_stateid */
2925         rcu_read_lock();
2926         delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2927         if (delegation == NULL) {
2928                 rcu_read_unlock();
2929                 nfs_state_clear_delegation(state);
2930                 return NFS_OK;
2931         }
2932
2933         spin_lock(&delegation->lock);
2934         nfs4_stateid_copy(&stateid, &delegation->stateid);
2935
2936         if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2937                                 &delegation->flags)) {
2938                 spin_unlock(&delegation->lock);
2939                 rcu_read_unlock();
2940                 return NFS_OK;
2941         }
2942
2943         if (delegation->cred)
2944                 cred = get_cred(delegation->cred);
2945         spin_unlock(&delegation->lock);
2946         rcu_read_unlock();
2947         status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2948         trace_nfs4_test_delegation_stateid(state, NULL, status);
2949         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2950                 nfs_finish_clear_delegation_stateid(state, &stateid);
2951         else
2952                 ret = status;
2953
2954         put_cred(cred);
2955         return ret;
2956 }
2957
2958 static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2959 {
2960         nfs4_stateid tmp;
2961
2962         if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2963             nfs4_copy_delegation_stateid(state->inode, state->state,
2964                                 &tmp, NULL) &&
2965             nfs4_stateid_match_other(&state->stateid, &tmp))
2966                 nfs_state_set_delegation(state, &tmp, state->state);
2967         else
2968                 nfs_state_clear_delegation(state);
2969 }
2970
2971 /**
2972  * nfs41_check_expired_locks - possibly free a lock stateid
2973  *
2974  * @state: NFSv4 state for an inode
2975  *
2976  * Returns NFS_OK if recovery for this stateid is now finished.
2977  * Otherwise a negative NFS4ERR value is returned.
2978  */
2979 static int nfs41_check_expired_locks(struct nfs4_state *state)
2980 {
2981         int status, ret = NFS_OK;
2982         struct nfs4_lock_state *lsp, *prev = NULL;
2983         struct nfs_server *server = NFS_SERVER(state->inode);
2984
2985         if (!test_bit(LK_STATE_IN_USE, &state->flags))
2986                 goto out;
2987
2988         spin_lock(&state->state_lock);
2989         list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2990                 if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2991                         const struct cred *cred = lsp->ls_state->owner->so_cred;
2992
2993                         refcount_inc(&lsp->ls_count);
2994                         spin_unlock(&state->state_lock);
2995
2996                         nfs4_put_lock_state(prev);
2997                         prev = lsp;
2998
2999                         status = nfs41_test_and_free_expired_stateid(server,
3000                                         &lsp->ls_stateid,
3001                                         cred);
3002                         trace_nfs4_test_lock_stateid(state, lsp, status);
3003                         if (status == -NFS4ERR_EXPIRED ||
3004                             status == -NFS4ERR_BAD_STATEID) {
3005                                 clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
3006                                 lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
3007                                 if (!recover_lost_locks)
3008                                         set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
3009                         } else if (status != NFS_OK) {
3010                                 ret = status;
3011                                 nfs4_put_lock_state(prev);
3012                                 goto out;
3013                         }
3014                         spin_lock(&state->state_lock);
3015                 }
3016         }
3017         spin_unlock(&state->state_lock);
3018         nfs4_put_lock_state(prev);
3019 out:
3020         return ret;
3021 }
3022
3023 /**
3024  * nfs41_check_open_stateid - possibly free an open stateid
3025  *
3026  * @state: NFSv4 state for an inode
3027  *
3028  * Returns NFS_OK if recovery for this stateid is now finished.
3029  * Otherwise a negative NFS4ERR value is returned.
3030  */
3031 static int nfs41_check_open_stateid(struct nfs4_state *state)
3032 {
3033         struct nfs_server *server = NFS_SERVER(state->inode);
3034         nfs4_stateid *stateid = &state->open_stateid;
3035         const struct cred *cred = state->owner->so_cred;
3036         int status;
3037
3038         if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
3039                 return -NFS4ERR_BAD_STATEID;
3040         status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
3041         trace_nfs4_test_open_stateid(state, NULL, status);
3042         if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
3043                 nfs_state_clear_open_state_flags(state);
3044                 stateid->type = NFS4_INVALID_STATEID_TYPE;
3045                 return status;
3046         }
3047         if (nfs_open_stateid_recover_openmode(state))
3048                 return -NFS4ERR_OPENMODE;
3049         return NFS_OK;
3050 }
3051
3052 static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
3053 {
3054         int status;
3055
3056         status = nfs41_check_delegation_stateid(state);
3057         if (status != NFS_OK)
3058                 return status;
3059         nfs41_delegation_recover_stateid(state);
3060
3061         status = nfs41_check_expired_locks(state);
3062         if (status != NFS_OK)
3063                 return status;
3064         status = nfs41_check_open_stateid(state);
3065         if (status != NFS_OK)
3066                 status = nfs4_open_expired(sp, state);
3067         return status;
3068 }
3069 #endif
3070
3071 /*
3072  * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
3073  * fields corresponding to attributes that were used to store the verifier.
3074  * Make sure we clobber those fields in the later setattr call
3075  */
3076 static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3077                                 struct iattr *sattr, struct nfs4_label **label)
3078 {
3079         const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3080         __u32 attrset[3];
3081         unsigned ret;
3082         unsigned i;
3083
3084         for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3085                 attrset[i] = opendata->o_res.attrset[i];
3086                 if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3087                         attrset[i] &= ~bitmask[i];
3088         }
3089
3090         ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3091                 sattr->ia_valid : 0;
3092
3093         if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3094                 if (sattr->ia_valid & ATTR_ATIME_SET)
3095                         ret |= ATTR_ATIME_SET;
3096                 else
3097                         ret |= ATTR_ATIME;
3098         }
3099
3100         if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3101                 if (sattr->ia_valid & ATTR_MTIME_SET)
3102                         ret |= ATTR_MTIME_SET;
3103                 else
3104                         ret |= ATTR_MTIME;
3105         }
3106
3107         if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3108                 *label = NULL;
3109         return ret;
3110 }
3111
3112 static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3113                 struct nfs_open_context *ctx)
3114 {
3115         struct nfs4_state_owner *sp = opendata->owner;
3116         struct nfs_server *server = sp->so_server;
3117         struct dentry *dentry;
3118         struct nfs4_state *state;
3119         fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3120         struct inode *dir = d_inode(opendata->dir);
3121         unsigned long dir_verifier;
3122         int ret;
3123
3124         dir_verifier = nfs_save_change_attribute(dir);
3125
3126         ret = _nfs4_proc_open(opendata, ctx);
3127         if (ret != 0)
3128                 goto out;
3129
3130         state = _nfs4_opendata_to_nfs4_state(opendata);
3131         ret = PTR_ERR(state);
3132         if (IS_ERR(state))
3133                 goto out;
3134         ctx->state = state;
3135         if (server->caps & NFS_CAP_POSIX_LOCK)
3136                 set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3137         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3138                 set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3139         if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3140                 set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3141
3142         dentry = opendata->dentry;
3143         if (d_really_is_negative(dentry)) {
3144                 struct dentry *alias;
3145                 d_drop(dentry);
3146                 alias = d_exact_alias(dentry, state->inode);
3147                 if (!alias)
3148                         alias = d_splice_alias(igrab(state->inode), dentry);
3149                 /* d_splice_alias() can't fail here - it's a non-directory */
3150                 if (alias) {
3151                         dput(ctx->dentry);
3152                         ctx->dentry = dentry = alias;
3153                 }
3154         }
3155
3156         switch(opendata->o_arg.claim) {
3157         default:
3158                 break;
3159         case NFS4_OPEN_CLAIM_NULL:
3160         case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3161         case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3162                 if (!opendata->rpc_done)
3163                         break;
3164                 if (opendata->o_res.delegation.type != 0)
3165                         dir_verifier = nfs_save_change_attribute(dir);
3166                 nfs_set_verifier(dentry, dir_verifier);
3167         }
3168
3169         /* Parse layoutget results before we check for access */
3170         pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3171
3172         ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3173         if (ret != 0)
3174                 goto out;
3175
3176         if (d_inode(dentry) == state->inode)
3177                 nfs_inode_attach_open_context(ctx);
3178
3179 out:
3180         if (!opendata->cancelled) {
3181                 if (opendata->lgp) {
3182                         nfs4_lgopen_release(opendata->lgp);
3183                         opendata->lgp = NULL;
3184                 }
3185                 nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3186         }
3187         return ret;
3188 }
3189
3190 /*
3191  * Returns a referenced nfs4_state
3192  */
3193 static int _nfs4_do_open(struct inode *dir,
3194                         struct nfs_open_context *ctx,
3195                         int flags,
3196                         const struct nfs4_open_createattrs *c,
3197                         int *opened)
3198 {
3199         struct nfs4_state_owner  *sp;
3200         struct nfs4_state     *state = NULL;
3201         struct nfs_server       *server = NFS_SERVER(dir);
3202         struct nfs4_opendata *opendata;
3203         struct dentry *dentry = ctx->dentry;
3204         const struct cred *cred = ctx->cred;
3205         struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3206         fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3207         enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3208         struct iattr *sattr = c->sattr;
3209         struct nfs4_label *label = c->label;
3210         int status;
3211
3212         /* Protect against reboot recovery conflicts */
3213         status = -ENOMEM;
3214         sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3215         if (sp == NULL) {
3216                 dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3217                 goto out_err;
3218         }
3219         status = nfs4_client_recover_expired_lease(server->nfs_client);
3220         if (status != 0)
3221                 goto err_put_state_owner;
3222         if (d_really_is_positive(dentry))
3223                 nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3224         status = -ENOMEM;
3225         if (d_really_is_positive(dentry))
3226                 claim = NFS4_OPEN_CLAIM_FH;
3227         opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3228                         c, claim, GFP_KERNEL);
3229         if (opendata == NULL)
3230                 goto err_put_state_owner;
3231
3232         if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3233                 if (!opendata->f_attr.mdsthreshold) {
3234                         opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3235                         if (!opendata->f_attr.mdsthreshold)
3236                                 goto err_opendata_put;
3237                 }
3238                 opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3239         }
3240         if (d_really_is_positive(dentry))
3241                 opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3242
3243         status = _nfs4_open_and_get_state(opendata, ctx);
3244         if (status != 0)
3245                 goto err_opendata_put;
3246         state = ctx->state;
3247
3248         if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3249             (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3250                 unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3251                 /*
3252                  * send create attributes which was not set by open
3253                  * with an extra setattr.
3254                  */
3255                 if (attrs || label) {
3256                         unsigned ia_old = sattr->ia_valid;
3257
3258                         sattr->ia_valid = attrs;
3259                         nfs_fattr_init(opendata->o_res.f_attr);
3260                         status = nfs4_do_setattr(state->inode, cred,
3261                                         opendata->o_res.f_attr, sattr,
3262                                         ctx, label);
3263                         if (status == 0) {
3264                                 nfs_setattr_update_inode(state->inode, sattr,
3265                                                 opendata->o_res.f_attr);
3266                                 nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3267                         }
3268                         sattr->ia_valid = ia_old;
3269                 }
3270         }
3271         if (opened && opendata->file_created)
3272                 *opened = 1;
3273
3274         if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3275                 *ctx_th = opendata->f_attr.mdsthreshold;
3276                 opendata->f_attr.mdsthreshold = NULL;
3277         }
3278
3279         nfs4_opendata_put(opendata);
3280         nfs4_put_state_owner(sp);
3281         return 0;
3282 err_opendata_put:
3283         nfs4_opendata_put(opendata);
3284 err_put_state_owner:
3285         nfs4_put_state_owner(sp);
3286 out_err:
3287         return status;
3288 }
3289
3290
3291 static struct nfs4_state *nfs4_do_open(struct inode *dir,
3292                                         struct nfs_open_context *ctx,
3293                                         int flags,
3294                                         struct iattr *sattr,
3295                                         struct nfs4_label *label,
3296                                         int *opened)
3297 {
3298         struct nfs_server *server = NFS_SERVER(dir);
3299         struct nfs4_exception exception = {
3300                 .interruptible = true,
3301         };
3302         struct nfs4_state *res;
3303         struct nfs4_open_createattrs c = {
3304                 .label = label,
3305                 .sattr = sattr,
3306                 .verf = {
3307                         [0] = (__u32)jiffies,
3308                         [1] = (__u32)current->pid,
3309                 },
3310         };
3311         int status;
3312
3313         do {
3314                 status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3315                 res = ctx->state;
3316                 trace_nfs4_open_file(ctx, flags, status);
3317                 if (status == 0)
3318                         break;
3319                 /* NOTE: BAD_SEQID means the server and client disagree about the
3320                  * book-keeping w.r.t. state-changing operations
3321                  * (OPEN/CLOSE/LOCK/LOCKU...)
3322                  * It is actually a sign of a bug on the client or on the server.
3323                  *
3324                  * If we receive a BAD_SEQID error in the particular case of
3325                  * doing an OPEN, we assume that nfs_increment_open_seqid() will
3326                  * have unhashed the old state_owner for us, and that we can
3327                  * therefore safely retry using a new one. We should still warn
3328                  * the user though...
3329                  */
3330                 if (status == -NFS4ERR_BAD_SEQID) {
3331                         pr_warn_ratelimited("NFS: v4 server %s "
3332                                         " returned a bad sequence-id error!\n",
3333                                         NFS_SERVER(dir)->nfs_client->cl_hostname);
3334                         exception.retry = 1;
3335                         continue;
3336                 }
3337                 /*
3338                  * BAD_STATEID on OPEN means that the server cancelled our
3339                  * state before it received the OPEN_CONFIRM.
3340                  * Recover by retrying the request as per the discussion
3341                  * on Page 181 of RFC3530.
3342                  */
3343                 if (status == -NFS4ERR_BAD_STATEID) {
3344                         exception.retry = 1;
3345                         continue;
3346                 }
3347                 if (status == -NFS4ERR_EXPIRED) {
3348                         nfs4_schedule_lease_recovery(server->nfs_client);
3349                         exception.retry = 1;
3350                         continue;
3351                 }
3352                 if (status == -EAGAIN) {
3353                         /* We must have found a delegation */
3354                         exception.retry = 1;
3355                         continue;
3356                 }
3357                 if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3358                         continue;
3359                 res = ERR_PTR(nfs4_handle_exception(server,
3360                                         status, &exception));
3361         } while (exception.retry);
3362         return res;
3363 }
3364
3365 static int _nfs4_do_setattr(struct inode *inode,
3366                             struct nfs_setattrargs *arg,
3367                             struct nfs_setattrres *res,
3368                             const struct cred *cred,
3369                             struct nfs_open_context *ctx)
3370 {
3371         struct nfs_server *server = NFS_SERVER(inode);
3372         struct rpc_message msg = {
3373                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3374                 .rpc_argp       = arg,
3375                 .rpc_resp       = res,
3376                 .rpc_cred       = cred,
3377         };
3378         const struct cred *delegation_cred = NULL;
3379         unsigned long timestamp = jiffies;
3380         bool truncate;
3381         int status;
3382
3383         nfs_fattr_init(res->fattr);
3384
3385         /* Servers should only apply open mode checks for file size changes */
3386         truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3387         if (!truncate) {
3388                 nfs4_inode_make_writeable(inode);
3389                 goto zero_stateid;
3390         }
3391
3392         if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3393                 /* Use that stateid */
3394         } else if (ctx != NULL && ctx->state) {
3395                 struct nfs_lock_context *l_ctx;
3396                 if (!nfs4_valid_open_stateid(ctx->state))
3397                         return -EBADF;
3398                 l_ctx = nfs_get_lock_context(ctx);
3399                 if (IS_ERR(l_ctx))
3400                         return PTR_ERR(l_ctx);
3401                 status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3402                                                 &arg->stateid, &delegation_cred);
3403                 nfs_put_lock_context(l_ctx);
3404                 if (status == -EIO)
3405                         return -EBADF;
3406                 else if (status == -EAGAIN)
3407                         goto zero_stateid;
3408         } else {
3409 zero_stateid:
3410                 nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3411         }
3412         if (delegation_cred)
3413                 msg.rpc_cred = delegation_cred;
3414
3415         status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3416
3417         put_cred(delegation_cred);
3418         if (status == 0 && ctx != NULL)
3419                 renew_lease(server, timestamp);
3420         trace_nfs4_setattr(inode, &arg->stateid, status);
3421         return status;
3422 }
3423
3424 static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3425                            struct nfs_fattr *fattr, struct iattr *sattr,
3426                            struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3427 {
3428         struct nfs_server *server = NFS_SERVER(inode);
3429         __u32 bitmask[NFS4_BITMASK_SZ];
3430         struct nfs4_state *state = ctx ? ctx->state : NULL;
3431         struct nfs_setattrargs  arg = {
3432                 .fh             = NFS_FH(inode),
3433                 .iap            = sattr,
3434                 .server         = server,
3435                 .bitmask = bitmask,
3436                 .label          = ilabel,
3437         };
3438         struct nfs_setattrres  res = {
3439                 .fattr          = fattr,
3440                 .server         = server,
3441         };
3442         struct nfs4_exception exception = {
3443                 .state = state,
3444                 .inode = inode,
3445                 .stateid = &arg.stateid,
3446         };
3447         unsigned long adjust_flags = NFS_INO_INVALID_CHANGE |
3448                                      NFS_INO_INVALID_CTIME;
3449         int err;
3450
3451         if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3452                 adjust_flags |= NFS_INO_INVALID_MODE;
3453         if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3454                 adjust_flags |= NFS_INO_INVALID_OTHER;
3455
3456         do {
3457                 nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3458                                         inode, adjust_flags);
3459
3460                 err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3461                 switch (err) {
3462                 case -NFS4ERR_OPENMODE:
3463                         if (!(sattr->ia_valid & ATTR_SIZE)) {
3464                                 pr_warn_once("NFSv4: server %s is incorrectly "
3465                                                 "applying open mode checks to "
3466                                                 "a SETATTR that is not "
3467                                                 "changing file size.\n",
3468                                                 server->nfs_client->cl_hostname);
3469                         }
3470                         if (state && !(state->state & FMODE_WRITE)) {
3471                                 err = -EBADF;
3472                                 if (sattr->ia_valid & ATTR_OPEN)
3473                                         err = -EACCES;
3474                                 goto out;
3475                         }
3476                 }
3477                 err = nfs4_handle_exception(server, err, &exception);
3478         } while (exception.retry);
3479 out:
3480         return err;
3481 }
3482
3483 static bool
3484 nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3485 {
3486         if (inode == NULL || !nfs_have_layout(inode))
3487                 return false;
3488
3489         return pnfs_wait_on_layoutreturn(inode, task);
3490 }
3491
3492 /*
3493  * Update the seqid of an open stateid
3494  */
3495 static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3496                 struct nfs4_state *state)
3497 {
3498         __be32 seqid_open;
3499         u32 dst_seqid;
3500         int seq;
3501
3502         for (;;) {
3503                 if (!nfs4_valid_open_stateid(state))
3504                         break;
3505                 seq = read_seqbegin(&state->seqlock);
3506                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3507                         nfs4_stateid_copy(dst, &state->open_stateid);
3508                         if (read_seqretry(&state->seqlock, seq))
3509                                 continue;
3510                         break;
3511                 }
3512                 seqid_open = state->open_stateid.seqid;
3513                 if (read_seqretry(&state->seqlock, seq))
3514                         continue;
3515
3516                 dst_seqid = be32_to_cpu(dst->seqid);
3517                 if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3518                         dst->seqid = seqid_open;
3519                 break;
3520         }
3521 }
3522
3523 /*
3524  * Update the seqid of an open stateid after receiving
3525  * NFS4ERR_OLD_STATEID
3526  */
3527 static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3528                 struct nfs4_state *state)
3529 {
3530         __be32 seqid_open;
3531         u32 dst_seqid;
3532         bool ret;
3533         int seq, status = -EAGAIN;
3534         DEFINE_WAIT(wait);
3535
3536         for (;;) {
3537                 ret = false;
3538                 if (!nfs4_valid_open_stateid(state))
3539                         break;
3540                 seq = read_seqbegin(&state->seqlock);
3541                 if (!nfs4_state_match_open_stateid_other(state, dst)) {
3542                         if (read_seqretry(&state->seqlock, seq))
3543                                 continue;
3544                         break;
3545                 }
3546
3547                 write_seqlock(&state->seqlock);
3548                 seqid_open = state->open_stateid.seqid;
3549
3550                 dst_seqid = be32_to_cpu(dst->seqid);
3551
3552                 /* Did another OPEN bump the state's seqid?  try again: */
3553                 if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3554                         dst->seqid = seqid_open;
3555                         write_sequnlock(&state->seqlock);
3556                         ret = true;
3557                         break;
3558                 }
3559
3560                 /* server says we're behind but we haven't seen the update yet */
3561                 set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3562                 prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3563                 write_sequnlock(&state->seqlock);
3564                 trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3565
3566                 if (fatal_signal_pending(current))
3567                         status = -EINTR;
3568                 else
3569                         if (schedule_timeout(5*HZ) != 0)
3570                                 status = 0;
3571
3572                 finish_wait(&state->waitq, &wait);
3573
3574                 if (!status)
3575                         continue;
3576                 if (status == -EINTR)
3577                         break;
3578
3579                 /* we slept the whole 5 seconds, we must have lost a seqid */
3580                 dst->seqid = cpu_to_be32(dst_seqid + 1);
3581                 ret = true;
3582                 break;
3583         }
3584
3585         return ret;
3586 }
3587
3588 struct nfs4_closedata {
3589         struct inode *inode;
3590         struct nfs4_state *state;
3591         struct nfs_closeargs arg;
3592         struct nfs_closeres res;
3593         struct {
3594                 struct nfs4_layoutreturn_args arg;
3595                 struct nfs4_layoutreturn_res res;
3596                 struct nfs4_xdr_opaque_data ld_private;
3597                 u32 roc_barrier;
3598                 bool roc;
3599         } lr;
3600         struct nfs_fattr fattr;
3601         unsigned long timestamp;
3602 };
3603
3604 static void nfs4_free_closedata(void *data)
3605 {
3606         struct nfs4_closedata *calldata = data;
3607         struct nfs4_state_owner *sp = calldata->state->owner;
3608         struct super_block *sb = calldata->state->inode->i_sb;
3609
3610         if (calldata->lr.roc)
3611                 pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3612                                 calldata->res.lr_ret);
3613         nfs4_put_open_state(calldata->state);
3614         nfs_free_seqid(calldata->arg.seqid);
3615         nfs4_put_state_owner(sp);
3616         nfs_sb_deactive(sb);
3617         kfree(calldata);
3618 }
3619
3620 static void nfs4_close_done(struct rpc_task *task, void *data)
3621 {
3622         struct nfs4_closedata *calldata = data;
3623         struct nfs4_state *state = calldata->state;
3624         struct nfs_server *server = NFS_SERVER(calldata->inode);
3625         nfs4_stateid *res_stateid = NULL;
3626         struct nfs4_exception exception = {
3627                 .state = state,
3628                 .inode = calldata->inode,
3629                 .stateid = &calldata->arg.stateid,
3630         };
3631
3632         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3633                 return;
3634         trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3635
3636         /* Handle Layoutreturn errors */
3637         if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3638                           &calldata->res.lr_ret) == -EAGAIN)
3639                 goto out_restart;
3640
3641         /* hmm. we are done with the inode, and in the process of freeing
3642          * the state_owner. we keep this around to process errors
3643          */
3644         switch (task->tk_status) {
3645                 case 0:
3646                         res_stateid = &calldata->res.stateid;
3647                         renew_lease(server, calldata->timestamp);
3648                         break;
3649                 case -NFS4ERR_ACCESS:
3650                         if (calldata->arg.bitmask != NULL) {
3651                                 calldata->arg.bitmask = NULL;
3652                                 calldata->res.fattr = NULL;
3653                                 goto out_restart;
3654
3655                         }
3656                         break;
3657                 case -NFS4ERR_OLD_STATEID:
3658                         /* Did we race with OPEN? */
3659                         if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3660                                                 state))
3661                                 goto out_restart;
3662                         goto out_release;
3663                 case -NFS4ERR_ADMIN_REVOKED:
3664                 case -NFS4ERR_STALE_STATEID:
3665                 case -NFS4ERR_EXPIRED:
3666                         nfs4_free_revoked_stateid(server,
3667                                         &calldata->arg.stateid,
3668                                         task->tk_msg.rpc_cred);
3669                         fallthrough;
3670                 case -NFS4ERR_BAD_STATEID:
3671                         if (calldata->arg.fmode == 0)
3672                                 break;
3673                         fallthrough;
3674                 default:
3675                         task->tk_status = nfs4_async_handle_exception(task,
3676                                         server, task->tk_status, &exception);
3677                         if (exception.retry)
3678                                 goto out_restart;
3679         }
3680         nfs_clear_open_stateid(state, &calldata->arg.stateid,
3681                         res_stateid, calldata->arg.fmode);
3682 out_release:
3683         task->tk_status = 0;
3684         nfs_release_seqid(calldata->arg.seqid);
3685         nfs_refresh_inode(calldata->inode, &calldata->fattr);
3686         dprintk("%s: ret = %d\n", __func__, task->tk_status);
3687         return;
3688 out_restart:
3689         task->tk_status = 0;
3690         rpc_restart_call_prepare(task);
3691         goto out_release;
3692 }
3693
3694 static void nfs4_close_prepare(struct rpc_task *task, void *data)
3695 {
3696         struct nfs4_closedata *calldata = data;
3697         struct nfs4_state *state = calldata->state;
3698         struct inode *inode = calldata->inode;
3699         struct nfs_server *server = NFS_SERVER(inode);
3700         struct pnfs_layout_hdr *lo;
3701         bool is_rdonly, is_wronly, is_rdwr;
3702         int call_close = 0;
3703
3704         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3705                 goto out_wait;
3706
3707         task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3708         spin_lock(&state->owner->so_lock);
3709         is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3710         is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3711         is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3712         /* Calculate the change in open mode */
3713         calldata->arg.fmode = 0;
3714         if (state->n_rdwr == 0) {
3715                 if (state->n_rdonly == 0)
3716                         call_close |= is_rdonly;
3717                 else if (is_rdonly)
3718                         calldata->arg.fmode |= FMODE_READ;
3719                 if (state->n_wronly == 0)
3720                         call_close |= is_wronly;
3721                 else if (is_wronly)
3722                         calldata->arg.fmode |= FMODE_WRITE;
3723                 if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3724                         call_close |= is_rdwr;
3725         } else if (is_rdwr)
3726                 calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3727
3728         nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3729         if (!nfs4_valid_open_stateid(state))
3730                 call_close = 0;
3731         spin_unlock(&state->owner->so_lock);
3732
3733         if (!call_close) {
3734                 /* Note: exit _without_ calling nfs4_close_done */
3735                 goto out_no_action;
3736         }
3737
3738         if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3739                 nfs_release_seqid(calldata->arg.seqid);
3740                 goto out_wait;
3741         }
3742
3743         lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3744         if (lo && !pnfs_layout_is_valid(lo)) {
3745                 calldata->arg.lr_args = NULL;
3746                 calldata->res.lr_res = NULL;
3747         }
3748
3749         if (calldata->arg.fmode == 0)
3750                 task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3751
3752         if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3753                 /* Close-to-open cache consistency revalidation */
3754                 if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
3755                         nfs4_bitmask_set(calldata->arg.bitmask_store,
3756                                          server->cache_consistency_bitmask,
3757                                          inode, 0);
3758                         calldata->arg.bitmask = calldata->arg.bitmask_store;
3759                 } else
3760                         calldata->arg.bitmask = NULL;
3761         }
3762
3763         calldata->arg.share_access =
3764                 nfs4_fmode_to_share_access(calldata->arg.fmode);
3765
3766         if (calldata->res.fattr == NULL)
3767                 calldata->arg.bitmask = NULL;
3768         else if (calldata->arg.bitmask == NULL)
3769                 calldata->res.fattr = NULL;
3770         calldata->timestamp = jiffies;
3771         if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3772                                 &calldata->arg.seq_args,
3773                                 &calldata->res.seq_res,
3774                                 task) != 0)
3775                 nfs_release_seqid(calldata->arg.seqid);
3776         return;
3777 out_no_action:
3778         task->tk_action = NULL;
3779 out_wait:
3780         nfs4_sequence_done(task, &calldata->res.seq_res);
3781 }
3782
3783 static const struct rpc_call_ops nfs4_close_ops = {
3784         .rpc_call_prepare = nfs4_close_prepare,
3785         .rpc_call_done = nfs4_close_done,
3786         .rpc_release = nfs4_free_closedata,
3787 };
3788
3789 /*
3790  * It is possible for data to be read/written from a mem-mapped file
3791  * after the sys_close call (which hits the vfs layer as a flush).
3792  * This means that we can't safely call nfsv4 close on a file until
3793  * the inode is cleared. This in turn means that we are not good
3794  * NFSv4 citizens - we do not indicate to the server to update the file's
3795  * share state even when we are done with one of the three share
3796  * stateid's in the inode.
3797  *
3798  * NOTE: Caller must be holding the sp->so_owner semaphore!
3799  */
3800 int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3801 {
3802         struct nfs_server *server = NFS_SERVER(state->inode);
3803         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3804         struct nfs4_closedata *calldata;
3805         struct nfs4_state_owner *sp = state->owner;
3806         struct rpc_task *task;
3807         struct rpc_message msg = {
3808                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3809                 .rpc_cred = state->owner->so_cred,
3810         };
3811         struct rpc_task_setup task_setup_data = {
3812                 .rpc_client = server->client,
3813                 .rpc_message = &msg,
3814                 .callback_ops = &nfs4_close_ops,
3815                 .workqueue = nfsiod_workqueue,
3816                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3817         };
3818         int status = -ENOMEM;
3819
3820         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3821                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
3822
3823         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3824                 &task_setup_data.rpc_client, &msg);
3825
3826         calldata = kzalloc(sizeof(*calldata), gfp_mask);
3827         if (calldata == NULL)
3828                 goto out;
3829         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3830         calldata->inode = state->inode;
3831         calldata->state = state;
3832         calldata->arg.fh = NFS_FH(state->inode);
3833         if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3834                 goto out_free_calldata;
3835         /* Serialization for the sequence id */
3836         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3837         calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3838         if (IS_ERR(calldata->arg.seqid))
3839                 goto out_free_calldata;
3840         nfs_fattr_init(&calldata->fattr);
3841         calldata->arg.fmode = 0;
3842         calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3843         calldata->res.fattr = &calldata->fattr;
3844         calldata->res.seqid = calldata->arg.seqid;
3845         calldata->res.server = server;
3846         calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3847         calldata->lr.roc = pnfs_roc(state->inode,
3848                         &calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3849         if (calldata->lr.roc) {
3850                 calldata->arg.lr_args = &calldata->lr.arg;
3851                 calldata->res.lr_res = &calldata->lr.res;
3852         }
3853         nfs_sb_active(calldata->inode->i_sb);
3854
3855         msg.rpc_argp = &calldata->arg;
3856         msg.rpc_resp = &calldata->res;
3857         task_setup_data.callback_data = calldata;
3858         task = rpc_run_task(&task_setup_data);
3859         if (IS_ERR(task))
3860                 return PTR_ERR(task);
3861         status = 0;
3862         if (wait)
3863                 status = rpc_wait_for_completion_task(task);
3864         rpc_put_task(task);
3865         return status;
3866 out_free_calldata:
3867         kfree(calldata);
3868 out:
3869         nfs4_put_open_state(state);
3870         nfs4_put_state_owner(sp);
3871         return status;
3872 }
3873
3874 static struct inode *
3875 nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3876                 int open_flags, struct iattr *attr, int *opened)
3877 {
3878         struct nfs4_state *state;
3879         struct nfs4_label l, *label;
3880
3881         label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3882
3883         /* Protect against concurrent sillydeletes */
3884         state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3885
3886         nfs4_label_release_security(label);
3887
3888         if (IS_ERR(state))
3889                 return ERR_CAST(state);
3890         return state->inode;
3891 }
3892
3893 static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3894 {
3895         if (ctx->state == NULL)
3896                 return;
3897         if (is_sync)
3898                 nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3899         else
3900                 nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3901 }
3902
3903 #define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3904 #define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3905 #define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_OPEN_ARGUMENTS - 1UL)
3906
3907 #define FATTR4_WORD2_NFS42_TIME_DELEG_MASK \
3908         (FATTR4_WORD2_TIME_DELEG_MODIFY|FATTR4_WORD2_TIME_DELEG_ACCESS)
3909 static bool nfs4_server_delegtime_capable(struct nfs4_server_caps_res *res)
3910 {
3911         u32 share_access_want = res->open_caps.oa_share_access_want[0];
3912         u32 attr_bitmask = res->attr_bitmask[2];
3913
3914         return (share_access_want & NFS4_SHARE_WANT_DELEG_TIMESTAMPS) &&
3915                ((attr_bitmask & FATTR4_WORD2_NFS42_TIME_DELEG_MASK) ==
3916                                         FATTR4_WORD2_NFS42_TIME_DELEG_MASK);
3917 }
3918
3919 static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3920 {
3921         u32 minorversion = server->nfs_client->cl_minorversion;
3922         u32 bitmask[3] = {
3923                 [0] = FATTR4_WORD0_SUPPORTED_ATTRS,
3924         };
3925         struct nfs4_server_caps_arg args = {
3926                 .fhandle = fhandle,
3927                 .bitmask = bitmask,
3928         };
3929         struct nfs4_server_caps_res res = {};
3930         struct rpc_message msg = {
3931                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3932                 .rpc_argp = &args,
3933                 .rpc_resp = &res,
3934         };
3935         int status;
3936         int i;
3937
3938         bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3939                      FATTR4_WORD0_FH_EXPIRE_TYPE |
3940                      FATTR4_WORD0_LINK_SUPPORT |
3941                      FATTR4_WORD0_SYMLINK_SUPPORT |
3942                      FATTR4_WORD0_ACLSUPPORT |
3943                      FATTR4_WORD0_CASE_INSENSITIVE |
3944                      FATTR4_WORD0_CASE_PRESERVING;
3945         if (minorversion)
3946                 bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT |
3947                              FATTR4_WORD2_OPEN_ARGUMENTS;
3948
3949         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3950         if (status == 0) {
3951                 bitmask[0] = (FATTR4_WORD0_SUPPORTED_ATTRS |
3952                               FATTR4_WORD0_FH_EXPIRE_TYPE |
3953                               FATTR4_WORD0_LINK_SUPPORT |
3954                               FATTR4_WORD0_SYMLINK_SUPPORT |
3955                               FATTR4_WORD0_ACLSUPPORT |
3956                               FATTR4_WORD0_CASE_INSENSITIVE |
3957                               FATTR4_WORD0_CASE_PRESERVING) &
3958                              res.attr_bitmask[0];
3959                 /* Sanity check the server answers */
3960                 switch (minorversion) {
3961                 case 0:
3962                         res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3963                         res.attr_bitmask[2] = 0;
3964                         break;
3965                 case 1:
3966                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3967                         bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT &
3968                                      res.attr_bitmask[2];
3969                         break;
3970                 case 2:
3971                         res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3972                         bitmask[2] = (FATTR4_WORD2_SUPPATTR_EXCLCREAT |
3973                                       FATTR4_WORD2_OPEN_ARGUMENTS) &
3974                                      res.attr_bitmask[2];
3975                 }
3976                 memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3977                 server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3978                                   NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3979                 server->fattr_valid = NFS_ATTR_FATTR_V4;
3980                 if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3981                                 res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3982                         server->caps |= NFS_CAP_ACLS;
3983                 if (res.has_links != 0)
3984                         server->caps |= NFS_CAP_HARDLINKS;
3985                 if (res.has_symlinks != 0)
3986                         server->caps |= NFS_CAP_SYMLINKS;
3987                 if (res.case_insensitive)
3988                         server->caps |= NFS_CAP_CASE_INSENSITIVE;
3989                 if (res.case_preserving)
3990                         server->caps |= NFS_CAP_CASE_PRESERVING;
3991 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
3992                 if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3993                         server->caps |= NFS_CAP_SECURITY_LABEL;
3994 #endif
3995                 if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3996                         server->caps |= NFS_CAP_FS_LOCATIONS;
3997                 if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3998                         server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3999                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
4000                         server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
4001                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
4002                         server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
4003                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
4004                         server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
4005                                 NFS_ATTR_FATTR_OWNER_NAME);
4006                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
4007                         server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
4008                                 NFS_ATTR_FATTR_GROUP_NAME);
4009                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
4010                         server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
4011                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
4012                         server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
4013                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
4014                         server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
4015                 if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
4016                         server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
4017                 memcpy(server->attr_bitmask_nl, res.attr_bitmask,
4018                                 sizeof(server->attr_bitmask));
4019                 server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
4020
4021                 if (res.open_caps.oa_share_access_want[0] &
4022                     NFS4_SHARE_WANT_OPEN_XOR_DELEGATION)
4023                         server->caps |= NFS_CAP_OPEN_XOR;
4024                 if (nfs4_server_delegtime_capable(&res))
4025                         server->caps |= NFS_CAP_DELEGTIME;
4026
4027                 memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
4028                 server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
4029                 server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
4030                 server->cache_consistency_bitmask[2] = 0;
4031
4032                 /* Avoid a regression due to buggy server */
4033                 for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
4034                         res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
4035                 memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
4036                         sizeof(server->exclcreat_bitmask));
4037
4038                 server->acl_bitmask = res.acl_bitmask;
4039                 server->fh_expire_type = res.fh_expire_type;
4040         }
4041
4042         return status;
4043 }
4044
4045 int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
4046 {
4047         struct nfs4_exception exception = {
4048                 .interruptible = true,
4049         };
4050         int err;
4051
4052         nfs4_server_set_init_caps(server);
4053         do {
4054                 err = nfs4_handle_exception(server,
4055                                 _nfs4_server_capabilities(server, fhandle),
4056                                 &exception);
4057         } while (exception.retry);
4058         return err;
4059 }
4060
4061 static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
4062                                           struct nfs_client *clp,
4063                                           struct nfs_server *server)
4064 {
4065         int i;
4066
4067         for (i = 0; i < location->nservers; i++) {
4068                 struct nfs4_string *srv_loc = &location->servers[i];
4069                 struct sockaddr_storage addr;
4070                 size_t addrlen;
4071                 struct xprt_create xprt_args = {
4072                         .ident = 0,
4073                         .net = clp->cl_net,
4074                 };
4075                 struct nfs4_add_xprt_data xprtdata = {
4076                         .clp = clp,
4077                 };
4078                 struct rpc_add_xprt_test rpcdata = {
4079                         .add_xprt_test = clp->cl_mvops->session_trunk,
4080                         .data = &xprtdata,
4081                 };
4082                 char *servername = NULL;
4083
4084                 if (!srv_loc->len)
4085                         continue;
4086
4087                 addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
4088                                                 &addr, sizeof(addr),
4089                                                 clp->cl_net, server->port);
4090                 if (!addrlen)
4091                         return;
4092                 xprt_args.dstaddr = (struct sockaddr *)&addr;
4093                 xprt_args.addrlen = addrlen;
4094                 servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
4095                 if (!servername)
4096                         return;
4097                 memcpy(servername, srv_loc->data, srv_loc->len);
4098                 servername[srv_loc->len] = '\0';
4099                 xprt_args.servername = servername;
4100
4101                 xprtdata.cred = nfs4_get_clid_cred(clp);
4102                 rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
4103                                   rpc_clnt_setup_test_and_add_xprt,
4104                                   &rpcdata);
4105                 if (xprtdata.cred)
4106                         put_cred(xprtdata.cred);
4107                 kfree(servername);
4108         }
4109 }
4110
4111 static bool _is_same_nfs4_pathname(struct nfs4_pathname *path1,
4112                                    struct nfs4_pathname *path2)
4113 {
4114         int i;
4115
4116         if (path1->ncomponents != path2->ncomponents)
4117                 return false;
4118         for (i = 0; i < path1->ncomponents; i++) {
4119                 if (path1->components[i].len != path2->components[i].len)
4120                         return false;
4121                 if (memcmp(path1->components[i].data, path2->components[i].data,
4122                                 path1->components[i].len))
4123                         return false;
4124         }
4125         return true;
4126 }
4127
4128 static int _nfs4_discover_trunking(struct nfs_server *server,
4129                                    struct nfs_fh *fhandle)
4130 {
4131         struct nfs4_fs_locations *locations = NULL;
4132         struct page *page;
4133         const struct cred *cred;
4134         struct nfs_client *clp = server->nfs_client;
4135         const struct nfs4_state_maintenance_ops *ops =
4136                 clp->cl_mvops->state_renewal_ops;
4137         int status = -ENOMEM, i;
4138
4139         cred = ops->get_state_renewal_cred(clp);
4140         if (cred == NULL) {
4141                 cred = nfs4_get_clid_cred(clp);
4142                 if (cred == NULL)
4143                         return -ENOKEY;
4144         }
4145
4146         page = alloc_page(GFP_KERNEL);
4147         if (!page)
4148                 goto out_put_cred;
4149         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4150         if (!locations)
4151                 goto out_free;
4152         locations->fattr = nfs_alloc_fattr();
4153         if (!locations->fattr)
4154                 goto out_free_2;
4155
4156         status = nfs4_proc_get_locations(server, fhandle, locations, page,
4157                                          cred);
4158         if (status)
4159                 goto out_free_3;
4160
4161         for (i = 0; i < locations->nlocations; i++) {
4162                 if (!_is_same_nfs4_pathname(&locations->fs_path,
4163                                         &locations->locations[i].rootpath))
4164                         continue;
4165                 test_fs_location_for_trunking(&locations->locations[i], clp,
4166                                               server);
4167         }
4168 out_free_3:
4169         kfree(locations->fattr);
4170 out_free_2:
4171         kfree(locations);
4172 out_free:
4173         __free_page(page);
4174 out_put_cred:
4175         put_cred(cred);
4176         return status;
4177 }
4178
4179 static int nfs4_discover_trunking(struct nfs_server *server,
4180                                   struct nfs_fh *fhandle)
4181 {
4182         struct nfs4_exception exception = {
4183                 .interruptible = true,
4184         };
4185         struct nfs_client *clp = server->nfs_client;
4186         int err = 0;
4187
4188         if (!nfs4_has_session(clp))
4189                 goto out;
4190         do {
4191                 err = nfs4_handle_exception(server,
4192                                 _nfs4_discover_trunking(server, fhandle),
4193                                 &exception);
4194         } while (exception.retry);
4195 out:
4196         return err;
4197 }
4198
4199 static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4200                 struct nfs_fsinfo *info)
4201 {
4202         u32 bitmask[3];
4203         struct nfs4_lookup_root_arg args = {
4204                 .bitmask = bitmask,
4205         };
4206         struct nfs4_lookup_res res = {
4207                 .server = server,
4208                 .fattr = info->fattr,
4209                 .fh = fhandle,
4210         };
4211         struct rpc_message msg = {
4212                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4213                 .rpc_argp = &args,
4214                 .rpc_resp = &res,
4215         };
4216
4217         bitmask[0] = nfs4_fattr_bitmap[0];
4218         bitmask[1] = nfs4_fattr_bitmap[1];
4219         /*
4220          * Process the label in the upcoming getfattr
4221          */
4222         bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4223
4224         nfs_fattr_init(info->fattr);
4225         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4226 }
4227
4228 static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4229                 struct nfs_fsinfo *info)
4230 {
4231         struct nfs4_exception exception = {
4232                 .interruptible = true,
4233         };
4234         int err;
4235         do {
4236                 err = _nfs4_lookup_root(server, fhandle, info);
4237                 trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4238                 switch (err) {
4239                 case 0:
4240                 case -NFS4ERR_WRONGSEC:
4241                         goto out;
4242                 default:
4243                         err = nfs4_handle_exception(server, err, &exception);
4244                 }
4245         } while (exception.retry);
4246 out:
4247         return err;
4248 }
4249
4250 static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4251                                 struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4252 {
4253         struct rpc_auth_create_args auth_args = {
4254                 .pseudoflavor = flavor,
4255         };
4256         struct rpc_auth *auth;
4257
4258         auth = rpcauth_create(&auth_args, server->client);
4259         if (IS_ERR(auth))
4260                 return -EACCES;
4261         return nfs4_lookup_root(server, fhandle, info);
4262 }
4263
4264 /*
4265  * Retry pseudoroot lookup with various security flavors.  We do this when:
4266  *
4267  *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4268  *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4269  *
4270  * Returns zero on success, or a negative NFS4ERR value, or a
4271  * negative errno value.
4272  */
4273 static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4274                               struct nfs_fsinfo *info)
4275 {
4276         /* Per 3530bis 15.33.5 */
4277         static const rpc_authflavor_t flav_array[] = {
4278                 RPC_AUTH_GSS_KRB5P,
4279                 RPC_AUTH_GSS_KRB5I,
4280                 RPC_AUTH_GSS_KRB5,
4281                 RPC_AUTH_UNIX,                  /* courtesy */
4282                 RPC_AUTH_NULL,
4283         };
4284         int status = -EPERM;
4285         size_t i;
4286
4287         if (server->auth_info.flavor_len > 0) {
4288                 /* try each flavor specified by user */
4289                 for (i = 0; i < server->auth_info.flavor_len; i++) {
4290                         status = nfs4_lookup_root_sec(server, fhandle, info,
4291                                                 server->auth_info.flavors[i]);
4292                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4293                                 continue;
4294                         break;
4295                 }
4296         } else {
4297                 /* no flavors specified by user, try default list */
4298                 for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4299                         status = nfs4_lookup_root_sec(server, fhandle, info,
4300                                                       flav_array[i]);
4301                         if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4302                                 continue;
4303                         break;
4304                 }
4305         }
4306
4307         /*
4308          * -EACCES could mean that the user doesn't have correct permissions
4309          * to access the mount.  It could also mean that we tried to mount
4310          * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4311          * existing mount programs don't handle -EACCES very well so it should
4312          * be mapped to -EPERM instead.
4313          */
4314         if (status == -EACCES)
4315                 status = -EPERM;
4316         return status;
4317 }
4318
4319 /**
4320  * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4321  * @server: initialized nfs_server handle
4322  * @fhandle: we fill in the pseudo-fs root file handle
4323  * @info: we fill in an FSINFO struct
4324  * @auth_probe: probe the auth flavours
4325  *
4326  * Returns zero on success, or a negative errno.
4327  */
4328 int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4329                          struct nfs_fsinfo *info,
4330                          bool auth_probe)
4331 {
4332         int status = 0;
4333
4334         if (!auth_probe)
4335                 status = nfs4_lookup_root(server, fhandle, info);
4336
4337         if (auth_probe || status == NFS4ERR_WRONGSEC)
4338                 status = server->nfs_client->cl_mvops->find_root_sec(server,
4339                                 fhandle, info);
4340
4341         if (status == 0)
4342                 status = nfs4_server_capabilities(server, fhandle);
4343         if (status == 0)
4344                 status = nfs4_do_fsinfo(server, fhandle, info);
4345
4346         return nfs4_map_errors(status);
4347 }
4348
4349 static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4350                               struct nfs_fsinfo *info)
4351 {
4352         int error;
4353         struct nfs_fattr *fattr = info->fattr;
4354
4355         error = nfs4_server_capabilities(server, mntfh);
4356         if (error < 0) {
4357                 dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4358                 return error;
4359         }
4360
4361         error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4362         if (error < 0) {
4363                 dprintk("nfs4_get_root: getattr error = %d\n", -error);
4364                 goto out;
4365         }
4366
4367         if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4368             !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4369                 memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4370
4371 out:
4372         return error;
4373 }
4374
4375 /*
4376  * Get locations and (maybe) other attributes of a referral.
4377  * Note that we'll actually follow the referral later when
4378  * we detect fsid mismatch in inode revalidation
4379  */
4380 static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4381                              const struct qstr *name, struct nfs_fattr *fattr,
4382                              struct nfs_fh *fhandle)
4383 {
4384         int status = -ENOMEM;
4385         struct page *page = NULL;
4386         struct nfs4_fs_locations *locations = NULL;
4387
4388         page = alloc_page(GFP_KERNEL);
4389         if (page == NULL)
4390                 goto out;
4391         locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4392         if (locations == NULL)
4393                 goto out;
4394
4395         locations->fattr = fattr;
4396
4397         status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4398         if (status != 0)
4399                 goto out;
4400
4401         /*
4402          * If the fsid didn't change, this is a migration event, not a
4403          * referral.  Cause us to drop into the exception handler, which
4404          * will kick off migration recovery.
4405          */
4406         if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4407                 dprintk("%s: server did not return a different fsid for"
4408                         " a referral at %s\n", __func__, name->name);
4409                 status = -NFS4ERR_MOVED;
4410                 goto out;
4411         }
4412         /* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4413         nfs_fixup_referral_attributes(fattr);
4414         memset(fhandle, 0, sizeof(struct nfs_fh));
4415 out:
4416         if (page)
4417                 __free_page(page);
4418         kfree(locations);
4419         return status;
4420 }
4421
4422 static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4423                                 struct nfs_fattr *fattr, struct inode *inode)
4424 {
4425         __u32 bitmask[NFS4_BITMASK_SZ];
4426         struct nfs4_getattr_arg args = {
4427                 .fh = fhandle,
4428                 .bitmask = bitmask,
4429         };
4430         struct nfs4_getattr_res res = {
4431                 .fattr = fattr,
4432                 .server = server,
4433         };
4434         struct rpc_message msg = {
4435                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4436                 .rpc_argp = &args,
4437                 .rpc_resp = &res,
4438         };
4439         unsigned short task_flags = 0;
4440
4441         if (nfs4_has_session(server->nfs_client))
4442                 task_flags = RPC_TASK_MOVEABLE;
4443
4444         /* Is this is an attribute revalidation, subject to softreval? */
4445         if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4446                 task_flags |= RPC_TASK_TIMEOUT;
4447
4448         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4449         nfs_fattr_init(fattr);
4450         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4451         return nfs4_do_call_sync(server->client, server, &msg,
4452                         &args.seq_args, &res.seq_res, task_flags);
4453 }
4454
4455 int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4456                                 struct nfs_fattr *fattr, struct inode *inode)
4457 {
4458         struct nfs4_exception exception = {
4459                 .interruptible = true,
4460         };
4461         int err;
4462         do {
4463                 err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4464                 trace_nfs4_getattr(server, fhandle, fattr, err);
4465                 err = nfs4_handle_exception(server, err,
4466                                 &exception);
4467         } while (exception.retry);
4468         return err;
4469 }
4470
4471 /*
4472  * The file is not closed if it is opened due to the a request to change
4473  * the size of the file. The open call will not be needed once the
4474  * VFS layer lookup-intents are implemented.
4475  *
4476  * Close is called when the inode is destroyed.
4477  * If we haven't opened the file for O_WRONLY, we
4478  * need to in the size_change case to obtain a stateid.
4479  *
4480  * Got race?
4481  * Because OPEN is always done by name in nfsv4, it is
4482  * possible that we opened a different file by the same
4483  * name.  We can recognize this race condition, but we
4484  * can't do anything about it besides returning an error.
4485  *
4486  * This will be fixed with VFS changes (lookup-intent).
4487  */
4488 static int
4489 nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4490                   struct iattr *sattr)
4491 {
4492         struct inode *inode = d_inode(dentry);
4493         const struct cred *cred = NULL;
4494         struct nfs_open_context *ctx = NULL;
4495         int status;
4496
4497         if (pnfs_ld_layoutret_on_setattr(inode) &&
4498             sattr->ia_valid & ATTR_SIZE &&
4499             sattr->ia_size < i_size_read(inode))
4500                 pnfs_commit_and_return_layout(inode);
4501
4502         nfs_fattr_init(fattr);
4503
4504         /* Deal with open(O_TRUNC) */
4505         if (sattr->ia_valid & ATTR_OPEN)
4506                 sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4507
4508         /* Optimization: if the end result is no change, don't RPC */
4509         if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4510                 return 0;
4511
4512         /* Search for an existing open(O_WRITE) file */
4513         if (sattr->ia_valid & ATTR_FILE) {
4514
4515                 ctx = nfs_file_open_context(sattr->ia_file);
4516                 if (ctx)
4517                         cred = ctx->cred;
4518         }
4519
4520         /* Return any delegations if we're going to change ACLs */
4521         if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4522                 nfs4_inode_make_writeable(inode);
4523
4524         status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4525         if (status == 0) {
4526                 nfs_setattr_update_inode(inode, sattr, fattr);
4527                 nfs_setsecurity(inode, fattr);
4528         }
4529         return status;
4530 }
4531
4532 static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4533                 struct dentry *dentry, struct nfs_fh *fhandle,
4534                 struct nfs_fattr *fattr)
4535 {
4536         struct nfs_server *server = NFS_SERVER(dir);
4537         int                    status;
4538         struct nfs4_lookup_arg args = {
4539                 .bitmask = server->attr_bitmask,
4540                 .dir_fh = NFS_FH(dir),
4541                 .name = &dentry->d_name,
4542         };
4543         struct nfs4_lookup_res res = {
4544                 .server = server,
4545                 .fattr = fattr,
4546                 .fh = fhandle,
4547         };
4548         struct rpc_message msg = {
4549                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4550                 .rpc_argp = &args,
4551                 .rpc_resp = &res,
4552         };
4553         unsigned short task_flags = 0;
4554
4555         if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4556                 task_flags = RPC_TASK_MOVEABLE;
4557
4558         /* Is this is an attribute revalidation, subject to softreval? */
4559         if (nfs_lookup_is_soft_revalidate(dentry))
4560                 task_flags |= RPC_TASK_TIMEOUT;
4561
4562         args.bitmask = nfs4_bitmask(server, fattr->label);
4563
4564         nfs_fattr_init(fattr);
4565
4566         dprintk("NFS call  lookup %pd2\n", dentry);
4567         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4568         status = nfs4_do_call_sync(clnt, server, &msg,
4569                         &args.seq_args, &res.seq_res, task_flags);
4570         dprintk("NFS reply lookup: %d\n", status);
4571         return status;
4572 }
4573
4574 static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4575 {
4576         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4577                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4578         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4579         fattr->nlink = 2;
4580 }
4581
4582 static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4583                                    struct dentry *dentry, struct nfs_fh *fhandle,
4584                                    struct nfs_fattr *fattr)
4585 {
4586         struct nfs4_exception exception = {
4587                 .interruptible = true,
4588         };
4589         struct rpc_clnt *client = *clnt;
4590         const struct qstr *name = &dentry->d_name;
4591         int err;
4592         do {
4593                 err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4594                 trace_nfs4_lookup(dir, name, err);
4595                 switch (err) {
4596                 case -NFS4ERR_BADNAME:
4597                         err = -ENOENT;
4598                         goto out;
4599                 case -NFS4ERR_MOVED:
4600                         err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4601                         if (err == -NFS4ERR_MOVED)
4602                                 err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4603                         goto out;
4604                 case -NFS4ERR_WRONGSEC:
4605                         err = -EPERM;
4606                         if (client != *clnt)
4607                                 goto out;
4608                         client = nfs4_negotiate_security(client, dir, name);
4609                         if (IS_ERR(client))
4610                                 return PTR_ERR(client);
4611
4612                         exception.retry = 1;
4613                         break;
4614                 default:
4615                         err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4616                 }
4617         } while (exception.retry);
4618
4619 out:
4620         if (err == 0)
4621                 *clnt = client;
4622         else if (client != *clnt)
4623                 rpc_shutdown_client(client);
4624
4625         return err;
4626 }
4627
4628 static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4629                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4630 {
4631         int status;
4632         struct rpc_clnt *client = NFS_CLIENT(dir);
4633
4634         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4635         if (client != NFS_CLIENT(dir)) {
4636                 rpc_shutdown_client(client);
4637                 nfs_fixup_secinfo_attributes(fattr);
4638         }
4639         return status;
4640 }
4641
4642 struct rpc_clnt *
4643 nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4644                             struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4645 {
4646         struct rpc_clnt *client = NFS_CLIENT(dir);
4647         int status;
4648
4649         status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4650         if (status < 0)
4651                 return ERR_PTR(status);
4652         return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4653 }
4654
4655 static int _nfs4_proc_lookupp(struct inode *inode,
4656                 struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4657 {
4658         struct rpc_clnt *clnt = NFS_CLIENT(inode);
4659         struct nfs_server *server = NFS_SERVER(inode);
4660         int                    status;
4661         struct nfs4_lookupp_arg args = {
4662                 .bitmask = server->attr_bitmask,
4663                 .fh = NFS_FH(inode),
4664         };
4665         struct nfs4_lookupp_res res = {
4666                 .server = server,
4667                 .fattr = fattr,
4668                 .fh = fhandle,
4669         };
4670         struct rpc_message msg = {
4671                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4672                 .rpc_argp = &args,
4673                 .rpc_resp = &res,
4674         };
4675         unsigned short task_flags = 0;
4676
4677         if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4678                 task_flags |= RPC_TASK_TIMEOUT;
4679
4680         args.bitmask = nfs4_bitmask(server, fattr->label);
4681
4682         nfs_fattr_init(fattr);
4683
4684         dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4685         status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4686                                 &res.seq_res, task_flags);
4687         dprintk("NFS reply lookupp: %d\n", status);
4688         return status;
4689 }
4690
4691 static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4692                              struct nfs_fattr *fattr)
4693 {
4694         struct nfs4_exception exception = {
4695                 .interruptible = true,
4696         };
4697         int err;
4698         do {
4699                 err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4700                 trace_nfs4_lookupp(inode, err);
4701                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4702                                 &exception);
4703         } while (exception.retry);
4704         return err;
4705 }
4706
4707 static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4708                              const struct cred *cred)
4709 {
4710         struct nfs_server *server = NFS_SERVER(inode);
4711         struct nfs4_accessargs args = {
4712                 .fh = NFS_FH(inode),
4713                 .access = entry->mask,
4714         };
4715         struct nfs4_accessres res = {
4716                 .server = server,
4717         };
4718         struct rpc_message msg = {
4719                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4720                 .rpc_argp = &args,
4721                 .rpc_resp = &res,
4722                 .rpc_cred = cred,
4723         };
4724         int status = 0;
4725
4726         if (!nfs4_have_delegation(inode, FMODE_READ, 0)) {
4727                 res.fattr = nfs_alloc_fattr();
4728                 if (res.fattr == NULL)
4729                         return -ENOMEM;
4730                 args.bitmask = server->cache_consistency_bitmask;
4731         }
4732         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4733         if (!status) {
4734                 nfs_access_set_mask(entry, res.access);
4735                 if (res.fattr)
4736                         nfs_refresh_inode(inode, res.fattr);
4737         }
4738         nfs_free_fattr(res.fattr);
4739         return status;
4740 }
4741
4742 static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4743                             const struct cred *cred)
4744 {
4745         struct nfs4_exception exception = {
4746                 .interruptible = true,
4747         };
4748         int err;
4749         do {
4750                 err = _nfs4_proc_access(inode, entry, cred);
4751                 trace_nfs4_access(inode, err);
4752                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4753                                 &exception);
4754         } while (exception.retry);
4755         return err;
4756 }
4757
4758 /*
4759  * TODO: For the time being, we don't try to get any attributes
4760  * along with any of the zero-copy operations READ, READDIR,
4761  * READLINK, WRITE.
4762  *
4763  * In the case of the first three, we want to put the GETATTR
4764  * after the read-type operation -- this is because it is hard
4765  * to predict the length of a GETATTR response in v4, and thus
4766  * align the READ data correctly.  This means that the GETATTR
4767  * may end up partially falling into the page cache, and we should
4768  * shift it into the 'tail' of the xdr_buf before processing.
4769  * To do this efficiently, we need to know the total length
4770  * of data received, which doesn't seem to be available outside
4771  * of the RPC layer.
4772  *
4773  * In the case of WRITE, we also want to put the GETATTR after
4774  * the operation -- in this case because we want to make sure
4775  * we get the post-operation mtime and size.
4776  *
4777  * Both of these changes to the XDR layer would in fact be quite
4778  * minor, but I decided to leave them for a subsequent patch.
4779  */
4780 static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4781                 unsigned int pgbase, unsigned int pglen)
4782 {
4783         struct nfs4_readlink args = {
4784                 .fh       = NFS_FH(inode),
4785                 .pgbase   = pgbase,
4786                 .pglen    = pglen,
4787                 .pages    = &page,
4788         };
4789         struct nfs4_readlink_res res;
4790         struct rpc_message msg = {
4791                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4792                 .rpc_argp = &args,
4793                 .rpc_resp = &res,
4794         };
4795
4796         return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4797 }
4798
4799 static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4800                 unsigned int pgbase, unsigned int pglen)
4801 {
4802         struct nfs4_exception exception = {
4803                 .interruptible = true,
4804         };
4805         int err;
4806         do {
4807                 err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4808                 trace_nfs4_readlink(inode, err);
4809                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
4810                                 &exception);
4811         } while (exception.retry);
4812         return err;
4813 }
4814
4815 /*
4816  * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4817  */
4818 static int
4819 nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4820                  int flags)
4821 {
4822         struct nfs_server *server = NFS_SERVER(dir);
4823         struct nfs4_label l, *ilabel;
4824         struct nfs_open_context *ctx;
4825         struct nfs4_state *state;
4826         int status = 0;
4827
4828         ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4829         if (IS_ERR(ctx))
4830                 return PTR_ERR(ctx);
4831
4832         ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4833
4834         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4835                 sattr->ia_mode &= ~current_umask();
4836         state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4837         if (IS_ERR(state)) {
4838                 status = PTR_ERR(state);
4839                 goto out;
4840         }
4841 out:
4842         nfs4_label_release_security(ilabel);
4843         put_nfs_open_context(ctx);
4844         return status;
4845 }
4846
4847 static int
4848 _nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4849 {
4850         struct nfs_server *server = NFS_SERVER(dir);
4851         struct nfs_removeargs args = {
4852                 .fh = NFS_FH(dir),
4853                 .name = *name,
4854         };
4855         struct nfs_removeres res = {
4856                 .server = server,
4857         };
4858         struct rpc_message msg = {
4859                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4860                 .rpc_argp = &args,
4861                 .rpc_resp = &res,
4862         };
4863         unsigned long timestamp = jiffies;
4864         int status;
4865
4866         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4867         if (status == 0) {
4868                 spin_lock(&dir->i_lock);
4869                 /* Removing a directory decrements nlink in the parent */
4870                 if (ftype == NF4DIR && dir->i_nlink > 2)
4871                         nfs4_dec_nlink_locked(dir);
4872                 nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4873                                               NFS_INO_INVALID_DATA);
4874                 spin_unlock(&dir->i_lock);
4875         }
4876         return status;
4877 }
4878
4879 static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4880 {
4881         struct nfs4_exception exception = {
4882                 .interruptible = true,
4883         };
4884         struct inode *inode = d_inode(dentry);
4885         int err;
4886
4887         if (inode) {
4888                 if (inode->i_nlink == 1)
4889                         nfs4_inode_return_delegation(inode);
4890                 else
4891                         nfs4_inode_make_writeable(inode);
4892         }
4893         do {
4894                 err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4895                 trace_nfs4_remove(dir, &dentry->d_name, err);
4896                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4897                                 &exception);
4898         } while (exception.retry);
4899         return err;
4900 }
4901
4902 static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4903 {
4904         struct nfs4_exception exception = {
4905                 .interruptible = true,
4906         };
4907         int err;
4908
4909         do {
4910                 err = _nfs4_proc_remove(dir, name, NF4DIR);
4911                 trace_nfs4_remove(dir, name, err);
4912                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
4913                                 &exception);
4914         } while (exception.retry);
4915         return err;
4916 }
4917
4918 static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4919                 struct dentry *dentry,
4920                 struct inode *inode)
4921 {
4922         struct nfs_removeargs *args = msg->rpc_argp;
4923         struct nfs_removeres *res = msg->rpc_resp;
4924
4925         res->server = NFS_SB(dentry->d_sb);
4926         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4927         nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4928
4929         nfs_fattr_init(res->dir_attr);
4930
4931         if (inode) {
4932                 nfs4_inode_return_delegation(inode);
4933                 nfs_d_prune_case_insensitive_aliases(inode);
4934         }
4935 }
4936
4937 static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4938 {
4939         nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4940                         &data->args.seq_args,
4941                         &data->res.seq_res,
4942                         task);
4943 }
4944
4945 static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4946 {
4947         struct nfs_unlinkdata *data = task->tk_calldata;
4948         struct nfs_removeres *res = &data->res;
4949
4950         if (!nfs4_sequence_done(task, &res->seq_res))
4951                 return 0;
4952         if (nfs4_async_handle_error(task, res->server, NULL,
4953                                     &data->timeout) == -EAGAIN)
4954                 return 0;
4955         if (task->tk_status == 0)
4956                 nfs4_update_changeattr(dir, &res->cinfo,
4957                                 res->dir_attr->time_start,
4958                                 NFS_INO_INVALID_DATA);
4959         return 1;
4960 }
4961
4962 static void nfs4_proc_rename_setup(struct rpc_message *msg,
4963                 struct dentry *old_dentry,
4964                 struct dentry *new_dentry)
4965 {
4966         struct nfs_renameargs *arg = msg->rpc_argp;
4967         struct nfs_renameres *res = msg->rpc_resp;
4968         struct inode *old_inode = d_inode(old_dentry);
4969         struct inode *new_inode = d_inode(new_dentry);
4970
4971         if (old_inode)
4972                 nfs4_inode_make_writeable(old_inode);
4973         if (new_inode)
4974                 nfs4_inode_return_delegation(new_inode);
4975         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4976         res->server = NFS_SB(old_dentry->d_sb);
4977         nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4978 }
4979
4980 static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4981 {
4982         nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4983                         &data->args.seq_args,
4984                         &data->res.seq_res,
4985                         task);
4986 }
4987
4988 static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4989                                  struct inode *new_dir)
4990 {
4991         struct nfs_renamedata *data = task->tk_calldata;
4992         struct nfs_renameres *res = &data->res;
4993
4994         if (!nfs4_sequence_done(task, &res->seq_res))
4995                 return 0;
4996         if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4997                 return 0;
4998
4999         if (task->tk_status == 0) {
5000                 nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
5001                 if (new_dir != old_dir) {
5002                         /* Note: If we moved a directory, nlink will change */
5003                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
5004                                         res->old_fattr->time_start,
5005                                         NFS_INO_INVALID_NLINK |
5006                                             NFS_INO_INVALID_DATA);
5007                         nfs4_update_changeattr(new_dir, &res->new_cinfo,
5008                                         res->new_fattr->time_start,
5009                                         NFS_INO_INVALID_NLINK |
5010                                             NFS_INO_INVALID_DATA);
5011                 } else
5012                         nfs4_update_changeattr(old_dir, &res->old_cinfo,
5013                                         res->old_fattr->time_start,
5014                                         NFS_INO_INVALID_DATA);
5015         }
5016         return 1;
5017 }
5018
5019 static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5020 {
5021         struct nfs_server *server = NFS_SERVER(inode);
5022         __u32 bitmask[NFS4_BITMASK_SZ];
5023         struct nfs4_link_arg arg = {
5024                 .fh     = NFS_FH(inode),
5025                 .dir_fh = NFS_FH(dir),
5026                 .name   = name,
5027                 .bitmask = bitmask,
5028         };
5029         struct nfs4_link_res res = {
5030                 .server = server,
5031         };
5032         struct rpc_message msg = {
5033                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
5034                 .rpc_argp = &arg,
5035                 .rpc_resp = &res,
5036         };
5037         int status = -ENOMEM;
5038
5039         res.fattr = nfs_alloc_fattr_with_label(server);
5040         if (res.fattr == NULL)
5041                 goto out;
5042
5043         nfs4_inode_make_writeable(inode);
5044         nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label),
5045                                 inode,
5046                                 NFS_INO_INVALID_CHANGE | NFS_INO_INVALID_CTIME);
5047         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
5048         if (!status) {
5049                 nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
5050                                        NFS_INO_INVALID_DATA);
5051                 nfs4_inc_nlink(inode);
5052                 status = nfs_post_op_update_inode(inode, res.fattr);
5053                 if (!status)
5054                         nfs_setsecurity(inode, res.fattr);
5055         }
5056
5057 out:
5058         nfs_free_fattr(res.fattr);
5059         return status;
5060 }
5061
5062 static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
5063 {
5064         struct nfs4_exception exception = {
5065                 .interruptible = true,
5066         };
5067         int err;
5068         do {
5069                 err = nfs4_handle_exception(NFS_SERVER(inode),
5070                                 _nfs4_proc_link(inode, dir, name),
5071                                 &exception);
5072         } while (exception.retry);
5073         return err;
5074 }
5075
5076 struct nfs4_createdata {
5077         struct rpc_message msg;
5078         struct nfs4_create_arg arg;
5079         struct nfs4_create_res res;
5080         struct nfs_fh fh;
5081         struct nfs_fattr fattr;
5082 };
5083
5084 static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
5085                 const struct qstr *name, struct iattr *sattr, u32 ftype)
5086 {
5087         struct nfs4_createdata *data;
5088
5089         data = kzalloc(sizeof(*data), GFP_KERNEL);
5090         if (data != NULL) {
5091                 struct nfs_server *server = NFS_SERVER(dir);
5092
5093                 data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
5094                 if (IS_ERR(data->fattr.label))
5095                         goto out_free;
5096
5097                 data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
5098                 data->msg.rpc_argp = &data->arg;
5099                 data->msg.rpc_resp = &data->res;
5100                 data->arg.dir_fh = NFS_FH(dir);
5101                 data->arg.server = server;
5102                 data->arg.name = name;
5103                 data->arg.attrs = sattr;
5104                 data->arg.ftype = ftype;
5105                 data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
5106                 data->arg.umask = current_umask();
5107                 data->res.server = server;
5108                 data->res.fh = &data->fh;
5109                 data->res.fattr = &data->fattr;
5110                 nfs_fattr_init(data->res.fattr);
5111         }
5112         return data;
5113 out_free:
5114         kfree(data);
5115         return NULL;
5116 }
5117
5118 static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
5119 {
5120         int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
5121                                     &data->arg.seq_args, &data->res.seq_res, 1);
5122         if (status == 0) {
5123                 spin_lock(&dir->i_lock);
5124                 /* Creating a directory bumps nlink in the parent */
5125                 if (data->arg.ftype == NF4DIR)
5126                         nfs4_inc_nlink_locked(dir);
5127                 nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5128                                               data->res.fattr->time_start,
5129                                               NFS_INO_INVALID_DATA);
5130                 spin_unlock(&dir->i_lock);
5131                 status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5132         }
5133         return status;
5134 }
5135
5136 static void nfs4_free_createdata(struct nfs4_createdata *data)
5137 {
5138         nfs4_label_free(data->fattr.label);
5139         kfree(data);
5140 }
5141
5142 static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5143                 struct folio *folio, unsigned int len, struct iattr *sattr,
5144                 struct nfs4_label *label)
5145 {
5146         struct page *page = &folio->page;
5147         struct nfs4_createdata *data;
5148         int status = -ENAMETOOLONG;
5149
5150         if (len > NFS4_MAXPATHLEN)
5151                 goto out;
5152
5153         status = -ENOMEM;
5154         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5155         if (data == NULL)
5156                 goto out;
5157
5158         data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5159         data->arg.u.symlink.pages = &page;
5160         data->arg.u.symlink.len = len;
5161         data->arg.label = label;
5162
5163         status = nfs4_do_create(dir, dentry, data);
5164
5165         nfs4_free_createdata(data);
5166 out:
5167         return status;
5168 }
5169
5170 static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5171                 struct folio *folio, unsigned int len, struct iattr *sattr)
5172 {
5173         struct nfs4_exception exception = {
5174                 .interruptible = true,
5175         };
5176         struct nfs4_label l, *label;
5177         int err;
5178
5179         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5180
5181         do {
5182                 err = _nfs4_proc_symlink(dir, dentry, folio, len, sattr, label);
5183                 trace_nfs4_symlink(dir, &dentry->d_name, err);
5184                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5185                                 &exception);
5186         } while (exception.retry);
5187
5188         nfs4_label_release_security(label);
5189         return err;
5190 }
5191
5192 static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5193                 struct iattr *sattr, struct nfs4_label *label)
5194 {
5195         struct nfs4_createdata *data;
5196         int status = -ENOMEM;
5197
5198         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5199         if (data == NULL)
5200                 goto out;
5201
5202         data->arg.label = label;
5203         status = nfs4_do_create(dir, dentry, data);
5204
5205         nfs4_free_createdata(data);
5206 out:
5207         return status;
5208 }
5209
5210 static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5211                 struct iattr *sattr)
5212 {
5213         struct nfs_server *server = NFS_SERVER(dir);
5214         struct nfs4_exception exception = {
5215                 .interruptible = true,
5216         };
5217         struct nfs4_label l, *label;
5218         int err;
5219
5220         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5221
5222         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5223                 sattr->ia_mode &= ~current_umask();
5224         do {
5225                 err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5226                 trace_nfs4_mkdir(dir, &dentry->d_name, err);
5227                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5228                                 &exception);
5229         } while (exception.retry);
5230         nfs4_label_release_security(label);
5231
5232         return err;
5233 }
5234
5235 static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5236                               struct nfs_readdir_res *nr_res)
5237 {
5238         struct inode            *dir = d_inode(nr_arg->dentry);
5239         struct nfs_server       *server = NFS_SERVER(dir);
5240         struct nfs4_readdir_arg args = {
5241                 .fh = NFS_FH(dir),
5242                 .pages = nr_arg->pages,
5243                 .pgbase = 0,
5244                 .count = nr_arg->page_len,
5245                 .plus = nr_arg->plus,
5246         };
5247         struct nfs4_readdir_res res;
5248         struct rpc_message msg = {
5249                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5250                 .rpc_argp = &args,
5251                 .rpc_resp = &res,
5252                 .rpc_cred = nr_arg->cred,
5253         };
5254         int                     status;
5255
5256         dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5257                 nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5258         if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5259                 args.bitmask = server->attr_bitmask_nl;
5260         else
5261                 args.bitmask = server->attr_bitmask;
5262
5263         nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5264         res.pgbase = args.pgbase;
5265         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5266                         &res.seq_res, 0);
5267         if (status >= 0) {
5268                 memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5269                 status += args.pgbase;
5270         }
5271
5272         nfs_invalidate_atime(dir);
5273
5274         dprintk("%s: returns %d\n", __func__, status);
5275         return status;
5276 }
5277
5278 static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5279                              struct nfs_readdir_res *res)
5280 {
5281         struct nfs4_exception exception = {
5282                 .interruptible = true,
5283         };
5284         int err;
5285         do {
5286                 err = _nfs4_proc_readdir(arg, res);
5287                 trace_nfs4_readdir(d_inode(arg->dentry), err);
5288                 err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5289                                             err, &exception);
5290         } while (exception.retry);
5291         return err;
5292 }
5293
5294 static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5295                 struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5296 {
5297         struct nfs4_createdata *data;
5298         int mode = sattr->ia_mode;
5299         int status = -ENOMEM;
5300
5301         data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5302         if (data == NULL)
5303                 goto out;
5304
5305         if (S_ISFIFO(mode))
5306                 data->arg.ftype = NF4FIFO;
5307         else if (S_ISBLK(mode)) {
5308                 data->arg.ftype = NF4BLK;
5309                 data->arg.u.device.specdata1 = MAJOR(rdev);
5310                 data->arg.u.device.specdata2 = MINOR(rdev);
5311         }
5312         else if (S_ISCHR(mode)) {
5313                 data->arg.ftype = NF4CHR;
5314                 data->arg.u.device.specdata1 = MAJOR(rdev);
5315                 data->arg.u.device.specdata2 = MINOR(rdev);
5316         } else if (!S_ISSOCK(mode)) {
5317                 status = -EINVAL;
5318                 goto out_free;
5319         }
5320
5321         data->arg.label = label;
5322         status = nfs4_do_create(dir, dentry, data);
5323 out_free:
5324         nfs4_free_createdata(data);
5325 out:
5326         return status;
5327 }
5328
5329 static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5330                 struct iattr *sattr, dev_t rdev)
5331 {
5332         struct nfs_server *server = NFS_SERVER(dir);
5333         struct nfs4_exception exception = {
5334                 .interruptible = true,
5335         };
5336         struct nfs4_label l, *label;
5337         int err;
5338
5339         label = nfs4_label_init_security(dir, dentry, sattr, &l);
5340
5341         if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5342                 sattr->ia_mode &= ~current_umask();
5343         do {
5344                 err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5345                 trace_nfs4_mknod(dir, &dentry->d_name, err);
5346                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
5347                                 &exception);
5348         } while (exception.retry);
5349
5350         nfs4_label_release_security(label);
5351
5352         return err;
5353 }
5354
5355 static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5356                  struct nfs_fsstat *fsstat)
5357 {
5358         struct nfs4_statfs_arg args = {
5359                 .fh = fhandle,
5360                 .bitmask = server->attr_bitmask,
5361         };
5362         struct nfs4_statfs_res res = {
5363                 .fsstat = fsstat,
5364         };
5365         struct rpc_message msg = {
5366                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5367                 .rpc_argp = &args,
5368                 .rpc_resp = &res,
5369         };
5370
5371         nfs_fattr_init(fsstat->fattr);
5372         return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5373 }
5374
5375 static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5376 {
5377         struct nfs4_exception exception = {
5378                 .interruptible = true,
5379         };
5380         int err;
5381         do {
5382                 err = nfs4_handle_exception(server,
5383                                 _nfs4_proc_statfs(server, fhandle, fsstat),
5384                                 &exception);
5385         } while (exception.retry);
5386         return err;
5387 }
5388
5389 static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5390                 struct nfs_fsinfo *fsinfo)
5391 {
5392         struct nfs4_fsinfo_arg args = {
5393                 .fh = fhandle,
5394                 .bitmask = server->attr_bitmask,
5395         };
5396         struct nfs4_fsinfo_res res = {
5397                 .fsinfo = fsinfo,
5398         };
5399         struct rpc_message msg = {
5400                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5401                 .rpc_argp = &args,
5402                 .rpc_resp = &res,
5403         };
5404
5405         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5406 }
5407
5408 static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5409 {
5410         struct nfs4_exception exception = {
5411                 .interruptible = true,
5412         };
5413         int err;
5414
5415         do {
5416                 err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5417                 trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5418                 if (err == 0) {
5419                         nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5420                         break;
5421                 }
5422                 err = nfs4_handle_exception(server, err, &exception);
5423         } while (exception.retry);
5424         return err;
5425 }
5426
5427 static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5428 {
5429         int error;
5430
5431         nfs_fattr_init(fsinfo->fattr);
5432         error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5433         if (error == 0) {
5434                 /* block layout checks this! */
5435                 server->pnfs_blksize = fsinfo->blksize;
5436                 set_pnfs_layoutdriver(server, fhandle, fsinfo);
5437         }
5438
5439         return error;
5440 }
5441
5442 static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5443                 struct nfs_pathconf *pathconf)
5444 {
5445         struct nfs4_pathconf_arg args = {
5446                 .fh = fhandle,
5447                 .bitmask = server->attr_bitmask,
5448         };
5449         struct nfs4_pathconf_res res = {
5450                 .pathconf = pathconf,
5451         };
5452         struct rpc_message msg = {
5453                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5454                 .rpc_argp = &args,
5455                 .rpc_resp = &res,
5456         };
5457
5458         /* None of the pathconf attributes are mandatory to implement */
5459         if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5460                 memset(pathconf, 0, sizeof(*pathconf));
5461                 return 0;
5462         }
5463
5464         nfs_fattr_init(pathconf->fattr);
5465         return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5466 }
5467
5468 static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5469                 struct nfs_pathconf *pathconf)
5470 {
5471         struct nfs4_exception exception = {
5472                 .interruptible = true,
5473         };
5474         int err;
5475
5476         do {
5477                 err = nfs4_handle_exception(server,
5478                                 _nfs4_proc_pathconf(server, fhandle, pathconf),
5479                                 &exception);
5480         } while (exception.retry);
5481         return err;
5482 }
5483
5484 int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5485                 const struct nfs_open_context *ctx,
5486                 const struct nfs_lock_context *l_ctx,
5487                 fmode_t fmode)
5488 {
5489         return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5490 }
5491 EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5492
5493 static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5494                 const struct nfs_open_context *ctx,
5495                 const struct nfs_lock_context *l_ctx,
5496                 fmode_t fmode)
5497 {
5498         nfs4_stateid _current_stateid;
5499
5500         /* If the current stateid represents a lost lock, then exit */
5501         if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5502                 return true;
5503         return nfs4_stateid_match(stateid, &_current_stateid);
5504 }
5505
5506 static bool nfs4_error_stateid_expired(int err)
5507 {
5508         switch (err) {
5509         case -NFS4ERR_DELEG_REVOKED:
5510         case -NFS4ERR_ADMIN_REVOKED:
5511         case -NFS4ERR_BAD_STATEID:
5512         case -NFS4ERR_STALE_STATEID:
5513         case -NFS4ERR_OLD_STATEID:
5514         case -NFS4ERR_OPENMODE:
5515         case -NFS4ERR_EXPIRED:
5516                 return true;
5517         }
5518         return false;
5519 }
5520
5521 static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5522 {
5523         struct nfs_server *server = NFS_SERVER(hdr->inode);
5524
5525         trace_nfs4_read(hdr, task->tk_status);
5526         if (task->tk_status < 0) {
5527                 struct nfs4_exception exception = {
5528                         .inode = hdr->inode,
5529                         .state = hdr->args.context->state,
5530                         .stateid = &hdr->args.stateid,
5531                 };
5532                 task->tk_status = nfs4_async_handle_exception(task,
5533                                 server, task->tk_status, &exception);
5534                 if (exception.retry) {
5535                         rpc_restart_call_prepare(task);
5536                         return -EAGAIN;
5537                 }
5538         }
5539
5540         if (task->tk_status > 0)
5541                 renew_lease(server, hdr->timestamp);
5542         return 0;
5543 }
5544
5545 static bool nfs4_read_stateid_changed(struct rpc_task *task,
5546                 struct nfs_pgio_args *args)
5547 {
5548
5549         if (!nfs4_error_stateid_expired(task->tk_status) ||
5550                 nfs4_stateid_is_current(&args->stateid,
5551                                 args->context,
5552                                 args->lock_context,
5553                                 FMODE_READ))
5554                 return false;
5555         rpc_restart_call_prepare(task);
5556         return true;
5557 }
5558
5559 static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5560                                          struct nfs_pgio_header *hdr)
5561 {
5562         struct nfs_server *server = NFS_SERVER(hdr->inode);
5563         struct rpc_message *msg = &task->tk_msg;
5564
5565         if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5566             task->tk_status == -ENOTSUPP) {
5567                 server->caps &= ~NFS_CAP_READ_PLUS;
5568                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5569                 rpc_restart_call_prepare(task);
5570                 return true;
5571         }
5572         return false;
5573 }
5574
5575 static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5576 {
5577         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5578                 return -EAGAIN;
5579         if (nfs4_read_stateid_changed(task, &hdr->args))
5580                 return -EAGAIN;
5581         if (nfs4_read_plus_not_supported(task, hdr))
5582                 return -EAGAIN;
5583         if (task->tk_status > 0)
5584                 nfs_invalidate_atime(hdr->inode);
5585         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5586                                     nfs4_read_done_cb(task, hdr);
5587 }
5588
5589 #if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5590 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5591                                     struct rpc_message *msg)
5592 {
5593         /* Note: We don't use READ_PLUS with pNFS yet */
5594         if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5595                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5596                 return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5597         }
5598         return false;
5599 }
5600 #else
5601 static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5602                                     struct rpc_message *msg)
5603 {
5604         return false;
5605 }
5606 #endif /* CONFIG_NFS_V4_2 */
5607
5608 static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5609                                  struct rpc_message *msg)
5610 {
5611         hdr->timestamp   = jiffies;
5612         if (!hdr->pgio_done_cb)
5613                 hdr->pgio_done_cb = nfs4_read_done_cb;
5614         if (!nfs42_read_plus_support(hdr, msg))
5615                 msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5616         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5617 }
5618
5619 static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5620                                       struct nfs_pgio_header *hdr)
5621 {
5622         if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5623                         &hdr->args.seq_args,
5624                         &hdr->res.seq_res,
5625                         task))
5626                 return 0;
5627         if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5628                                 hdr->args.lock_context,
5629                                 hdr->rw_mode) == -EIO)
5630                 return -EIO;
5631         if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5632                 return -EIO;
5633         return 0;
5634 }
5635
5636 static int nfs4_write_done_cb(struct rpc_task *task,
5637                               struct nfs_pgio_header *hdr)
5638 {
5639         struct inode *inode = hdr->inode;
5640
5641         trace_nfs4_write(hdr, task->tk_status);
5642         if (task->tk_status < 0) {
5643                 struct nfs4_exception exception = {
5644                         .inode = hdr->inode,
5645                         .state = hdr->args.context->state,
5646                         .stateid = &hdr->args.stateid,
5647                 };
5648                 task->tk_status = nfs4_async_handle_exception(task,
5649                                 NFS_SERVER(inode), task->tk_status,
5650                                 &exception);
5651                 if (exception.retry) {
5652                         rpc_restart_call_prepare(task);
5653                         return -EAGAIN;
5654                 }
5655         }
5656         if (task->tk_status >= 0) {
5657                 renew_lease(NFS_SERVER(inode), hdr->timestamp);
5658                 nfs_writeback_update_inode(hdr);
5659         }
5660         return 0;
5661 }
5662
5663 static bool nfs4_write_stateid_changed(struct rpc_task *task,
5664                 struct nfs_pgio_args *args)
5665 {
5666
5667         if (!nfs4_error_stateid_expired(task->tk_status) ||
5668                 nfs4_stateid_is_current(&args->stateid,
5669                                 args->context,
5670                                 args->lock_context,
5671                                 FMODE_WRITE))
5672                 return false;
5673         rpc_restart_call_prepare(task);
5674         return true;
5675 }
5676
5677 static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5678 {
5679         if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5680                 return -EAGAIN;
5681         if (nfs4_write_stateid_changed(task, &hdr->args))
5682                 return -EAGAIN;
5683         return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5684                 nfs4_write_done_cb(task, hdr);
5685 }
5686
5687 static
5688 bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5689 {
5690         /* Don't request attributes for pNFS or O_DIRECT writes */
5691         if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5692                 return false;
5693         /* Otherwise, request attributes if and only if we don't hold
5694          * a delegation
5695          */
5696         return nfs4_have_delegation(hdr->inode, FMODE_READ, 0) == 0;
5697 }
5698
5699 void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5700                       struct inode *inode, unsigned long cache_validity)
5701 {
5702         struct nfs_server *server = NFS_SERVER(inode);
5703         unsigned int i;
5704
5705         memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5706         cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5707
5708         if (cache_validity & NFS_INO_INVALID_CHANGE)
5709                 bitmask[0] |= FATTR4_WORD0_CHANGE;
5710         if (cache_validity & NFS_INO_INVALID_ATIME)
5711                 bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5712         if (cache_validity & NFS_INO_INVALID_MODE)
5713                 bitmask[1] |= FATTR4_WORD1_MODE;
5714         if (cache_validity & NFS_INO_INVALID_OTHER)
5715                 bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5716         if (cache_validity & NFS_INO_INVALID_NLINK)
5717                 bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5718         if (cache_validity & NFS_INO_INVALID_CTIME)
5719                 bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5720         if (cache_validity & NFS_INO_INVALID_MTIME)
5721                 bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5722         if (cache_validity & NFS_INO_INVALID_BLOCKS)
5723                 bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5724
5725         if (cache_validity & NFS_INO_INVALID_SIZE)
5726                 bitmask[0] |= FATTR4_WORD0_SIZE;
5727
5728         for (i = 0; i < NFS4_BITMASK_SZ; i++)
5729                 bitmask[i] &= server->attr_bitmask[i];
5730 }
5731
5732 static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5733                                   struct rpc_message *msg,
5734                                   struct rpc_clnt **clnt)
5735 {
5736         struct nfs_server *server = NFS_SERVER(hdr->inode);
5737
5738         if (!nfs4_write_need_cache_consistency_data(hdr)) {
5739                 hdr->args.bitmask = NULL;
5740                 hdr->res.fattr = NULL;
5741         } else {
5742                 nfs4_bitmask_set(hdr->args.bitmask_store,
5743                                  server->cache_consistency_bitmask,
5744                                  hdr->inode, NFS_INO_INVALID_BLOCKS);
5745                 hdr->args.bitmask = hdr->args.bitmask_store;
5746         }
5747
5748         if (!hdr->pgio_done_cb)
5749                 hdr->pgio_done_cb = nfs4_write_done_cb;
5750         hdr->res.server = server;
5751         hdr->timestamp   = jiffies;
5752
5753         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5754         nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5755         nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5756 }
5757
5758 static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5759 {
5760         nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5761                         &data->args.seq_args,
5762                         &data->res.seq_res,
5763                         task);
5764 }
5765
5766 static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5767 {
5768         struct inode *inode = data->inode;
5769
5770         trace_nfs4_commit(data, task->tk_status);
5771         if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5772                                     NULL, NULL) == -EAGAIN) {
5773                 rpc_restart_call_prepare(task);
5774                 return -EAGAIN;
5775         }
5776         return 0;
5777 }
5778
5779 static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5780 {
5781         if (!nfs4_sequence_done(task, &data->res.seq_res))
5782                 return -EAGAIN;
5783         return data->commit_done_cb(task, data);
5784 }
5785
5786 static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5787                                    struct rpc_clnt **clnt)
5788 {
5789         struct nfs_server *server = NFS_SERVER(data->inode);
5790
5791         if (data->commit_done_cb == NULL)
5792                 data->commit_done_cb = nfs4_commit_done_cb;
5793         data->res.server = server;
5794         msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5795         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5796         nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5797                         NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5798 }
5799
5800 static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5801                                 struct nfs_commitres *res)
5802 {
5803         struct inode *dst_inode = file_inode(dst);
5804         struct nfs_server *server = NFS_SERVER(dst_inode);
5805         struct rpc_message msg = {
5806                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5807                 .rpc_argp = args,
5808                 .rpc_resp = res,
5809         };
5810
5811         args->fh = NFS_FH(dst_inode);
5812         return nfs4_call_sync(server->client, server, &msg,
5813                         &args->seq_args, &res->seq_res, 1);
5814 }
5815
5816 int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5817 {
5818         struct nfs_commitargs args = {
5819                 .offset = offset,
5820                 .count = count,
5821         };
5822         struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5823         struct nfs4_exception exception = { };
5824         int status;
5825
5826         do {
5827                 status = _nfs4_proc_commit(dst, &args, res);
5828                 status = nfs4_handle_exception(dst_server, status, &exception);
5829         } while (exception.retry);
5830
5831         return status;
5832 }
5833
5834 struct nfs4_renewdata {
5835         struct nfs_client       *client;
5836         unsigned long           timestamp;
5837 };
5838
5839 /*
5840  * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5841  * standalone procedure for queueing an asynchronous RENEW.
5842  */
5843 static void nfs4_renew_release(void *calldata)
5844 {
5845         struct nfs4_renewdata *data = calldata;
5846         struct nfs_client *clp = data->client;
5847
5848         if (refcount_read(&clp->cl_count) > 1)
5849                 nfs4_schedule_state_renewal(clp);
5850         nfs_put_client(clp);
5851         kfree(data);
5852 }
5853
5854 static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5855 {
5856         struct nfs4_renewdata *data = calldata;
5857         struct nfs_client *clp = data->client;
5858         unsigned long timestamp = data->timestamp;
5859
5860         trace_nfs4_renew_async(clp, task->tk_status);
5861         switch (task->tk_status) {
5862         case 0:
5863                 break;
5864         case -NFS4ERR_LEASE_MOVED:
5865                 nfs4_schedule_lease_moved_recovery(clp);
5866                 break;
5867         default:
5868                 /* Unless we're shutting down, schedule state recovery! */
5869                 if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5870                         return;
5871                 if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5872                         nfs4_schedule_lease_recovery(clp);
5873                         return;
5874                 }
5875                 nfs4_schedule_path_down_recovery(clp);
5876         }
5877         do_renew_lease(clp, timestamp);
5878 }
5879
5880 static const struct rpc_call_ops nfs4_renew_ops = {
5881         .rpc_call_done = nfs4_renew_done,
5882         .rpc_release = nfs4_renew_release,
5883 };
5884
5885 static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5886 {
5887         struct rpc_message msg = {
5888                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5889                 .rpc_argp       = clp,
5890                 .rpc_cred       = cred,
5891         };
5892         struct nfs4_renewdata *data;
5893
5894         if (renew_flags == 0)
5895                 return 0;
5896         if (!refcount_inc_not_zero(&clp->cl_count))
5897                 return -EIO;
5898         data = kmalloc(sizeof(*data), GFP_NOFS);
5899         if (data == NULL) {
5900                 nfs_put_client(clp);
5901                 return -ENOMEM;
5902         }
5903         data->client = clp;
5904         data->timestamp = jiffies;
5905         return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5906                         &nfs4_renew_ops, data);
5907 }
5908
5909 static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5910 {
5911         struct rpc_message msg = {
5912                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5913                 .rpc_argp       = clp,
5914                 .rpc_cred       = cred,
5915         };
5916         unsigned long now = jiffies;
5917         int status;
5918
5919         status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5920         if (status < 0)
5921                 return status;
5922         do_renew_lease(clp, now);
5923         return 0;
5924 }
5925
5926 static bool nfs4_server_supports_acls(const struct nfs_server *server,
5927                                       enum nfs4_acl_type type)
5928 {
5929         switch (type) {
5930         default:
5931                 return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5932         case NFS4ACL_DACL:
5933                 return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5934         case NFS4ACL_SACL:
5935                 return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5936         }
5937 }
5938
5939 /* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5940  * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5941  * the stack.
5942  */
5943 #define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5944
5945 int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5946                 struct page **pages)
5947 {
5948         struct page *newpage, **spages;
5949         int rc = 0;
5950         size_t len;
5951         spages = pages;
5952
5953         do {
5954                 len = min_t(size_t, PAGE_SIZE, buflen);
5955                 newpage = alloc_page(GFP_KERNEL);
5956
5957                 if (newpage == NULL)
5958                         goto unwind;
5959                 memcpy(page_address(newpage), buf, len);
5960                 buf += len;
5961                 buflen -= len;
5962                 *pages++ = newpage;
5963                 rc++;
5964         } while (buflen != 0);
5965
5966         return rc;
5967
5968 unwind:
5969         for(; rc > 0; rc--)
5970                 __free_page(spages[rc-1]);
5971         return -ENOMEM;
5972 }
5973
5974 struct nfs4_cached_acl {
5975         enum nfs4_acl_type type;
5976         int cached;
5977         size_t len;
5978         char data[];
5979 };
5980
5981 static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5982 {
5983         struct nfs_inode *nfsi = NFS_I(inode);
5984
5985         spin_lock(&inode->i_lock);
5986         kfree(nfsi->nfs4_acl);
5987         nfsi->nfs4_acl = acl;
5988         spin_unlock(&inode->i_lock);
5989 }
5990
5991 static void nfs4_zap_acl_attr(struct inode *inode)
5992 {
5993         nfs4_set_cached_acl(inode, NULL);
5994 }
5995
5996 static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5997                                     size_t buflen, enum nfs4_acl_type type)
5998 {
5999         struct nfs_inode *nfsi = NFS_I(inode);
6000         struct nfs4_cached_acl *acl;
6001         int ret = -ENOENT;
6002
6003         spin_lock(&inode->i_lock);
6004         acl = nfsi->nfs4_acl;
6005         if (acl == NULL)
6006                 goto out;
6007         if (acl->type != type)
6008                 goto out;
6009         if (buf == NULL) /* user is just asking for length */
6010                 goto out_len;
6011         if (acl->cached == 0)
6012                 goto out;
6013         ret = -ERANGE; /* see getxattr(2) man page */
6014         if (acl->len > buflen)
6015                 goto out;
6016         memcpy(buf, acl->data, acl->len);
6017 out_len:
6018         ret = acl->len;
6019 out:
6020         spin_unlock(&inode->i_lock);
6021         return ret;
6022 }
6023
6024 static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
6025                                   size_t pgbase, size_t acl_len,
6026                                   enum nfs4_acl_type type)
6027 {
6028         struct nfs4_cached_acl *acl;
6029         size_t buflen = sizeof(*acl) + acl_len;
6030
6031         if (buflen <= PAGE_SIZE) {
6032                 acl = kmalloc(buflen, GFP_KERNEL);
6033                 if (acl == NULL)
6034                         goto out;
6035                 acl->cached = 1;
6036                 _copy_from_pages(acl->data, pages, pgbase, acl_len);
6037         } else {
6038                 acl = kmalloc(sizeof(*acl), GFP_KERNEL);
6039                 if (acl == NULL)
6040                         goto out;
6041                 acl->cached = 0;
6042         }
6043         acl->type = type;
6044         acl->len = acl_len;
6045 out:
6046         nfs4_set_cached_acl(inode, acl);
6047 }
6048
6049 /*
6050  * The getxattr API returns the required buffer length when called with a
6051  * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
6052  * the required buf.  On a NULL buf, we send a page of data to the server
6053  * guessing that the ACL request can be serviced by a page. If so, we cache
6054  * up to the page of ACL data, and the 2nd call to getxattr is serviced by
6055  * the cache. If not so, we throw away the page, and cache the required
6056  * length. The next getxattr call will then produce another round trip to
6057  * the server, this time with the input buf of the required size.
6058  */
6059 static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
6060                                        size_t buflen, enum nfs4_acl_type type)
6061 {
6062         struct page **pages;
6063         struct nfs_getaclargs args = {
6064                 .fh = NFS_FH(inode),
6065                 .acl_type = type,
6066                 .acl_len = buflen,
6067         };
6068         struct nfs_getaclres res = {
6069                 .acl_type = type,
6070                 .acl_len = buflen,
6071         };
6072         struct rpc_message msg = {
6073                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
6074                 .rpc_argp = &args,
6075                 .rpc_resp = &res,
6076         };
6077         unsigned int npages;
6078         int ret = -ENOMEM, i;
6079         struct nfs_server *server = NFS_SERVER(inode);
6080
6081         if (buflen == 0)
6082                 buflen = server->rsize;
6083
6084         npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
6085         pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
6086         if (!pages)
6087                 return -ENOMEM;
6088
6089         args.acl_pages = pages;
6090
6091         for (i = 0; i < npages; i++) {
6092                 pages[i] = alloc_page(GFP_KERNEL);
6093                 if (!pages[i])
6094                         goto out_free;
6095         }
6096
6097         /* for decoding across pages */
6098         res.acl_scratch = alloc_page(GFP_KERNEL);
6099         if (!res.acl_scratch)
6100                 goto out_free;
6101
6102         args.acl_len = npages * PAGE_SIZE;
6103
6104         dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
6105                 __func__, buf, buflen, npages, args.acl_len);
6106         ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
6107                              &msg, &args.seq_args, &res.seq_res, 0);
6108         if (ret)
6109                 goto out_free;
6110
6111         /* Handle the case where the passed-in buffer is too short */
6112         if (res.acl_flags & NFS4_ACL_TRUNC) {
6113                 /* Did the user only issue a request for the acl length? */
6114                 if (buf == NULL)
6115                         goto out_ok;
6116                 ret = -ERANGE;
6117                 goto out_free;
6118         }
6119         nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
6120                               type);
6121         if (buf) {
6122                 if (res.acl_len > buflen) {
6123                         ret = -ERANGE;
6124                         goto out_free;
6125                 }
6126                 _copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
6127         }
6128 out_ok:
6129         ret = res.acl_len;
6130 out_free:
6131         while (--i >= 0)
6132                 __free_page(pages[i]);
6133         if (res.acl_scratch)
6134                 __free_page(res.acl_scratch);
6135         kfree(pages);
6136         return ret;
6137 }
6138
6139 static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6140                                      size_t buflen, enum nfs4_acl_type type)
6141 {
6142         struct nfs4_exception exception = {
6143                 .interruptible = true,
6144         };
6145         ssize_t ret;
6146         do {
6147                 ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6148                 trace_nfs4_get_acl(inode, ret);
6149                 if (ret >= 0)
6150                         break;
6151                 ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6152         } while (exception.retry);
6153         return ret;
6154 }
6155
6156 static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6157                                  enum nfs4_acl_type type)
6158 {
6159         struct nfs_server *server = NFS_SERVER(inode);
6160         int ret;
6161
6162         if (!nfs4_server_supports_acls(server, type))
6163                 return -EOPNOTSUPP;
6164         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6165         if (ret < 0)
6166                 return ret;
6167         if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6168                 nfs_zap_acl_cache(inode);
6169         ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6170         if (ret != -ENOENT)
6171                 /* -ENOENT is returned if there is no ACL or if there is an ACL
6172                  * but no cached acl data, just the acl length */
6173                 return ret;
6174         return nfs4_get_acl_uncached(inode, buf, buflen, type);
6175 }
6176
6177 static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6178                                size_t buflen, enum nfs4_acl_type type)
6179 {
6180         struct nfs_server *server = NFS_SERVER(inode);
6181         struct page *pages[NFS4ACL_MAXPAGES];
6182         struct nfs_setaclargs arg = {
6183                 .fh = NFS_FH(inode),
6184                 .acl_type = type,
6185                 .acl_len = buflen,
6186                 .acl_pages = pages,
6187         };
6188         struct nfs_setaclres res;
6189         struct rpc_message msg = {
6190                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6191                 .rpc_argp       = &arg,
6192                 .rpc_resp       = &res,
6193         };
6194         unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6195         int ret, i;
6196
6197         /* You can't remove system.nfs4_acl: */
6198         if (buflen == 0)
6199                 return -EINVAL;
6200         if (!nfs4_server_supports_acls(server, type))
6201                 return -EOPNOTSUPP;
6202         if (npages > ARRAY_SIZE(pages))
6203                 return -ERANGE;
6204         i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6205         if (i < 0)
6206                 return i;
6207         nfs4_inode_make_writeable(inode);
6208         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6209
6210         /*
6211          * Free each page after tx, so the only ref left is
6212          * held by the network stack
6213          */
6214         for (; i > 0; i--)
6215                 put_page(pages[i-1]);
6216
6217         /*
6218          * Acl update can result in inode attribute update.
6219          * so mark the attribute cache invalid.
6220          */
6221         spin_lock(&inode->i_lock);
6222         nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6223                                              NFS_INO_INVALID_CTIME |
6224                                              NFS_INO_REVAL_FORCED);
6225         spin_unlock(&inode->i_lock);
6226         nfs_access_zap_cache(inode);
6227         nfs_zap_acl_cache(inode);
6228         return ret;
6229 }
6230
6231 static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6232                              size_t buflen, enum nfs4_acl_type type)
6233 {
6234         struct nfs4_exception exception = { };
6235         int err;
6236         do {
6237                 err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6238                 trace_nfs4_set_acl(inode, err);
6239                 if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6240                         /*
6241                          * no need to retry since the kernel
6242                          * isn't involved in encoding the ACEs.
6243                          */
6244                         err = -EINVAL;
6245                         break;
6246                 }
6247                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6248                                 &exception);
6249         } while (exception.retry);
6250         return err;
6251 }
6252
6253 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
6254 static int _nfs4_get_security_label(struct inode *inode, void *buf,
6255                                         size_t buflen)
6256 {
6257         struct nfs_server *server = NFS_SERVER(inode);
6258         struct nfs4_label label = {0, 0, buflen, buf};
6259
6260         u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6261         struct nfs_fattr fattr = {
6262                 .label = &label,
6263         };
6264         struct nfs4_getattr_arg arg = {
6265                 .fh             = NFS_FH(inode),
6266                 .bitmask        = bitmask,
6267         };
6268         struct nfs4_getattr_res res = {
6269                 .fattr          = &fattr,
6270                 .server         = server,
6271         };
6272         struct rpc_message msg = {
6273                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6274                 .rpc_argp       = &arg,
6275                 .rpc_resp       = &res,
6276         };
6277         int ret;
6278
6279         nfs_fattr_init(&fattr);
6280
6281         ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6282         if (ret)
6283                 return ret;
6284         if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6285                 return -ENOENT;
6286         return label.len;
6287 }
6288
6289 static int nfs4_get_security_label(struct inode *inode, void *buf,
6290                                         size_t buflen)
6291 {
6292         struct nfs4_exception exception = {
6293                 .interruptible = true,
6294         };
6295         int err;
6296
6297         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6298                 return -EOPNOTSUPP;
6299
6300         do {
6301                 err = _nfs4_get_security_label(inode, buf, buflen);
6302                 trace_nfs4_get_security_label(inode, err);
6303                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6304                                 &exception);
6305         } while (exception.retry);
6306         return err;
6307 }
6308
6309 static int _nfs4_do_set_security_label(struct inode *inode,
6310                 struct nfs4_label *ilabel,
6311                 struct nfs_fattr *fattr)
6312 {
6313
6314         struct iattr sattr = {0};
6315         struct nfs_server *server = NFS_SERVER(inode);
6316         const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6317         struct nfs_setattrargs arg = {
6318                 .fh             = NFS_FH(inode),
6319                 .iap            = &sattr,
6320                 .server         = server,
6321                 .bitmask        = bitmask,
6322                 .label          = ilabel,
6323         };
6324         struct nfs_setattrres res = {
6325                 .fattr          = fattr,
6326                 .server         = server,
6327         };
6328         struct rpc_message msg = {
6329                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6330                 .rpc_argp       = &arg,
6331                 .rpc_resp       = &res,
6332         };
6333         int status;
6334
6335         nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6336
6337         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6338         if (status)
6339                 dprintk("%s failed: %d\n", __func__, status);
6340
6341         return status;
6342 }
6343
6344 static int nfs4_do_set_security_label(struct inode *inode,
6345                 struct nfs4_label *ilabel,
6346                 struct nfs_fattr *fattr)
6347 {
6348         struct nfs4_exception exception = { };
6349         int err;
6350
6351         do {
6352                 err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6353                 trace_nfs4_set_security_label(inode, err);
6354                 err = nfs4_handle_exception(NFS_SERVER(inode), err,
6355                                 &exception);
6356         } while (exception.retry);
6357         return err;
6358 }
6359
6360 static int
6361 nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6362 {
6363         struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6364         struct nfs_fattr *fattr;
6365         int status;
6366
6367         if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6368                 return -EOPNOTSUPP;
6369
6370         fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6371         if (fattr == NULL)
6372                 return -ENOMEM;
6373
6374         status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6375         if (status == 0)
6376                 nfs_setsecurity(inode, fattr);
6377
6378         nfs_free_fattr(fattr);
6379         return status;
6380 }
6381 #endif  /* CONFIG_NFS_V4_SECURITY_LABEL */
6382
6383
6384 static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6385                                     nfs4_verifier *bootverf)
6386 {
6387         __be32 verf[2];
6388
6389         if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6390                 /* An impossible timestamp guarantees this value
6391                  * will never match a generated boot time. */
6392                 verf[0] = cpu_to_be32(U32_MAX);
6393                 verf[1] = cpu_to_be32(U32_MAX);
6394         } else {
6395                 struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6396                 u64 ns = ktime_to_ns(nn->boot_time);
6397
6398                 verf[0] = cpu_to_be32(ns >> 32);
6399                 verf[1] = cpu_to_be32(ns);
6400         }
6401         memcpy(bootverf->data, verf, sizeof(bootverf->data));
6402 }
6403
6404 static size_t
6405 nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6406 {
6407         struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6408         struct nfs_netns_client *nn_clp = nn->nfs_client;
6409         const char *id;
6410
6411         buf[0] = '\0';
6412
6413         if (nn_clp) {
6414                 rcu_read_lock();
6415                 id = rcu_dereference(nn_clp->identifier);
6416                 if (id)
6417                         strscpy(buf, id, buflen);
6418                 rcu_read_unlock();
6419         }
6420
6421         if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6422                 strscpy(buf, nfs4_client_id_uniquifier, buflen);
6423
6424         return strlen(buf);
6425 }
6426
6427 static int
6428 nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6429 {
6430         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6431         size_t buflen;
6432         size_t len;
6433         char *str;
6434
6435         if (clp->cl_owner_id != NULL)
6436                 return 0;
6437
6438         rcu_read_lock();
6439         len = 14 +
6440                 strlen(clp->cl_rpcclient->cl_nodename) +
6441                 1 +
6442                 strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6443                 1;
6444         rcu_read_unlock();
6445
6446         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6447         if (buflen)
6448                 len += buflen + 1;
6449
6450         if (len > NFS4_OPAQUE_LIMIT + 1)
6451                 return -EINVAL;
6452
6453         /*
6454          * Since this string is allocated at mount time, and held until the
6455          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6456          * about a memory-reclaim deadlock.
6457          */
6458         str = kmalloc(len, GFP_KERNEL);
6459         if (!str)
6460                 return -ENOMEM;
6461
6462         rcu_read_lock();
6463         if (buflen)
6464                 scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6465                           clp->cl_rpcclient->cl_nodename, buf,
6466                           rpc_peeraddr2str(clp->cl_rpcclient,
6467                                            RPC_DISPLAY_ADDR));
6468         else
6469                 scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6470                           clp->cl_rpcclient->cl_nodename,
6471                           rpc_peeraddr2str(clp->cl_rpcclient,
6472                                            RPC_DISPLAY_ADDR));
6473         rcu_read_unlock();
6474
6475         clp->cl_owner_id = str;
6476         return 0;
6477 }
6478
6479 static int
6480 nfs4_init_uniform_client_string(struct nfs_client *clp)
6481 {
6482         char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6483         size_t buflen;
6484         size_t len;
6485         char *str;
6486
6487         if (clp->cl_owner_id != NULL)
6488                 return 0;
6489
6490         len = 10 + 10 + 1 + 10 + 1 +
6491                 strlen(clp->cl_rpcclient->cl_nodename) + 1;
6492
6493         buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6494         if (buflen)
6495                 len += buflen + 1;
6496
6497         if (len > NFS4_OPAQUE_LIMIT + 1)
6498                 return -EINVAL;
6499
6500         /*
6501          * Since this string is allocated at mount time, and held until the
6502          * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6503          * about a memory-reclaim deadlock.
6504          */
6505         str = kmalloc(len, GFP_KERNEL);
6506         if (!str)
6507                 return -ENOMEM;
6508
6509         if (buflen)
6510                 scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6511                           clp->rpc_ops->version, clp->cl_minorversion,
6512                           buf, clp->cl_rpcclient->cl_nodename);
6513         else
6514                 scnprintf(str, len, "Linux NFSv%u.%u %s",
6515                           clp->rpc_ops->version, clp->cl_minorversion,
6516                           clp->cl_rpcclient->cl_nodename);
6517         clp->cl_owner_id = str;
6518         return 0;
6519 }
6520
6521 /*
6522  * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6523  * services.  Advertise one based on the address family of the
6524  * clientaddr.
6525  */
6526 static unsigned int
6527 nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6528 {
6529         if (strchr(clp->cl_ipaddr, ':') != NULL)
6530                 return scnprintf(buf, len, "tcp6");
6531         else
6532                 return scnprintf(buf, len, "tcp");
6533 }
6534
6535 static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6536 {
6537         struct nfs4_setclientid *sc = calldata;
6538
6539         if (task->tk_status == 0)
6540                 sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6541 }
6542
6543 static const struct rpc_call_ops nfs4_setclientid_ops = {
6544         .rpc_call_done = nfs4_setclientid_done,
6545 };
6546
6547 /**
6548  * nfs4_proc_setclientid - Negotiate client ID
6549  * @clp: state data structure
6550  * @program: RPC program for NFSv4 callback service
6551  * @port: IP port number for NFS4 callback service
6552  * @cred: credential to use for this call
6553  * @res: where to place the result
6554  *
6555  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6556  */
6557 int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6558                 unsigned short port, const struct cred *cred,
6559                 struct nfs4_setclientid_res *res)
6560 {
6561         nfs4_verifier sc_verifier;
6562         struct nfs4_setclientid setclientid = {
6563                 .sc_verifier = &sc_verifier,
6564                 .sc_prog = program,
6565                 .sc_clnt = clp,
6566         };
6567         struct rpc_message msg = {
6568                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6569                 .rpc_argp = &setclientid,
6570                 .rpc_resp = res,
6571                 .rpc_cred = cred,
6572         };
6573         struct rpc_task_setup task_setup_data = {
6574                 .rpc_client = clp->cl_rpcclient,
6575                 .rpc_message = &msg,
6576                 .callback_ops = &nfs4_setclientid_ops,
6577                 .callback_data = &setclientid,
6578                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6579         };
6580         unsigned long now = jiffies;
6581         int status;
6582
6583         /* nfs_client_id4 */
6584         nfs4_init_boot_verifier(clp, &sc_verifier);
6585
6586         if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6587                 status = nfs4_init_uniform_client_string(clp);
6588         else
6589                 status = nfs4_init_nonuniform_client_string(clp);
6590
6591         if (status)
6592                 goto out;
6593
6594         /* cb_client4 */
6595         setclientid.sc_netid_len =
6596                                 nfs4_init_callback_netid(clp,
6597                                                 setclientid.sc_netid,
6598                                                 sizeof(setclientid.sc_netid));
6599         setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6600                                 sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6601                                 clp->cl_ipaddr, port >> 8, port & 255);
6602
6603         dprintk("NFS call  setclientid auth=%s, '%s'\n",
6604                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6605                 clp->cl_owner_id);
6606
6607         status = nfs4_call_sync_custom(&task_setup_data);
6608         if (setclientid.sc_cred) {
6609                 kfree(clp->cl_acceptor);
6610                 clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6611                 put_rpccred(setclientid.sc_cred);
6612         }
6613
6614         if (status == 0)
6615                 do_renew_lease(clp, now);
6616 out:
6617         trace_nfs4_setclientid(clp, status);
6618         dprintk("NFS reply setclientid: %d\n", status);
6619         return status;
6620 }
6621
6622 /**
6623  * nfs4_proc_setclientid_confirm - Confirm client ID
6624  * @clp: state data structure
6625  * @arg: result of a previous SETCLIENTID
6626  * @cred: credential to use for this call
6627  *
6628  * Returns zero, a negative errno, or a negative NFS4ERR status code.
6629  */
6630 int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6631                 struct nfs4_setclientid_res *arg,
6632                 const struct cred *cred)
6633 {
6634         struct rpc_message msg = {
6635                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6636                 .rpc_argp = arg,
6637                 .rpc_cred = cred,
6638         };
6639         int status;
6640
6641         dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6642                 clp->cl_rpcclient->cl_auth->au_ops->au_name,
6643                 clp->cl_clientid);
6644         status = rpc_call_sync(clp->cl_rpcclient, &msg,
6645                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6646         trace_nfs4_setclientid_confirm(clp, status);
6647         dprintk("NFS reply setclientid_confirm: %d\n", status);
6648         return status;
6649 }
6650
6651 struct nfs4_delegreturndata {
6652         struct nfs4_delegreturnargs args;
6653         struct nfs4_delegreturnres res;
6654         struct nfs_fh fh;
6655         nfs4_stateid stateid;
6656         unsigned long timestamp;
6657         struct {
6658                 struct nfs4_layoutreturn_args arg;
6659                 struct nfs4_layoutreturn_res res;
6660                 struct nfs4_xdr_opaque_data ld_private;
6661                 u32 roc_barrier;
6662                 bool roc;
6663         } lr;
6664         struct nfs4_delegattr sattr;
6665         struct nfs_fattr fattr;
6666         int rpc_status;
6667         struct inode *inode;
6668 };
6669
6670 static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6671 {
6672         struct nfs4_delegreturndata *data = calldata;
6673         struct nfs4_exception exception = {
6674                 .inode = data->inode,
6675                 .stateid = &data->stateid,
6676                 .task_is_privileged = data->args.seq_args.sa_privileged,
6677         };
6678
6679         if (!nfs4_sequence_done(task, &data->res.seq_res))
6680                 return;
6681
6682         trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6683
6684         /* Handle Layoutreturn errors */
6685         if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6686                           &data->res.lr_ret) == -EAGAIN)
6687                 goto out_restart;
6688
6689         if (data->args.sattr_args && task->tk_status != 0) {
6690                 switch(data->res.sattr_ret) {
6691                 case 0:
6692                         data->args.sattr_args = NULL;
6693                         data->res.sattr_res = false;
6694                         break;
6695                 case -NFS4ERR_ADMIN_REVOKED:
6696                 case -NFS4ERR_DELEG_REVOKED:
6697                 case -NFS4ERR_EXPIRED:
6698                 case -NFS4ERR_BAD_STATEID:
6699                         /* Let the main handler below do stateid recovery */
6700                         break;
6701                 case -NFS4ERR_OLD_STATEID:
6702                         if (nfs4_refresh_delegation_stateid(&data->stateid,
6703                                                 data->inode))
6704                                 goto out_restart;
6705                         fallthrough;
6706                 default:
6707                         data->args.sattr_args = NULL;
6708                         data->res.sattr_res = false;
6709                         goto out_restart;
6710                 }
6711         }
6712
6713         switch (task->tk_status) {
6714         case 0:
6715                 renew_lease(data->res.server, data->timestamp);
6716                 break;
6717         case -NFS4ERR_ADMIN_REVOKED:
6718         case -NFS4ERR_DELEG_REVOKED:
6719         case -NFS4ERR_EXPIRED:
6720                 nfs4_free_revoked_stateid(data->res.server,
6721                                 data->args.stateid,
6722                                 task->tk_msg.rpc_cred);
6723                 fallthrough;
6724         case -NFS4ERR_BAD_STATEID:
6725         case -NFS4ERR_STALE_STATEID:
6726         case -ETIMEDOUT:
6727                 task->tk_status = 0;
6728                 break;
6729         case -NFS4ERR_OLD_STATEID:
6730                 if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6731                         nfs4_stateid_seqid_inc(&data->stateid);
6732                 if (data->args.bitmask) {
6733                         data->args.bitmask = NULL;
6734                         data->res.fattr = NULL;
6735                 }
6736                 goto out_restart;
6737         case -NFS4ERR_ACCESS:
6738                 if (data->args.bitmask) {
6739                         data->args.bitmask = NULL;
6740                         data->res.fattr = NULL;
6741                         goto out_restart;
6742                 }
6743                 fallthrough;
6744         default:
6745                 task->tk_status = nfs4_async_handle_exception(task,
6746                                 data->res.server, task->tk_status,
6747                                 &exception);
6748                 if (exception.retry)
6749                         goto out_restart;
6750         }
6751         nfs_delegation_mark_returned(data->inode, data->args.stateid);
6752         data->rpc_status = task->tk_status;
6753         return;
6754 out_restart:
6755         task->tk_status = 0;
6756         rpc_restart_call_prepare(task);
6757 }
6758
6759 static void nfs4_delegreturn_release(void *calldata)
6760 {
6761         struct nfs4_delegreturndata *data = calldata;
6762         struct inode *inode = data->inode;
6763
6764         if (data->lr.roc)
6765                 pnfs_roc_release(&data->lr.arg, &data->lr.res,
6766                                  data->res.lr_ret);
6767         if (inode) {
6768                 nfs4_fattr_set_prechange(&data->fattr,
6769                                          inode_peek_iversion_raw(inode));
6770                 nfs_refresh_inode(inode, &data->fattr);
6771                 nfs_iput_and_deactive(inode);
6772         }
6773         kfree(calldata);
6774 }
6775
6776 static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6777 {
6778         struct nfs4_delegreturndata *d_data;
6779         struct pnfs_layout_hdr *lo;
6780
6781         d_data = data;
6782
6783         if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6784                 nfs4_sequence_done(task, &d_data->res.seq_res);
6785                 return;
6786         }
6787
6788         lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6789         if (lo && !pnfs_layout_is_valid(lo)) {
6790                 d_data->args.lr_args = NULL;
6791                 d_data->res.lr_res = NULL;
6792         }
6793
6794         nfs4_setup_sequence(d_data->res.server->nfs_client,
6795                         &d_data->args.seq_args,
6796                         &d_data->res.seq_res,
6797                         task);
6798 }
6799
6800 static const struct rpc_call_ops nfs4_delegreturn_ops = {
6801         .rpc_call_prepare = nfs4_delegreturn_prepare,
6802         .rpc_call_done = nfs4_delegreturn_done,
6803         .rpc_release = nfs4_delegreturn_release,
6804 };
6805
6806 static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6807                                   const nfs4_stateid *stateid,
6808                                   struct nfs_delegation *delegation,
6809                                   int issync)
6810 {
6811         struct nfs4_delegreturndata *data;
6812         struct nfs_server *server = NFS_SERVER(inode);
6813         struct rpc_task *task;
6814         struct rpc_message msg = {
6815                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6816                 .rpc_cred = cred,
6817         };
6818         struct rpc_task_setup task_setup_data = {
6819                 .rpc_client = server->client,
6820                 .rpc_message = &msg,
6821                 .callback_ops = &nfs4_delegreturn_ops,
6822                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6823         };
6824         int status = 0;
6825
6826         if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6827                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
6828
6829         data = kzalloc(sizeof(*data), GFP_KERNEL);
6830         if (data == NULL)
6831                 return -ENOMEM;
6832
6833         nfs4_state_protect(server->nfs_client,
6834                         NFS_SP4_MACH_CRED_CLEANUP,
6835                         &task_setup_data.rpc_client, &msg);
6836
6837         data->args.fhandle = &data->fh;
6838         data->args.stateid = &data->stateid;
6839         nfs4_bitmask_set(data->args.bitmask_store,
6840                          server->cache_consistency_bitmask, inode, 0);
6841         data->args.bitmask = data->args.bitmask_store;
6842         nfs_copy_fh(&data->fh, NFS_FH(inode));
6843         nfs4_stateid_copy(&data->stateid, stateid);
6844         data->res.fattr = &data->fattr;
6845         data->res.server = server;
6846         data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6847         data->lr.arg.ld_private = &data->lr.ld_private;
6848         nfs_fattr_init(data->res.fattr);
6849         data->timestamp = jiffies;
6850         data->rpc_status = 0;
6851         data->inode = nfs_igrab_and_active(inode);
6852         if (data->inode || issync) {
6853                 data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6854                                         cred);
6855                 if (data->lr.roc) {
6856                         data->args.lr_args = &data->lr.arg;
6857                         data->res.lr_res = &data->lr.res;
6858                 }
6859         }
6860
6861         if (delegation &&
6862             test_bit(NFS_DELEGATION_DELEGTIME, &delegation->flags)) {
6863                 if (delegation->type & FMODE_READ) {
6864                         data->sattr.atime = inode_get_atime(inode);
6865                         data->sattr.atime_set = true;
6866                 }
6867                 if (delegation->type & FMODE_WRITE) {
6868                         data->sattr.mtime = inode_get_mtime(inode);
6869                         data->sattr.mtime_set = true;
6870                 }
6871                 data->args.sattr_args = &data->sattr;
6872                 data->res.sattr_res = true;
6873         }
6874
6875         if (!data->inode)
6876                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6877                                    1);
6878         else
6879                 nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6880                                    0);
6881
6882         task_setup_data.callback_data = data;
6883         msg.rpc_argp = &data->args;
6884         msg.rpc_resp = &data->res;
6885         task = rpc_run_task(&task_setup_data);
6886         if (IS_ERR(task))
6887                 return PTR_ERR(task);
6888         if (!issync)
6889                 goto out;
6890         status = rpc_wait_for_completion_task(task);
6891         if (status != 0)
6892                 goto out;
6893         status = data->rpc_status;
6894 out:
6895         rpc_put_task(task);
6896         return status;
6897 }
6898
6899 int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred,
6900                           const nfs4_stateid *stateid,
6901                           struct nfs_delegation *delegation, int issync)
6902 {
6903         struct nfs_server *server = NFS_SERVER(inode);
6904         struct nfs4_exception exception = { };
6905         int err;
6906         do {
6907                 err = _nfs4_proc_delegreturn(inode, cred, stateid,
6908                                              delegation, issync);
6909                 trace_nfs4_delegreturn(inode, stateid, err);
6910                 switch (err) {
6911                         case -NFS4ERR_STALE_STATEID:
6912                         case -NFS4ERR_EXPIRED:
6913                         case 0:
6914                                 return 0;
6915                 }
6916                 err = nfs4_handle_exception(server, err, &exception);
6917         } while (exception.retry);
6918         return err;
6919 }
6920
6921 static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6922 {
6923         struct inode *inode = state->inode;
6924         struct nfs_server *server = NFS_SERVER(inode);
6925         struct nfs_client *clp = server->nfs_client;
6926         struct nfs_lockt_args arg = {
6927                 .fh = NFS_FH(inode),
6928                 .fl = request,
6929         };
6930         struct nfs_lockt_res res = {
6931                 .denied = request,
6932         };
6933         struct rpc_message msg = {
6934                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6935                 .rpc_argp       = &arg,
6936                 .rpc_resp       = &res,
6937                 .rpc_cred       = state->owner->so_cred,
6938         };
6939         struct nfs4_lock_state *lsp;
6940         int status;
6941
6942         arg.lock_owner.clientid = clp->cl_clientid;
6943         status = nfs4_set_lock_state(state, request);
6944         if (status != 0)
6945                 goto out;
6946         lsp = request->fl_u.nfs4_fl.owner;
6947         arg.lock_owner.id = lsp->ls_seqid.owner_id;
6948         arg.lock_owner.s_dev = server->s_dev;
6949         status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6950         switch (status) {
6951                 case 0:
6952                         request->c.flc_type = F_UNLCK;
6953                         break;
6954                 case -NFS4ERR_DENIED:
6955                         status = 0;
6956         }
6957         request->fl_ops->fl_release_private(request);
6958         request->fl_ops = NULL;
6959 out:
6960         return status;
6961 }
6962
6963 static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6964 {
6965         struct nfs4_exception exception = {
6966                 .interruptible = true,
6967         };
6968         int err;
6969
6970         do {
6971                 err = _nfs4_proc_getlk(state, cmd, request);
6972                 trace_nfs4_get_lock(request, state, cmd, err);
6973                 err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6974                                 &exception);
6975         } while (exception.retry);
6976         return err;
6977 }
6978
6979 /*
6980  * Update the seqid of a lock stateid after receiving
6981  * NFS4ERR_OLD_STATEID
6982  */
6983 static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6984                 struct nfs4_lock_state *lsp)
6985 {
6986         struct nfs4_state *state = lsp->ls_state;
6987         bool ret = false;
6988
6989         spin_lock(&state->state_lock);
6990         if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6991                 goto out;
6992         if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6993                 nfs4_stateid_seqid_inc(dst);
6994         else
6995                 dst->seqid = lsp->ls_stateid.seqid;
6996         ret = true;
6997 out:
6998         spin_unlock(&state->state_lock);
6999         return ret;
7000 }
7001
7002 static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
7003                 struct nfs4_lock_state *lsp)
7004 {
7005         struct nfs4_state *state = lsp->ls_state;
7006         bool ret;
7007
7008         spin_lock(&state->state_lock);
7009         ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
7010         nfs4_stateid_copy(dst, &lsp->ls_stateid);
7011         spin_unlock(&state->state_lock);
7012         return ret;
7013 }
7014
7015 struct nfs4_unlockdata {
7016         struct nfs_locku_args arg;
7017         struct nfs_locku_res res;
7018         struct nfs4_lock_state *lsp;
7019         struct nfs_open_context *ctx;
7020         struct nfs_lock_context *l_ctx;
7021         struct file_lock fl;
7022         struct nfs_server *server;
7023         unsigned long timestamp;
7024 };
7025
7026 static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
7027                 struct nfs_open_context *ctx,
7028                 struct nfs4_lock_state *lsp,
7029                 struct nfs_seqid *seqid)
7030 {
7031         struct nfs4_unlockdata *p;
7032         struct nfs4_state *state = lsp->ls_state;
7033         struct inode *inode = state->inode;
7034
7035         p = kzalloc(sizeof(*p), GFP_KERNEL);
7036         if (p == NULL)
7037                 return NULL;
7038         p->arg.fh = NFS_FH(inode);
7039         p->arg.fl = &p->fl;
7040         p->arg.seqid = seqid;
7041         p->res.seqid = seqid;
7042         p->lsp = lsp;
7043         /* Ensure we don't close file until we're done freeing locks! */
7044         p->ctx = get_nfs_open_context(ctx);
7045         p->l_ctx = nfs_get_lock_context(ctx);
7046         locks_init_lock(&p->fl);
7047         locks_copy_lock(&p->fl, fl);
7048         p->server = NFS_SERVER(inode);
7049         spin_lock(&state->state_lock);
7050         nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
7051         spin_unlock(&state->state_lock);
7052         return p;
7053 }
7054
7055 static void nfs4_locku_release_calldata(void *data)
7056 {
7057         struct nfs4_unlockdata *calldata = data;
7058         nfs_free_seqid(calldata->arg.seqid);
7059         nfs4_put_lock_state(calldata->lsp);
7060         nfs_put_lock_context(calldata->l_ctx);
7061         put_nfs_open_context(calldata->ctx);
7062         kfree(calldata);
7063 }
7064
7065 static void nfs4_locku_done(struct rpc_task *task, void *data)
7066 {
7067         struct nfs4_unlockdata *calldata = data;
7068         struct nfs4_exception exception = {
7069                 .inode = calldata->lsp->ls_state->inode,
7070                 .stateid = &calldata->arg.stateid,
7071         };
7072
7073         if (!nfs4_sequence_done(task, &calldata->res.seq_res))
7074                 return;
7075         switch (task->tk_status) {
7076                 case 0:
7077                         renew_lease(calldata->server, calldata->timestamp);
7078                         locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
7079                         if (nfs4_update_lock_stateid(calldata->lsp,
7080                                         &calldata->res.stateid))
7081                                 break;
7082                         fallthrough;
7083                 case -NFS4ERR_ADMIN_REVOKED:
7084                 case -NFS4ERR_EXPIRED:
7085                         nfs4_free_revoked_stateid(calldata->server,
7086                                         &calldata->arg.stateid,
7087                                         task->tk_msg.rpc_cred);
7088                         fallthrough;
7089                 case -NFS4ERR_BAD_STATEID:
7090                 case -NFS4ERR_STALE_STATEID:
7091                         if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
7092                                                 calldata->lsp))
7093                                 rpc_restart_call_prepare(task);
7094                         break;
7095                 case -NFS4ERR_OLD_STATEID:
7096                         if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
7097                                                 calldata->lsp))
7098                                 rpc_restart_call_prepare(task);
7099                         break;
7100                 default:
7101                         task->tk_status = nfs4_async_handle_exception(task,
7102                                         calldata->server, task->tk_status,
7103                                         &exception);
7104                         if (exception.retry)
7105                                 rpc_restart_call_prepare(task);
7106         }
7107         nfs_release_seqid(calldata->arg.seqid);
7108 }
7109
7110 static void nfs4_locku_prepare(struct rpc_task *task, void *data)
7111 {
7112         struct nfs4_unlockdata *calldata = data;
7113
7114         if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
7115                 nfs_async_iocounter_wait(task, calldata->l_ctx))
7116                 return;
7117
7118         if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
7119                 goto out_wait;
7120         if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
7121                 /* Note: exit _without_ running nfs4_locku_done */
7122                 goto out_no_action;
7123         }
7124         calldata->timestamp = jiffies;
7125         if (nfs4_setup_sequence(calldata->server->nfs_client,
7126                                 &calldata->arg.seq_args,
7127                                 &calldata->res.seq_res,
7128                                 task) != 0)
7129                 nfs_release_seqid(calldata->arg.seqid);
7130         return;
7131 out_no_action:
7132         task->tk_action = NULL;
7133 out_wait:
7134         nfs4_sequence_done(task, &calldata->res.seq_res);
7135 }
7136
7137 static const struct rpc_call_ops nfs4_locku_ops = {
7138         .rpc_call_prepare = nfs4_locku_prepare,
7139         .rpc_call_done = nfs4_locku_done,
7140         .rpc_release = nfs4_locku_release_calldata,
7141 };
7142
7143 static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
7144                 struct nfs_open_context *ctx,
7145                 struct nfs4_lock_state *lsp,
7146                 struct nfs_seqid *seqid)
7147 {
7148         struct nfs4_unlockdata *data;
7149         struct rpc_message msg = {
7150                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
7151                 .rpc_cred = ctx->cred,
7152         };
7153         struct rpc_task_setup task_setup_data = {
7154                 .rpc_client = NFS_CLIENT(lsp->ls_state->inode),
7155                 .rpc_message = &msg,
7156                 .callback_ops = &nfs4_locku_ops,
7157                 .workqueue = nfsiod_workqueue,
7158                 .flags = RPC_TASK_ASYNC,
7159         };
7160
7161         if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
7162                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7163
7164         nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
7165                 NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
7166
7167         /* Ensure this is an unlock - when canceling a lock, the
7168          * canceled lock is passed in, and it won't be an unlock.
7169          */
7170         fl->c.flc_type = F_UNLCK;
7171         if (fl->c.flc_flags & FL_CLOSE)
7172                 set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
7173
7174         data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
7175         if (data == NULL) {
7176                 nfs_free_seqid(seqid);
7177                 return ERR_PTR(-ENOMEM);
7178         }
7179
7180         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7181         msg.rpc_argp = &data->arg;
7182         msg.rpc_resp = &data->res;
7183         task_setup_data.callback_data = data;
7184         return rpc_run_task(&task_setup_data);
7185 }
7186
7187 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7188 {
7189         struct inode *inode = state->inode;
7190         struct nfs4_state_owner *sp = state->owner;
7191         struct nfs_inode *nfsi = NFS_I(inode);
7192         struct nfs_seqid *seqid;
7193         struct nfs4_lock_state *lsp;
7194         struct rpc_task *task;
7195         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7196         int status = 0;
7197         unsigned char saved_flags = request->c.flc_flags;
7198
7199         status = nfs4_set_lock_state(state, request);
7200         /* Unlock _before_ we do the RPC call */
7201         request->c.flc_flags |= FL_EXISTS;
7202         /* Exclude nfs_delegation_claim_locks() */
7203         mutex_lock(&sp->so_delegreturn_mutex);
7204         /* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7205         down_read(&nfsi->rwsem);
7206         if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7207                 up_read(&nfsi->rwsem);
7208                 mutex_unlock(&sp->so_delegreturn_mutex);
7209                 goto out;
7210         }
7211         lsp = request->fl_u.nfs4_fl.owner;
7212         set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7213         up_read(&nfsi->rwsem);
7214         mutex_unlock(&sp->so_delegreturn_mutex);
7215         if (status != 0)
7216                 goto out;
7217         /* Is this a delegated lock? */
7218         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7219                 goto out;
7220         alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7221         seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7222         status = -ENOMEM;
7223         if (IS_ERR(seqid))
7224                 goto out;
7225         task = nfs4_do_unlck(request,
7226                              nfs_file_open_context(request->c.flc_file),
7227                              lsp, seqid);
7228         status = PTR_ERR(task);
7229         if (IS_ERR(task))
7230                 goto out;
7231         status = rpc_wait_for_completion_task(task);
7232         rpc_put_task(task);
7233 out:
7234         request->c.flc_flags = saved_flags;
7235         trace_nfs4_unlock(request, state, F_SETLK, status);
7236         return status;
7237 }
7238
7239 struct nfs4_lockdata {
7240         struct nfs_lock_args arg;
7241         struct nfs_lock_res res;
7242         struct nfs4_lock_state *lsp;
7243         struct nfs_open_context *ctx;
7244         struct file_lock fl;
7245         unsigned long timestamp;
7246         int rpc_status;
7247         int cancelled;
7248         struct nfs_server *server;
7249 };
7250
7251 static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7252                 struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7253                 gfp_t gfp_mask)
7254 {
7255         struct nfs4_lockdata *p;
7256         struct inode *inode = lsp->ls_state->inode;
7257         struct nfs_server *server = NFS_SERVER(inode);
7258         struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7259
7260         p = kzalloc(sizeof(*p), gfp_mask);
7261         if (p == NULL)
7262                 return NULL;
7263
7264         p->arg.fh = NFS_FH(inode);
7265         p->arg.fl = &p->fl;
7266         p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7267         if (IS_ERR(p->arg.open_seqid))
7268                 goto out_free;
7269         alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7270         p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7271         if (IS_ERR(p->arg.lock_seqid))
7272                 goto out_free_seqid;
7273         p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7274         p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7275         p->arg.lock_owner.s_dev = server->s_dev;
7276         p->res.lock_seqid = p->arg.lock_seqid;
7277         p->lsp = lsp;
7278         p->server = server;
7279         p->ctx = get_nfs_open_context(ctx);
7280         locks_init_lock(&p->fl);
7281         locks_copy_lock(&p->fl, fl);
7282         return p;
7283 out_free_seqid:
7284         nfs_free_seqid(p->arg.open_seqid);
7285 out_free:
7286         kfree(p);
7287         return NULL;
7288 }
7289
7290 static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7291 {
7292         struct nfs4_lockdata *data = calldata;
7293         struct nfs4_state *state = data->lsp->ls_state;
7294
7295         if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7296                 goto out_wait;
7297         /* Do we need to do an open_to_lock_owner? */
7298         if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7299                 if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7300                         goto out_release_lock_seqid;
7301                 }
7302                 nfs4_stateid_copy(&data->arg.open_stateid,
7303                                 &state->open_stateid);
7304                 data->arg.new_lock_owner = 1;
7305                 data->res.open_seqid = data->arg.open_seqid;
7306         } else {
7307                 data->arg.new_lock_owner = 0;
7308                 nfs4_stateid_copy(&data->arg.lock_stateid,
7309                                 &data->lsp->ls_stateid);
7310         }
7311         if (!nfs4_valid_open_stateid(state)) {
7312                 data->rpc_status = -EBADF;
7313                 task->tk_action = NULL;
7314                 goto out_release_open_seqid;
7315         }
7316         data->timestamp = jiffies;
7317         if (nfs4_setup_sequence(data->server->nfs_client,
7318                                 &data->arg.seq_args,
7319                                 &data->res.seq_res,
7320                                 task) == 0)
7321                 return;
7322 out_release_open_seqid:
7323         nfs_release_seqid(data->arg.open_seqid);
7324 out_release_lock_seqid:
7325         nfs_release_seqid(data->arg.lock_seqid);
7326 out_wait:
7327         nfs4_sequence_done(task, &data->res.seq_res);
7328         dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7329 }
7330
7331 static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7332 {
7333         struct nfs4_lockdata *data = calldata;
7334         struct nfs4_lock_state *lsp = data->lsp;
7335
7336         if (!nfs4_sequence_done(task, &data->res.seq_res))
7337                 return;
7338
7339         data->rpc_status = task->tk_status;
7340         switch (task->tk_status) {
7341         case 0:
7342                 renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7343                                 data->timestamp);
7344                 if (data->arg.new_lock && !data->cancelled) {
7345                         data->fl.c.flc_flags &= ~(FL_SLEEP | FL_ACCESS);
7346                         if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7347                                 goto out_restart;
7348                 }
7349                 if (data->arg.new_lock_owner != 0) {
7350                         nfs_confirm_seqid(&lsp->ls_seqid, 0);
7351                         nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7352                         set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7353                 } else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7354                         goto out_restart;
7355                 break;
7356         case -NFS4ERR_OLD_STATEID:
7357                 if (data->arg.new_lock_owner != 0 &&
7358                         nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7359                                         lsp->ls_state))
7360                         goto out_restart;
7361                 if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7362                         goto out_restart;
7363                 fallthrough;
7364         case -NFS4ERR_BAD_STATEID:
7365         case -NFS4ERR_STALE_STATEID:
7366         case -NFS4ERR_EXPIRED:
7367                 if (data->arg.new_lock_owner != 0) {
7368                         if (!nfs4_stateid_match(&data->arg.open_stateid,
7369                                                 &lsp->ls_state->open_stateid))
7370                                 goto out_restart;
7371                 } else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7372                                                 &lsp->ls_stateid))
7373                                 goto out_restart;
7374         }
7375 out_done:
7376         dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7377         return;
7378 out_restart:
7379         if (!data->cancelled)
7380                 rpc_restart_call_prepare(task);
7381         goto out_done;
7382 }
7383
7384 static void nfs4_lock_release(void *calldata)
7385 {
7386         struct nfs4_lockdata *data = calldata;
7387
7388         nfs_free_seqid(data->arg.open_seqid);
7389         if (data->cancelled && data->rpc_status == 0) {
7390                 struct rpc_task *task;
7391                 task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7392                                 data->arg.lock_seqid);
7393                 if (!IS_ERR(task))
7394                         rpc_put_task_async(task);
7395                 dprintk("%s: cancelling lock!\n", __func__);
7396         } else
7397                 nfs_free_seqid(data->arg.lock_seqid);
7398         nfs4_put_lock_state(data->lsp);
7399         put_nfs_open_context(data->ctx);
7400         kfree(data);
7401 }
7402
7403 static const struct rpc_call_ops nfs4_lock_ops = {
7404         .rpc_call_prepare = nfs4_lock_prepare,
7405         .rpc_call_done = nfs4_lock_done,
7406         .rpc_release = nfs4_lock_release,
7407 };
7408
7409 static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7410 {
7411         switch (error) {
7412         case -NFS4ERR_ADMIN_REVOKED:
7413         case -NFS4ERR_EXPIRED:
7414         case -NFS4ERR_BAD_STATEID:
7415                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7416                 if (new_lock_owner != 0 ||
7417                    test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7418                         nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7419                 break;
7420         case -NFS4ERR_STALE_STATEID:
7421                 lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7422                 nfs4_schedule_lease_recovery(server->nfs_client);
7423         }
7424 }
7425
7426 static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7427 {
7428         struct nfs4_lockdata *data;
7429         struct rpc_task *task;
7430         struct rpc_message msg = {
7431                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7432                 .rpc_cred = state->owner->so_cred,
7433         };
7434         struct rpc_task_setup task_setup_data = {
7435                 .rpc_client = NFS_CLIENT(state->inode),
7436                 .rpc_message = &msg,
7437                 .callback_ops = &nfs4_lock_ops,
7438                 .workqueue = nfsiod_workqueue,
7439                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7440         };
7441         int ret;
7442
7443         if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7444                 task_setup_data.flags |= RPC_TASK_MOVEABLE;
7445
7446         data = nfs4_alloc_lockdata(fl,
7447                                    nfs_file_open_context(fl->c.flc_file),
7448                                    fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7449         if (data == NULL)
7450                 return -ENOMEM;
7451         if (IS_SETLKW(cmd))
7452                 data->arg.block = 1;
7453         nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7454                                 recovery_type > NFS_LOCK_NEW);
7455         msg.rpc_argp = &data->arg;
7456         msg.rpc_resp = &data->res;
7457         task_setup_data.callback_data = data;
7458         if (recovery_type > NFS_LOCK_NEW) {
7459                 if (recovery_type == NFS_LOCK_RECLAIM)
7460                         data->arg.reclaim = NFS_LOCK_RECLAIM;
7461         } else
7462                 data->arg.new_lock = 1;
7463         task = rpc_run_task(&task_setup_data);
7464         if (IS_ERR(task))
7465                 return PTR_ERR(task);
7466         ret = rpc_wait_for_completion_task(task);
7467         if (ret == 0) {
7468                 ret = data->rpc_status;
7469                 if (ret)
7470                         nfs4_handle_setlk_error(data->server, data->lsp,
7471                                         data->arg.new_lock_owner, ret);
7472         } else
7473                 data->cancelled = true;
7474         trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7475         rpc_put_task(task);
7476         dprintk("%s: ret = %d\n", __func__, ret);
7477         return ret;
7478 }
7479
7480 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7481 {
7482         struct nfs_server *server = NFS_SERVER(state->inode);
7483         struct nfs4_exception exception = {
7484                 .inode = state->inode,
7485         };
7486         int err;
7487
7488         do {
7489                 /* Cache the lock if possible... */
7490                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7491                         return 0;
7492                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7493                 if (err != -NFS4ERR_DELAY)
7494                         break;
7495                 nfs4_handle_exception(server, err, &exception);
7496         } while (exception.retry);
7497         return err;
7498 }
7499
7500 static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7501 {
7502         struct nfs_server *server = NFS_SERVER(state->inode);
7503         struct nfs4_exception exception = {
7504                 .inode = state->inode,
7505         };
7506         int err;
7507
7508         err = nfs4_set_lock_state(state, request);
7509         if (err != 0)
7510                 return err;
7511         if (!recover_lost_locks) {
7512                 set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7513                 return 0;
7514         }
7515         do {
7516                 if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7517                         return 0;
7518                 err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7519                 switch (err) {
7520                 default:
7521                         goto out;
7522                 case -NFS4ERR_GRACE:
7523                 case -NFS4ERR_DELAY:
7524                         nfs4_handle_exception(server, err, &exception);
7525                         err = 0;
7526                 }
7527         } while (exception.retry);
7528 out:
7529         return err;
7530 }
7531
7532 #if defined(CONFIG_NFS_V4_1)
7533 static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7534 {
7535         struct nfs4_lock_state *lsp;
7536         int status;
7537
7538         status = nfs4_set_lock_state(state, request);
7539         if (status != 0)
7540                 return status;
7541         lsp = request->fl_u.nfs4_fl.owner;
7542         if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7543             test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7544                 return 0;
7545         return nfs4_lock_expired(state, request);
7546 }
7547 #endif
7548
7549 static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7550 {
7551         struct nfs_inode *nfsi = NFS_I(state->inode);
7552         struct nfs4_state_owner *sp = state->owner;
7553         unsigned char flags = request->c.flc_flags;
7554         int status;
7555
7556         request->c.flc_flags |= FL_ACCESS;
7557         status = locks_lock_inode_wait(state->inode, request);
7558         if (status < 0)
7559                 goto out;
7560         mutex_lock(&sp->so_delegreturn_mutex);
7561         down_read(&nfsi->rwsem);
7562         if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7563                 /* Yes: cache locks! */
7564                 /* ...but avoid races with delegation recall... */
7565                 request->c.flc_flags = flags & ~FL_SLEEP;
7566                 status = locks_lock_inode_wait(state->inode, request);
7567                 up_read(&nfsi->rwsem);
7568                 mutex_unlock(&sp->so_delegreturn_mutex);
7569                 goto out;
7570         }
7571         up_read(&nfsi->rwsem);
7572         mutex_unlock(&sp->so_delegreturn_mutex);
7573         status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7574 out:
7575         request->c.flc_flags = flags;
7576         return status;
7577 }
7578
7579 static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7580 {
7581         struct nfs4_exception exception = {
7582                 .state = state,
7583                 .inode = state->inode,
7584                 .interruptible = true,
7585         };
7586         int err;
7587
7588         do {
7589                 err = _nfs4_proc_setlk(state, cmd, request);
7590                 if (err == -NFS4ERR_DENIED)
7591                         err = -EAGAIN;
7592                 err = nfs4_handle_exception(NFS_SERVER(state->inode),
7593                                 err, &exception);
7594         } while (exception.retry);
7595         return err;
7596 }
7597
7598 #define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7599 #define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7600
7601 static int
7602 nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7603                         struct file_lock *request)
7604 {
7605         int             status = -ERESTARTSYS;
7606         unsigned long   timeout = NFS4_LOCK_MINTIMEOUT;
7607
7608         while(!signalled()) {
7609                 status = nfs4_proc_setlk(state, cmd, request);
7610                 if ((status != -EAGAIN) || IS_SETLK(cmd))
7611                         break;
7612                 __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7613                 schedule_timeout(timeout);
7614                 timeout *= 2;
7615                 timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7616                 status = -ERESTARTSYS;
7617         }
7618         return status;
7619 }
7620
7621 #ifdef CONFIG_NFS_V4_1
7622 struct nfs4_lock_waiter {
7623         struct inode            *inode;
7624         struct nfs_lowner       owner;
7625         wait_queue_entry_t      wait;
7626 };
7627
7628 static int
7629 nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7630 {
7631         struct nfs4_lock_waiter *waiter =
7632                 container_of(wait, struct nfs4_lock_waiter, wait);
7633
7634         /* NULL key means to wake up everyone */
7635         if (key) {
7636                 struct cb_notify_lock_args      *cbnl = key;
7637                 struct nfs_lowner               *lowner = &cbnl->cbnl_owner,
7638                                                 *wowner = &waiter->owner;
7639
7640                 /* Only wake if the callback was for the same owner. */
7641                 if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7642                         return 0;
7643
7644                 /* Make sure it's for the right inode */
7645                 if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7646                         return 0;
7647         }
7648
7649         return woken_wake_function(wait, mode, flags, key);
7650 }
7651
7652 static int
7653 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7654 {
7655         struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7656         struct nfs_server *server = NFS_SERVER(state->inode);
7657         struct nfs_client *clp = server->nfs_client;
7658         wait_queue_head_t *q = &clp->cl_lock_waitq;
7659         struct nfs4_lock_waiter waiter = {
7660                 .inode = state->inode,
7661                 .owner = { .clientid = clp->cl_clientid,
7662                            .id = lsp->ls_seqid.owner_id,
7663                            .s_dev = server->s_dev },
7664         };
7665         int status;
7666
7667         /* Don't bother with waitqueue if we don't expect a callback */
7668         if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7669                 return nfs4_retry_setlk_simple(state, cmd, request);
7670
7671         init_wait(&waiter.wait);
7672         waiter.wait.func = nfs4_wake_lock_waiter;
7673         add_wait_queue(q, &waiter.wait);
7674
7675         do {
7676                 status = nfs4_proc_setlk(state, cmd, request);
7677                 if (status != -EAGAIN || IS_SETLK(cmd))
7678                         break;
7679
7680                 status = -ERESTARTSYS;
7681                 wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7682                            NFS4_LOCK_MAXTIMEOUT);
7683         } while (!signalled());
7684
7685         remove_wait_queue(q, &waiter.wait);
7686
7687         return status;
7688 }
7689 #else /* !CONFIG_NFS_V4_1 */
7690 static inline int
7691 nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7692 {
7693         return nfs4_retry_setlk_simple(state, cmd, request);
7694 }
7695 #endif
7696
7697 static int
7698 nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7699 {
7700         struct nfs_open_context *ctx;
7701         struct nfs4_state *state;
7702         int status;
7703
7704         /* verify open state */
7705         ctx = nfs_file_open_context(filp);
7706         state = ctx->state;
7707
7708         if (IS_GETLK(cmd)) {
7709                 if (state != NULL)
7710                         return nfs4_proc_getlk(state, F_GETLK, request);
7711                 return 0;
7712         }
7713
7714         if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7715                 return -EINVAL;
7716
7717         if (lock_is_unlock(request)) {
7718                 if (state != NULL)
7719                         return nfs4_proc_unlck(state, cmd, request);
7720                 return 0;
7721         }
7722
7723         if (state == NULL)
7724                 return -ENOLCK;
7725
7726         if ((request->c.flc_flags & FL_POSIX) &&
7727             !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7728                 return -ENOLCK;
7729
7730         /*
7731          * Don't rely on the VFS having checked the file open mode,
7732          * since it won't do this for flock() locks.
7733          */
7734         switch (request->c.flc_type) {
7735         case F_RDLCK:
7736                 if (!(filp->f_mode & FMODE_READ))
7737                         return -EBADF;
7738                 break;
7739         case F_WRLCK:
7740                 if (!(filp->f_mode & FMODE_WRITE))
7741                         return -EBADF;
7742         }
7743
7744         status = nfs4_set_lock_state(state, request);
7745         if (status != 0)
7746                 return status;
7747
7748         return nfs4_retry_setlk(state, cmd, request);
7749 }
7750
7751 static int nfs4_delete_lease(struct file *file, void **priv)
7752 {
7753         return generic_setlease(file, F_UNLCK, NULL, priv);
7754 }
7755
7756 static int nfs4_add_lease(struct file *file, int arg, struct file_lease **lease,
7757                           void **priv)
7758 {
7759         struct inode *inode = file_inode(file);
7760         fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7761         int ret;
7762
7763         /* No delegation, no lease */
7764         if (!nfs4_have_delegation(inode, type, 0))
7765                 return -EAGAIN;
7766         ret = generic_setlease(file, arg, lease, priv);
7767         if (ret || nfs4_have_delegation(inode, type, 0))
7768                 return ret;
7769         /* We raced with a delegation return */
7770         nfs4_delete_lease(file, priv);
7771         return -EAGAIN;
7772 }
7773
7774 int nfs4_proc_setlease(struct file *file, int arg, struct file_lease **lease,
7775                        void **priv)
7776 {
7777         switch (arg) {
7778         case F_RDLCK:
7779         case F_WRLCK:
7780                 return nfs4_add_lease(file, arg, lease, priv);
7781         case F_UNLCK:
7782                 return nfs4_delete_lease(file, priv);
7783         default:
7784                 return -EINVAL;
7785         }
7786 }
7787
7788 int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7789 {
7790         struct nfs_server *server = NFS_SERVER(state->inode);
7791         int err;
7792
7793         err = nfs4_set_lock_state(state, fl);
7794         if (err != 0)
7795                 return err;
7796         do {
7797                 err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7798                 if (err != -NFS4ERR_DELAY)
7799                         break;
7800                 ssleep(1);
7801         } while (err == -NFS4ERR_DELAY);
7802         return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7803 }
7804
7805 struct nfs_release_lockowner_data {
7806         struct nfs4_lock_state *lsp;
7807         struct nfs_server *server;
7808         struct nfs_release_lockowner_args args;
7809         struct nfs_release_lockowner_res res;
7810         unsigned long timestamp;
7811 };
7812
7813 static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7814 {
7815         struct nfs_release_lockowner_data *data = calldata;
7816         struct nfs_server *server = data->server;
7817         nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7818                            &data->res.seq_res, task);
7819         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7820         data->timestamp = jiffies;
7821 }
7822
7823 static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7824 {
7825         struct nfs_release_lockowner_data *data = calldata;
7826         struct nfs_server *server = data->server;
7827
7828         nfs40_sequence_done(task, &data->res.seq_res);
7829
7830         switch (task->tk_status) {
7831         case 0:
7832                 renew_lease(server, data->timestamp);
7833                 break;
7834         case -NFS4ERR_STALE_CLIENTID:
7835         case -NFS4ERR_EXPIRED:
7836                 nfs4_schedule_lease_recovery(server->nfs_client);
7837                 break;
7838         case -NFS4ERR_LEASE_MOVED:
7839         case -NFS4ERR_DELAY:
7840                 if (nfs4_async_handle_error(task, server,
7841                                             NULL, NULL) == -EAGAIN)
7842                         rpc_restart_call_prepare(task);
7843         }
7844 }
7845
7846 static void nfs4_release_lockowner_release(void *calldata)
7847 {
7848         struct nfs_release_lockowner_data *data = calldata;
7849         nfs4_free_lock_state(data->server, data->lsp);
7850         kfree(calldata);
7851 }
7852
7853 static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7854         .rpc_call_prepare = nfs4_release_lockowner_prepare,
7855         .rpc_call_done = nfs4_release_lockowner_done,
7856         .rpc_release = nfs4_release_lockowner_release,
7857 };
7858
7859 static void
7860 nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7861 {
7862         struct nfs_release_lockowner_data *data;
7863         struct rpc_message msg = {
7864                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7865         };
7866
7867         if (server->nfs_client->cl_mvops->minor_version != 0)
7868                 return;
7869
7870         data = kmalloc(sizeof(*data), GFP_KERNEL);
7871         if (!data)
7872                 return;
7873         data->lsp = lsp;
7874         data->server = server;
7875         data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7876         data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7877         data->args.lock_owner.s_dev = server->s_dev;
7878
7879         msg.rpc_argp = &data->args;
7880         msg.rpc_resp = &data->res;
7881         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7882         rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7883 }
7884
7885 #define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7886
7887 static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7888                                    struct mnt_idmap *idmap,
7889                                    struct dentry *unused, struct inode *inode,
7890                                    const char *key, const void *buf,
7891                                    size_t buflen, int flags)
7892 {
7893         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7894 }
7895
7896 static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7897                                    struct dentry *unused, struct inode *inode,
7898                                    const char *key, void *buf, size_t buflen)
7899 {
7900         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7901 }
7902
7903 static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7904 {
7905         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7906 }
7907
7908 #if defined(CONFIG_NFS_V4_1)
7909 #define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7910
7911 static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7912                                     struct mnt_idmap *idmap,
7913                                     struct dentry *unused, struct inode *inode,
7914                                     const char *key, const void *buf,
7915                                     size_t buflen, int flags)
7916 {
7917         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7918 }
7919
7920 static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7921                                     struct dentry *unused, struct inode *inode,
7922                                     const char *key, void *buf, size_t buflen)
7923 {
7924         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7925 }
7926
7927 static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7928 {
7929         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7930 }
7931
7932 #define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7933
7934 static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7935                                     struct mnt_idmap *idmap,
7936                                     struct dentry *unused, struct inode *inode,
7937                                     const char *key, const void *buf,
7938                                     size_t buflen, int flags)
7939 {
7940         return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7941 }
7942
7943 static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7944                                     struct dentry *unused, struct inode *inode,
7945                                     const char *key, void *buf, size_t buflen)
7946 {
7947         return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7948 }
7949
7950 static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7951 {
7952         return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7953 }
7954
7955 #endif
7956
7957 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
7958
7959 static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7960                                      struct mnt_idmap *idmap,
7961                                      struct dentry *unused, struct inode *inode,
7962                                      const char *key, const void *buf,
7963                                      size_t buflen, int flags)
7964 {
7965         if (security_ismaclabel(key))
7966                 return nfs4_set_security_label(inode, buf, buflen);
7967
7968         return -EOPNOTSUPP;
7969 }
7970
7971 static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7972                                      struct dentry *unused, struct inode *inode,
7973                                      const char *key, void *buf, size_t buflen)
7974 {
7975         if (security_ismaclabel(key))
7976                 return nfs4_get_security_label(inode, buf, buflen);
7977         return -EOPNOTSUPP;
7978 }
7979
7980 static ssize_t
7981 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7982 {
7983         int len = 0;
7984
7985         if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7986                 len = security_inode_listsecurity(inode, list, list_len);
7987                 if (len >= 0 && list_len && len > list_len)
7988                         return -ERANGE;
7989         }
7990         return len;
7991 }
7992
7993 static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7994         .prefix = XATTR_SECURITY_PREFIX,
7995         .get    = nfs4_xattr_get_nfs4_label,
7996         .set    = nfs4_xattr_set_nfs4_label,
7997 };
7998
7999 #else
8000
8001 static ssize_t
8002 nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
8003 {
8004         return 0;
8005 }
8006
8007 #endif
8008
8009 #ifdef CONFIG_NFS_V4_2
8010 static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
8011                                     struct mnt_idmap *idmap,
8012                                     struct dentry *unused, struct inode *inode,
8013                                     const char *key, const void *buf,
8014                                     size_t buflen, int flags)
8015 {
8016         u32 mask;
8017         int ret;
8018
8019         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8020                 return -EOPNOTSUPP;
8021
8022         /*
8023          * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
8024          * flags right now. Handling of xattr operations use the normal
8025          * file read/write permissions.
8026          *
8027          * Just in case the server has other ideas (which RFC 8276 allows),
8028          * do a cached access check for the XA* flags to possibly avoid
8029          * doing an RPC and getting EACCES back.
8030          */
8031         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8032                 if (!(mask & NFS_ACCESS_XAWRITE))
8033                         return -EACCES;
8034         }
8035
8036         if (buf == NULL) {
8037                 ret = nfs42_proc_removexattr(inode, key);
8038                 if (!ret)
8039                         nfs4_xattr_cache_remove(inode, key);
8040         } else {
8041                 ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
8042                 if (!ret)
8043                         nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
8044         }
8045
8046         return ret;
8047 }
8048
8049 static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
8050                                     struct dentry *unused, struct inode *inode,
8051                                     const char *key, void *buf, size_t buflen)
8052 {
8053         u32 mask;
8054         ssize_t ret;
8055
8056         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8057                 return -EOPNOTSUPP;
8058
8059         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8060                 if (!(mask & NFS_ACCESS_XAREAD))
8061                         return -EACCES;
8062         }
8063
8064         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
8065         if (ret)
8066                 return ret;
8067
8068         ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
8069         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
8070                 return ret;
8071
8072         ret = nfs42_proc_getxattr(inode, key, buf, buflen);
8073
8074         return ret;
8075 }
8076
8077 static ssize_t
8078 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
8079 {
8080         u64 cookie;
8081         bool eof;
8082         ssize_t ret, size;
8083         char *buf;
8084         size_t buflen;
8085         u32 mask;
8086
8087         if (!nfs_server_capable(inode, NFS_CAP_XATTR))
8088                 return 0;
8089
8090         if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
8091                 if (!(mask & NFS_ACCESS_XALIST))
8092                         return 0;
8093         }
8094
8095         ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
8096         if (ret)
8097                 return ret;
8098
8099         ret = nfs4_xattr_cache_list(inode, list, list_len);
8100         if (ret >= 0 || (ret < 0 && ret != -ENOENT))
8101                 return ret;
8102
8103         cookie = 0;
8104         eof = false;
8105         buflen = list_len ? list_len : XATTR_LIST_MAX;
8106         buf = list_len ? list : NULL;
8107         size = 0;
8108
8109         while (!eof) {
8110                 ret = nfs42_proc_listxattrs(inode, buf, buflen,
8111                     &cookie, &eof);
8112                 if (ret < 0)
8113                         return ret;
8114
8115                 if (list_len) {
8116                         buf += ret;
8117                         buflen -= ret;
8118                 }
8119                 size += ret;
8120         }
8121
8122         if (list_len)
8123                 nfs4_xattr_cache_set_list(inode, list, size);
8124
8125         return size;
8126 }
8127
8128 #else
8129
8130 static ssize_t
8131 nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
8132 {
8133         return 0;
8134 }
8135 #endif /* CONFIG_NFS_V4_2 */
8136
8137 /*
8138  * nfs_fhget will use either the mounted_on_fileid or the fileid
8139  */
8140 static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
8141 {
8142         if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
8143                (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
8144               (fattr->valid & NFS_ATTR_FATTR_FSID) &&
8145               (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
8146                 return;
8147
8148         fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
8149                 NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
8150         fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
8151         fattr->nlink = 2;
8152 }
8153
8154 static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8155                                    const struct qstr *name,
8156                                    struct nfs4_fs_locations *fs_locations,
8157                                    struct page *page)
8158 {
8159         struct nfs_server *server = NFS_SERVER(dir);
8160         u32 bitmask[3];
8161         struct nfs4_fs_locations_arg args = {
8162                 .dir_fh = NFS_FH(dir),
8163                 .name = name,
8164                 .page = page,
8165                 .bitmask = bitmask,
8166         };
8167         struct nfs4_fs_locations_res res = {
8168                 .fs_locations = fs_locations,
8169         };
8170         struct rpc_message msg = {
8171                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8172                 .rpc_argp = &args,
8173                 .rpc_resp = &res,
8174         };
8175         int status;
8176
8177         dprintk("%s: start\n", __func__);
8178
8179         bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
8180         bitmask[1] = nfs4_fattr_bitmap[1];
8181
8182         /* Ask for the fileid of the absent filesystem if mounted_on_fileid
8183          * is not supported */
8184         if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8185                 bitmask[0] &= ~FATTR4_WORD0_FILEID;
8186         else
8187                 bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8188
8189         nfs_fattr_init(fs_locations->fattr);
8190         fs_locations->server = server;
8191         fs_locations->nlocations = 0;
8192         status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8193         dprintk("%s: returned status = %d\n", __func__, status);
8194         return status;
8195 }
8196
8197 int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8198                            const struct qstr *name,
8199                            struct nfs4_fs_locations *fs_locations,
8200                            struct page *page)
8201 {
8202         struct nfs4_exception exception = {
8203                 .interruptible = true,
8204         };
8205         int err;
8206         do {
8207                 err = _nfs4_proc_fs_locations(client, dir, name,
8208                                 fs_locations, page);
8209                 trace_nfs4_get_fs_locations(dir, name, err);
8210                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8211                                 &exception);
8212         } while (exception.retry);
8213         return err;
8214 }
8215
8216 /*
8217  * This operation also signals the server that this client is
8218  * performing migration recovery.  The server can stop returning
8219  * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8220  * appended to this compound to identify the client ID which is
8221  * performing recovery.
8222  */
8223 static int _nfs40_proc_get_locations(struct nfs_server *server,
8224                                      struct nfs_fh *fhandle,
8225                                      struct nfs4_fs_locations *locations,
8226                                      struct page *page, const struct cred *cred)
8227 {
8228         struct rpc_clnt *clnt = server->client;
8229         u32 bitmask[2] = {
8230                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8231         };
8232         struct nfs4_fs_locations_arg args = {
8233                 .clientid       = server->nfs_client->cl_clientid,
8234                 .fh             = fhandle,
8235                 .page           = page,
8236                 .bitmask        = bitmask,
8237                 .migration      = 1,            /* skip LOOKUP */
8238                 .renew          = 1,            /* append RENEW */
8239         };
8240         struct nfs4_fs_locations_res res = {
8241                 .fs_locations   = locations,
8242                 .migration      = 1,
8243                 .renew          = 1,
8244         };
8245         struct rpc_message msg = {
8246                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8247                 .rpc_argp       = &args,
8248                 .rpc_resp       = &res,
8249                 .rpc_cred       = cred,
8250         };
8251         unsigned long now = jiffies;
8252         int status;
8253
8254         nfs_fattr_init(locations->fattr);
8255         locations->server = server;
8256         locations->nlocations = 0;
8257
8258         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8259         status = nfs4_call_sync_sequence(clnt, server, &msg,
8260                                         &args.seq_args, &res.seq_res);
8261         if (status)
8262                 return status;
8263
8264         renew_lease(server, now);
8265         return 0;
8266 }
8267
8268 #ifdef CONFIG_NFS_V4_1
8269
8270 /*
8271  * This operation also signals the server that this client is
8272  * performing migration recovery.  The server can stop asserting
8273  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8274  * performing this operation is identified in the SEQUENCE
8275  * operation in this compound.
8276  *
8277  * When the client supports GETATTR(fs_locations_info), it can
8278  * be plumbed in here.
8279  */
8280 static int _nfs41_proc_get_locations(struct nfs_server *server,
8281                                      struct nfs_fh *fhandle,
8282                                      struct nfs4_fs_locations *locations,
8283                                      struct page *page, const struct cred *cred)
8284 {
8285         struct rpc_clnt *clnt = server->client;
8286         u32 bitmask[2] = {
8287                 [0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8288         };
8289         struct nfs4_fs_locations_arg args = {
8290                 .fh             = fhandle,
8291                 .page           = page,
8292                 .bitmask        = bitmask,
8293                 .migration      = 1,            /* skip LOOKUP */
8294         };
8295         struct nfs4_fs_locations_res res = {
8296                 .fs_locations   = locations,
8297                 .migration      = 1,
8298         };
8299         struct rpc_message msg = {
8300                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8301                 .rpc_argp       = &args,
8302                 .rpc_resp       = &res,
8303                 .rpc_cred       = cred,
8304         };
8305         struct nfs4_call_sync_data data = {
8306                 .seq_server = server,
8307                 .seq_args = &args.seq_args,
8308                 .seq_res = &res.seq_res,
8309         };
8310         struct rpc_task_setup task_setup_data = {
8311                 .rpc_client = clnt,
8312                 .rpc_message = &msg,
8313                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8314                 .callback_data = &data,
8315                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8316         };
8317         int status;
8318
8319         nfs_fattr_init(locations->fattr);
8320         locations->server = server;
8321         locations->nlocations = 0;
8322
8323         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8324         status = nfs4_call_sync_custom(&task_setup_data);
8325         if (status == NFS4_OK &&
8326             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8327                 status = -NFS4ERR_LEASE_MOVED;
8328         return status;
8329 }
8330
8331 #endif  /* CONFIG_NFS_V4_1 */
8332
8333 /**
8334  * nfs4_proc_get_locations - discover locations for a migrated FSID
8335  * @server: pointer to nfs_server to process
8336  * @fhandle: pointer to the kernel NFS client file handle
8337  * @locations: result of query
8338  * @page: buffer
8339  * @cred: credential to use for this operation
8340  *
8341  * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8342  * operation failed, or a negative errno if a local error occurred.
8343  *
8344  * On success, "locations" is filled in, but if the server has
8345  * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8346  * asserted.
8347  *
8348  * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8349  * from this client that require migration recovery.
8350  */
8351 int nfs4_proc_get_locations(struct nfs_server *server,
8352                             struct nfs_fh *fhandle,
8353                             struct nfs4_fs_locations *locations,
8354                             struct page *page, const struct cred *cred)
8355 {
8356         struct nfs_client *clp = server->nfs_client;
8357         const struct nfs4_mig_recovery_ops *ops =
8358                                         clp->cl_mvops->mig_recovery_ops;
8359         struct nfs4_exception exception = {
8360                 .interruptible = true,
8361         };
8362         int status;
8363
8364         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8365                 (unsigned long long)server->fsid.major,
8366                 (unsigned long long)server->fsid.minor,
8367                 clp->cl_hostname);
8368         nfs_display_fhandle(fhandle, __func__);
8369
8370         do {
8371                 status = ops->get_locations(server, fhandle, locations, page,
8372                                             cred);
8373                 if (status != -NFS4ERR_DELAY)
8374                         break;
8375                 nfs4_handle_exception(server, status, &exception);
8376         } while (exception.retry);
8377         return status;
8378 }
8379
8380 /*
8381  * This operation also signals the server that this client is
8382  * performing "lease moved" recovery.  The server can stop
8383  * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8384  * is appended to this compound to identify the client ID which is
8385  * performing recovery.
8386  */
8387 static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8388 {
8389         struct nfs_server *server = NFS_SERVER(inode);
8390         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8391         struct rpc_clnt *clnt = server->client;
8392         struct nfs4_fsid_present_arg args = {
8393                 .fh             = NFS_FH(inode),
8394                 .clientid       = clp->cl_clientid,
8395                 .renew          = 1,            /* append RENEW */
8396         };
8397         struct nfs4_fsid_present_res res = {
8398                 .renew          = 1,
8399         };
8400         struct rpc_message msg = {
8401                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8402                 .rpc_argp       = &args,
8403                 .rpc_resp       = &res,
8404                 .rpc_cred       = cred,
8405         };
8406         unsigned long now = jiffies;
8407         int status;
8408
8409         res.fh = nfs_alloc_fhandle();
8410         if (res.fh == NULL)
8411                 return -ENOMEM;
8412
8413         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8414         status = nfs4_call_sync_sequence(clnt, server, &msg,
8415                                                 &args.seq_args, &res.seq_res);
8416         nfs_free_fhandle(res.fh);
8417         if (status)
8418                 return status;
8419
8420         do_renew_lease(clp, now);
8421         return 0;
8422 }
8423
8424 #ifdef CONFIG_NFS_V4_1
8425
8426 /*
8427  * This operation also signals the server that this client is
8428  * performing "lease moved" recovery.  The server can stop asserting
8429  * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8430  * this operation is identified in the SEQUENCE operation in this
8431  * compound.
8432  */
8433 static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8434 {
8435         struct nfs_server *server = NFS_SERVER(inode);
8436         struct rpc_clnt *clnt = server->client;
8437         struct nfs4_fsid_present_arg args = {
8438                 .fh             = NFS_FH(inode),
8439         };
8440         struct nfs4_fsid_present_res res = {
8441         };
8442         struct rpc_message msg = {
8443                 .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8444                 .rpc_argp       = &args,
8445                 .rpc_resp       = &res,
8446                 .rpc_cred       = cred,
8447         };
8448         int status;
8449
8450         res.fh = nfs_alloc_fhandle();
8451         if (res.fh == NULL)
8452                 return -ENOMEM;
8453
8454         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8455         status = nfs4_call_sync_sequence(clnt, server, &msg,
8456                                                 &args.seq_args, &res.seq_res);
8457         nfs_free_fhandle(res.fh);
8458         if (status == NFS4_OK &&
8459             res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8460                 status = -NFS4ERR_LEASE_MOVED;
8461         return status;
8462 }
8463
8464 #endif  /* CONFIG_NFS_V4_1 */
8465
8466 /**
8467  * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8468  * @inode: inode on FSID to check
8469  * @cred: credential to use for this operation
8470  *
8471  * Server indicates whether the FSID is present, moved, or not
8472  * recognized.  This operation is necessary to clear a LEASE_MOVED
8473  * condition for this client ID.
8474  *
8475  * Returns NFS4_OK if the FSID is present on this server,
8476  * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8477  *  NFS4ERR code if some error occurred on the server, or a
8478  *  negative errno if a local failure occurred.
8479  */
8480 int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8481 {
8482         struct nfs_server *server = NFS_SERVER(inode);
8483         struct nfs_client *clp = server->nfs_client;
8484         const struct nfs4_mig_recovery_ops *ops =
8485                                         clp->cl_mvops->mig_recovery_ops;
8486         struct nfs4_exception exception = {
8487                 .interruptible = true,
8488         };
8489         int status;
8490
8491         dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8492                 (unsigned long long)server->fsid.major,
8493                 (unsigned long long)server->fsid.minor,
8494                 clp->cl_hostname);
8495         nfs_display_fhandle(NFS_FH(inode), __func__);
8496
8497         do {
8498                 status = ops->fsid_present(inode, cred);
8499                 if (status != -NFS4ERR_DELAY)
8500                         break;
8501                 nfs4_handle_exception(server, status, &exception);
8502         } while (exception.retry);
8503         return status;
8504 }
8505
8506 /*
8507  * If 'use_integrity' is true and the state managment nfs_client
8508  * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8509  * and the machine credential as per RFC3530bis and RFC5661 Security
8510  * Considerations sections. Otherwise, just use the user cred with the
8511  * filesystem's rpc_client.
8512  */
8513 static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8514 {
8515         int status;
8516         struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8517         struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8518         struct nfs4_secinfo_arg args = {
8519                 .dir_fh = NFS_FH(dir),
8520                 .name   = name,
8521         };
8522         struct nfs4_secinfo_res res = {
8523                 .flavors     = flavors,
8524         };
8525         struct rpc_message msg = {
8526                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8527                 .rpc_argp = &args,
8528                 .rpc_resp = &res,
8529         };
8530         struct nfs4_call_sync_data data = {
8531                 .seq_server = NFS_SERVER(dir),
8532                 .seq_args = &args.seq_args,
8533                 .seq_res = &res.seq_res,
8534         };
8535         struct rpc_task_setup task_setup = {
8536                 .rpc_client = clnt,
8537                 .rpc_message = &msg,
8538                 .callback_ops = clp->cl_mvops->call_sync_ops,
8539                 .callback_data = &data,
8540                 .flags = RPC_TASK_NO_ROUND_ROBIN,
8541         };
8542         const struct cred *cred = NULL;
8543
8544         if (use_integrity) {
8545                 clnt = clp->cl_rpcclient;
8546                 task_setup.rpc_client = clnt;
8547
8548                 cred = nfs4_get_clid_cred(clp);
8549                 msg.rpc_cred = cred;
8550         }
8551
8552         dprintk("NFS call  secinfo %s\n", name->name);
8553
8554         nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8555         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8556         status = nfs4_call_sync_custom(&task_setup);
8557
8558         dprintk("NFS reply  secinfo: %d\n", status);
8559
8560         put_cred(cred);
8561         return status;
8562 }
8563
8564 int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8565                       struct nfs4_secinfo_flavors *flavors)
8566 {
8567         struct nfs4_exception exception = {
8568                 .interruptible = true,
8569         };
8570         int err;
8571         do {
8572                 err = -NFS4ERR_WRONGSEC;
8573
8574                 /* try to use integrity protection with machine cred */
8575                 if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8576                         err = _nfs4_proc_secinfo(dir, name, flavors, true);
8577
8578                 /*
8579                  * if unable to use integrity protection, or SECINFO with
8580                  * integrity protection returns NFS4ERR_WRONGSEC (which is
8581                  * disallowed by spec, but exists in deployed servers) use
8582                  * the current filesystem's rpc_client and the user cred.
8583                  */
8584                 if (err == -NFS4ERR_WRONGSEC)
8585                         err = _nfs4_proc_secinfo(dir, name, flavors, false);
8586
8587                 trace_nfs4_secinfo(dir, name, err);
8588                 err = nfs4_handle_exception(NFS_SERVER(dir), err,
8589                                 &exception);
8590         } while (exception.retry);
8591         return err;
8592 }
8593
8594 #ifdef CONFIG_NFS_V4_1
8595 /*
8596  * Check the exchange flags returned by the server for invalid flags, having
8597  * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8598  * DS flags set.
8599  */
8600 static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8601 {
8602         if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8603                 goto out_inval;
8604         else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8605                 goto out_inval;
8606         if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8607             (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8608                 goto out_inval;
8609         if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8610                 goto out_inval;
8611         return NFS_OK;
8612 out_inval:
8613         return -NFS4ERR_INVAL;
8614 }
8615
8616 static bool
8617 nfs41_same_server_scope(struct nfs41_server_scope *a,
8618                         struct nfs41_server_scope *b)
8619 {
8620         if (a->server_scope_sz != b->server_scope_sz)
8621                 return false;
8622         return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8623 }
8624
8625 static void
8626 nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8627 {
8628         struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8629         struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8630         struct nfs_client *clp = args->client;
8631
8632         switch (task->tk_status) {
8633         case -NFS4ERR_BADSESSION:
8634         case -NFS4ERR_DEADSESSION:
8635                 nfs4_schedule_session_recovery(clp->cl_session,
8636                                 task->tk_status);
8637                 return;
8638         }
8639         if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8640                         res->dir != NFS4_CDFS4_BOTH) {
8641                 rpc_task_close_connection(task);
8642                 if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8643                         rpc_restart_call(task);
8644         }
8645 }
8646
8647 static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8648         .rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8649 };
8650
8651 /*
8652  * nfs4_proc_bind_one_conn_to_session()
8653  *
8654  * The 4.1 client currently uses the same TCP connection for the
8655  * fore and backchannel.
8656  */
8657 static
8658 int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8659                 struct rpc_xprt *xprt,
8660                 struct nfs_client *clp,
8661                 const struct cred *cred)
8662 {
8663         int status;
8664         struct nfs41_bind_conn_to_session_args args = {
8665                 .client = clp,
8666                 .dir = NFS4_CDFC4_FORE_OR_BOTH,
8667                 .retries = 0,
8668         };
8669         struct nfs41_bind_conn_to_session_res res;
8670         struct rpc_message msg = {
8671                 .rpc_proc =
8672                         &nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8673                 .rpc_argp = &args,
8674                 .rpc_resp = &res,
8675                 .rpc_cred = cred,
8676         };
8677         struct rpc_task_setup task_setup_data = {
8678                 .rpc_client = clnt,
8679                 .rpc_xprt = xprt,
8680                 .callback_ops = &nfs4_bind_one_conn_to_session_ops,
8681                 .rpc_message = &msg,
8682                 .flags = RPC_TASK_TIMEOUT,
8683         };
8684         struct rpc_task *task;
8685
8686         nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8687         if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8688                 args.dir = NFS4_CDFC4_FORE;
8689
8690         /* Do not set the backchannel flag unless this is clnt->cl_xprt */
8691         if (xprt != rcu_access_pointer(clnt->cl_xprt))
8692                 args.dir = NFS4_CDFC4_FORE;
8693
8694         task = rpc_run_task(&task_setup_data);
8695         if (!IS_ERR(task)) {
8696                 status = task->tk_status;
8697                 rpc_put_task(task);
8698         } else
8699                 status = PTR_ERR(task);
8700         trace_nfs4_bind_conn_to_session(clp, status);
8701         if (status == 0) {
8702                 if (memcmp(res.sessionid.data,
8703                     clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8704                         dprintk("NFS: %s: Session ID mismatch\n", __func__);
8705                         return -EIO;
8706                 }
8707                 if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8708                         dprintk("NFS: %s: Unexpected direction from server\n",
8709                                 __func__);
8710                         return -EIO;
8711                 }
8712                 if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8713                         dprintk("NFS: %s: Server returned RDMA mode = true\n",
8714                                 __func__);
8715                         return -EIO;
8716                 }
8717         }
8718
8719         return status;
8720 }
8721
8722 struct rpc_bind_conn_calldata {
8723         struct nfs_client *clp;
8724         const struct cred *cred;
8725 };
8726
8727 static int
8728 nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8729                 struct rpc_xprt *xprt,
8730                 void *calldata)
8731 {
8732         struct rpc_bind_conn_calldata *p = calldata;
8733
8734         return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8735 }
8736
8737 int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8738 {
8739         struct rpc_bind_conn_calldata data = {
8740                 .clp = clp,
8741                 .cred = cred,
8742         };
8743         return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8744                         nfs4_proc_bind_conn_to_session_callback, &data);
8745 }
8746
8747 /*
8748  * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8749  * and operations we'd like to see to enable certain features in the allow map
8750  */
8751 static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8752         .how = SP4_MACH_CRED,
8753         .enforce.u.words = {
8754                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8755                       1 << (OP_EXCHANGE_ID - 32) |
8756                       1 << (OP_CREATE_SESSION - 32) |
8757                       1 << (OP_DESTROY_SESSION - 32) |
8758                       1 << (OP_DESTROY_CLIENTID - 32)
8759         },
8760         .allow.u.words = {
8761                 [0] = 1 << (OP_CLOSE) |
8762                       1 << (OP_OPEN_DOWNGRADE) |
8763                       1 << (OP_LOCKU) |
8764                       1 << (OP_DELEGRETURN) |
8765                       1 << (OP_COMMIT),
8766                 [1] = 1 << (OP_SECINFO - 32) |
8767                       1 << (OP_SECINFO_NO_NAME - 32) |
8768                       1 << (OP_LAYOUTRETURN - 32) |
8769                       1 << (OP_TEST_STATEID - 32) |
8770                       1 << (OP_FREE_STATEID - 32) |
8771                       1 << (OP_WRITE - 32)
8772         }
8773 };
8774
8775 /*
8776  * Select the state protection mode for client `clp' given the server results
8777  * from exchange_id in `sp'.
8778  *
8779  * Returns 0 on success, negative errno otherwise.
8780  */
8781 static int nfs4_sp4_select_mode(struct nfs_client *clp,
8782                                  struct nfs41_state_protection *sp)
8783 {
8784         static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8785                 [1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8786                       1 << (OP_EXCHANGE_ID - 32) |
8787                       1 << (OP_CREATE_SESSION - 32) |
8788                       1 << (OP_DESTROY_SESSION - 32) |
8789                       1 << (OP_DESTROY_CLIENTID - 32)
8790         };
8791         unsigned long flags = 0;
8792         unsigned int i;
8793         int ret = 0;
8794
8795         if (sp->how == SP4_MACH_CRED) {
8796                 /* Print state protect result */
8797                 dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8798                 for (i = 0; i <= LAST_NFS4_OP; i++) {
8799                         if (test_bit(i, sp->enforce.u.longs))
8800                                 dfprintk(MOUNT, "  enforce op %d\n", i);
8801                         if (test_bit(i, sp->allow.u.longs))
8802                                 dfprintk(MOUNT, "  allow op %d\n", i);
8803                 }
8804
8805                 /* make sure nothing is on enforce list that isn't supported */
8806                 for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8807                         if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8808                                 dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8809                                 ret = -EINVAL;
8810                                 goto out;
8811                         }
8812                 }
8813
8814                 /*
8815                  * Minimal mode - state operations are allowed to use machine
8816                  * credential.  Note this already happens by default, so the
8817                  * client doesn't have to do anything more than the negotiation.
8818                  *
8819                  * NOTE: we don't care if EXCHANGE_ID is in the list -
8820                  *       we're already using the machine cred for exchange_id
8821                  *       and will never use a different cred.
8822                  */
8823                 if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8824                     test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8825                     test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8826                     test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8827                         dfprintk(MOUNT, "sp4_mach_cred:\n");
8828                         dfprintk(MOUNT, "  minimal mode enabled\n");
8829                         __set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8830                 } else {
8831                         dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8832                         ret = -EINVAL;
8833                         goto out;
8834                 }
8835
8836                 if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8837                     test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8838                     test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8839                     test_bit(OP_LOCKU, sp->allow.u.longs)) {
8840                         dfprintk(MOUNT, "  cleanup mode enabled\n");
8841                         __set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8842                 }
8843
8844                 if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8845                         dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8846                         __set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8847                 }
8848
8849                 if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8850                     test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8851                         dfprintk(MOUNT, "  secinfo mode enabled\n");
8852                         __set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8853                 }
8854
8855                 if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8856                     test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8857                         dfprintk(MOUNT, "  stateid mode enabled\n");
8858                         __set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8859                 }
8860
8861                 if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8862                         dfprintk(MOUNT, "  write mode enabled\n");
8863                         __set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8864                 }
8865
8866                 if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8867                         dfprintk(MOUNT, "  commit mode enabled\n");
8868                         __set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8869                 }
8870         }
8871 out:
8872         clp->cl_sp4_flags = flags;
8873         return ret;
8874 }
8875
8876 struct nfs41_exchange_id_data {
8877         struct nfs41_exchange_id_res res;
8878         struct nfs41_exchange_id_args args;
8879 };
8880
8881 static void nfs4_exchange_id_release(void *data)
8882 {
8883         struct nfs41_exchange_id_data *cdata =
8884                                         (struct nfs41_exchange_id_data *)data;
8885
8886         nfs_put_client(cdata->args.client);
8887         kfree(cdata->res.impl_id);
8888         kfree(cdata->res.server_scope);
8889         kfree(cdata->res.server_owner);
8890         kfree(cdata);
8891 }
8892
8893 static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8894         .rpc_release = nfs4_exchange_id_release,
8895 };
8896
8897 /*
8898  * _nfs4_proc_exchange_id()
8899  *
8900  * Wrapper for EXCHANGE_ID operation.
8901  */
8902 static struct rpc_task *
8903 nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8904                         u32 sp4_how, struct rpc_xprt *xprt)
8905 {
8906         struct rpc_message msg = {
8907                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8908                 .rpc_cred = cred,
8909         };
8910         struct rpc_task_setup task_setup_data = {
8911                 .rpc_client = clp->cl_rpcclient,
8912                 .callback_ops = &nfs4_exchange_id_call_ops,
8913                 .rpc_message = &msg,
8914                 .flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8915         };
8916         struct nfs41_exchange_id_data *calldata;
8917         int status;
8918
8919         if (!refcount_inc_not_zero(&clp->cl_count))
8920                 return ERR_PTR(-EIO);
8921
8922         status = -ENOMEM;
8923         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8924         if (!calldata)
8925                 goto out;
8926
8927         nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8928
8929         status = nfs4_init_uniform_client_string(clp);
8930         if (status)
8931                 goto out_calldata;
8932
8933         calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8934                                                 GFP_NOFS);
8935         status = -ENOMEM;
8936         if (unlikely(calldata->res.server_owner == NULL))
8937                 goto out_calldata;
8938
8939         calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8940                                         GFP_NOFS);
8941         if (unlikely(calldata->res.server_scope == NULL))
8942                 goto out_server_owner;
8943
8944         calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8945         if (unlikely(calldata->res.impl_id == NULL))
8946                 goto out_server_scope;
8947
8948         switch (sp4_how) {
8949         case SP4_NONE:
8950                 calldata->args.state_protect.how = SP4_NONE;
8951                 break;
8952
8953         case SP4_MACH_CRED:
8954                 calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8955                 break;
8956
8957         default:
8958                 /* unsupported! */
8959                 WARN_ON_ONCE(1);
8960                 status = -EINVAL;
8961                 goto out_impl_id;
8962         }
8963         if (xprt) {
8964                 task_setup_data.rpc_xprt = xprt;
8965                 task_setup_data.flags |= RPC_TASK_SOFTCONN;
8966                 memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8967                                 sizeof(calldata->args.verifier.data));
8968         }
8969         calldata->args.client = clp;
8970         calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8971         EXCHGID4_FLAG_BIND_PRINC_STATEID;
8972 #ifdef CONFIG_NFS_V4_1_MIGRATION
8973         calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8974 #endif
8975         if (test_bit(NFS_CS_PNFS, &clp->cl_flags))
8976                 calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8977         msg.rpc_argp = &calldata->args;
8978         msg.rpc_resp = &calldata->res;
8979         task_setup_data.callback_data = calldata;
8980
8981         return rpc_run_task(&task_setup_data);
8982
8983 out_impl_id:
8984         kfree(calldata->res.impl_id);
8985 out_server_scope:
8986         kfree(calldata->res.server_scope);
8987 out_server_owner:
8988         kfree(calldata->res.server_owner);
8989 out_calldata:
8990         kfree(calldata);
8991 out:
8992         nfs_put_client(clp);
8993         return ERR_PTR(status);
8994 }
8995
8996 /*
8997  * _nfs4_proc_exchange_id()
8998  *
8999  * Wrapper for EXCHANGE_ID operation.
9000  */
9001 static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
9002                         u32 sp4_how)
9003 {
9004         struct rpc_task *task;
9005         struct nfs41_exchange_id_args *argp;
9006         struct nfs41_exchange_id_res *resp;
9007         unsigned long now = jiffies;
9008         int status;
9009
9010         task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
9011         if (IS_ERR(task))
9012                 return PTR_ERR(task);
9013
9014         argp = task->tk_msg.rpc_argp;
9015         resp = task->tk_msg.rpc_resp;
9016         status = task->tk_status;
9017         if (status  != 0)
9018                 goto out;
9019
9020         status = nfs4_check_cl_exchange_flags(resp->flags,
9021                         clp->cl_mvops->minor_version);
9022         if (status  != 0)
9023                 goto out;
9024
9025         status = nfs4_sp4_select_mode(clp, &resp->state_protect);
9026         if (status != 0)
9027                 goto out;
9028
9029         do_renew_lease(clp, now);
9030
9031         clp->cl_clientid = resp->clientid;
9032         clp->cl_exchange_flags = resp->flags;
9033         clp->cl_seqid = resp->seqid;
9034         /* Client ID is not confirmed */
9035         if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
9036                 clear_bit(NFS4_SESSION_ESTABLISHED,
9037                           &clp->cl_session->session_state);
9038
9039         if (clp->cl_serverscope != NULL &&
9040             !nfs41_same_server_scope(clp->cl_serverscope,
9041                                 resp->server_scope)) {
9042                 dprintk("%s: server_scope mismatch detected\n",
9043                         __func__);
9044                 set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
9045         }
9046
9047         swap(clp->cl_serverowner, resp->server_owner);
9048         swap(clp->cl_serverscope, resp->server_scope);
9049         swap(clp->cl_implid, resp->impl_id);
9050
9051         /* Save the EXCHANGE_ID verifier session trunk tests */
9052         memcpy(clp->cl_confirm.data, argp->verifier.data,
9053                sizeof(clp->cl_confirm.data));
9054 out:
9055         trace_nfs4_exchange_id(clp, status);
9056         rpc_put_task(task);
9057         return status;
9058 }
9059
9060 /*
9061  * nfs4_proc_exchange_id()
9062  *
9063  * Returns zero, a negative errno, or a negative NFS4ERR status code.
9064  *
9065  * Since the clientid has expired, all compounds using sessions
9066  * associated with the stale clientid will be returning
9067  * NFS4ERR_BADSESSION in the sequence operation, and will therefore
9068  * be in some phase of session reset.
9069  *
9070  * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
9071  */
9072 int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
9073 {
9074         rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
9075         int status;
9076
9077         /* try SP4_MACH_CRED if krb5i/p */
9078         if (authflavor == RPC_AUTH_GSS_KRB5I ||
9079             authflavor == RPC_AUTH_GSS_KRB5P) {
9080                 status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
9081                 if (!status)
9082                         return 0;
9083         }
9084
9085         /* try SP4_NONE */
9086         return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
9087 }
9088
9089 /**
9090  * nfs4_test_session_trunk
9091  *
9092  * This is an add_xprt_test() test function called from
9093  * rpc_clnt_setup_test_and_add_xprt.
9094  *
9095  * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
9096  * and is dereferrenced in nfs4_exchange_id_release
9097  *
9098  * Upon success, add the new transport to the rpc_clnt
9099  *
9100  * @clnt: struct rpc_clnt to get new transport
9101  * @xprt: the rpc_xprt to test
9102  * @data: call data for _nfs4_proc_exchange_id.
9103  */
9104 void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
9105                             void *data)
9106 {
9107         struct nfs4_add_xprt_data *adata = data;
9108         struct rpc_task *task;
9109         int status;
9110
9111         u32 sp4_how;
9112
9113         dprintk("--> %s try %s\n", __func__,
9114                 xprt->address_strings[RPC_DISPLAY_ADDR]);
9115
9116         sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
9117
9118 try_again:
9119         /* Test connection for session trunking. Async exchange_id call */
9120         task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
9121         if (IS_ERR(task))
9122                 return;
9123
9124         status = task->tk_status;
9125         if (status == 0) {
9126                 status = nfs4_detect_session_trunking(adata->clp,
9127                                 task->tk_msg.rpc_resp, xprt);
9128                 trace_nfs4_trunked_exchange_id(adata->clp,
9129                         xprt->address_strings[RPC_DISPLAY_ADDR], status);
9130         }
9131         if (status == 0)
9132                 rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
9133         else if (status != -NFS4ERR_DELAY && rpc_clnt_xprt_switch_has_addr(clnt,
9134                                 (struct sockaddr *)&xprt->addr))
9135                 rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
9136
9137         rpc_put_task(task);
9138         if (status == -NFS4ERR_DELAY) {
9139                 ssleep(1);
9140                 goto try_again;
9141         }
9142 }
9143 EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
9144
9145 static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
9146                 const struct cred *cred)
9147 {
9148         struct rpc_message msg = {
9149                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
9150                 .rpc_argp = clp,
9151                 .rpc_cred = cred,
9152         };
9153         int status;
9154
9155         status = rpc_call_sync(clp->cl_rpcclient, &msg,
9156                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9157         trace_nfs4_destroy_clientid(clp, status);
9158         if (status)
9159                 dprintk("NFS: Got error %d from the server %s on "
9160                         "DESTROY_CLIENTID.", status, clp->cl_hostname);
9161         return status;
9162 }
9163
9164 static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
9165                 const struct cred *cred)
9166 {
9167         unsigned int loop;
9168         int ret;
9169
9170         for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
9171                 ret = _nfs4_proc_destroy_clientid(clp, cred);
9172                 switch (ret) {
9173                 case -NFS4ERR_DELAY:
9174                 case -NFS4ERR_CLIENTID_BUSY:
9175                         ssleep(1);
9176                         break;
9177                 default:
9178                         return ret;
9179                 }
9180         }
9181         return 0;
9182 }
9183
9184 int nfs4_destroy_clientid(struct nfs_client *clp)
9185 {
9186         const struct cred *cred;
9187         int ret = 0;
9188
9189         if (clp->cl_mvops->minor_version < 1)
9190                 goto out;
9191         if (clp->cl_exchange_flags == 0)
9192                 goto out;
9193         if (clp->cl_preserve_clid)
9194                 goto out;
9195         cred = nfs4_get_clid_cred(clp);
9196         ret = nfs4_proc_destroy_clientid(clp, cred);
9197         put_cred(cred);
9198         switch (ret) {
9199         case 0:
9200         case -NFS4ERR_STALE_CLIENTID:
9201                 clp->cl_exchange_flags = 0;
9202         }
9203 out:
9204         return ret;
9205 }
9206
9207 #endif /* CONFIG_NFS_V4_1 */
9208
9209 struct nfs4_get_lease_time_data {
9210         struct nfs4_get_lease_time_args *args;
9211         struct nfs4_get_lease_time_res *res;
9212         struct nfs_client *clp;
9213 };
9214
9215 static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9216                                         void *calldata)
9217 {
9218         struct nfs4_get_lease_time_data *data =
9219                         (struct nfs4_get_lease_time_data *)calldata;
9220
9221         /* just setup sequence, do not trigger session recovery
9222            since we're invoked within one */
9223         nfs4_setup_sequence(data->clp,
9224                         &data->args->la_seq_args,
9225                         &data->res->lr_seq_res,
9226                         task);
9227 }
9228
9229 /*
9230  * Called from nfs4_state_manager thread for session setup, so don't recover
9231  * from sequence operation or clientid errors.
9232  */
9233 static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9234 {
9235         struct nfs4_get_lease_time_data *data =
9236                         (struct nfs4_get_lease_time_data *)calldata;
9237
9238         if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9239                 return;
9240         switch (task->tk_status) {
9241         case -NFS4ERR_DELAY:
9242         case -NFS4ERR_GRACE:
9243                 rpc_delay(task, NFS4_POLL_RETRY_MIN);
9244                 task->tk_status = 0;
9245                 fallthrough;
9246         case -NFS4ERR_RETRY_UNCACHED_REP:
9247                 rpc_restart_call_prepare(task);
9248                 return;
9249         }
9250 }
9251
9252 static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9253         .rpc_call_prepare = nfs4_get_lease_time_prepare,
9254         .rpc_call_done = nfs4_get_lease_time_done,
9255 };
9256
9257 int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9258 {
9259         struct nfs4_get_lease_time_args args;
9260         struct nfs4_get_lease_time_res res = {
9261                 .lr_fsinfo = fsinfo,
9262         };
9263         struct nfs4_get_lease_time_data data = {
9264                 .args = &args,
9265                 .res = &res,
9266                 .clp = clp,
9267         };
9268         struct rpc_message msg = {
9269                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9270                 .rpc_argp = &args,
9271                 .rpc_resp = &res,
9272         };
9273         struct rpc_task_setup task_setup = {
9274                 .rpc_client = clp->cl_rpcclient,
9275                 .rpc_message = &msg,
9276                 .callback_ops = &nfs4_get_lease_time_ops,
9277                 .callback_data = &data,
9278                 .flags = RPC_TASK_TIMEOUT,
9279         };
9280
9281         nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9282         return nfs4_call_sync_custom(&task_setup);
9283 }
9284
9285 #ifdef CONFIG_NFS_V4_1
9286
9287 /*
9288  * Initialize the values to be used by the client in CREATE_SESSION
9289  * If nfs4_init_session set the fore channel request and response sizes,
9290  * use them.
9291  *
9292  * Set the back channel max_resp_sz_cached to zero to force the client to
9293  * always set csa_cachethis to FALSE because the current implementation
9294  * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9295  */
9296 static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9297                                     struct rpc_clnt *clnt)
9298 {
9299         unsigned int max_rqst_sz, max_resp_sz;
9300         unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9301         unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9302
9303         max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9304         max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9305
9306         /* Fore channel attributes */
9307         args->fc_attrs.max_rqst_sz = max_rqst_sz;
9308         args->fc_attrs.max_resp_sz = max_resp_sz;
9309         args->fc_attrs.max_ops = NFS4_MAX_OPS;
9310         args->fc_attrs.max_reqs = max_session_slots;
9311
9312         dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9313                 "max_ops=%u max_reqs=%u\n",
9314                 __func__,
9315                 args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9316                 args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9317
9318         /* Back channel attributes */
9319         args->bc_attrs.max_rqst_sz = max_bc_payload;
9320         args->bc_attrs.max_resp_sz = max_bc_payload;
9321         args->bc_attrs.max_resp_sz_cached = 0;
9322         args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9323         args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9324         if (args->bc_attrs.max_reqs > max_bc_slots)
9325                 args->bc_attrs.max_reqs = max_bc_slots;
9326
9327         dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9328                 "max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9329                 __func__,
9330                 args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9331                 args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9332                 args->bc_attrs.max_reqs);
9333 }
9334
9335 static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9336                 struct nfs41_create_session_res *res)
9337 {
9338         struct nfs4_channel_attrs *sent = &args->fc_attrs;
9339         struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9340
9341         if (rcvd->max_resp_sz > sent->max_resp_sz)
9342                 return -EINVAL;
9343         /*
9344          * Our requested max_ops is the minimum we need; we're not
9345          * prepared to break up compounds into smaller pieces than that.
9346          * So, no point even trying to continue if the server won't
9347          * cooperate:
9348          */
9349         if (rcvd->max_ops < sent->max_ops)
9350                 return -EINVAL;
9351         if (rcvd->max_reqs == 0)
9352                 return -EINVAL;
9353         if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9354                 rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9355         return 0;
9356 }
9357
9358 static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9359                 struct nfs41_create_session_res *res)
9360 {
9361         struct nfs4_channel_attrs *sent = &args->bc_attrs;
9362         struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9363
9364         if (!(res->flags & SESSION4_BACK_CHAN))
9365                 goto out;
9366         if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9367                 return -EINVAL;
9368         if (rcvd->max_resp_sz < sent->max_resp_sz)
9369                 return -EINVAL;
9370         if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9371                 return -EINVAL;
9372         if (rcvd->max_ops > sent->max_ops)
9373                 return -EINVAL;
9374         if (rcvd->max_reqs > sent->max_reqs)
9375                 return -EINVAL;
9376 out:
9377         return 0;
9378 }
9379
9380 static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9381                                      struct nfs41_create_session_res *res)
9382 {
9383         int ret;
9384
9385         ret = nfs4_verify_fore_channel_attrs(args, res);
9386         if (ret)
9387                 return ret;
9388         return nfs4_verify_back_channel_attrs(args, res);
9389 }
9390
9391 static void nfs4_update_session(struct nfs4_session *session,
9392                 struct nfs41_create_session_res *res)
9393 {
9394         nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9395         /* Mark client id and session as being confirmed */
9396         session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9397         set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9398         session->flags = res->flags;
9399         memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9400         if (res->flags & SESSION4_BACK_CHAN)
9401                 memcpy(&session->bc_attrs, &res->bc_attrs,
9402                                 sizeof(session->bc_attrs));
9403 }
9404
9405 static int _nfs4_proc_create_session(struct nfs_client *clp,
9406                 const struct cred *cred)
9407 {
9408         struct nfs4_session *session = clp->cl_session;
9409         struct nfs41_create_session_args args = {
9410                 .client = clp,
9411                 .clientid = clp->cl_clientid,
9412                 .seqid = clp->cl_seqid,
9413                 .cb_program = NFS4_CALLBACK,
9414         };
9415         struct nfs41_create_session_res res;
9416
9417         struct rpc_message msg = {
9418                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9419                 .rpc_argp = &args,
9420                 .rpc_resp = &res,
9421                 .rpc_cred = cred,
9422         };
9423         int status;
9424
9425         nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9426         args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9427
9428         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9429                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9430         trace_nfs4_create_session(clp, status);
9431
9432         switch (status) {
9433         case -NFS4ERR_STALE_CLIENTID:
9434         case -NFS4ERR_DELAY:
9435         case -ETIMEDOUT:
9436         case -EACCES:
9437         case -EAGAIN:
9438                 goto out;
9439         }
9440
9441         clp->cl_seqid++;
9442         if (!status) {
9443                 /* Verify the session's negotiated channel_attrs values */
9444                 status = nfs4_verify_channel_attrs(&args, &res);
9445                 /* Increment the clientid slot sequence id */
9446                 if (status)
9447                         goto out;
9448                 nfs4_update_session(session, &res);
9449         }
9450 out:
9451         return status;
9452 }
9453
9454 /*
9455  * Issues a CREATE_SESSION operation to the server.
9456  * It is the responsibility of the caller to verify the session is
9457  * expired before calling this routine.
9458  */
9459 int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9460 {
9461         int status;
9462         unsigned *ptr;
9463         struct nfs4_session *session = clp->cl_session;
9464         struct nfs4_add_xprt_data xprtdata = {
9465                 .clp = clp,
9466         };
9467         struct rpc_add_xprt_test rpcdata = {
9468                 .add_xprt_test = clp->cl_mvops->session_trunk,
9469                 .data = &xprtdata,
9470         };
9471
9472         dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9473
9474         status = _nfs4_proc_create_session(clp, cred);
9475         if (status)
9476                 goto out;
9477
9478         /* Init or reset the session slot tables */
9479         status = nfs4_setup_session_slot_tables(session);
9480         dprintk("slot table setup returned %d\n", status);
9481         if (status)
9482                 goto out;
9483
9484         ptr = (unsigned *)&session->sess_id.data[0];
9485         dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9486                 clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9487         rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9488 out:
9489         return status;
9490 }
9491
9492 /*
9493  * Issue the over-the-wire RPC DESTROY_SESSION.
9494  * The caller must serialize access to this routine.
9495  */
9496 int nfs4_proc_destroy_session(struct nfs4_session *session,
9497                 const struct cred *cred)
9498 {
9499         struct rpc_message msg = {
9500                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9501                 .rpc_argp = session,
9502                 .rpc_cred = cred,
9503         };
9504         int status = 0;
9505
9506         /* session is still being setup */
9507         if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9508                 return 0;
9509
9510         status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9511                                RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9512         trace_nfs4_destroy_session(session->clp, status);
9513
9514         if (status)
9515                 dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9516                         "Session has been destroyed regardless...\n", status);
9517         rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9518         return status;
9519 }
9520
9521 /*
9522  * Renew the cl_session lease.
9523  */
9524 struct nfs4_sequence_data {
9525         struct nfs_client *clp;
9526         struct nfs4_sequence_args args;
9527         struct nfs4_sequence_res res;
9528 };
9529
9530 static void nfs41_sequence_release(void *data)
9531 {
9532         struct nfs4_sequence_data *calldata = data;
9533         struct nfs_client *clp = calldata->clp;
9534
9535         if (refcount_read(&clp->cl_count) > 1)
9536                 nfs4_schedule_state_renewal(clp);
9537         nfs_put_client(clp);
9538         kfree(calldata);
9539 }
9540
9541 static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9542 {
9543         switch(task->tk_status) {
9544         case -NFS4ERR_DELAY:
9545                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9546                 return -EAGAIN;
9547         default:
9548                 nfs4_schedule_lease_recovery(clp);
9549         }
9550         return 0;
9551 }
9552
9553 static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9554 {
9555         struct nfs4_sequence_data *calldata = data;
9556         struct nfs_client *clp = calldata->clp;
9557
9558         if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9559                 return;
9560
9561         trace_nfs4_sequence(clp, task->tk_status);
9562         if (task->tk_status < 0 && !task->tk_client->cl_shutdown) {
9563                 dprintk("%s ERROR %d\n", __func__, task->tk_status);
9564                 if (refcount_read(&clp->cl_count) == 1)
9565                         return;
9566
9567                 if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9568                         rpc_restart_call_prepare(task);
9569                         return;
9570                 }
9571         }
9572         dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9573 }
9574
9575 static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9576 {
9577         struct nfs4_sequence_data *calldata = data;
9578         struct nfs_client *clp = calldata->clp;
9579         struct nfs4_sequence_args *args;
9580         struct nfs4_sequence_res *res;
9581
9582         args = task->tk_msg.rpc_argp;
9583         res = task->tk_msg.rpc_resp;
9584
9585         nfs4_setup_sequence(clp, args, res, task);
9586 }
9587
9588 static const struct rpc_call_ops nfs41_sequence_ops = {
9589         .rpc_call_done = nfs41_sequence_call_done,
9590         .rpc_call_prepare = nfs41_sequence_prepare,
9591         .rpc_release = nfs41_sequence_release,
9592 };
9593
9594 static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9595                 const struct cred *cred,
9596                 struct nfs4_slot *slot,
9597                 bool is_privileged)
9598 {
9599         struct nfs4_sequence_data *calldata;
9600         struct rpc_message msg = {
9601                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9602                 .rpc_cred = cred,
9603         };
9604         struct rpc_task_setup task_setup_data = {
9605                 .rpc_client = clp->cl_rpcclient,
9606                 .rpc_message = &msg,
9607                 .callback_ops = &nfs41_sequence_ops,
9608                 .flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9609         };
9610         struct rpc_task *ret;
9611
9612         ret = ERR_PTR(-EIO);
9613         if (!refcount_inc_not_zero(&clp->cl_count))
9614                 goto out_err;
9615
9616         ret = ERR_PTR(-ENOMEM);
9617         calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9618         if (calldata == NULL)
9619                 goto out_put_clp;
9620         nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9621         nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9622         msg.rpc_argp = &calldata->args;
9623         msg.rpc_resp = &calldata->res;
9624         calldata->clp = clp;
9625         task_setup_data.callback_data = calldata;
9626
9627         ret = rpc_run_task(&task_setup_data);
9628         if (IS_ERR(ret))
9629                 goto out_err;
9630         return ret;
9631 out_put_clp:
9632         nfs_put_client(clp);
9633 out_err:
9634         nfs41_release_slot(slot);
9635         return ret;
9636 }
9637
9638 static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9639 {
9640         struct rpc_task *task;
9641         int ret = 0;
9642
9643         if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9644                 return -EAGAIN;
9645         task = _nfs41_proc_sequence(clp, cred, NULL, false);
9646         if (IS_ERR(task))
9647                 ret = PTR_ERR(task);
9648         else
9649                 rpc_put_task_async(task);
9650         dprintk("<-- %s status=%d\n", __func__, ret);
9651         return ret;
9652 }
9653
9654 static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9655 {
9656         struct rpc_task *task;
9657         int ret;
9658
9659         task = _nfs41_proc_sequence(clp, cred, NULL, true);
9660         if (IS_ERR(task)) {
9661                 ret = PTR_ERR(task);
9662                 goto out;
9663         }
9664         ret = rpc_wait_for_completion_task(task);
9665         if (!ret)
9666                 ret = task->tk_status;
9667         rpc_put_task(task);
9668 out:
9669         dprintk("<-- %s status=%d\n", __func__, ret);
9670         return ret;
9671 }
9672
9673 struct nfs4_reclaim_complete_data {
9674         struct nfs_client *clp;
9675         struct nfs41_reclaim_complete_args arg;
9676         struct nfs41_reclaim_complete_res res;
9677 };
9678
9679 static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9680 {
9681         struct nfs4_reclaim_complete_data *calldata = data;
9682
9683         nfs4_setup_sequence(calldata->clp,
9684                         &calldata->arg.seq_args,
9685                         &calldata->res.seq_res,
9686                         task);
9687 }
9688
9689 static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9690 {
9691         switch(task->tk_status) {
9692         case 0:
9693                 wake_up_all(&clp->cl_lock_waitq);
9694                 fallthrough;
9695         case -NFS4ERR_COMPLETE_ALREADY:
9696         case -NFS4ERR_WRONG_CRED: /* What to do here? */
9697                 break;
9698         case -NFS4ERR_DELAY:
9699                 rpc_delay(task, NFS4_POLL_RETRY_MAX);
9700                 fallthrough;
9701         case -NFS4ERR_RETRY_UNCACHED_REP:
9702         case -EACCES:
9703                 dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9704                         __func__, task->tk_status, clp->cl_hostname);
9705                 return -EAGAIN;
9706         case -NFS4ERR_BADSESSION:
9707         case -NFS4ERR_DEADSESSION:
9708         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9709                 break;
9710         default:
9711                 nfs4_schedule_lease_recovery(clp);
9712         }
9713         return 0;
9714 }
9715
9716 static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9717 {
9718         struct nfs4_reclaim_complete_data *calldata = data;
9719         struct nfs_client *clp = calldata->clp;
9720         struct nfs4_sequence_res *res = &calldata->res.seq_res;
9721
9722         if (!nfs41_sequence_done(task, res))
9723                 return;
9724
9725         trace_nfs4_reclaim_complete(clp, task->tk_status);
9726         if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9727                 rpc_restart_call_prepare(task);
9728                 return;
9729         }
9730 }
9731
9732 static void nfs4_free_reclaim_complete_data(void *data)
9733 {
9734         struct nfs4_reclaim_complete_data *calldata = data;
9735
9736         kfree(calldata);
9737 }
9738
9739 static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9740         .rpc_call_prepare = nfs4_reclaim_complete_prepare,
9741         .rpc_call_done = nfs4_reclaim_complete_done,
9742         .rpc_release = nfs4_free_reclaim_complete_data,
9743 };
9744
9745 /*
9746  * Issue a global reclaim complete.
9747  */
9748 static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9749                 const struct cred *cred)
9750 {
9751         struct nfs4_reclaim_complete_data *calldata;
9752         struct rpc_message msg = {
9753                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9754                 .rpc_cred = cred,
9755         };
9756         struct rpc_task_setup task_setup_data = {
9757                 .rpc_client = clp->cl_rpcclient,
9758                 .rpc_message = &msg,
9759                 .callback_ops = &nfs4_reclaim_complete_call_ops,
9760                 .flags = RPC_TASK_NO_ROUND_ROBIN,
9761         };
9762         int status = -ENOMEM;
9763
9764         calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9765         if (calldata == NULL)
9766                 goto out;
9767         calldata->clp = clp;
9768         calldata->arg.one_fs = 0;
9769
9770         nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9771         msg.rpc_argp = &calldata->arg;
9772         msg.rpc_resp = &calldata->res;
9773         task_setup_data.callback_data = calldata;
9774         status = nfs4_call_sync_custom(&task_setup_data);
9775 out:
9776         dprintk("<-- %s status=%d\n", __func__, status);
9777         return status;
9778 }
9779
9780 static void
9781 nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9782 {
9783         struct nfs4_layoutget *lgp = calldata;
9784         struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9785
9786         nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9787                                 &lgp->res.seq_res, task);
9788 }
9789
9790 static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9791 {
9792         struct nfs4_layoutget *lgp = calldata;
9793
9794         nfs41_sequence_process(task, &lgp->res.seq_res);
9795 }
9796
9797 static int
9798 nfs4_layoutget_handle_exception(struct rpc_task *task,
9799                 struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9800 {
9801         struct inode *inode = lgp->args.inode;
9802         struct nfs_server *server = NFS_SERVER(inode);
9803         struct pnfs_layout_hdr *lo = lgp->lo;
9804         int nfs4err = task->tk_status;
9805         int err, status = 0;
9806         LIST_HEAD(head);
9807
9808         dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9809
9810         nfs4_sequence_free_slot(&lgp->res.seq_res);
9811
9812         exception->state = NULL;
9813         exception->stateid = NULL;
9814
9815         switch (nfs4err) {
9816         case 0:
9817                 goto out;
9818
9819         /*
9820          * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9821          * on the file. set tk_status to -ENODATA to tell upper layer to
9822          * retry go inband.
9823          */
9824         case -NFS4ERR_LAYOUTUNAVAILABLE:
9825                 status = -ENODATA;
9826                 goto out;
9827         /*
9828          * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9829          * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9830          */
9831         case -NFS4ERR_BADLAYOUT:
9832                 status = -EOVERFLOW;
9833                 goto out;
9834         /*
9835          * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9836          * (or clients) writing to the same RAID stripe except when
9837          * the minlength argument is 0 (see RFC5661 section 18.43.3).
9838          *
9839          * Treat it like we would RECALLCONFLICT -- we retry for a little
9840          * while, and then eventually give up.
9841          */
9842         case -NFS4ERR_LAYOUTTRYLATER:
9843                 if (lgp->args.minlength == 0) {
9844                         status = -EOVERFLOW;
9845                         goto out;
9846                 }
9847                 status = -EBUSY;
9848                 break;
9849         case -NFS4ERR_RECALLCONFLICT:
9850         case -NFS4ERR_RETURNCONFLICT:
9851                 status = -ERECALLCONFLICT;
9852                 break;
9853         case -NFS4ERR_DELEG_REVOKED:
9854         case -NFS4ERR_ADMIN_REVOKED:
9855         case -NFS4ERR_EXPIRED:
9856         case -NFS4ERR_BAD_STATEID:
9857                 exception->timeout = 0;
9858                 spin_lock(&inode->i_lock);
9859                 /* If the open stateid was bad, then recover it. */
9860                 if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9861                     !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9862                         spin_unlock(&inode->i_lock);
9863                         exception->state = lgp->args.ctx->state;
9864                         exception->stateid = &lgp->args.stateid;
9865                         break;
9866                 }
9867
9868                 /*
9869                  * Mark the bad layout state as invalid, then retry
9870                  */
9871                 pnfs_mark_layout_stateid_invalid(lo, &head);
9872                 spin_unlock(&inode->i_lock);
9873                 nfs_commit_inode(inode, 0);
9874                 pnfs_free_lseg_list(&head);
9875                 status = -EAGAIN;
9876                 goto out;
9877         }
9878
9879         err = nfs4_handle_exception(server, nfs4err, exception);
9880         if (!status) {
9881                 if (exception->retry)
9882                         status = -EAGAIN;
9883                 else
9884                         status = err;
9885         }
9886 out:
9887         return status;
9888 }
9889
9890 size_t max_response_pages(struct nfs_server *server)
9891 {
9892         u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9893         return nfs_page_array_len(0, max_resp_sz);
9894 }
9895
9896 static void nfs4_layoutget_release(void *calldata)
9897 {
9898         struct nfs4_layoutget *lgp = calldata;
9899
9900         nfs4_sequence_free_slot(&lgp->res.seq_res);
9901         pnfs_layoutget_free(lgp);
9902 }
9903
9904 static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9905         .rpc_call_prepare = nfs4_layoutget_prepare,
9906         .rpc_call_done = nfs4_layoutget_done,
9907         .rpc_release = nfs4_layoutget_release,
9908 };
9909
9910 struct pnfs_layout_segment *
9911 nfs4_proc_layoutget(struct nfs4_layoutget *lgp,
9912                     struct nfs4_exception *exception)
9913 {
9914         struct inode *inode = lgp->args.inode;
9915         struct nfs_server *server = NFS_SERVER(inode);
9916         struct rpc_task *task;
9917         struct rpc_message msg = {
9918                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9919                 .rpc_argp = &lgp->args,
9920                 .rpc_resp = &lgp->res,
9921                 .rpc_cred = lgp->cred,
9922         };
9923         struct rpc_task_setup task_setup_data = {
9924                 .rpc_client = server->client,
9925                 .rpc_message = &msg,
9926                 .callback_ops = &nfs4_layoutget_call_ops,
9927                 .callback_data = lgp,
9928                 .flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9929                          RPC_TASK_MOVEABLE,
9930         };
9931         struct pnfs_layout_segment *lseg = NULL;
9932         int status = 0;
9933
9934         nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9935         exception->retry = 0;
9936
9937         task = rpc_run_task(&task_setup_data);
9938         if (IS_ERR(task))
9939                 return ERR_CAST(task);
9940
9941         status = rpc_wait_for_completion_task(task);
9942         if (status != 0)
9943                 goto out;
9944
9945         if (task->tk_status < 0) {
9946                 exception->retry = 1;
9947                 status = nfs4_layoutget_handle_exception(task, lgp, exception);
9948         } else if (lgp->res.layoutp->len == 0) {
9949                 exception->retry = 1;
9950                 status = -EAGAIN;
9951                 nfs4_update_delay(&exception->timeout);
9952         } else
9953                 lseg = pnfs_layout_process(lgp);
9954 out:
9955         trace_nfs4_layoutget(lgp->args.ctx,
9956                         &lgp->args.range,
9957                         &lgp->res.range,
9958                         &lgp->res.stateid,
9959                         status);
9960
9961         rpc_put_task(task);
9962         dprintk("<-- %s status=%d\n", __func__, status);
9963         if (status)
9964                 return ERR_PTR(status);
9965         return lseg;
9966 }
9967
9968 static void
9969 nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9970 {
9971         struct nfs4_layoutreturn *lrp = calldata;
9972
9973         nfs4_setup_sequence(lrp->clp,
9974                         &lrp->args.seq_args,
9975                         &lrp->res.seq_res,
9976                         task);
9977         if (!pnfs_layout_is_valid(lrp->args.layout))
9978                 rpc_exit(task, 0);
9979 }
9980
9981 static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9982 {
9983         struct nfs4_layoutreturn *lrp = calldata;
9984         struct nfs_server *server;
9985
9986         if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9987                 return;
9988
9989         if (task->tk_rpc_status == -ETIMEDOUT) {
9990                 lrp->rpc_status = -EAGAIN;
9991                 lrp->res.lrs_present = 0;
9992                 return;
9993         }
9994         /*
9995          * Was there an RPC level error? Assume the call succeeded,
9996          * and that we need to release the layout
9997          */
9998         if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9999                 lrp->res.lrs_present = 0;
10000                 return;
10001         }
10002
10003         server = NFS_SERVER(lrp->args.inode);
10004         switch (task->tk_status) {
10005         case -NFS4ERR_OLD_STATEID:
10006                 if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
10007                                         &lrp->args.range,
10008                                         lrp->args.inode))
10009                         goto out_restart;
10010                 fallthrough;
10011         default:
10012                 task->tk_status = 0;
10013                 lrp->res.lrs_present = 0;
10014                 fallthrough;
10015         case 0:
10016                 break;
10017         case -NFS4ERR_BADSESSION:
10018         case -NFS4ERR_DEADSESSION:
10019         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10020                 nfs4_schedule_session_recovery(server->nfs_client->cl_session,
10021                                                task->tk_status);
10022                 lrp->res.lrs_present = 0;
10023                 lrp->rpc_status = -EAGAIN;
10024                 task->tk_status = 0;
10025                 break;
10026         case -NFS4ERR_DELAY:
10027                 if (nfs4_async_handle_error(task, server, NULL, NULL) ==
10028                     -EAGAIN)
10029                         goto out_restart;
10030                 lrp->res.lrs_present = 0;
10031                 break;
10032         }
10033         return;
10034 out_restart:
10035         task->tk_status = 0;
10036         nfs4_sequence_free_slot(&lrp->res.seq_res);
10037         rpc_restart_call_prepare(task);
10038 }
10039
10040 static void nfs4_layoutreturn_release(void *calldata)
10041 {
10042         struct nfs4_layoutreturn *lrp = calldata;
10043         struct pnfs_layout_hdr *lo = lrp->args.layout;
10044
10045         if (lrp->rpc_status == 0 || !lrp->inode)
10046                 pnfs_layoutreturn_free_lsegs(
10047                         lo, &lrp->args.stateid, &lrp->args.range,
10048                         lrp->res.lrs_present ? &lrp->res.stateid : NULL);
10049         else
10050                 pnfs_layoutreturn_retry_later(lo, &lrp->args.stateid,
10051                                               &lrp->args.range);
10052         nfs4_sequence_free_slot(&lrp->res.seq_res);
10053         if (lrp->ld_private.ops && lrp->ld_private.ops->free)
10054                 lrp->ld_private.ops->free(&lrp->ld_private);
10055         pnfs_put_layout_hdr(lrp->args.layout);
10056         nfs_iput_and_deactive(lrp->inode);
10057         put_cred(lrp->cred);
10058         kfree(calldata);
10059 }
10060
10061 static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
10062         .rpc_call_prepare = nfs4_layoutreturn_prepare,
10063         .rpc_call_done = nfs4_layoutreturn_done,
10064         .rpc_release = nfs4_layoutreturn_release,
10065 };
10066
10067 int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, unsigned int flags)
10068 {
10069         struct rpc_task *task;
10070         struct rpc_message msg = {
10071                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
10072                 .rpc_argp = &lrp->args,
10073                 .rpc_resp = &lrp->res,
10074                 .rpc_cred = lrp->cred,
10075         };
10076         struct rpc_task_setup task_setup_data = {
10077                 .rpc_client = NFS_SERVER(lrp->args.inode)->client,
10078                 .rpc_message = &msg,
10079                 .callback_ops = &nfs4_layoutreturn_call_ops,
10080                 .callback_data = lrp,
10081                 .flags = RPC_TASK_MOVEABLE,
10082         };
10083         int status = 0;
10084
10085         nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
10086                         NFS_SP4_MACH_CRED_PNFS_CLEANUP,
10087                         &task_setup_data.rpc_client, &msg);
10088
10089         lrp->inode = nfs_igrab_and_active(lrp->args.inode);
10090         if (flags & PNFS_FL_LAYOUTRETURN_ASYNC) {
10091                 if (!lrp->inode) {
10092                         nfs4_layoutreturn_release(lrp);
10093                         return -EAGAIN;
10094                 }
10095                 task_setup_data.flags |= RPC_TASK_ASYNC;
10096         }
10097         if (!lrp->inode)
10098                 flags |= PNFS_FL_LAYOUTRETURN_PRIVILEGED;
10099         if (flags & PNFS_FL_LAYOUTRETURN_PRIVILEGED)
10100                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
10101                                    1);
10102         else
10103                 nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
10104                                    0);
10105         task = rpc_run_task(&task_setup_data);
10106         if (IS_ERR(task))
10107                 return PTR_ERR(task);
10108         if (!(flags & PNFS_FL_LAYOUTRETURN_ASYNC))
10109                 status = task->tk_status;
10110         trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
10111         dprintk("<-- %s status=%d\n", __func__, status);
10112         rpc_put_task(task);
10113         return status;
10114 }
10115
10116 static int
10117 _nfs4_proc_getdeviceinfo(struct nfs_server *server,
10118                 struct pnfs_device *pdev,
10119                 const struct cred *cred)
10120 {
10121         struct nfs4_getdeviceinfo_args args = {
10122                 .pdev = pdev,
10123                 .notify_types = NOTIFY_DEVICEID4_CHANGE |
10124                         NOTIFY_DEVICEID4_DELETE,
10125         };
10126         struct nfs4_getdeviceinfo_res res = {
10127                 .pdev = pdev,
10128         };
10129         struct rpc_message msg = {
10130                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
10131                 .rpc_argp = &args,
10132                 .rpc_resp = &res,
10133                 .rpc_cred = cred,
10134         };
10135         int status;
10136
10137         status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
10138         if (res.notification & ~args.notify_types)
10139                 dprintk("%s: unsupported notification\n", __func__);
10140         if (res.notification != args.notify_types)
10141                 pdev->nocache = 1;
10142
10143         trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
10144
10145         dprintk("<-- %s status=%d\n", __func__, status);
10146
10147         return status;
10148 }
10149
10150 int nfs4_proc_getdeviceinfo(struct nfs_server *server,
10151                 struct pnfs_device *pdev,
10152                 const struct cred *cred)
10153 {
10154         struct nfs4_exception exception = { };
10155         int err;
10156
10157         do {
10158                 err = nfs4_handle_exception(server,
10159                                         _nfs4_proc_getdeviceinfo(server, pdev, cred),
10160                                         &exception);
10161         } while (exception.retry);
10162         return err;
10163 }
10164 EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
10165
10166 static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
10167 {
10168         struct nfs4_layoutcommit_data *data = calldata;
10169         struct nfs_server *server = NFS_SERVER(data->args.inode);
10170
10171         nfs4_setup_sequence(server->nfs_client,
10172                         &data->args.seq_args,
10173                         &data->res.seq_res,
10174                         task);
10175 }
10176
10177 static void
10178 nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
10179 {
10180         struct nfs4_layoutcommit_data *data = calldata;
10181         struct nfs_server *server = NFS_SERVER(data->args.inode);
10182
10183         if (!nfs41_sequence_done(task, &data->res.seq_res))
10184                 return;
10185
10186         switch (task->tk_status) { /* Just ignore these failures */
10187         case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
10188         case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
10189         case -NFS4ERR_BADLAYOUT:     /* no layout */
10190         case -NFS4ERR_GRACE:        /* loca_recalim always false */
10191                 task->tk_status = 0;
10192                 break;
10193         case 0:
10194                 break;
10195         default:
10196                 if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
10197                         rpc_restart_call_prepare(task);
10198                         return;
10199                 }
10200         }
10201 }
10202
10203 static void nfs4_layoutcommit_release(void *calldata)
10204 {
10205         struct nfs4_layoutcommit_data *data = calldata;
10206
10207         pnfs_cleanup_layoutcommit(data);
10208         nfs_post_op_update_inode_force_wcc(data->args.inode,
10209                                            data->res.fattr);
10210         put_cred(data->cred);
10211         nfs_iput_and_deactive(data->inode);
10212         kfree(data);
10213 }
10214
10215 static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10216         .rpc_call_prepare = nfs4_layoutcommit_prepare,
10217         .rpc_call_done = nfs4_layoutcommit_done,
10218         .rpc_release = nfs4_layoutcommit_release,
10219 };
10220
10221 int
10222 nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10223 {
10224         struct rpc_message msg = {
10225                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10226                 .rpc_argp = &data->args,
10227                 .rpc_resp = &data->res,
10228                 .rpc_cred = data->cred,
10229         };
10230         struct rpc_task_setup task_setup_data = {
10231                 .task = &data->task,
10232                 .rpc_client = NFS_CLIENT(data->args.inode),
10233                 .rpc_message = &msg,
10234                 .callback_ops = &nfs4_layoutcommit_ops,
10235                 .callback_data = data,
10236                 .flags = RPC_TASK_MOVEABLE,
10237         };
10238         struct rpc_task *task;
10239         int status = 0;
10240
10241         dprintk("NFS: initiating layoutcommit call. sync %d "
10242                 "lbw: %llu inode %lu\n", sync,
10243                 data->args.lastbytewritten,
10244                 data->args.inode->i_ino);
10245
10246         if (!sync) {
10247                 data->inode = nfs_igrab_and_active(data->args.inode);
10248                 if (data->inode == NULL) {
10249                         nfs4_layoutcommit_release(data);
10250                         return -EAGAIN;
10251                 }
10252                 task_setup_data.flags = RPC_TASK_ASYNC;
10253         }
10254         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10255         task = rpc_run_task(&task_setup_data);
10256         if (IS_ERR(task))
10257                 return PTR_ERR(task);
10258         if (sync)
10259                 status = task->tk_status;
10260         trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10261         dprintk("%s: status %d\n", __func__, status);
10262         rpc_put_task(task);
10263         return status;
10264 }
10265
10266 /*
10267  * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10268  * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10269  */
10270 static int
10271 _nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10272                     struct nfs_fsinfo *info,
10273                     struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10274 {
10275         struct nfs41_secinfo_no_name_args args = {
10276                 .style = SECINFO_STYLE_CURRENT_FH,
10277         };
10278         struct nfs4_secinfo_res res = {
10279                 .flavors = flavors,
10280         };
10281         struct rpc_message msg = {
10282                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10283                 .rpc_argp = &args,
10284                 .rpc_resp = &res,
10285         };
10286         struct nfs4_call_sync_data data = {
10287                 .seq_server = server,
10288                 .seq_args = &args.seq_args,
10289                 .seq_res = &res.seq_res,
10290         };
10291         struct rpc_task_setup task_setup = {
10292                 .rpc_client = server->client,
10293                 .rpc_message = &msg,
10294                 .callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10295                 .callback_data = &data,
10296                 .flags = RPC_TASK_NO_ROUND_ROBIN,
10297         };
10298         const struct cred *cred = NULL;
10299         int status;
10300
10301         if (use_integrity) {
10302                 task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10303
10304                 cred = nfs4_get_clid_cred(server->nfs_client);
10305                 msg.rpc_cred = cred;
10306         }
10307
10308         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10309         status = nfs4_call_sync_custom(&task_setup);
10310         dprintk("<-- %s status=%d\n", __func__, status);
10311
10312         put_cred(cred);
10313
10314         return status;
10315 }
10316
10317 static int
10318 nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10319                            struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10320 {
10321         struct nfs4_exception exception = {
10322                 .interruptible = true,
10323         };
10324         int err;
10325         do {
10326                 /* first try using integrity protection */
10327                 err = -NFS4ERR_WRONGSEC;
10328
10329                 /* try to use integrity protection with machine cred */
10330                 if (_nfs4_is_integrity_protected(server->nfs_client))
10331                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10332                                                           flavors, true);
10333
10334                 /*
10335                  * if unable to use integrity protection, or SECINFO with
10336                  * integrity protection returns NFS4ERR_WRONGSEC (which is
10337                  * disallowed by spec, but exists in deployed servers) use
10338                  * the current filesystem's rpc_client and the user cred.
10339                  */
10340                 if (err == -NFS4ERR_WRONGSEC)
10341                         err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10342                                                           flavors, false);
10343
10344                 switch (err) {
10345                 case 0:
10346                 case -NFS4ERR_WRONGSEC:
10347                 case -ENOTSUPP:
10348                         goto out;
10349                 default:
10350                         err = nfs4_handle_exception(server, err, &exception);
10351                 }
10352         } while (exception.retry);
10353 out:
10354         return err;
10355 }
10356
10357 static int
10358 nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10359                     struct nfs_fsinfo *info)
10360 {
10361         int err;
10362         struct page *page;
10363         rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10364         struct nfs4_secinfo_flavors *flavors;
10365         struct nfs4_secinfo4 *secinfo;
10366         int i;
10367
10368         page = alloc_page(GFP_KERNEL);
10369         if (!page) {
10370                 err = -ENOMEM;
10371                 goto out;
10372         }
10373
10374         flavors = page_address(page);
10375         err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10376
10377         /*
10378          * Fall back on "guess and check" method if
10379          * the server doesn't support SECINFO_NO_NAME
10380          */
10381         if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10382                 err = nfs4_find_root_sec(server, fhandle, info);
10383                 goto out_freepage;
10384         }
10385         if (err)
10386                 goto out_freepage;
10387
10388         for (i = 0; i < flavors->num_flavors; i++) {
10389                 secinfo = &flavors->flavors[i];
10390
10391                 switch (secinfo->flavor) {
10392                 case RPC_AUTH_NULL:
10393                 case RPC_AUTH_UNIX:
10394                 case RPC_AUTH_GSS:
10395                         flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10396                                         &secinfo->flavor_info);
10397                         break;
10398                 default:
10399                         flavor = RPC_AUTH_MAXFLAVOR;
10400                         break;
10401                 }
10402
10403                 if (!nfs_auth_info_match(&server->auth_info, flavor))
10404                         flavor = RPC_AUTH_MAXFLAVOR;
10405
10406                 if (flavor != RPC_AUTH_MAXFLAVOR) {
10407                         err = nfs4_lookup_root_sec(server, fhandle,
10408                                                    info, flavor);
10409                         if (!err)
10410                                 break;
10411                 }
10412         }
10413
10414         if (flavor == RPC_AUTH_MAXFLAVOR)
10415                 err = -EPERM;
10416
10417 out_freepage:
10418         put_page(page);
10419         if (err == -EACCES)
10420                 return -EPERM;
10421 out:
10422         return err;
10423 }
10424
10425 static int _nfs41_test_stateid(struct nfs_server *server,
10426                                const nfs4_stateid *stateid,
10427                                const struct cred *cred)
10428 {
10429         int status;
10430         struct nfs41_test_stateid_args args = {
10431                 .stateid = *stateid,
10432         };
10433         struct nfs41_test_stateid_res res;
10434         struct rpc_message msg = {
10435                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10436                 .rpc_argp = &args,
10437                 .rpc_resp = &res,
10438                 .rpc_cred = cred,
10439         };
10440         struct rpc_clnt *rpc_client = server->client;
10441
10442         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10443                 &rpc_client, &msg);
10444
10445         dprintk("NFS call  test_stateid %p\n", stateid);
10446         nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10447         status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10448                         &args.seq_args, &res.seq_res);
10449         if (status != NFS_OK) {
10450                 dprintk("NFS reply test_stateid: failed, %d\n", status);
10451                 return status;
10452         }
10453         dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10454         return -res.status;
10455 }
10456
10457 static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10458                 int err, struct nfs4_exception *exception)
10459 {
10460         exception->retry = 0;
10461         switch(err) {
10462         case -NFS4ERR_DELAY:
10463         case -NFS4ERR_RETRY_UNCACHED_REP:
10464                 nfs4_handle_exception(server, err, exception);
10465                 break;
10466         case -NFS4ERR_BADSESSION:
10467         case -NFS4ERR_BADSLOT:
10468         case -NFS4ERR_BAD_HIGH_SLOT:
10469         case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10470         case -NFS4ERR_DEADSESSION:
10471                 nfs4_do_handle_exception(server, err, exception);
10472         }
10473 }
10474
10475 /**
10476  * nfs41_test_stateid - perform a TEST_STATEID operation
10477  *
10478  * @server: server / transport on which to perform the operation
10479  * @stateid: state ID to test
10480  * @cred: credential
10481  *
10482  * Returns NFS_OK if the server recognizes that "stateid" is valid.
10483  * Otherwise a negative NFS4ERR value is returned if the operation
10484  * failed or the state ID is not currently valid.
10485  */
10486 static int nfs41_test_stateid(struct nfs_server *server,
10487                               const nfs4_stateid *stateid,
10488                               const struct cred *cred)
10489 {
10490         struct nfs4_exception exception = {
10491                 .interruptible = true,
10492         };
10493         int err;
10494         do {
10495                 err = _nfs41_test_stateid(server, stateid, cred);
10496                 nfs4_handle_delay_or_session_error(server, err, &exception);
10497         } while (exception.retry);
10498         return err;
10499 }
10500
10501 struct nfs_free_stateid_data {
10502         struct nfs_server *server;
10503         struct nfs41_free_stateid_args args;
10504         struct nfs41_free_stateid_res res;
10505 };
10506
10507 static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10508 {
10509         struct nfs_free_stateid_data *data = calldata;
10510         nfs4_setup_sequence(data->server->nfs_client,
10511                         &data->args.seq_args,
10512                         &data->res.seq_res,
10513                         task);
10514 }
10515
10516 static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10517 {
10518         struct nfs_free_stateid_data *data = calldata;
10519
10520         nfs41_sequence_done(task, &data->res.seq_res);
10521
10522         switch (task->tk_status) {
10523         case -NFS4ERR_DELAY:
10524                 if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10525                         rpc_restart_call_prepare(task);
10526         }
10527 }
10528
10529 static void nfs41_free_stateid_release(void *calldata)
10530 {
10531         struct nfs_free_stateid_data *data = calldata;
10532         struct nfs_client *clp = data->server->nfs_client;
10533
10534         nfs_put_client(clp);
10535         kfree(calldata);
10536 }
10537
10538 static const struct rpc_call_ops nfs41_free_stateid_ops = {
10539         .rpc_call_prepare = nfs41_free_stateid_prepare,
10540         .rpc_call_done = nfs41_free_stateid_done,
10541         .rpc_release = nfs41_free_stateid_release,
10542 };
10543
10544 /**
10545  * nfs41_free_stateid - perform a FREE_STATEID operation
10546  *
10547  * @server: server / transport on which to perform the operation
10548  * @stateid: state ID to release
10549  * @cred: credential
10550  * @privileged: set to true if this call needs to be privileged
10551  *
10552  * Note: this function is always asynchronous.
10553  */
10554 static int nfs41_free_stateid(struct nfs_server *server,
10555                 const nfs4_stateid *stateid,
10556                 const struct cred *cred,
10557                 bool privileged)
10558 {
10559         struct rpc_message msg = {
10560                 .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10561                 .rpc_cred = cred,
10562         };
10563         struct rpc_task_setup task_setup = {
10564                 .rpc_client = server->client,
10565                 .rpc_message = &msg,
10566                 .callback_ops = &nfs41_free_stateid_ops,
10567                 .flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10568         };
10569         struct nfs_free_stateid_data *data;
10570         struct rpc_task *task;
10571         struct nfs_client *clp = server->nfs_client;
10572
10573         if (!refcount_inc_not_zero(&clp->cl_count))
10574                 return -EIO;
10575
10576         nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10577                 &task_setup.rpc_client, &msg);
10578
10579         dprintk("NFS call  free_stateid %p\n", stateid);
10580         data = kmalloc(sizeof(*data), GFP_KERNEL);
10581         if (!data)
10582                 return -ENOMEM;
10583         data->server = server;
10584         nfs4_stateid_copy(&data->args.stateid, stateid);
10585
10586         task_setup.callback_data = data;
10587
10588         msg.rpc_argp = &data->args;
10589         msg.rpc_resp = &data->res;
10590         nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10591         task = rpc_run_task(&task_setup);
10592         if (IS_ERR(task))
10593                 return PTR_ERR(task);
10594         rpc_put_task(task);
10595         return 0;
10596 }
10597
10598 static void
10599 nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10600 {
10601         const struct cred *cred = lsp->ls_state->owner->so_cred;
10602
10603         nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10604         nfs4_free_lock_state(server, lsp);
10605 }
10606
10607 static bool nfs41_match_stateid(const nfs4_stateid *s1,
10608                 const nfs4_stateid *s2)
10609 {
10610         if (s1->type != s2->type)
10611                 return false;
10612
10613         if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10614                 return false;
10615
10616         if (s1->seqid == s2->seqid)
10617                 return true;
10618
10619         return s1->seqid == 0 || s2->seqid == 0;
10620 }
10621
10622 #endif /* CONFIG_NFS_V4_1 */
10623
10624 static bool nfs4_match_stateid(const nfs4_stateid *s1,
10625                 const nfs4_stateid *s2)
10626 {
10627         return nfs4_stateid_match(s1, s2);
10628 }
10629
10630
10631 static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10632         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10633         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10634         .recover_open   = nfs4_open_reclaim,
10635         .recover_lock   = nfs4_lock_reclaim,
10636         .establish_clid = nfs4_init_clientid,
10637         .detect_trunking = nfs40_discover_server_trunking,
10638 };
10639
10640 #if defined(CONFIG_NFS_V4_1)
10641 static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10642         .owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10643         .state_flag_bit = NFS_STATE_RECLAIM_REBOOT,
10644         .recover_open   = nfs4_open_reclaim,
10645         .recover_lock   = nfs4_lock_reclaim,
10646         .establish_clid = nfs41_init_clientid,
10647         .reclaim_complete = nfs41_proc_reclaim_complete,
10648         .detect_trunking = nfs41_discover_server_trunking,
10649 };
10650 #endif /* CONFIG_NFS_V4_1 */
10651
10652 static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10653         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10654         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10655         .recover_open   = nfs40_open_expired,
10656         .recover_lock   = nfs4_lock_expired,
10657         .establish_clid = nfs4_init_clientid,
10658 };
10659
10660 #if defined(CONFIG_NFS_V4_1)
10661 static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10662         .owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10663         .state_flag_bit = NFS_STATE_RECLAIM_NOGRACE,
10664         .recover_open   = nfs41_open_expired,
10665         .recover_lock   = nfs41_lock_expired,
10666         .establish_clid = nfs41_init_clientid,
10667 };
10668 #endif /* CONFIG_NFS_V4_1 */
10669
10670 static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10671         .sched_state_renewal = nfs4_proc_async_renew,
10672         .get_state_renewal_cred = nfs4_get_renew_cred,
10673         .renew_lease = nfs4_proc_renew,
10674 };
10675
10676 #if defined(CONFIG_NFS_V4_1)
10677 static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10678         .sched_state_renewal = nfs41_proc_async_sequence,
10679         .get_state_renewal_cred = nfs4_get_machine_cred,
10680         .renew_lease = nfs4_proc_sequence,
10681 };
10682 #endif
10683
10684 static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10685         .get_locations = _nfs40_proc_get_locations,
10686         .fsid_present = _nfs40_proc_fsid_present,
10687 };
10688
10689 #if defined(CONFIG_NFS_V4_1)
10690 static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10691         .get_locations = _nfs41_proc_get_locations,
10692         .fsid_present = _nfs41_proc_fsid_present,
10693 };
10694 #endif  /* CONFIG_NFS_V4_1 */
10695
10696 static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10697         .minor_version = 0,
10698         .init_caps = NFS_CAP_READDIRPLUS
10699                 | NFS_CAP_ATOMIC_OPEN
10700                 | NFS_CAP_POSIX_LOCK,
10701         .init_client = nfs40_init_client,
10702         .shutdown_client = nfs40_shutdown_client,
10703         .match_stateid = nfs4_match_stateid,
10704         .find_root_sec = nfs4_find_root_sec,
10705         .free_lock_state = nfs4_release_lockowner,
10706         .test_and_free_expired = nfs40_test_and_free_expired_stateid,
10707         .alloc_seqid = nfs_alloc_seqid,
10708         .call_sync_ops = &nfs40_call_sync_ops,
10709         .reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10710         .nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10711         .state_renewal_ops = &nfs40_state_renewal_ops,
10712         .mig_recovery_ops = &nfs40_mig_recovery_ops,
10713 };
10714
10715 #if defined(CONFIG_NFS_V4_1)
10716 static struct nfs_seqid *
10717 nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10718 {
10719         return NULL;
10720 }
10721
10722 static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10723         .minor_version = 1,
10724         .init_caps = NFS_CAP_READDIRPLUS
10725                 | NFS_CAP_ATOMIC_OPEN
10726                 | NFS_CAP_POSIX_LOCK
10727                 | NFS_CAP_STATEID_NFSV41
10728                 | NFS_CAP_ATOMIC_OPEN_V1
10729                 | NFS_CAP_LGOPEN
10730                 | NFS_CAP_MOVEABLE,
10731         .init_client = nfs41_init_client,
10732         .shutdown_client = nfs41_shutdown_client,
10733         .match_stateid = nfs41_match_stateid,
10734         .find_root_sec = nfs41_find_root_sec,
10735         .free_lock_state = nfs41_free_lock_state,
10736         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10737         .alloc_seqid = nfs_alloc_no_seqid,
10738         .session_trunk = nfs4_test_session_trunk,
10739         .call_sync_ops = &nfs41_call_sync_ops,
10740         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10741         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10742         .state_renewal_ops = &nfs41_state_renewal_ops,
10743         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10744 };
10745 #endif
10746
10747 #if defined(CONFIG_NFS_V4_2)
10748 static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10749         .minor_version = 2,
10750         .init_caps = NFS_CAP_READDIRPLUS
10751                 | NFS_CAP_ATOMIC_OPEN
10752                 | NFS_CAP_POSIX_LOCK
10753                 | NFS_CAP_STATEID_NFSV41
10754                 | NFS_CAP_ATOMIC_OPEN_V1
10755                 | NFS_CAP_LGOPEN
10756                 | NFS_CAP_ALLOCATE
10757                 | NFS_CAP_COPY
10758                 | NFS_CAP_OFFLOAD_CANCEL
10759                 | NFS_CAP_COPY_NOTIFY
10760                 | NFS_CAP_DEALLOCATE
10761                 | NFS_CAP_SEEK
10762                 | NFS_CAP_LAYOUTSTATS
10763                 | NFS_CAP_CLONE
10764                 | NFS_CAP_LAYOUTERROR
10765                 | NFS_CAP_READ_PLUS
10766                 | NFS_CAP_MOVEABLE,
10767         .init_client = nfs41_init_client,
10768         .shutdown_client = nfs41_shutdown_client,
10769         .match_stateid = nfs41_match_stateid,
10770         .find_root_sec = nfs41_find_root_sec,
10771         .free_lock_state = nfs41_free_lock_state,
10772         .call_sync_ops = &nfs41_call_sync_ops,
10773         .test_and_free_expired = nfs41_test_and_free_expired_stateid,
10774         .alloc_seqid = nfs_alloc_no_seqid,
10775         .session_trunk = nfs4_test_session_trunk,
10776         .reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10777         .nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10778         .state_renewal_ops = &nfs41_state_renewal_ops,
10779         .mig_recovery_ops = &nfs41_mig_recovery_ops,
10780 };
10781 #endif
10782
10783 const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10784         [0] = &nfs_v4_0_minor_ops,
10785 #if defined(CONFIG_NFS_V4_1)
10786         [1] = &nfs_v4_1_minor_ops,
10787 #endif
10788 #if defined(CONFIG_NFS_V4_2)
10789         [2] = &nfs_v4_2_minor_ops,
10790 #endif
10791 };
10792
10793 static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10794 {
10795         ssize_t error, error2, error3;
10796         size_t left = size;
10797
10798         error = generic_listxattr(dentry, list, left);
10799         if (error < 0)
10800                 return error;
10801         if (list) {
10802                 list += error;
10803                 left -= error;
10804         }
10805
10806         error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, left);
10807         if (error2 < 0)
10808                 return error2;
10809
10810         if (list) {
10811                 list += error2;
10812                 left -= error2;
10813         }
10814
10815         error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10816         if (error3 < 0)
10817                 return error3;
10818
10819         error += error2 + error3;
10820         if (size && error > size)
10821                 return -ERANGE;
10822         return error;
10823 }
10824
10825 static void nfs4_enable_swap(struct inode *inode)
10826 {
10827         /* The state manager thread must always be running.
10828          * It will notice the client is a swapper, and stay put.
10829          */
10830         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10831
10832         nfs4_schedule_state_manager(clp);
10833 }
10834
10835 static void nfs4_disable_swap(struct inode *inode)
10836 {
10837         /* The state manager thread will now exit once it is
10838          * woken.
10839          */
10840         struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10841
10842         set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10843         clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10844         wake_up_var(&clp->cl_state);
10845 }
10846
10847 static const struct inode_operations nfs4_dir_inode_operations = {
10848         .create         = nfs_create,
10849         .lookup         = nfs_lookup,
10850         .atomic_open    = nfs_atomic_open,
10851         .link           = nfs_link,
10852         .unlink         = nfs_unlink,
10853         .symlink        = nfs_symlink,
10854         .mkdir          = nfs_mkdir,
10855         .rmdir          = nfs_rmdir,
10856         .mknod          = nfs_mknod,
10857         .rename         = nfs_rename,
10858         .permission     = nfs_permission,
10859         .getattr        = nfs_getattr,
10860         .setattr        = nfs_setattr,
10861         .listxattr      = nfs4_listxattr,
10862 };
10863
10864 static const struct inode_operations nfs4_file_inode_operations = {
10865         .permission     = nfs_permission,
10866         .getattr        = nfs_getattr,
10867         .setattr        = nfs_setattr,
10868         .listxattr      = nfs4_listxattr,
10869 };
10870
10871 const struct nfs_rpc_ops nfs_v4_clientops = {
10872         .version        = 4,                    /* protocol version */
10873         .dentry_ops     = &nfs4_dentry_operations,
10874         .dir_inode_ops  = &nfs4_dir_inode_operations,
10875         .file_inode_ops = &nfs4_file_inode_operations,
10876         .file_ops       = &nfs4_file_operations,
10877         .getroot        = nfs4_proc_get_root,
10878         .submount       = nfs4_submount,
10879         .try_get_tree   = nfs4_try_get_tree,
10880         .getattr        = nfs4_proc_getattr,
10881         .setattr        = nfs4_proc_setattr,
10882         .lookup         = nfs4_proc_lookup,
10883         .lookupp        = nfs4_proc_lookupp,
10884         .access         = nfs4_proc_access,
10885         .readlink       = nfs4_proc_readlink,
10886         .create         = nfs4_proc_create,
10887         .remove         = nfs4_proc_remove,
10888         .unlink_setup   = nfs4_proc_unlink_setup,
10889         .unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10890         .unlink_done    = nfs4_proc_unlink_done,
10891         .rename_setup   = nfs4_proc_rename_setup,
10892         .rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10893         .rename_done    = nfs4_proc_rename_done,
10894         .link           = nfs4_proc_link,
10895         .symlink        = nfs4_proc_symlink,
10896         .mkdir          = nfs4_proc_mkdir,
10897         .rmdir          = nfs4_proc_rmdir,
10898         .readdir        = nfs4_proc_readdir,
10899         .mknod          = nfs4_proc_mknod,
10900         .statfs         = nfs4_proc_statfs,
10901         .fsinfo         = nfs4_proc_fsinfo,
10902         .pathconf       = nfs4_proc_pathconf,
10903         .set_capabilities = nfs4_server_capabilities,
10904         .decode_dirent  = nfs4_decode_dirent,
10905         .pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10906         .read_setup     = nfs4_proc_read_setup,
10907         .read_done      = nfs4_read_done,
10908         .write_setup    = nfs4_proc_write_setup,
10909         .write_done     = nfs4_write_done,
10910         .commit_setup   = nfs4_proc_commit_setup,
10911         .commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10912         .commit_done    = nfs4_commit_done,
10913         .lock           = nfs4_proc_lock,
10914         .clear_acl_cache = nfs4_zap_acl_attr,
10915         .close_context  = nfs4_close_context,
10916         .open_context   = nfs4_atomic_open,
10917         .have_delegation = nfs4_have_delegation,
10918         .return_delegation = nfs4_inode_return_delegation,
10919         .alloc_client   = nfs4_alloc_client,
10920         .init_client    = nfs4_init_client,
10921         .free_client    = nfs4_free_client,
10922         .create_server  = nfs4_create_server,
10923         .clone_server   = nfs_clone_server,
10924         .discover_trunking = nfs4_discover_trunking,
10925         .enable_swap    = nfs4_enable_swap,
10926         .disable_swap   = nfs4_disable_swap,
10927 };
10928
10929 static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10930         .name   = XATTR_NAME_NFSV4_ACL,
10931         .list   = nfs4_xattr_list_nfs4_acl,
10932         .get    = nfs4_xattr_get_nfs4_acl,
10933         .set    = nfs4_xattr_set_nfs4_acl,
10934 };
10935
10936 #if defined(CONFIG_NFS_V4_1)
10937 static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10938         .name   = XATTR_NAME_NFSV4_DACL,
10939         .list   = nfs4_xattr_list_nfs4_dacl,
10940         .get    = nfs4_xattr_get_nfs4_dacl,
10941         .set    = nfs4_xattr_set_nfs4_dacl,
10942 };
10943
10944 static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10945         .name   = XATTR_NAME_NFSV4_SACL,
10946         .list   = nfs4_xattr_list_nfs4_sacl,
10947         .get    = nfs4_xattr_get_nfs4_sacl,
10948         .set    = nfs4_xattr_set_nfs4_sacl,
10949 };
10950 #endif
10951
10952 #ifdef CONFIG_NFS_V4_2
10953 static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10954         .prefix = XATTR_USER_PREFIX,
10955         .get    = nfs4_xattr_get_nfs4_user,
10956         .set    = nfs4_xattr_set_nfs4_user,
10957 };
10958 #endif
10959
10960 const struct xattr_handler * const nfs4_xattr_handlers[] = {
10961         &nfs4_xattr_nfs4_acl_handler,
10962 #if defined(CONFIG_NFS_V4_1)
10963         &nfs4_xattr_nfs4_dacl_handler,
10964         &nfs4_xattr_nfs4_sacl_handler,
10965 #endif
10966 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
10967         &nfs4_xattr_nfs4_label_handler,
10968 #endif
10969 #ifdef CONFIG_NFS_V4_2
10970         &nfs4_xattr_nfs4_user_handler,
10971 #endif
10972         NULL
10973 };