projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFS: Fix typo in nfs_do_clone_mount()
[linux-2.6-block.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
4ece3a2d 29#include <linux/sunrpc/metrics.h>
1da177e4
LT		 30#include <linux/nfs_fs.h>
31#include <linux/nfs_mount.h>
32#include <linux/nfs4_mount.h>
33#include <linux/lockd/bind.h>
34#include <linux/smp_lock.h>
35#include <linux/seq_file.h>
36#include <linux/mount.h>
37#include <linux/nfs_idmap.h>
38#include <linux/vfs.h>
9cdb3883
MN		 39#include <linux/inet.h>
40#include <linux/nfs_xdr.h>
1da177e4
LT		 41
42#include <asm/system.h>
43#include <asm/uaccess.h>
44
4ce79717 45#include "nfs4_fs.h"
a72b4422 46#include "callback.h"
1da177e4 47#include "delegation.h"
d9ef5a8c 48#include "iostat.h"
1da177e4
LT		 49
50#define NFSDBG_FACILITY NFSDBG_VFS
51#define NFS_PARANOIA 1
52
53/* Maximum number of readahead requests
54 * FIXME: this should really be a sysctl so that users may tune it to suit
55 * their needs. People that do NFS over a slow network, might for
56 * instance want to reduce it to something closer to 1 for improved
57 * interactive response.
58 */
59#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
60
61static void nfs_invalidate_inode(struct inode *);
24aa1fe6 62static int nfs_update_inode(struct inode *, struct nfs_fattr *);
1da177e4
LT		 63
64static struct inode *nfs_alloc_inode(struct super_block *sb);
65static void nfs_destroy_inode(struct inode *);
66static int nfs_write_inode(struct inode *,int);
1da177e4 67static void nfs_clear_inode(struct inode *);
8b512d9a 68static void nfs_umount_begin(struct vfsmount *, int);
1da177e4
LT		 69static int nfs_statfs(struct super_block *, struct kstatfs *);
70static int nfs_show_options(struct seq_file *, struct vfsmount *);
d9ef5a8c 71static int nfs_show_stats(struct seq_file *, struct vfsmount *);
ada70d94 72static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT		 73
74static struct rpc_program nfs_program;
75
76static struct super_operations nfs_sops = {
77 .alloc_inode = nfs_alloc_inode,
78 .destroy_inode = nfs_destroy_inode,
79 .write_inode = nfs_write_inode,
1da177e4
LT		 80 .statfs = nfs_statfs,
81 .clear_inode = nfs_clear_inode,
82 .umount_begin = nfs_umount_begin,
83 .show_options = nfs_show_options,
d9ef5a8c 84 .show_stats = nfs_show_stats,
1da177e4
LT		 85};
86
87/*
88 * RPC cruft for NFS
89 */
90static struct rpc_stat nfs_rpcstat = {
91 .program = &nfs_program
92};
93static struct rpc_version * nfs_version[] = {
94 NULL,
95 NULL,
96 &nfs_version2,
97#if defined(CONFIG_NFS_V3)
98 &nfs_version3,
99#elif defined(CONFIG_NFS_V4)
100 NULL,
101#endif
102#if defined(CONFIG_NFS_V4)
103 &nfs_version4,
104#endif
105};
106
107static struct rpc_program nfs_program = {
108 .name = "nfs",
109 .number = NFS_PROGRAM,
e8c96f8c 110 .nrvers = ARRAY_SIZE(nfs_version),
1da177e4
LT		 111 .version = nfs_version,
112 .stats = &nfs_rpcstat,
113 .pipe_dir_name = "/nfs",
114};
115
b7fa0554
AG		 116#ifdef CONFIG_NFS_V3_ACL
117static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
118static struct rpc_version * nfsacl_version[] = {
119 [3] = &nfsacl_version3,
120};
121
122struct rpc_program nfsacl_program = {
123 .name = "nfsacl",
124 .number = NFS_ACL_PROGRAM,
e8c96f8c 125 .nrvers = ARRAY_SIZE(nfsacl_version),
b7fa0554
AG		 126 .version = nfsacl_version,
127 .stats = &nfsacl_rpcstat,
128};
129#endif /* CONFIG_NFS_V3_ACL */
130
1da177e4
LT		 131static inline unsigned long
132nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
133{
134 return nfs_fileid_to_ino_t(fattr->fileid);
135}
136
137static int
138nfs_write_inode(struct inode *inode, int sync)
139{
c42de9dd 140 int flags = sync ? FLUSH_SYNC : 0;
1da177e4
LT		 141 int ret;
142
3da28eb1 143 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 144 if (ret < 0)
145 return ret;
146 return 0;
147}
148
1da177e4
LT		 149static void
150nfs_clear_inode(struct inode *inode)
151{
152 struct nfs_inode *nfsi = NFS_I(inode);
153 struct rpc_cred *cred;
154
da6d503a
TM		 155 /*
156 * The following should never happen...
157 */
158 BUG_ON(nfs_have_writebacks(inode));
1da177e4 159 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 160 nfs_zap_acl_cache(inode);
1da177e4
LT		 161 cred = nfsi->cache_access.cred;
162 if (cred)
163 put_rpccred(cred);
164 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
165}
166
8b512d9a 167static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags)
1da177e4 168{
8b512d9a
TM		 169 struct nfs_server *server;
170 struct rpc_clnt *rpc;
1da177e4 171
51d8fa6a 172 shrink_submounts(vfsmnt, &nfs_automount_list);
8b512d9a
TM		 173 if (!(flags & MNT_FORCE))
174 return;
1da177e4 175 /* -EIO all pending I/O */
8b512d9a
TM		 176 server = NFS_SB(vfsmnt->mnt_sb);
177 rpc = server->client;
6a19275a 178 if (!IS_ERR(rpc))
1da177e4 179 rpc_killall_tasks(rpc);
8b512d9a 180 rpc = server->client_acl;
b7fa0554
AG		 181 if (!IS_ERR(rpc))
182 rpc_killall_tasks(rpc);
1da177e4
LT		 183}
184
185
186static inline unsigned long
187nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
188{
189 /* make sure blocksize is a power of two */
190 if ((bsize & (bsize - 1)) || nrbitsp) {
191 unsigned char nrbits;
192
193 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
194 ;
195 bsize = 1 << nrbits;
196 if (nrbitsp)
197 *nrbitsp = nrbits;
198 }
199
200 return bsize;
201}
202
203/*
204 * Calculate the number of 512byte blocks used.
205 */
206static inline unsigned long
207nfs_calc_block_size(u64 tsize)
208{
209 loff_t used = (tsize + 511) >> 9;
210 return (used > ULONG_MAX) ? ULONG_MAX : used;
211}
212
213/*
214 * Compute and set NFS server blocksize
215 */
216static inline unsigned long
217nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
218{
40859d7e
CL		 219 if (bsize < NFS_MIN_FILE_IO_SIZE)
220 bsize = NFS_DEF_FILE_IO_SIZE;
221 else if (bsize >= NFS_MAX_FILE_IO_SIZE)
222 bsize = NFS_MAX_FILE_IO_SIZE;
1da177e4
LT		 223
224 return nfs_block_bits(bsize, nrbitsp);
225}
226
55a97593
TM		 227static inline void
228nfs_super_set_maxbytes(struct super_block *sb, __u64 maxfilesize)
229{
230 sb->s_maxbytes = (loff_t)maxfilesize;
231 if (sb->s_maxbytes > MAX_LFS_FILESIZE || sb->s_maxbytes <= 0)
232 sb->s_maxbytes = MAX_LFS_FILESIZE;
233}
234
1da177e4
LT		 235/*
236 * Obtain the root inode of the file system.
237 */
238static struct inode *
239nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
240{
241 struct nfs_server *server = NFS_SB(sb);
1da177e4
LT		 242 int error;
243
244 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
245 if (error < 0) {
246 dprintk("nfs_get_root: getattr error = %d\n", -error);
247 return ERR_PTR(error);
248 }
249
8b4bdcf8 250 server->fsid = fsinfo->fattr->fsid;
03f28e3a 251 return nfs_fhget(sb, rootfh, fsinfo->fattr);
1da177e4
LT		 252}
253
254/*
255 * Do NFS version-independent mount processing, and sanity checking
256 */
257static int
258nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
259{
260 struct nfs_server *server;
261 struct inode *root_inode;
262 struct nfs_fattr fattr;
263 struct nfs_fsinfo fsinfo = {
264 .fattr = &fattr,
265 };
266 struct nfs_pathconf pathinfo = {
267 .fattr = &fattr,
268 };
269 int no_root_error = 0;
270 unsigned long max_rpc_payload;
271
272 /* We probably want something more informative here */
273 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
274
275 server = NFS_SB(sb);
276
277 sb->s_magic = NFS_SUPER_MAGIC;
278
01d0ae8b
TM		 279 server->io_stats = nfs_alloc_iostats();
280 if (server->io_stats == NULL)
281 return -ENOMEM;
282
1da177e4
LT		 283 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
284 /* Did getting the root inode fail? */
285 if (IS_ERR(root_inode)) {
286 no_root_error = PTR_ERR(root_inode);
287 goto out_no_root;
288 }
289 sb->s_root = d_alloc_root(root_inode);
290 if (!sb->s_root) {
291 no_root_error = -ENOMEM;
292 goto out_no_root;
293 }
294 sb->s_root->d_op = server->rpc_ops->dentry_ops;
295
67ec9f46
CL		 296 /* mount time stamp, in seconds */
297 server->mount_time = jiffies;
298
1da177e4
LT		 299 /* Get some general file system info */
300 if (server->namelen == 0 &&
301 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
302 server->namelen = pathinfo.max_namelen;
303 /* Work out a lot of parameters */
304 if (server->rsize == 0)
305 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
306 if (server->wsize == 0)
307 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
308
309 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
310 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
311 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
312 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
313
314 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
315 if (server->rsize > max_rpc_payload)
316 server->rsize = max_rpc_payload;
40859d7e
CL		 317 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
318 server->rsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 319 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4 320
40859d7e
CL		 321 if (server->wsize > max_rpc_payload)
322 server->wsize = max_rpc_payload;
323 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
324 server->wsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 325 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4
LT		 326
327 if (sb->s_blocksize == 0)
328 sb->s_blocksize = nfs_block_bits(server->wsize,
329 &sb->s_blocksize_bits);
330 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
331
332 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
333 if (server->dtsize > PAGE_CACHE_SIZE)
334 server->dtsize = PAGE_CACHE_SIZE;
335 if (server->dtsize > server->rsize)
336 server->dtsize = server->rsize;
337
338 if (server->flags & NFS_MOUNT_NOAC) {
339 server->acregmin = server->acregmax = 0;
340 server->acdirmin = server->acdirmax = 0;
341 sb->s_flags |= MS_SYNCHRONOUS;
342 }
343 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
344
55a97593 345 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
1da177e4
LT		 346
347 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
348 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
349
350 /* We're airborne Set socket buffersize */
351 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
352 return 0;
353 /* Yargs. It didn't work out. */
354out_no_root:
355 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
356 if (!IS_ERR(root_inode))
357 iput(root_inode);
358 return no_root_error;
359}
360
eab5c084
CL		 361static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
362{
363 to->to_initval = timeo * HZ / 10;
364 to->to_retries = retrans;
365 if (!to->to_retries)
366 to->to_retries = 2;
367
368 switch (proto) {
369 case IPPROTO_TCP:
370 if (!to->to_initval)
371 to->to_initval = 60 * HZ;
03bf4b70
CL		 372 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
373 to->to_initval = NFS_MAX_TCP_TIMEOUT;
eab5c084
CL		 374 to->to_increment = to->to_initval;
375 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
376 to->to_exponential = 0;
377 break;
378 case IPPROTO_UDP:
379 default:
380 if (!to->to_initval)
381 to->to_initval = 11 * HZ / 10;
03bf4b70
CL		 382 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
383 to->to_initval = NFS_MAX_UDP_TIMEOUT;
384 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
eab5c084
CL		 385 to->to_exponential = 1;
386 break;
387 }
388}
389
1da177e4
LT		 390/*
391 * Create an RPC client handle.
392 */
393static struct rpc_clnt *
394nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
395{
396 struct rpc_timeout timeparms;
397 struct rpc_xprt *xprt = NULL;
398 struct rpc_clnt *clnt = NULL;
eab5c084 399 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
1da177e4 400
eab5c084 401 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
1da177e4 402
7a480e25
CL		 403 server->retrans_timeo = timeparms.to_initval;
404 server->retrans_count = timeparms.to_retries;
405
1da177e4 406 /* create transport and client */
eab5c084 407 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 408 if (IS_ERR(xprt)) {
9085bbcb
TM		 409 dprintk("%s: cannot create RPC transport. Error = %ld\n",
410 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 411 return (struct rpc_clnt *)xprt;
412 }
413 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
414 server->rpc_ops->version, data->pseudoflavor);
415 if (IS_ERR(clnt)) {
9085bbcb
TM		 416 dprintk("%s: cannot create RPC client. Error = %ld\n",
417 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 418 goto out_fail;
419 }
420
421 clnt->cl_intr = 1;
422 clnt->cl_softrtry = 1;
1da177e4
LT		 423
424 return clnt;
425
426out_fail:
1da177e4
LT		 427 return clnt;
428}
429
430/*
431 * The way this works is that the mount process passes a structure
432 * in the data argument which contains the server's IP address
433 * and the root file handle obtained from the server's mount
434 * daemon. We stash these away in the private superblock fields.
435 */
436static int
437nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
438{
439 struct nfs_server *server;
440 rpc_authflavor_t authflavor;
441
442 server = NFS_SB(sb);
443 sb->s_blocksize_bits = 0;
444 sb->s_blocksize = 0;
445 if (data->bsize)
446 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
447 if (data->rsize)
448 server->rsize = nfs_block_size(data->rsize, NULL);
449 if (data->wsize)
450 server->wsize = nfs_block_size(data->wsize, NULL);
451 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
452
453 server->acregmin = data->acregmin*HZ;
454 server->acregmax = data->acregmax*HZ;
455 server->acdirmin = data->acdirmin*HZ;
456 server->acdirmax = data->acdirmax*HZ;
457
458 /* Start lockd here, before we might error out */
459 if (!(server->flags & NFS_MOUNT_NONLM))
460 lockd_up();
461
462 server->namelen = data->namlen;
463 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
464 if (!server->hostname)
465 return -ENOMEM;
466 strcpy(server->hostname, data->hostname);
467
468 /* Check NFS protocol revision and initialize RPC op vector
469 * and file handle pool. */
1da177e4 470#ifdef CONFIG_NFS_V3
9085bbcb 471 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 472 server->rpc_ops = &nfs_v3_clientops;
473 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 474 } else {
475 server->rpc_ops = &nfs_v2_clientops;
476 }
9085bbcb
TM		 477#else
478 server->rpc_ops = &nfs_v2_clientops;
479#endif
1da177e4
LT		 480
481 /* Fill in pseudoflavor for mount version < 5 */
482 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
483 data->pseudoflavor = RPC_AUTH_UNIX;
484 authflavor = data->pseudoflavor; /* save for sb_init() */
485 /* XXX maybe we want to add a server->pseudoflavor field */
486
487 /* Create RPC client handles */
488 server->client = nfs_create_client(server, data);
489 if (IS_ERR(server->client))
490 return PTR_ERR(server->client);
491 /* RFC 2623, sec 2.3.2 */
492 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 493 struct rpc_auth *auth;
494
1da177e4
LT		 495 server->client_sys = rpc_clone_client(server->client);
496 if (IS_ERR(server->client_sys))
497 return PTR_ERR(server->client_sys);
6a19275a
BF		 498 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
499 if (IS_ERR(auth))
500 return PTR_ERR(auth);
1da177e4
LT		 501 } else {
502 atomic_inc(&server->client->cl_count);
503 server->client_sys = server->client;
504 }
1da177e4 505 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 506#ifdef CONFIG_NFS_V3_ACL
507 if (!(server->flags & NFS_MOUNT_NOACL)) {
508 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
509 /* No errors! Assume that Sun nfsacls are supported */
510 if (!IS_ERR(server->client_acl))
511 server->caps |= NFS_CAP_ACLS;
512 }
513#else
514 server->flags &= ~NFS_MOUNT_NOACL;
515#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 516 /*
517 * The VFS shouldn't apply the umask to mode bits. We will
518 * do so ourselves when necessary.
519 */
520 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 521 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
522 server->namelen = NFS3_MAXNAMLEN;
523 sb->s_time_gran = 1;
524 } else {
525 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
526 server->namelen = NFS2_MAXNAMLEN;
527 }
528
529 sb->s_op = &nfs_sops;
530 return nfs_sb_init(sb, authflavor);
531}
532
533static int
534nfs_statfs(struct super_block *sb, struct kstatfs *buf)
535{
536 struct nfs_server *server = NFS_SB(sb);
537 unsigned char blockbits;
538 unsigned long blockres;
539 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
540 struct nfs_fattr fattr;
541 struct nfs_fsstat res = {
542 .fattr = &fattr,
543 };
544 int error;
545
546 lock_kernel();
547
548 error = server->rpc_ops->statfs(server, rootfh, &res);
549 buf->f_type = NFS_SUPER_MAGIC;
550 if (error < 0)
551 goto out_err;
552
553 /*
554 * Current versions of glibc do not correctly handle the
555 * case where f_frsize != f_bsize. Eventually we want to
556 * report the value of wtmult in this field.
557 */
558 buf->f_frsize = sb->s_blocksize;
559
560 /*
561 * On most *nix systems, f_blocks, f_bfree, and f_bavail
562 * are reported in units of f_frsize. Linux hasn't had
563 * an f_frsize field in its statfs struct until recently,
564 * thus historically Linux's sys_statfs reports these
565 * fields in units of f_bsize.
566 */
567 buf->f_bsize = sb->s_blocksize;
568 blockbits = sb->s_blocksize_bits;
569 blockres = (1 << blockbits) - 1;
570 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
571 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
572 buf->f_bavail = (res.abytes + blockres) >> blockbits;
573
574 buf->f_files = res.tfiles;
575 buf->f_ffree = res.afiles;
576
577 buf->f_namelen = server->namelen;
578 out:
579 unlock_kernel();
1da177e4
LT		 580 return 0;
581
582 out_err:
dc20f803 583 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
1da177e4
LT		 584 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
585 goto out;
586
587}
588
d9ef5a8c 589static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
1da177e4
LT		 590{
591 static struct proc_nfs_info {
592 int flag;
593 char *str;
594 char *nostr;
595 } nfs_info[] = {
596 { NFS_MOUNT_SOFT, ",soft", ",hard" },
597 { NFS_MOUNT_INTR, ",intr", "" },
1da177e4
LT		 598 { NFS_MOUNT_NOCTO, ",nocto", "" },
599 { NFS_MOUNT_NOAC, ",noac", "" },
c8bded96 600 { NFS_MOUNT_NONLM, ",nolock", "" },
b7fa0554 601 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 602 { 0, NULL, NULL }
603 };
604 struct proc_nfs_info *nfs_infop;
3063d8a1
TM		 605 char buf[12];
606 char *proto;
1da177e4 607
c8bded96 608 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
1da177e4
LT		 609 seq_printf(m, ",rsize=%d", nfss->rsize);
610 seq_printf(m, ",wsize=%d", nfss->wsize);
d9ef5a8c 611 if (nfss->acregmin != 3*HZ || showdefaults)
1da177e4 612 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
d9ef5a8c 613 if (nfss->acregmax != 60*HZ || showdefaults)
1da177e4 614 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
d9ef5a8c 615 if (nfss->acdirmin != 30*HZ || showdefaults)
1da177e4 616 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
d9ef5a8c 617 if (nfss->acdirmax != 60*HZ || showdefaults)
1da177e4
LT		 618 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
619 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
620 if (nfss->flags & nfs_infop->flag)
621 seq_puts(m, nfs_infop->str);
622 else
623 seq_puts(m, nfs_infop->nostr);
624 }
3063d8a1
TM		 625 switch (nfss->client->cl_xprt->prot) {
626 case IPPROTO_TCP:
627 proto = "tcp";
628 break;
629 case IPPROTO_UDP:
630 proto = "udp";
631 break;
632 default:
633 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
634 proto = buf;
635 }
636 seq_printf(m, ",proto=%s", proto);
7a480e25
CL		 637 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
638 seq_printf(m, ",retrans=%u", nfss->retrans_count);
d9ef5a8c
CL		 639}
640
641static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
642{
643 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
644
645 nfs_show_mount_options(m, nfss, 0);
646
1da177e4
LT		 647 seq_puts(m, ",addr=");
648 seq_escape(m, nfss->hostname, " \t\n\\");
d9ef5a8c
CL		 649
650 return 0;
651}
652
653static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
654{
655 int i, cpu;
656 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
657 struct rpc_auth *auth = nfss->client->cl_auth;
658 struct nfs_iostats totals = { };
659
660 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
661
662 /*
663 * Display all mount option settings
664 */
665 seq_printf(m, "\n\topts:\t");
666 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
667 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
668 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
669 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
670 nfs_show_mount_options(m, nfss, 1);
671
67ec9f46
CL		 672 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
673
d9ef5a8c
CL		 674 seq_printf(m, "\n\tcaps:\t");
675 seq_printf(m, "caps=0x%x", nfss->caps);
676 seq_printf(m, ",wtmult=%d", nfss->wtmult);
677 seq_printf(m, ",dtsize=%d", nfss->dtsize);
678 seq_printf(m, ",bsize=%d", nfss->bsize);
679 seq_printf(m, ",namelen=%d", nfss->namelen);
680
681#ifdef CONFIG_NFS_V4
682 if (nfss->rpc_ops->version == 4) {
683 seq_printf(m, "\n\tnfsv4:\t");
684 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
685 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
686 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
687 }
688#endif
689
690 /*
691 * Display security flavor in effect for this mount
692 */
693 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
694 if (auth->au_flavor)
695 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
696
697 /*
698 * Display superblock I/O counters
699 */
b9d9506d 700 for_each_possible_cpu(cpu) {
d9ef5a8c
CL		 701 struct nfs_iostats *stats;
702
d9ef5a8c
CL		 703 preempt_disable();
704 stats = per_cpu_ptr(nfss->io_stats, cpu);
705
706 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
707 totals.events[i] += stats->events[i];
708 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
709 totals.bytes[i] += stats->bytes[i];
710
711 preempt_enable();
712 }
713
714 seq_printf(m, "\n\tevents:\t");
715 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
716 seq_printf(m, "%lu ", totals.events[i]);
717 seq_printf(m, "\n\tbytes:\t");
718 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
719 seq_printf(m, "%Lu ", totals.bytes[i]);
4ece3a2d
CL		 720 seq_printf(m, "\n");
721
722 rpc_print_iostats(m, nfss->client);
d9ef5a8c 723
1da177e4
LT		 724 return 0;
725}
726
29884df0
TM		 727/**
728 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
729 */
730int nfs_sync_mapping(struct address_space *mapping)
731{
732 int ret;
733
734 if (mapping->nrpages == 0)
735 return 0;
736 unmap_mapping_range(mapping, 0, 0, 0);
28fd1298 737 ret = filemap_write_and_wait(mapping);
29884df0
TM		 738 if (ret != 0)
739 goto out;
740 ret = nfs_wb_all(mapping->host);
741out:
742 return ret;
743}
744
1da177e4
LT		 745/*
746 * Invalidate the local caches
747 */
b37b03b7 748static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT		 749{
750 struct nfs_inode *nfsi = NFS_I(inode);
751 int mode = inode->i_mode;
752
91d5b470
CL		 753 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
754
1da177e4
LT		 755 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
756 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
757
758 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
759 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 760 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 761 else
55296809 762 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 763}
dc59250c 764
b37b03b7
TM		 765void nfs_zap_caches(struct inode *inode)
766{
767 spin_lock(&inode->i_lock);
768 nfs_zap_caches_locked(inode);
dc59250c 769 spin_unlock(&inode->i_lock);
ada70d94
TM		 770}
771
772static void nfs_zap_acl_cache(struct inode *inode)
773{
774 void (*clear_acl_cache)(struct inode *);
775
776 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
777 if (clear_acl_cache != NULL)
778 clear_acl_cache(inode);
dc59250c 779 spin_lock(&inode->i_lock);
55296809 780 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 781 spin_unlock(&inode->i_lock);
1da177e4
LT		 782}
783
784/*
b37b03b7
TM		 785 * Invalidate, but do not unhash, the inode.
786 * NB: must be called with inode->i_lock held!
1da177e4 787 */
b37b03b7 788static void nfs_invalidate_inode(struct inode *inode)
1da177e4 789{
b37b03b7
TM		 790 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
791 nfs_zap_caches_locked(inode);
1da177e4
LT		 792}
793
794struct nfs_find_desc {
795 struct nfs_fh *fh;
796 struct nfs_fattr *fattr;
797};
798
799/*
800 * In NFSv3 we can have 64bit inode numbers. In order to support
801 * this, and re-exported directories (also seen in NFSv2)
802 * we are forced to allow 2 different inodes to have the same
803 * i_ino.
804 */
805static int
806nfs_find_actor(struct inode *inode, void *opaque)
807{
808 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
809 struct nfs_fh *fh = desc->fh;
810 struct nfs_fattr *fattr = desc->fattr;
811
812 if (NFS_FILEID(inode) != fattr->fileid)
813 return 0;
814 if (nfs_compare_fh(NFS_FH(inode), fh))
815 return 0;
816 if (is_bad_inode(inode) || NFS_STALE(inode))
817 return 0;
818 return 1;
819}
820
821static int
822nfs_init_locked(struct inode *inode, void *opaque)
823{
824 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
825 struct nfs_fattr *fattr = desc->fattr;
826
827 NFS_FILEID(inode) = fattr->fileid;
828 nfs_copy_fh(NFS_FH(inode), desc->fh);
829 return 0;
830}
831
832/* Don't use READDIRPLUS on directories that we believe are too large */
833#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
834
835/*
836 * This is our front-end to iget that looks up inodes by file handle
837 * instead of inode number.
838 */
839struct inode *
840nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
841{
842 struct nfs_find_desc desc = {
843 .fh = fh,
844 .fattr = fattr
845 };
03f28e3a 846 struct inode *inode = ERR_PTR(-ENOENT);
1da177e4
LT		 847 unsigned long hash;
848
849 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
850 goto out_no_inode;
851
852 if (!fattr->nlink) {
853 printk("NFS: Buggy server - nlink == 0!\n");
854 goto out_no_inode;
855 }
856
857 hash = nfs_fattr_to_ino_t(fattr);
858
03f28e3a
TM		 859 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
860 if (inode == NULL) {
861 inode = ERR_PTR(-ENOMEM);
1da177e4 862 goto out_no_inode;
03f28e3a 863 }
1da177e4
LT		 864
865 if (inode->i_state & I_NEW) {
866 struct nfs_inode *nfsi = NFS_I(inode);
867
868 /* We set i_ino for the few things that still rely on it,
869 * such as stat(2) */
870 inode->i_ino = hash;
871
872 /* We can't support update_atime(), since the server will reset it */
873 inode->i_flags |= S_NOATIME|S_NOCMTIME;
874 inode->i_mode = fattr->mode;
875 /* Why so? Because we want revalidate for devices/FIFOs, and
876 * that's precisely what we have in nfs_file_inode_operations.
877 */
92cfc62c 878 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 879 if (S_ISREG(inode->i_mode)) {
880 inode->i_fop = &nfs_file_operations;
881 inode->i_data.a_ops = &nfs_file_aops;
882 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
883 } else if (S_ISDIR(inode->i_mode)) {
884 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
885 inode->i_fop = &nfs_dir_operations;
886 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
887 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 888 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
55a97593
TM		 889 /* Deal with crossing mountpoints */
890 if (!nfs_fsid_equal(&NFS_SB(sb)->fsid, &fattr->fsid)) {
6b97fd3d
MN		 891 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)
892 inode->i_op = &nfs_referral_inode_operations;
893 else
894 inode->i_op = &nfs_mountpoint_inode_operations;
55a97593
TM		 895 inode->i_fop = NULL;
896 }
1da177e4
LT		 897 } else if (S_ISLNK(inode->i_mode))
898 inode->i_op = &nfs_symlink_inode_operations;
899 else
900 init_special_inode(inode, inode->i_mode, fattr->rdev);
901
33801147
TM		 902 nfsi->read_cache_jiffies = fattr->time_start;
903 nfsi->last_updated = jiffies;
1da177e4
LT		 904 inode->i_atime = fattr->atime;
905 inode->i_mtime = fattr->mtime;
906 inode->i_ctime = fattr->ctime;
907 if (fattr->valid & NFS_ATTR_FATTR_V4)
908 nfsi->change_attr = fattr->change_attr;
909 inode->i_size = nfs_size_to_loff_t(fattr->size);
910 inode->i_nlink = fattr->nlink;
911 inode->i_uid = fattr->uid;
912 inode->i_gid = fattr->gid;
913 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
914 /*
915 * report the blocks in 512byte units
916 */
917 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
918 inode->i_blksize = inode->i_sb->s_blocksize;
919 } else {
920 inode->i_blocks = fattr->du.nfs2.blocks;
921 inode->i_blksize = fattr->du.nfs2.blocksize;
922 }
923 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
924 nfsi->attrtimeo_timestamp = jiffies;
925 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
926 nfsi->cache_access.cred = NULL;
927
928 unlock_new_inode(inode);
929 } else
930 nfs_refresh_inode(inode, fattr);
931 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
932 inode->i_sb->s_id,
933 (long long)NFS_FILEID(inode),
934 atomic_read(&inode->i_count));
935
936out:
937 return inode;
938
939out_no_inode:
03f28e3a 940 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
1da177e4
LT		 941 goto out;
942}
943
944#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
945
946int
947nfs_setattr(struct dentry *dentry, struct iattr *attr)
948{
949 struct inode *inode = dentry->d_inode;
950 struct nfs_fattr fattr;
951 int error;
952
91d5b470
CL		 953 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
954
1da177e4
LT		 955 if (attr->ia_valid & ATTR_SIZE) {
956 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
957 attr->ia_valid &= ~ATTR_SIZE;
958 }
959
960 /* Optimization: if the end result is no change, don't RPC */
961 attr->ia_valid &= NFS_VALID_ATTRS;
962 if (attr->ia_valid == 0)
963 return 0;
964
965 lock_kernel();
966 nfs_begin_data_update(inode);
755c1e20
TM		 967 /* Write all dirty data */
968 filemap_write_and_wait(inode->i_mapping);
969 nfs_wb_all(inode);
642ac549
TM		 970 /*
971 * Return any delegations if we're going to change ACLs
972 */
973 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
974 nfs_inode_return_delegation(inode);
1da177e4 975 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 976 if (error == 0)
1da177e4 977 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 978 nfs_end_data_update(inode);
979 unlock_kernel();
980 return error;
981}
982
983/**
984 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
985 * @inode: pointer to struct inode
986 * @attr: pointer to struct iattr
987 *
988 * Note: we do this in the *proc.c in order to ensure that
989 * it works for things like exclusive creates too.
990 */
991void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
992{
993 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 994 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 995 int mode = attr->ia_mode & S_IALLUGO;
996 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 997 inode->i_mode = mode;
998 }
999 if ((attr->ia_valid & ATTR_UID) != 0)
1000 inode->i_uid = attr->ia_uid;
1001 if ((attr->ia_valid & ATTR_GID) != 0)
1002 inode->i_gid = attr->ia_gid;
dc59250c 1003 spin_lock(&inode->i_lock);
55296809 1004 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 1005 spin_unlock(&inode->i_lock);
65e4308d
TM		 1006 }
1007 if ((attr->ia_valid & ATTR_SIZE) != 0) {
91d5b470 1008 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
65e4308d
TM		 1009 inode->i_size = attr->ia_size;
1010 vmtruncate(inode, attr->ia_size);
1011 }
1da177e4
LT		 1012}
1013
412d582e
CL		 1014static int nfs_wait_schedule(void *word)
1015{
1016 if (signal_pending(current))
1017 return -ERESTARTSYS;
1018 schedule();
1019 return 0;
1020}
1021
1da177e4
LT		 1022/*
1023 * Wait for the inode to get unlocked.
1da177e4 1024 */
412d582e 1025static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 1026{
1027 struct rpc_clnt *clnt = NFS_CLIENT(inode);
1028 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 1029 sigset_t oldmask;
1da177e4 1030 int error;
412d582e 1031
412d582e
CL		 1032 rpc_clnt_sigmask(clnt, &oldmask);
1033 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
1034 nfs_wait_schedule, TASK_INTERRUPTIBLE);
1035 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 1036
1da177e4
LT		 1037 return error;
1038}
1039
412d582e
CL		 1040static void nfs_wake_up_inode(struct inode *inode)
1041{
1042 struct nfs_inode *nfsi = NFS_I(inode);
1043
1044 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
1045 smp_mb__after_clear_bit();
1046 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
1047}
1048
1da177e4
LT		 1049int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1050{
1051 struct inode *inode = dentry->d_inode;
55296809 1052 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 1053 int err;
1054
70b9ecbd 1055 /* Flush out writes to the server in order to update c/mtime */
c42de9dd 1056 nfs_sync_inode_wait(inode, 0, 0, FLUSH_NOCOMMIT);
fc33a7bb
CH		 1057
1058 /*
1059 * We may force a getattr if the user cares about atime.
1060 *
1061 * Note that we only have to check the vfsmount flags here:
1062 * - NFS always sets S_NOATIME by so checking it would give a
1063 * bogus result
1064 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1065 * no point in checking those.
1066 */
1067 if ((mnt->mnt_flags & MNT_NOATIME) ||
1068 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
1da177e4 1069 need_atime = 0;
fc33a7bb 1070
1da177e4
LT		 1071 if (need_atime)
1072 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
1073 else
1074 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
1075 if (!err)
1076 generic_fillattr(inode, stat);
1077 return err;
1078}
1079
b92dccf6 1080static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
1da177e4
LT		 1081{
1082 struct nfs_open_context *ctx;
1083
1084 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
1085 if (ctx != NULL) {
1086 atomic_set(&ctx->count, 1);
1087 ctx->dentry = dget(dentry);
b92dccf6 1088 ctx->vfsmnt = mntget(mnt);
1da177e4
LT		 1089 ctx->cred = get_rpccred(cred);
1090 ctx->state = NULL;
1091 ctx->lockowner = current->files;
1092 ctx->error = 0;
00a92642 1093 ctx->dir_cookie = 0;
1da177e4
LT		 1094 }
1095 return ctx;
1096}
1097
1098struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
1099{
1100 if (ctx != NULL)
1101 atomic_inc(&ctx->count);
1102 return ctx;
1103}
1104
1105void put_nfs_open_context(struct nfs_open_context *ctx)
1106{
1107 if (atomic_dec_and_test(&ctx->count)) {
1108 if (!list_empty(&ctx->list)) {
1109 struct inode *inode = ctx->dentry->d_inode;
1110 spin_lock(&inode->i_lock);
1111 list_del(&ctx->list);
1112 spin_unlock(&inode->i_lock);
1113 }
1114 if (ctx->state != NULL)
1115 nfs4_close_state(ctx->state, ctx->mode);
1116 if (ctx->cred != NULL)
1117 put_rpccred(ctx->cred);
1118 dput(ctx->dentry);
b92dccf6 1119 mntput(ctx->vfsmnt);
1da177e4
LT		 1120 kfree(ctx);
1121 }
1122}
1123
1124/*
1125 * Ensure that mmap has a recent RPC credential for use when writing out
1126 * shared pages
1127 */
b92dccf6 1128static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1da177e4
LT		 1129{
1130 struct inode *inode = filp->f_dentry->d_inode;
1131 struct nfs_inode *nfsi = NFS_I(inode);
1132
1133 filp->private_data = get_nfs_open_context(ctx);
1134 spin_lock(&inode->i_lock);
1135 list_add(&ctx->list, &nfsi->open_files);
1136 spin_unlock(&inode->i_lock);
1137}
1138
d530838b
TM		 1139/*
1140 * Given an inode, search for an open context with the desired characteristics
1141 */
1142struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1143{
1144 struct nfs_inode *nfsi = NFS_I(inode);
1145 struct nfs_open_context *pos, *ctx = NULL;
1146
1147 spin_lock(&inode->i_lock);
1148 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1149 if (cred != NULL && pos->cred != cred)
1150 continue;
1da177e4
LT		 1151 if ((pos->mode & mode) == mode) {
1152 ctx = get_nfs_open_context(pos);
1153 break;
1154 }
1155 }
1156 spin_unlock(&inode->i_lock);
1157 return ctx;
1158}
1159
b92dccf6 1160static void nfs_file_clear_open_context(struct file *filp)
1da177e4
LT		 1161{
1162 struct inode *inode = filp->f_dentry->d_inode;
1163 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1164
1165 if (ctx) {
1166 filp->private_data = NULL;
1167 spin_lock(&inode->i_lock);
1168 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1169 spin_unlock(&inode->i_lock);
1170 put_nfs_open_context(ctx);
1171 }
1172}
1173
1174/*
1175 * These allocate and release file read/write context information.
1176 */
1177int nfs_open(struct inode *inode, struct file *filp)
1178{
1179 struct nfs_open_context *ctx;
1180 struct rpc_cred *cred;
1181
1182 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1183 if (IS_ERR(cred))
1184 return PTR_ERR(cred);
b92dccf6 1185 ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
1da177e4
LT		 1186 put_rpccred(cred);
1187 if (ctx == NULL)
1188 return -ENOMEM;
1189 ctx->mode = filp->f_mode;
1190 nfs_file_set_open_context(filp, ctx);
1191 put_nfs_open_context(ctx);
1da177e4
LT		 1192 return 0;
1193}
1194
1195int nfs_release(struct inode *inode, struct file *filp)
1196{
1da177e4
LT		 1197 nfs_file_clear_open_context(filp);
1198 return 0;
1199}
1200
1201/*
1202 * This function is called whenever some part of NFS notices that
1203 * the cached attributes have to be refreshed.
1204 */
1205int
1206__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1207{
1208 int status = -ESTALE;
1209 struct nfs_fattr fattr;
1210 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4
LT		 1211
1212 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1213 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1214
1842bfb4 1215 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
1da177e4
LT		 1216 lock_kernel();
1217 if (!inode || is_bad_inode(inode))
1218 goto out_nowait;
1219 if (NFS_STALE(inode))
1220 goto out_nowait;
1221
412d582e
CL		 1222 status = nfs_wait_on_inode(inode);
1223 if (status < 0)
1224 goto out;
1225 if (NFS_STALE(inode)) {
1226 status = -ESTALE;
1227 /* Do we trust the cached ESTALE? */
1228 if (NFS_ATTRTIMEO(inode) != 0) {
44b11874 1229 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME)) {
412d582e
CL		 1230 /* no */
1231 } else
1232 goto out;
1233 }
1da177e4 1234 }
1da177e4 1235
1da177e4
LT		 1236 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1237 if (status != 0) {
1238 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1239 inode->i_sb->s_id,
1240 (long long)NFS_FILEID(inode), status);
1241 if (status == -ESTALE) {
1242 nfs_zap_caches(inode);
1243 if (!S_ISDIR(inode->i_mode))
412d582e 1244 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1245 }
1246 goto out;
1247 }
1248
33801147 1249 spin_lock(&inode->i_lock);
24aa1fe6 1250 status = nfs_update_inode(inode, &fattr);
1da177e4 1251 if (status) {
33801147 1252 spin_unlock(&inode->i_lock);
1da177e4
LT		 1253 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1254 inode->i_sb->s_id,
1255 (long long)NFS_FILEID(inode), status);
1256 goto out;
1257 }
dc59250c 1258 spin_unlock(&inode->i_lock);
55296809 1259
24aa1fe6 1260 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1261 nfs_zap_acl_cache(inode);
55296809 1262
1da177e4
LT		 1263 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1264 inode->i_sb->s_id,
1265 (long long)NFS_FILEID(inode));
1266
412d582e
CL		 1267 out:
1268 nfs_wake_up_inode(inode);
1269
1da177e4
LT		 1270 out_nowait:
1271 unlock_kernel();
1272 return status;
1273}
1274
1275int nfs_attribute_timeout(struct inode *inode)
1276{
1277 struct nfs_inode *nfsi = NFS_I(inode);
1278
1279 if (nfs_have_delegation(inode, FMODE_READ))
1280 return 0;
1281 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1282}
1283
1284/**
1285 * nfs_revalidate_inode - Revalidate the inode attributes
1286 * @server - pointer to nfs_server struct
1287 * @inode - pointer to inode struct
1288 *
1289 * Updates inode attribute information by retrieving the data from the server.
1290 */
1291int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1292{
44b11874 1293 if (!(NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATTR)
1da177e4
LT		 1294 && !nfs_attribute_timeout(inode))
1295 return NFS_STALE(inode) ? -ESTALE : 0;
1296 return __nfs_revalidate_inode(server, inode);
1297}
1298
7d52e862
TM		 1299/**
1300 * nfs_revalidate_mapping - Revalidate the pagecache
1301 * @inode - pointer to host inode
1302 * @mapping - pointer to mapping
1303 */
44b11874 1304int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
7d52e862
TM		 1305{
1306 struct nfs_inode *nfsi = NFS_I(inode);
44b11874
TM		 1307 int ret = 0;
1308
1309 if (NFS_STALE(inode))
1310 ret = -ESTALE;
1311 if ((nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
1312 || nfs_attribute_timeout(inode))
1313 ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
7d52e862 1314
55296809 1315 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
91d5b470 1316 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
29884df0
TM		 1317 if (S_ISREG(inode->i_mode))
1318 nfs_sync_mapping(mapping);
7d52e862 1319 invalidate_inode_pages2(mapping);
dc59250c
CL		 1320
1321 spin_lock(&inode->i_lock);
55296809 1322 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1323 if (S_ISDIR(inode->i_mode)) {
1324 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1325 /* This ensures we revalidate child dentries */
913a70fc 1326 nfsi->cache_change_attribute = jiffies;
7d52e862 1327 }
dc59250c
CL		 1328 spin_unlock(&inode->i_lock);
1329
7d52e862
TM		 1330 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1331 inode->i_sb->s_id,
1332 (long long)NFS_FILEID(inode));
1333 }
44b11874 1334 return ret;
7d52e862
TM		 1335}
1336
1da177e4
LT		 1337/**
1338 * nfs_begin_data_update
1339 * @inode - pointer to inode
1340 * Declare that a set of operations will update file data on the server
1341 */
1342void nfs_begin_data_update(struct inode *inode)
1343{
1344 atomic_inc(&NFS_I(inode)->data_updates);
1345}
1346
1347/**
1348 * nfs_end_data_update
1349 * @inode - pointer to inode
1350 * Declare end of the operations that will update file data
1351 * This will mark the inode as immediately needing revalidation
1352 * of its attribute cache.
1353 */
1354void nfs_end_data_update(struct inode *inode)
1355{
1356 struct nfs_inode *nfsi = NFS_I(inode);
1357
1358 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1359 /* Directories and symlinks: invalidate page cache */
1360 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1361 spin_lock(&inode->i_lock);
55296809 1362 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1363 spin_unlock(&inode->i_lock);
1364 }
1da177e4 1365 }
913a70fc 1366 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1367 atomic_dec(&nfsi->data_updates);
1368}
1369
a895b4a1
TM		 1370static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1371{
1372 struct nfs_inode *nfsi = NFS_I(inode);
1373
a895b4a1
TM		 1374 /* If we have atomic WCC data, we may update some attributes */
1375 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1376 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
1377 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1378 nfsi->cache_change_attribute = jiffies;
1379 }
1380 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
1381 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1382 nfsi->cache_change_attribute = jiffies;
1383 }
1384 if (inode->i_size == fattr->pre_size && nfsi->npages == 0) {
1385 inode->i_size = fattr->size;
1386 nfsi->cache_change_attribute = jiffies;
1387 }
1388 }
1389}
1390
1da177e4 1391/**
33801147 1392 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1393 * @inode - pointer to inode
1394 * @fattr - updated attributes
1395 *
1396 * Verifies the attribute cache. If we have just changed the attributes,
1397 * so that fattr carries weak cache consistency data, then it may
1398 * also update the ctime/mtime/change_attribute.
1399 */
33801147 1400static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1401{
1402 struct nfs_inode *nfsi = NFS_I(inode);
1403 loff_t cur_size, new_isize;
1404 int data_unstable;
1405
dc59250c 1406
ca62b9c3
TM		 1407 /* Has the inode gone and changed behind our back? */
1408 if (nfsi->fileid != fattr->fileid
1409 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1410 return -EIO;
1411 }
1412
1da177e4
LT		 1413 /* Are we in the process of updating data on the server? */
1414 data_unstable = nfs_caches_unstable(inode);
1415
a895b4a1
TM		 1416 /* Do atomic weak cache consistency updates */
1417 nfs_wcc_update_inode(inode, fattr);
1da177e4 1418
9d1e9232 1419 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 &&
f1bb0b92
TM		 1420 nfsi->change_attr != fattr->change_attr)
1421 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4 1422
1da177e4 1423 /* Verify a few of the more important attributes */
f1bb0b92
TM		 1424 if (!timespec_equal(&inode->i_mtime, &fattr->mtime))
1425 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
ca62b9c3
TM		 1426
1427 cur_size = i_size_read(inode);
1428 new_isize = nfs_size_to_loff_t(fattr->size);
fb374d24
TM		 1429 if (cur_size != new_isize && nfsi->npages == 0)
1430 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4
LT		 1431
1432 /* Have any file permissions changed? */
1433 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1434 || inode->i_uid != fattr->uid
1435 || inode->i_gid != fattr->gid)
55296809 1436 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1437
1438 /* Has the link count changed? */
1439 if (inode->i_nlink != fattr->nlink)
55296809 1440 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1441
1442 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1443 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1444
33801147 1445 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1446 return 0;
1447}
1448
33801147
TM		 1449/**
1450 * nfs_refresh_inode - try to update the inode attribute cache
1451 * @inode - pointer to inode
1452 * @fattr - updated attributes
1453 *
1454 * Check that an RPC call that returned attributes has not overlapped with
1455 * other recent updates of the inode metadata, then decide whether it is
1456 * safe to do a full update of the inode attributes, or whether just to
1457 * call nfs_check_inode_attributes.
1458 */
1459int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1460{
1461 struct nfs_inode *nfsi = NFS_I(inode);
1462 int status;
1463
1464 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1465 return 0;
1466 spin_lock(&inode->i_lock);
33801147 1467 if (time_after(fattr->time_start, nfsi->last_updated))
24aa1fe6 1468 status = nfs_update_inode(inode, fattr);
33801147
TM		 1469 else
1470 status = nfs_check_inode_attributes(inode, fattr);
1471
1472 spin_unlock(&inode->i_lock);
1473 return status;
1474}
1475
decf491f
TM		 1476/**
1477 * nfs_post_op_update_inode - try to update the inode attribute cache
1478 * @inode - pointer to inode
1479 * @fattr - updated attributes
1480 *
1481 * After an operation that has changed the inode metadata, mark the
1482 * attribute cache as being invalid, then try to update it.
1483 */
1484int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1485{
1486 struct nfs_inode *nfsi = NFS_I(inode);
1487 int status = 0;
1488
1489 spin_lock(&inode->i_lock);
1490 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
f1bb0b92 1491 nfsi->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
decf491f
TM		 1492 goto out;
1493 }
24aa1fe6 1494 status = nfs_update_inode(inode, fattr);
decf491f
TM		 1495out:
1496 spin_unlock(&inode->i_lock);
1497 return status;
1498}
1499
1da177e4
LT		 1500/*
1501 * Many nfs protocol calls return the new file attributes after
1502 * an operation. Here we update the inode to reflect the state
1503 * of the server's inode.
1504 *
1505 * This is a bit tricky because we have to make sure all dirty pages
1506 * have been sent off to the server before calling invalidate_inode_pages.
1507 * To make sure no other process adds more write requests while we try
1508 * our best to flush them, we make them sleep during the attribute refresh.
1509 *
1510 * A very similar scenario holds for the dir cache.
1511 */
24aa1fe6 1512static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1da177e4 1513{
8b4bdcf8 1514 struct nfs_server *server;
1da177e4 1515 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1516 loff_t cur_isize, new_isize;
1da177e4 1517 unsigned int invalid = 0;
24aa1fe6 1518 int data_stable;
1da177e4
LT		 1519
1520 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1521 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1522 atomic_read(&inode->i_count), fattr->valid);
1523
325cfed9
CL		 1524 if (nfsi->fileid != fattr->fileid)
1525 goto out_fileid;
1da177e4
LT		 1526
1527 /*
1528 * Make sure the inode's type hasn't changed.
1529 */
33801147 1530 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT		 1531 goto out_changed;
1532
8b4bdcf8
TM		 1533 server = NFS_SERVER(inode);
1534 /* Update the fsid if and only if this is the root directory */
1535 if (inode == inode->i_sb->s_root->d_inode
1536 && !nfs_fsid_equal(&server->fsid, &fattr->fsid))
1537 server->fsid = fattr->fsid;
1538
1da177e4
LT		 1539 /*
1540 * Update the read time so we don't revalidate too often.
1541 */
33801147
TM		 1542 nfsi->read_cache_jiffies = fattr->time_start;
1543 nfsi->last_updated = jiffies;
1da177e4
LT		 1544
1545 /* Are we racing with known updates of the metadata on the server? */
24aa1fe6
TM		 1546 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1547 if (data_stable)
f1bb0b92 1548 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATIME);
1da177e4 1549
a895b4a1
TM		 1550 /* Do atomic weak cache consistency updates */
1551 nfs_wcc_update_inode(inode, fattr);
1552
951a143b 1553 /* Check if our cached file size is stale */
1da177e4
LT		 1554 new_isize = nfs_size_to_loff_t(fattr->size);
1555 cur_isize = i_size_read(inode);
951a143b
TM		 1556 if (new_isize != cur_isize) {
1557 /* Do we perhaps have any outstanding writes? */
1558 if (nfsi->npages == 0) {
1559 /* No, but did we race with nfs_end_data_update()? */
24aa1fe6 1560 if (data_stable) {
1da177e4 1561 inode->i_size = new_isize;
951a143b 1562 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1563 }
951a143b
TM		 1564 invalid |= NFS_INO_INVALID_ATTR;
1565 } else if (new_isize > cur_isize) {
1da177e4
LT		 1566 inode->i_size = new_isize;
1567 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1568 }
24aa1fe6 1569 nfsi->cache_change_attribute = jiffies;
951a143b
TM		 1570 dprintk("NFS: isize change on server for file %s/%ld\n",
1571 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1572 }
1573
951a143b 1574 /* Check if the mtime agrees */
1da177e4
LT		 1575 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1576 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1577 dprintk("NFS: mtime change on server for file %s/%ld\n",
1578 inode->i_sb->s_id, inode->i_ino);
24aa1fe6
TM		 1579 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1580 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1581 }
1582
ada70d94
TM		 1583 /* If ctime has changed we should definitely clear access+acl caches */
1584 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
24aa1fe6 1585 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
ada70d94 1586 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
24aa1fe6 1587 nfsi->cache_change_attribute = jiffies;
ada70d94 1588 }
1da177e4
LT		 1589 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1590
1591 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1592 inode->i_uid != fattr->uid ||
1593 inode->i_gid != fattr->gid)
ada70d94 1594 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1595
1596 inode->i_mode = fattr->mode;
1597 inode->i_nlink = fattr->nlink;
1598 inode->i_uid = fattr->uid;
1599 inode->i_gid = fattr->gid;
1600
1601 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1602 /*
1603 * report the blocks in 512byte units
1604 */
1605 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1606 inode->i_blksize = inode->i_sb->s_blocksize;
1607 } else {
1608 inode->i_blocks = fattr->du.nfs2.blocks;
1609 inode->i_blksize = fattr->du.nfs2.blocksize;
1610 }
1611
9d1e9232
TM		 1612 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 &&
1613 nfsi->change_attr != fattr->change_attr) {
1614 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1615 inode->i_sb->s_id, inode->i_ino);
1616 nfsi->change_attr = fattr->change_attr;
1617 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1618 nfsi->cache_change_attribute = jiffies;
ca62b9c3
TM		 1619 }
1620
1da177e4
LT		 1621 /* Update attrtimeo value if we're out of the unstable period */
1622 if (invalid & NFS_INO_INVALID_ATTR) {
91d5b470 1623 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
1da177e4
LT		 1624 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1625 nfsi->attrtimeo_timestamp = jiffies;
1626 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1627 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1628 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1629 nfsi->attrtimeo_timestamp = jiffies;
1630 }
1631 /* Don't invalidate the data if we were to blame */
1632 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1633 || S_ISLNK(inode->i_mode)))
1634 invalid &= ~NFS_INO_INVALID_DATA;
24aa1fe6
TM		 1635 if (data_stable)
1636 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1da177e4 1637 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1638 nfsi->cache_validity |= invalid;
1da177e4
LT		 1639
1640 return 0;
1641 out_changed:
1642 /*
1643 * Big trouble! The inode has become a different object.
1644 */
1645#ifdef NFS_PARANOIA
1646 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1647 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1648#endif
b37b03b7 1649 out_err:
1da177e4
LT		 1650 /*
1651 * No need to worry about unhashing the dentry, as the
1652 * lookup validation will know that the inode is bad.
1653 * (But we fall through to invalidate the caches.)
1654 */
1655 nfs_invalidate_inode(inode);
1da177e4 1656 return -ESTALE;
325cfed9
CL		 1657
1658 out_fileid:
1659 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1660 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1661 NFS_SERVER(inode)->hostname, inode->i_sb->s_id,
1662 (long long)nfsi->fileid, (long long)fattr->fileid);
1663 goto out_err;
1da177e4
LT		 1664}
1665
1666/*
1667 * File system information
1668 */
1669
55a97593
TM		 1670/*
1671 * nfs_path - reconstruct the path given an arbitrary dentry
1672 * @base - arbitrary string to prepend to the path
1673 * @dentry - pointer to dentry
1674 * @buffer - result buffer
1675 * @buflen - length of buffer
1676 *
1677 * Helper function for constructing the path from the
1678 * root dentry to an arbitrary hashed dentry.
1679 *
1680 * This is mainly for use in figuring out the path on the
1681 * server side when automounting on top of an existing partition.
1682 */
1683static char *nfs_path(const char *base, const struct dentry *dentry,
1684 char *buffer, ssize_t buflen)
1685{
1686 char *end = buffer+buflen;
1687 int namelen;
1688
1689 *--end = '\0';
1690 buflen--;
1691 spin_lock(&dcache_lock);
1692 while (!IS_ROOT(dentry)) {
1693 namelen = dentry->d_name.len;
1694 buflen -= namelen + 1;
1695 if (buflen < 0)
1696 goto Elong;
1697 end -= namelen;
1698 memcpy(end, dentry->d_name.name, namelen);
1699 *--end = '/';
1700 dentry = dentry->d_parent;
1701 }
1702 spin_unlock(&dcache_lock);
1703 namelen = strlen(base);
1704 /* Strip off excess slashes in base string */
1705 while (namelen > 0 && base[namelen - 1] == '/')
1706 namelen--;
1707 buflen -= namelen;
1708 if (buflen < 0)
1709 goto Elong;
1710 end -= namelen;
1711 memcpy(end, base, namelen);
1712 return end;
1713Elong:
1714 return ERR_PTR(-ENAMETOOLONG);
1715}
1716
1717struct nfs_clone_mount {
1718 const struct super_block *sb;
1719 const struct dentry *dentry;
1720 struct nfs_fh *fh;
1721 struct nfs_fattr *fattr;
9cdb3883
MN		 1722 char *hostname;
1723 char *mnt_path;
1724 struct sockaddr_in *addr;
1725 rpc_authflavor_t authflavor;
55a97593
TM		 1726};
1727
1728static struct super_block *nfs_clone_generic_sb(struct nfs_clone_mount *data,
61f5164c
MN		 1729 struct super_block *(*fill_sb)(struct nfs_server *, struct nfs_clone_mount *),
1730 struct nfs_server *(*fill_server)(struct super_block *, struct nfs_clone_mount *))
55a97593
TM		 1731{
1732 struct nfs_server *server;
1733 struct nfs_server *parent = NFS_SB(data->sb);
1734 struct super_block *sb = ERR_PTR(-EINVAL);
1735 void *err = ERR_PTR(-ENOMEM);
9cdb3883 1736 char *hostname;
55a97593
TM		 1737 int len;
1738
1739 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1740 if (server == NULL)
1741 goto out_err;
1742 memcpy(server, parent, sizeof(*server));
9cdb3883
MN		 1743 hostname = (data->hostname != NULL) ? data->hostname : parent->hostname;
1744 len = strlen(hostname) + 1;
55a97593
TM		 1745 server->hostname = kmalloc(len, GFP_KERNEL);
1746 if (server->hostname == NULL)
1747 goto free_server;
9cdb3883 1748 memcpy(server->hostname, hostname, len);
55a97593
TM		 1749 if (rpciod_up() != 0)
1750 goto free_hostname;
1751
61f5164c 1752 sb = fill_sb(server, data);
55a97593
TM		 1753 if (IS_ERR((err = sb)) || sb->s_root)
1754 goto kill_rpciod;
1755
61f5164c
MN		 1756 server = fill_server(sb, data);
1757 if (IS_ERR((err = server)))
55a97593 1758 goto out_deactivate;
55a97593 1759 return sb;
55a97593
TM		 1760out_deactivate:
1761 up_write(&sb->s_umount);
1762 deactivate_super(sb);
1763 return (struct super_block *)err;
1764kill_rpciod:
1765 rpciod_down();
1766free_hostname:
1767 kfree(server->hostname);
1768free_server:
1769 kfree(server);
1770out_err:
1771 return (struct super_block *)err;
1772}
1773
1da177e4
LT		 1774static int nfs_set_super(struct super_block *s, void *data)
1775{
1776 s->s_fs_info = data;
1777 return set_anon_super(s, data);
1778}
1779
1780static int nfs_compare_super(struct super_block *sb, void *data)
1781{
1782 struct nfs_server *server = data;
1783 struct nfs_server *old = NFS_SB(sb);
1784
1785 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1786 return 0;
1787 if (old->addr.sin_port != server->addr.sin_port)
1788 return 0;
1789 return !nfs_compare_fh(&old->fh, &server->fh);
1790}
1791
1792static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1793 int flags, const char *dev_name, void *raw_data)
1794{
1795 int error;
9085bbcb 1796 struct nfs_server *server = NULL;
1da177e4
LT		 1797 struct super_block *s;
1798 struct nfs_fh *root;
1799 struct nfs_mount_data *data = raw_data;
1800
9085bbcb
TM		 1801 s = ERR_PTR(-EINVAL);
1802 if (data == NULL) {
1803 dprintk("%s: missing data argument\n", __FUNCTION__);
1804 goto out_err;
1805 }
1806 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1807 dprintk("%s: bad mount version\n", __FUNCTION__);
1808 goto out_err;
1da177e4 1809 }
9085bbcb
TM		 1810 switch (data->version) {
1811 case 1:
1812 data->namlen = 0;
1813 case 2:
1814 data->bsize = 0;
1815 case 3:
1816 if (data->flags & NFS_MOUNT_VER3) {
1817 dprintk("%s: mount structure version %d does not support NFSv3\n",
1818 __FUNCTION__,
1819 data->version);
1820 goto out_err;
1821 }
1822 data->root.size = NFS2_FHSIZE;
1823 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1824 case 4:
1825 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1826 dprintk("%s: mount structure version %d does not support strong security\n",
1827 __FUNCTION__,
1828 data->version);
1829 goto out_err;
1830 }
1831 case 5:
1832 memset(data->context, 0, sizeof(data->context));
1833 }
1834#ifndef CONFIG_NFS_V3
1835 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1836 s = ERR_PTR(-EPROTONOSUPPORT);
1837 if (data->flags & NFS_MOUNT_VER3) {
1838 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1839 goto out_err;
1840 }
1841#endif /* CONFIG_NFS_V3 */
1da177e4 1842
9085bbcb 1843 s = ERR_PTR(-ENOMEM);
bd647545 1844 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4 1845 if (!server)
9085bbcb 1846 goto out_err;
1da177e4
LT		 1847 /* Zero out the NFS state stuff */
1848 init_nfsv4_state(server);
b7fa0554 1849 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1850
1da177e4
LT		 1851 root = &server->fh;
1852 if (data->flags & NFS_MOUNT_VER3)
1853 root->size = data->root.size;
1854 else
1855 root->size = NFS2_FHSIZE;
9085bbcb 1856 s = ERR_PTR(-EINVAL);
1da177e4 1857 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1858 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1859 goto out_err;
1da177e4
LT		 1860 }
1861 memcpy(root->data, data->root.data, root->size);
1862
1863 /* We now require that the mount process passes the remote address */
1864 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1865 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1866 dprintk("%s: mount program didn't pass remote address!\n",
1867 __FUNCTION__);
1868 goto out_err;
1da177e4
LT		 1869 }
1870
9085bbcb
TM		 1871 /* Fire up rpciod if not yet running */
1872 s = ERR_PTR(rpciod_up());
1873 if (IS_ERR(s)) {
1874 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1875 __FUNCTION__, PTR_ERR(s));
1876 goto out_err;
1da177e4
LT		 1877 }
1878
9085bbcb
TM		 1879 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1880 if (IS_ERR(s) || s->s_root)
1881 goto out_rpciod_down;
1da177e4 1882
9085bbcb 1883 s->s_flags = flags;
1da177e4 1884
9b04c997 1885 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT		 1886 if (error) {
1887 up_write(&s->s_umount);
1888 deactivate_super(s);
1889 return ERR_PTR(error);
1890 }
1891 s->s_flags |= MS_ACTIVE;
1892 return s;
9085bbcb
TM		 1893out_rpciod_down:
1894 rpciod_down();
1895out_err:
1896 kfree(server);
1897 return s;
1da177e4
LT		 1898}
1899
1900static void nfs_kill_super(struct super_block *s)
1901{
1902 struct nfs_server *server = NFS_SB(s);
1903
1904 kill_anon_super(s);
1905
6a19275a 1906 if (!IS_ERR(server->client))
1da177e4 1907 rpc_shutdown_client(server->client);
6a19275a 1908 if (!IS_ERR(server->client_sys))
1da177e4 1909 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1910 if (!IS_ERR(server->client_acl))
1911 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1912
1913 if (!(server->flags & NFS_MOUNT_NONLM))
1914 lockd_down(); /* release rpc.lockd */
1915
1916 rpciod_down(); /* release rpciod */
1917
01d0ae8b 1918 nfs_free_iostats(server->io_stats);
f99d49ad 1919 kfree(server->hostname);
1da177e4 1920 kfree(server);
51d8fa6a 1921 nfs_release_automount_timer();
1da177e4
LT		 1922}
1923
1924static struct file_system_type nfs_fs_type = {
1925 .owner = THIS_MODULE,
1926 .name = "nfs",
1927 .get_sb = nfs_get_sb,
1928 .kill_sb = nfs_kill_super,
1929 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1930};
1931
61f5164c 1932static struct super_block *nfs_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
55a97593
TM		 1933{
1934 struct super_block *sb;
1935
61f5164c
MN		 1936 server->fsid = data->fattr->fsid;
1937 nfs_copy_fh(&server->fh, data->fh);
55a97593
TM		 1938 sb = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server);
1939 if (!IS_ERR(sb) && sb->s_root == NULL && !(server->flags & NFS_MOUNT_NONLM))
1940 lockd_up();
1941 return sb;
1942}
1943
61f5164c
MN		 1944static struct nfs_server *nfs_clone_server(struct super_block *sb, struct nfs_clone_mount *data)
1945{
1946 struct nfs_server *server = NFS_SB(sb);
1947 struct nfs_server *parent = NFS_SB(data->sb);
1948 struct inode *root_inode;
1949 struct nfs_fsinfo fsinfo;
1950 void *err = ERR_PTR(-ENOMEM);
1951
1952 sb->s_op = data->sb->s_op;
1953 sb->s_blocksize = data->sb->s_blocksize;
1954 sb->s_blocksize_bits = data->sb->s_blocksize_bits;
1955 sb->s_maxbytes = data->sb->s_maxbytes;
1956
1957 server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1958 server->io_stats = nfs_alloc_iostats();
1959 if (server->io_stats == NULL)
1960 goto out;
1961
1962 server->client = rpc_clone_client(parent->client);
1963 if (IS_ERR((err = server->client)))
1964 goto out;
1965
1966 if (!IS_ERR(parent->client_sys)) {
1967 server->client_sys = rpc_clone_client(parent->client_sys);
1968 if (IS_ERR((err = server->client_sys)))
1969 goto out;
1970 }
1971 if (!IS_ERR(parent->client_acl)) {
1972 server->client_acl = rpc_clone_client(parent->client_acl);
1973 if (IS_ERR((err = server->client_acl)))
1974 goto out;
1975 }
1976 root_inode = nfs_fhget(sb, data->fh, data->fattr);
1977 if (!root_inode)
1978 goto out;
1979 sb->s_root = d_alloc_root(root_inode);
1980 if (!sb->s_root)
1981 goto out_put_root;
1982 fsinfo.fattr = data->fattr;
1983 if (NFS_PROTO(root_inode)->fsinfo(server, data->fh, &fsinfo) == 0)
1984 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
1985 sb->s_root->d_op = server->rpc_ops->dentry_ops;
1986 sb->s_flags |= MS_ACTIVE;
1987 return server;
1988out_put_root:
1989 iput(root_inode);
1990out:
1991 return err;
1992}
1993
55a97593
TM		 1994static struct super_block *nfs_clone_nfs_sb(struct file_system_type *fs_type,
1995 int flags, const char *dev_name, void *raw_data)
1996{
1997 struct nfs_clone_mount *data = raw_data;
61f5164c 1998 return nfs_clone_generic_sb(data, nfs_clone_sb, nfs_clone_server);
55a97593
TM		 1999}
2000
2001static struct file_system_type clone_nfs_fs_type = {
2002 .owner = THIS_MODULE,
2003 .name = "nfs",
2004 .get_sb = nfs_clone_nfs_sb,
2005 .kill_sb = nfs_kill_super,
2006 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2007};
2008
1da177e4
LT		 2009#ifdef CONFIG_NFS_V4
2010
2011static void nfs4_clear_inode(struct inode *);
2012
2013
2014static struct super_operations nfs4_sops = {
2015 .alloc_inode = nfs_alloc_inode,
2016 .destroy_inode = nfs_destroy_inode,
2017 .write_inode = nfs_write_inode,
1da177e4
LT		 2018 .statfs = nfs_statfs,
2019 .clear_inode = nfs4_clear_inode,
2020 .umount_begin = nfs_umount_begin,
2021 .show_options = nfs_show_options,
d9ef5a8c 2022 .show_stats = nfs_show_stats,
1da177e4
LT		 2023};
2024
2025/*
2026 * Clean out any remaining NFSv4 state that might be left over due
2027 * to open() calls that passed nfs_atomic_lookup, but failed to call
2028 * nfs_open().
2029 */
2030static void nfs4_clear_inode(struct inode *inode)
2031{
2032 struct nfs_inode *nfsi = NFS_I(inode);
2033
2034 /* If we are holding a delegation, return it! */
cae7a073 2035 nfs_inode_return_delegation(inode);
1da177e4
LT		 2036 /* First call standard NFS clear_inode() code */
2037 nfs_clear_inode(inode);
2038 /* Now clear out any remaining state */
2039 while (!list_empty(&nfsi->open_states)) {
2040 struct nfs4_state *state;
2041
2042 state = list_entry(nfsi->open_states.next,
2043 struct nfs4_state,
2044 inode_states);
2045 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
2046 __FUNCTION__,
2047 inode->i_sb->s_id,
2048 (long long)NFS_FILEID(inode),
2049 state);
2050 BUG_ON(atomic_read(&state->count) != 1);
2051 nfs4_close_state(state, state->state);
2052 }
2053}
2054
2055
c818ba43
MN		 2056static struct rpc_clnt *nfs4_create_client(struct nfs_server *server,
2057 struct rpc_timeout *timeparms, int proto, rpc_authflavor_t flavor)
1da177e4 2058{
c818ba43 2059 struct nfs4_client *clp;
1da177e4
LT		 2060 struct rpc_xprt *xprt = NULL;
2061 struct rpc_clnt *clnt = NULL;
eab5c084 2062 int err = -EIO;
1da177e4 2063
1da177e4
LT		 2064 clp = nfs4_get_client(&server->addr.sin_addr);
2065 if (!clp) {
9085bbcb 2066 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
c818ba43 2067 return ERR_PTR(err);
1da177e4
LT		 2068 }
2069
2070 /* Now create transport and client */
1da177e4 2071 down_write(&clp->cl_sem);
6a19275a 2072 if (IS_ERR(clp->cl_rpcclient)) {
c818ba43 2073 xprt = xprt_create_proto(proto, &server->addr, timeparms);
1da177e4
LT		 2074 if (IS_ERR(xprt)) {
2075 up_write(&clp->cl_sem);
1da177e4 2076 err = PTR_ERR(xprt);
9085bbcb
TM		 2077 dprintk("%s: cannot create RPC transport. Error = %d\n",
2078 __FUNCTION__, err);
1da177e4
LT		 2079 goto out_fail;
2080 }
87e4ba1a
TM		 2081 /* Bind to a reserved port! */
2082 xprt->resvport = 1;
1da177e4 2083 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
c818ba43 2084 server->rpc_ops->version, flavor);
1da177e4
LT		 2085 if (IS_ERR(clnt)) {
2086 up_write(&clp->cl_sem);
1da177e4 2087 err = PTR_ERR(clnt);
9085bbcb
TM		 2088 dprintk("%s: cannot create RPC client. Error = %d\n",
2089 __FUNCTION__, err);
1da177e4
LT		 2090 goto out_fail;
2091 }
2092 clnt->cl_intr = 1;
2093 clnt->cl_softrtry = 1;
1da177e4 2094 clp->cl_rpcclient = clnt;
1da177e4
LT		 2095 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
2096 nfs_idmap_new(clp);
2097 }
1da177e4
LT		 2098 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
2099 clnt = rpc_clone_client(clp->cl_rpcclient);
2100 if (!IS_ERR(clnt))
c818ba43 2101 server->nfs4_state = clp;
1da177e4
LT		 2102 up_write(&clp->cl_sem);
2103 clp = NULL;
2104
2105 if (IS_ERR(clnt)) {
9085bbcb
TM		 2106 dprintk("%s: cannot create RPC client. Error = %d\n",
2107 __FUNCTION__, err);
c818ba43 2108 return clnt;
1da177e4
LT		 2109 }
2110
1da177e4 2111 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 2112 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
c818ba43 2113 return ERR_PTR(-ENOMEM);
1da177e4
LT		 2114 }
2115
c818ba43 2116 if (clnt->cl_auth->au_flavor != flavor) {
6a19275a
BF		 2117 struct rpc_auth *auth;
2118
c818ba43 2119 auth = rpcauth_create(flavor, clnt);
6a19275a 2120 if (IS_ERR(auth)) {
9085bbcb 2121 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
c818ba43 2122 return (struct rpc_clnt *)auth;
1da177e4
LT		 2123 }
2124 }
c818ba43
MN		 2125 return clnt;
2126
2127 out_fail:
2128 if (clp)
2129 nfs4_put_client(clp);
2130 return ERR_PTR(err);
2131}
2132
2133static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
2134{
2135 struct nfs_server *server;
2136 struct rpc_timeout timeparms;
2137 rpc_authflavor_t authflavour;
2138 int err = -EIO;
2139
2140 sb->s_blocksize_bits = 0;
2141 sb->s_blocksize = 0;
2142 server = NFS_SB(sb);
2143 if (data->rsize != 0)
2144 server->rsize = nfs_block_size(data->rsize, NULL);
2145 if (data->wsize != 0)
2146 server->wsize = nfs_block_size(data->wsize, NULL);
2147 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
2148 server->caps = NFS_CAP_ATOMIC_OPEN;
2149
2150 server->acregmin = data->acregmin*HZ;
2151 server->acregmax = data->acregmax*HZ;
2152 server->acdirmin = data->acdirmin*HZ;
2153 server->acdirmax = data->acdirmax*HZ;
2154
2155 server->rpc_ops = &nfs_v4_clientops;
2156
2157 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
2158
2159 server->retrans_timeo = timeparms.to_initval;
2160 server->retrans_count = timeparms.to_retries;
2161
2162 /* Now create transport and client */
2163 authflavour = RPC_AUTH_UNIX;
2164 if (data->auth_flavourlen != 0) {
2165 if (data->auth_flavourlen != 1) {
2166 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
2167 __FUNCTION__, data->auth_flavourlen);
2168 err = -EINVAL;
2169 goto out_fail;
2170 }
2171 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
2172 err = -EFAULT;
2173 goto out_fail;
2174 }
2175 }
2176
2177 server->client = nfs4_create_client(server, &timeparms, data->proto, authflavour);
2178 if (IS_ERR(server->client)) {
2179 err = PTR_ERR(server->client);
2180 dprintk("%s: cannot create RPC client. Error = %d\n",
2181 __FUNCTION__, err);
2182 goto out_fail;
2183 }
1da177e4
LT		 2184
2185 sb->s_time_gran = 1;
2186
2187 sb->s_op = &nfs4_sops;
2188 err = nfs_sb_init(sb, authflavour);
c818ba43
MN		 2189
2190 out_fail:
1da177e4
LT		 2191 return err;
2192}
2193
2194static int nfs4_compare_super(struct super_block *sb, void *data)
2195{
2196 struct nfs_server *server = data;
2197 struct nfs_server *old = NFS_SB(sb);
2198
2199 if (strcmp(server->hostname, old->hostname) != 0)
2200 return 0;
2201 if (strcmp(server->mnt_path, old->mnt_path) != 0)
2202 return 0;
2203 return 1;
2204}
2205
2206static void *
2207nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
2208{
2209 void *p = NULL;
2210
2211 if (!src->len)
2212 return ERR_PTR(-EINVAL);
2213 if (src->len < maxlen)
2214 maxlen = src->len;
2215 if (dst == NULL) {
2216 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
2217 if (p == NULL)
2218 return ERR_PTR(-ENOMEM);
2219 }
2220 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 2221 kfree(p);
1da177e4
LT		 2222 return ERR_PTR(-EFAULT);
2223 }
2224 dst[maxlen] = '\0';
2225 return dst;
2226}
2227
2228static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
2229 int flags, const char *dev_name, void *raw_data)
2230{
2231 int error;
2232 struct nfs_server *server;
2233 struct super_block *s;
2234 struct nfs4_mount_data *data = raw_data;
2235 void *p;
2236
9085bbcb
TM		 2237 if (data == NULL) {
2238 dprintk("%s: missing data argument\n", __FUNCTION__);
2239 return ERR_PTR(-EINVAL);
2240 }
2241 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
2242 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 2243 return ERR_PTR(-EINVAL);
2244 }
2245
bd647545 2246 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4
LT		 2247 if (!server)
2248 return ERR_PTR(-ENOMEM);
1da177e4
LT		 2249 /* Zero out the NFS state stuff */
2250 init_nfsv4_state(server);
b7fa0554 2251 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 2252
1da177e4
LT		 2253 p = nfs_copy_user_string(NULL, &data->hostname, 256);
2254 if (IS_ERR(p))
2255 goto out_err;
2256 server->hostname = p;
2257
2258 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
2259 if (IS_ERR(p))
2260 goto out_err;
2261 server->mnt_path = p;
2262
2263 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
2264 sizeof(server->ip_addr) - 1);
2265 if (IS_ERR(p))
2266 goto out_err;
2267
2268 /* We now require that the mount process passes the remote address */
2269 if (data->host_addrlen != sizeof(server->addr)) {
2270 s = ERR_PTR(-EINVAL);
2271 goto out_free;
2272 }
2273 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
2274 s = ERR_PTR(-EFAULT);
2275 goto out_free;
2276 }
2277 if (server->addr.sin_family != AF_INET ||
2278 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 2279 dprintk("%s: mount program didn't pass remote IP address!\n",
2280 __FUNCTION__);
1da177e4
LT		 2281 s = ERR_PTR(-EINVAL);
2282 goto out_free;
2283 }
2284
9085bbcb
TM		 2285 /* Fire up rpciod if not yet running */
2286 s = ERR_PTR(rpciod_up());
2287 if (IS_ERR(s)) {
2288 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2289 __FUNCTION__, PTR_ERR(s));
2290 goto out_free;
2291 }
2292
1da177e4
LT		 2293 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
2294
2295 if (IS_ERR(s) || s->s_root)
2296 goto out_free;
2297
2298 s->s_flags = flags;
2299
9b04c997 2300 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT		 2301 if (error) {
2302 up_write(&s->s_umount);
2303 deactivate_super(s);
2304 return ERR_PTR(error);
2305 }
2306 s->s_flags |= MS_ACTIVE;
2307 return s;
2308out_err:
2309 s = (struct super_block *)p;
2310out_free:
f99d49ad
JJ		 2311 kfree(server->mnt_path);
2312 kfree(server->hostname);
1da177e4
LT		 2313 kfree(server);
2314 return s;
2315}
2316
2317static void nfs4_kill_super(struct super_block *sb)
2318{
2319 struct nfs_server *server = NFS_SB(sb);
2320
2321 nfs_return_all_delegations(sb);
2322 kill_anon_super(sb);
2323
2324 nfs4_renewd_prepare_shutdown(server);
2325
2326 if (server->client != NULL && !IS_ERR(server->client))
2327 rpc_shutdown_client(server->client);
1da177e4
LT		 2328
2329 destroy_nfsv4_state(server);
2330
967b9281
TM		 2331 rpciod_down();
2332
01d0ae8b 2333 nfs_free_iostats(server->io_stats);
f99d49ad 2334 kfree(server->hostname);
1da177e4 2335 kfree(server);
51d8fa6a 2336 nfs_release_automount_timer();
1da177e4
LT		 2337}
2338
2339static struct file_system_type nfs4_fs_type = {
2340 .owner = THIS_MODULE,
2341 .name = "nfs4",
2342 .get_sb = nfs4_get_sb,
2343 .kill_sb = nfs4_kill_super,
2344 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2345};
2346
a72b4422
TM		 2347static const int nfs_set_port_min = 0;
2348static const int nfs_set_port_max = 65535;
2349static int param_set_port(const char *val, struct kernel_param *kp)
2350{
2351 char *endp;
2352 int num = simple_strtol(val, &endp, 0);
2353 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
2354 return -EINVAL;
2355 *((int *)kp->arg) = num;
2356 return 0;
2357}
2358
2359module_param_call(callback_tcpport, param_set_port, param_get_int,
2360 &nfs_callback_set_tcpport, 0644);
2361
58df095b
TM		 2362static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
2363{
2364 char *endp;
2365 int num = simple_strtol(val, &endp, 0);
2366 int jif = num * HZ;
2367 if (endp == val || *endp || num < 0 || jif < num)
2368 return -EINVAL;
2369 *((int *)kp->arg) = jif;
2370 return 0;
2371}
2372
2373module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
2374 &nfs_idmap_cache_timeout, 0644);
2375
55a97593
TM		 2376/* Constructs the SERVER-side path */
2377static inline char *nfs4_path(const struct dentry *dentry, char *buffer, ssize_t buflen)
2378{
2379 return nfs_path(NFS_SB(dentry->d_sb)->mnt_path, dentry, buffer, buflen);
2380}
2381
2382static inline char *nfs4_dup_path(const struct dentry *dentry)
2383{
2384 char *page = (char *) __get_free_page(GFP_USER);
2385 char *path;
2386
2387 path = nfs4_path(dentry, page, PAGE_SIZE);
2388 if (!IS_ERR(path)) {
2389 int len = PAGE_SIZE + page - path;
2390 char *tmp = path;
2391
2392 path = kmalloc(len, GFP_KERNEL);
2393 if (path)
2394 memcpy(path, tmp, len);
2395 else
2396 path = ERR_PTR(-ENOMEM);
2397 }
2398 free_page((unsigned long)page);
2399 return path;
2400}
2401
61f5164c 2402static struct super_block *nfs4_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
55a97593
TM		 2403{
2404 const struct dentry *dentry = data->dentry;
2405 struct nfs4_client *clp = server->nfs4_state;
2406 struct super_block *sb;
2407
61f5164c
MN		 2408 server->fsid = data->fattr->fsid;
2409 nfs_copy_fh(&server->fh, data->fh);
55a97593
TM		 2410 server->mnt_path = nfs4_dup_path(dentry);
2411 if (IS_ERR(server->mnt_path)) {
2412 sb = (struct super_block *)server->mnt_path;
2413 goto err;
2414 }
2415 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
2416 if (IS_ERR(sb) || sb->s_root)
2417 goto free_path;
2418 nfs4_server_capabilities(server, &server->fh);
2419
2420 down_write(&clp->cl_sem);
2421 atomic_inc(&clp->cl_count);
2422 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
2423 up_write(&clp->cl_sem);
2424 return sb;
2425free_path:
2426 kfree(server->mnt_path);
2427err:
2428 server->mnt_path = NULL;
2429 return sb;
2430}
2431
2432static struct super_block *nfs_clone_nfs4_sb(struct file_system_type *fs_type,
2433 int flags, const char *dev_name, void *raw_data)
2434{
2435 struct nfs_clone_mount *data = raw_data;
61f5164c 2436 return nfs_clone_generic_sb(data, nfs4_clone_sb, nfs_clone_server);
55a97593
TM		 2437}
2438
2439static struct file_system_type clone_nfs4_fs_type = {
2440 .owner = THIS_MODULE,
61f5164c 2441 .name = "nfs4",
55a97593
TM		 2442 .get_sb = nfs_clone_nfs4_sb,
2443 .kill_sb = nfs4_kill_super,
2444 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2445};
2446
860de071
TM		 2447static struct vfsmount *nfs_do_clone_mount(struct nfs_server *server, char *devname, struct nfs_clone_mount *mountdata)
2448{
2449 struct vfsmount *mnt = NULL;
2450 switch (server->rpc_ops->version) {
2451 case 2:
2452 case 3:
4e5ccf60 2453 mnt = vfs_kern_mount(&clone_nfs_fs_type, 0, devname, mountdata);
860de071
TM		 2454 break;
2455 case 4:
2456 mnt = vfs_kern_mount(&clone_nfs4_fs_type, 0, devname, mountdata);
2457 }
2458 return mnt;
2459}
2460
1da177e4
LT		 2461#define nfs4_init_once(nfsi) \
2462 do { \
2463 INIT_LIST_HEAD(&(nfsi)->open_states); \
2464 nfsi->delegation = NULL; \
2465 nfsi->delegation_state = 0; \
2466 init_rwsem(&nfsi->rwsem); \
2467 } while(0)
a72b4422
TM		 2468
2469static inline int register_nfs4fs(void)
2470{
2471 int ret;
2472
2473 ret = nfs_register_sysctl();
2474 if (ret != 0)
2475 return ret;
2476 ret = register_filesystem(&nfs4_fs_type);
2477 if (ret != 0)
2478 nfs_unregister_sysctl();
2479 return ret;
2480}
2481
2482static inline void unregister_nfs4fs(void)
2483{
2484 unregister_filesystem(&nfs4_fs_type);
2485 nfs_unregister_sysctl();
2486}
1da177e4 2487#else
9cdb3883
MN		 2488#define nfs4_fill_sb(a,b) ERR_PTR(-EINVAL)
2489#define nfs4_fill_super(a,b) ERR_PTR(-EINVAL)
1da177e4
LT		 2490#define nfs4_init_once(nfsi) \
2491 do { } while (0)
2492#define register_nfs4fs() (0)
2493#define unregister_nfs4fs()
860de071
TM		 2494static struct vfsmount *nfs_do_clone_mount(struct nfs_server *server, char *devname, struct nfs_clone_mount *mountdata)
2495{
4e5ccf60 2496 return vfs_kern_mount(&clone_nfs_fs_type, 0, devname, mountdata);
860de071 2497}
1da177e4
LT		 2498#endif
2499
55a97593
TM		 2500static inline char *nfs_devname(const struct vfsmount *mnt_parent,
2501 const struct dentry *dentry,
2502 char *buffer, ssize_t buflen)
2503{
2504 return nfs_path(mnt_parent->mnt_devname, dentry, buffer, buflen);
2505}
2506
2507/**
2508 * nfs_do_submount - set up mountpoint when crossing a filesystem boundary
2509 * @mnt_parent - mountpoint of parent directory
2510 * @dentry - parent directory
2511 * @fh - filehandle for new root dentry
2512 * @fattr - attributes for new root inode
2513 *
2514 */
2515struct vfsmount *nfs_do_submount(const struct vfsmount *mnt_parent,
2516 const struct dentry *dentry, struct nfs_fh *fh,
2517 struct nfs_fattr *fattr)
2518{
2519 struct nfs_clone_mount mountdata = {
2520 .sb = mnt_parent->mnt_sb,
2521 .dentry = dentry,
2522 .fh = fh,
2523 .fattr = fattr,
2524 };
2525 struct vfsmount *mnt = ERR_PTR(-ENOMEM);
2526 char *page = (char *) __get_free_page(GFP_USER);
2527 char *devname;
2528
2529 dprintk("%s: submounting on %s/%s\n", __FUNCTION__,
2530 dentry->d_parent->d_name.name,
2531 dentry->d_name.name);
2532 if (page == NULL)
2533 goto out;
2534 devname = nfs_devname(mnt_parent, dentry, page, PAGE_SIZE);
2535 mnt = (struct vfsmount *)devname;
2536 if (IS_ERR(devname))
2537 goto free_page;
860de071 2538 mnt = nfs_do_clone_mount(NFS_SB(mnt_parent->mnt_sb), devname, &mountdata);
55a97593
TM		 2539free_page:
2540 free_page((unsigned long)page);
2541out:
2542 dprintk("%s: done\n", __FUNCTION__);
2543 return mnt;
2544}
2545
860de071 2546#ifdef CONFIG_NFS_V4
9cdb3883
MN		 2547/* Check if fs_root is valid */
2548static inline char *nfs4_pathname_string(struct nfs4_pathname *pathname, char *buffer, ssize_t buflen)
2549{
2550 char *end = buffer + buflen;
2551 int n;
2552
2553 *--end = '\0';
2554 buflen--;
2555
2556 n = pathname->ncomponents;
2557 while (--n >= 0) {
2558 struct nfs4_string *component = &pathname->components[n];
2559 buflen -= component->len + 1;
2560 if (buflen < 0)
2561 goto Elong;
2562 end -= component->len;
2563 memcpy(end, component->data, component->len);
2564 *--end = '/';
2565 }
2566 return end;
2567Elong:
2568 return ERR_PTR(-ENAMETOOLONG);
2569}
2570
2571/* Check if the string represents a "valid" IPv4 address */
2572static inline int valid_ipaddr4(const char *buf)
2573{
2574 int rc, count, in[4];
2575
2576 rc = sscanf(buf, "%d.%d.%d.%d", &in[0], &in[1], &in[2], &in[3]);
2577 if (rc != 4)
2578 return -EINVAL;
2579 for (count = 0; count < 4; count++) {
2580 if (in[count] > 255)
2581 return -EINVAL;
2582 }
2583 return 0;
2584}
2585
2586static struct super_block *nfs4_referral_sb(struct nfs_server *server, struct nfs_clone_mount *data)
2587{
2588 struct super_block *sb = ERR_PTR(-ENOMEM);
2589 int len;
2590
2591 len = strlen(data->mnt_path) + 1;
2592 server->mnt_path = kmalloc(len, GFP_KERNEL);
2593 if (server->mnt_path == NULL)
2594 goto err;
2595 memcpy(server->mnt_path, data->mnt_path, len);
2596 memcpy(&server->addr, data->addr, sizeof(struct sockaddr_in));
2597
2598 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
2599 if (IS_ERR(sb) || sb->s_root)
2600 goto free_path;
2601 return sb;
2602free_path:
2603 kfree(server->mnt_path);
2604err:
2605 server->mnt_path = NULL;
2606 return sb;
2607}
2608
2609static struct nfs_server *nfs4_referral_server(struct super_block *sb, struct nfs_clone_mount *data)
2610{
2611 struct nfs_server *server = NFS_SB(sb);
2612 struct rpc_timeout timeparms;
2613 int proto, timeo, retrans;
2614 void *err;
2615
2616 proto = IPPROTO_TCP;
2617 /* Since we are following a referral and there may be alternatives,
2618 set the timeouts and retries to low values */
2619 timeo = 2;
2620 retrans = 1;
2621 nfs_init_timeout_values(&timeparms, proto, timeo, retrans);
2622
2623 server->client = nfs4_create_client(server, &timeparms, proto, data->authflavor);
2624 if (IS_ERR((err = server->client)))
2625 goto out_err;
2626
2627 sb->s_time_gran = 1;
2628 sb->s_op = &nfs4_sops;
2629 err = ERR_PTR(nfs_sb_init(sb, data->authflavor));
2630 if (!IS_ERR(err))
2631 return server;
2632out_err:
2633 return (struct nfs_server *)err;
2634}
2635
2636static struct super_block *nfs_referral_nfs4_sb(struct file_system_type *fs_type,
2637 int flags, const char *dev_name, void *raw_data)
2638{
2639 struct nfs_clone_mount *data = raw_data;
2640 return nfs_clone_generic_sb(data, nfs4_referral_sb, nfs4_referral_server);
2641}
2642
2643static struct file_system_type nfs_referral_nfs4_fs_type = {
2644 .owner = THIS_MODULE,
2645 .name = "nfs4",
2646 .get_sb = nfs_referral_nfs4_sb,
2647 .kill_sb = nfs4_kill_super,
2648 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2649};
2650
2651/**
2652 * nfs_follow_referral - set up mountpoint when hitting a referral on moved error
2653 * @mnt_parent - mountpoint of parent directory
2654 * @dentry - parent directory
2655 * @fspath - fs path returned in fs_locations
2656 * @mntpath - mount path to new server
2657 * @hostname - hostname of new server
2658 * @addr - host addr of new server
2659 *
2660 */
2661struct vfsmount *nfs_follow_referral(const struct vfsmount *mnt_parent,
2662 const struct dentry *dentry, struct nfs4_fs_locations *locations)
2663{
2664 struct vfsmount *mnt = ERR_PTR(-ENOENT);
2665 struct nfs_clone_mount mountdata = {
2666 .sb = mnt_parent->mnt_sb,
2667 .dentry = dentry,
2668 .authflavor = NFS_SB(mnt_parent->mnt_sb)->client->cl_auth->au_flavor,
2669 };
2670 char *page, *page2;
2671 char *path, *fs_path;
2672 char *devname;
2673 int loc, s;
2674
2675 if (locations == NULL || locations->nlocations <= 0)
2676 goto out;
2677
2678 dprintk("%s: referral at %s/%s\n", __FUNCTION__,
2679 dentry->d_parent->d_name.name, dentry->d_name.name);
2680
2681 /* Ensure fs path is a prefix of current dentry path */
2682 page = (char *) __get_free_page(GFP_USER);
2683 if (page == NULL)
2684 goto out;
2685 page2 = (char *) __get_free_page(GFP_USER);
2686 if (page2 == NULL)
2687 goto out;
2688
2689 path = nfs4_path(dentry, page, PAGE_SIZE);
2690 if (IS_ERR(path))
2691 goto out_free;
2692
2693 fs_path = nfs4_pathname_string(&locations->fs_path, page2, PAGE_SIZE);
2694 if (IS_ERR(fs_path))
2695 goto out_free;
2696
2697 if (strncmp(path, fs_path, strlen(fs_path)) != 0) {
2698 dprintk("%s: path %s does not begin with fsroot %s\n", __FUNCTION__, path, fs_path);
2699 goto out_free;
2700 }
2701
2702 devname = nfs_devname(mnt_parent, dentry, page, PAGE_SIZE);
2703 if (IS_ERR(devname)) {
2704 mnt = (struct vfsmount *)devname;
2705 goto out_free;
2706 }
2707
2708 loc = 0;
2709 while (loc < locations->nlocations && IS_ERR(mnt)) {
2710 struct nfs4_fs_location *location = &locations->locations[loc];
2711 char *mnt_path;
2712
2713 if (location == NULL || location->nservers <= 0 ||
2714 location->rootpath.ncomponents == 0) {
2715 loc++;
2716 continue;
2717 }
2718
2719 mnt_path = nfs4_pathname_string(&location->rootpath, page2, PAGE_SIZE);
2720 if (IS_ERR(mnt_path)) {
2721 loc++;
2722 continue;
2723 }
2724 mountdata.mnt_path = mnt_path;
2725
2726 s = 0;
2727 while (s < location->nservers) {
2728 struct sockaddr_in addr = {};
2729
2730 if (location->servers[s].len <= 0 ||
2731 valid_ipaddr4(location->servers[s].data) < 0) {
2732 s++;
2733 continue;
2734 }
2735
2736 mountdata.hostname = location->servers[s].data;
2737 addr.sin_addr.s_addr = in_aton(mountdata.hostname);
2738 addr.sin_family = AF_INET;
2739 addr.sin_port = htons(NFS_PORT);
2740 mountdata.addr = &addr;
2741
2742 mnt = vfs_kern_mount(&nfs_referral_nfs4_fs_type, 0, devname, &mountdata);
2743 if (!IS_ERR(mnt)) {
2744 break;
2745 }
2746 s++;
2747 }
2748 loc++;
2749 }
2750
2751out_free:
2752 free_page((unsigned long)page);
2753 free_page((unsigned long)page2);
2754out:
2755 dprintk("%s: done\n", __FUNCTION__);
2756 return mnt;
2757}
2758
2759/*
2760 * nfs_do_refmount - handle crossing a referral on server
2761 * @dentry - dentry of referral
2762 * @nd - nameidata info
2763 *
2764 */
2765struct vfsmount *nfs_do_refmount(const struct vfsmount *mnt_parent, struct dentry *dentry)
2766{
2767 struct vfsmount *mnt = ERR_PTR(-ENOENT);
2768 struct dentry *parent;
2769 struct nfs4_fs_locations *fs_locations = NULL;
2770 struct page *page;
2771 int err;
2772
2773 /* BUG_ON(IS_ROOT(dentry)); */
2774 dprintk("%s: enter\n", __FUNCTION__);
2775
2776 page = alloc_page(GFP_KERNEL);
2777 if (page == NULL)
2778 goto out;
2779
2780 fs_locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
2781 if (fs_locations == NULL)
2782 goto out_free;
2783
2784 /* Get locations */
2785 parent = dget_parent(dentry);
2786 dprintk("%s: getting locations for %s/%s\n", __FUNCTION__, parent->d_name.name, dentry->d_name.name);
2787 err = nfs4_proc_fs_locations(parent->d_inode, dentry, fs_locations, page);
2788 dput(parent);
2789 if (err != 0 || fs_locations->nlocations <= 0 ||
2790 fs_locations->fs_path.ncomponents <= 0)
2791 goto out_free;
2792
2793 mnt = nfs_follow_referral(mnt_parent, dentry, fs_locations);
2794out_free:
2795 __free_page(page);
2796 kfree(fs_locations);
2797out:
2798 dprintk("%s: done\n", __FUNCTION__);
2799 return mnt;
2800}
860de071
TM		 2801#else
2802struct vfsmount *nfs_do_refmount(const struct vfsmount *mnt_parent, struct dentry *dentry)
2803{
2804 return ERR_PTR(-ENOENT);
2805}
2806#endif
9cdb3883 2807
1da177e4
LT		 2808extern int nfs_init_nfspagecache(void);
2809extern void nfs_destroy_nfspagecache(void);
2810extern int nfs_init_readpagecache(void);
2811extern void nfs_destroy_readpagecache(void);
2812extern int nfs_init_writepagecache(void);
2813extern void nfs_destroy_writepagecache(void);
2814#ifdef CONFIG_NFS_DIRECTIO
2815extern int nfs_init_directcache(void);
2816extern void nfs_destroy_directcache(void);
2817#endif
2818
2819static kmem_cache_t * nfs_inode_cachep;
2820
2821static struct inode *nfs_alloc_inode(struct super_block *sb)
2822{
2823 struct nfs_inode *nfsi;
2824 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2825 if (!nfsi)
2826 return NULL;
55296809
CL		 2827 nfsi->flags = 0UL;
2828 nfsi->cache_validity = 0UL;
223db122 2829 nfsi->cache_change_attribute = jiffies;
458818ed
TM		 2830#ifdef CONFIG_NFS_V3_ACL
2831 nfsi->acl_access = ERR_PTR(-EAGAIN);
2832 nfsi->acl_default = ERR_PTR(-EAGAIN);
2833#endif
e50a1c2e
BF		 2834#ifdef CONFIG_NFS_V4
2835 nfsi->nfs4_acl = NULL;
2836#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2837 return &nfsi->vfs_inode;
2838}
2839
2840static void nfs_destroy_inode(struct inode *inode)
2841{
2842 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2843}
2844
2845static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2846{
2847 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2848
2849 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2850 SLAB_CTOR_CONSTRUCTOR) {
2851 inode_init_once(&nfsi->vfs_inode);
2852 spin_lock_init(&nfsi->req_lock);
2853 INIT_LIST_HEAD(&nfsi->dirty);
2854 INIT_LIST_HEAD(&nfsi->commit);
2855 INIT_LIST_HEAD(&nfsi->open_files);
2856 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2857 atomic_set(&nfsi->data_updates, 0);
2858 nfsi->ndirty = 0;
2859 nfsi->ncommit = 0;
2860 nfsi->npages = 0;
1da177e4
LT		 2861 nfs4_init_once(nfsi);
2862 }
2863}
2864
75c96f85 2865static int nfs_init_inodecache(void)
1da177e4
LT		 2866{
2867 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2868 sizeof(struct nfs_inode),
fffb60f9
PJ		 2869 0, (SLAB_RECLAIM_ACCOUNT|
2870 SLAB_MEM_SPREAD),
1da177e4
LT		 2871 init_once, NULL);
2872 if (nfs_inode_cachep == NULL)
2873 return -ENOMEM;
2874
2875 return 0;
2876}
2877
75c96f85 2878static void nfs_destroy_inodecache(void)
1da177e4
LT		 2879{
2880 if (kmem_cache_destroy(nfs_inode_cachep))
2881 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2882}
2883
2884/*
2885 * Initialize NFS
2886 */
2887static int __init init_nfs_fs(void)
2888{
2889 int err;
2890
2891 err = nfs_init_nfspagecache();
2892 if (err)
2893 goto out4;
2894
2895 err = nfs_init_inodecache();
2896 if (err)
2897 goto out3;
2898
2899 err = nfs_init_readpagecache();
2900 if (err)
2901 goto out2;
2902
2903 err = nfs_init_writepagecache();
2904 if (err)
2905 goto out1;
2906
2907#ifdef CONFIG_NFS_DIRECTIO
2908 err = nfs_init_directcache();
2909 if (err)
2910 goto out0;
2911#endif
2912
2913#ifdef CONFIG_PROC_FS
2914 rpc_proc_register(&nfs_rpcstat);
2915#endif
2916 err = register_filesystem(&nfs_fs_type);
2917 if (err)
2918 goto out;
2919 if ((err = register_nfs4fs()) != 0)
2920 goto out;
2921 return 0;
2922out:
2923#ifdef CONFIG_PROC_FS
2924 rpc_proc_unregister("nfs");
2925#endif
1da177e4 2926#ifdef CONFIG_NFS_DIRECTIO
1da177e4 2927 nfs_destroy_directcache();
6b59a754 2928out0:
1da177e4 2929#endif
6b59a754 2930 nfs_destroy_writepagecache();
1da177e4
LT		 2931out1:
2932 nfs_destroy_readpagecache();
2933out2:
2934 nfs_destroy_inodecache();
2935out3:
2936 nfs_destroy_nfspagecache();
2937out4:
2938 return err;
2939}
2940
2941static void __exit exit_nfs_fs(void)
2942{
2943#ifdef CONFIG_NFS_DIRECTIO
2944 nfs_destroy_directcache();
2945#endif
2946 nfs_destroy_writepagecache();
2947 nfs_destroy_readpagecache();
2948 nfs_destroy_inodecache();
2949 nfs_destroy_nfspagecache();
2950#ifdef CONFIG_PROC_FS
2951 rpc_proc_unregister("nfs");
2952#endif
2953 unregister_filesystem(&nfs_fs_type);
2954 unregister_nfs4fs();
2955}
2956
2957/* Not quite true; I just maintain it */
2958MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2959MODULE_LICENSE("GPL");
2960
2961module_init(init_nfs_fs)
2962module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)