projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFSv4: Fix typo in lock caching
[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>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
4ce79717 42#include "nfs4_fs.h"
1da177e4
LT		 43#include "delegation.h"
44
45#define NFSDBG_FACILITY NFSDBG_VFS
46#define NFS_PARANOIA 1
47
48/* Maximum number of readahead requests
49 * FIXME: this should really be a sysctl so that users may tune it to suit
50 * their needs. People that do NFS over a slow network, might for
51 * instance want to reduce it to something closer to 1 for improved
52 * interactive response.
53 */
54#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
55
56static void nfs_invalidate_inode(struct inode *);
57static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long);
58
59static struct inode *nfs_alloc_inode(struct super_block *sb);
60static void nfs_destroy_inode(struct inode *);
61static int nfs_write_inode(struct inode *,int);
62static void nfs_delete_inode(struct inode *);
63static void nfs_clear_inode(struct inode *);
64static void nfs_umount_begin(struct super_block *);
65static int nfs_statfs(struct super_block *, struct kstatfs *);
66static int nfs_show_options(struct seq_file *, struct vfsmount *);
ada70d94 67static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT		 68
69static struct rpc_program nfs_program;
70
71static struct super_operations nfs_sops = {
72 .alloc_inode = nfs_alloc_inode,
73 .destroy_inode = nfs_destroy_inode,
74 .write_inode = nfs_write_inode,
75 .delete_inode = nfs_delete_inode,
76 .statfs = nfs_statfs,
77 .clear_inode = nfs_clear_inode,
78 .umount_begin = nfs_umount_begin,
79 .show_options = nfs_show_options,
80};
81
82/*
83 * RPC cruft for NFS
84 */
85static struct rpc_stat nfs_rpcstat = {
86 .program = &nfs_program
87};
88static struct rpc_version * nfs_version[] = {
89 NULL,
90 NULL,
91 &nfs_version2,
92#if defined(CONFIG_NFS_V3)
93 &nfs_version3,
94#elif defined(CONFIG_NFS_V4)
95 NULL,
96#endif
97#if defined(CONFIG_NFS_V4)
98 &nfs_version4,
99#endif
100};
101
102static struct rpc_program nfs_program = {
103 .name = "nfs",
104 .number = NFS_PROGRAM,
105 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
106 .version = nfs_version,
107 .stats = &nfs_rpcstat,
108 .pipe_dir_name = "/nfs",
109};
110
b7fa0554
AG		 111#ifdef CONFIG_NFS_V3_ACL
112static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
113static struct rpc_version * nfsacl_version[] = {
114 [3] = &nfsacl_version3,
115};
116
117struct rpc_program nfsacl_program = {
118 .name = "nfsacl",
119 .number = NFS_ACL_PROGRAM,
120 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
121 .version = nfsacl_version,
122 .stats = &nfsacl_rpcstat,
123};
124#endif /* CONFIG_NFS_V3_ACL */
125
1da177e4
LT		 126static inline unsigned long
127nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
128{
129 return nfs_fileid_to_ino_t(fattr->fileid);
130}
131
132static int
133nfs_write_inode(struct inode *inode, int sync)
134{
135 int flags = sync ? FLUSH_WAIT : 0;
136 int ret;
137
3da28eb1 138 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 139 if (ret < 0)
140 return ret;
141 return 0;
142}
143
144static void
145nfs_delete_inode(struct inode * inode)
146{
147 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
148
fef26658
MF		 149 truncate_inode_pages(&inode->i_data, 0);
150
1da177e4
LT		 151 nfs_wb_all(inode);
152 /*
153 * The following should never happen...
154 */
155 if (nfs_have_writebacks(inode)) {
156 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
157 }
158
159 clear_inode(inode);
160}
161
1da177e4
LT		 162static void
163nfs_clear_inode(struct inode *inode)
164{
165 struct nfs_inode *nfsi = NFS_I(inode);
166 struct rpc_cred *cred;
167
168 nfs_wb_all(inode);
169 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 170 nfs_zap_acl_cache(inode);
1da177e4
LT		 171 cred = nfsi->cache_access.cred;
172 if (cred)
173 put_rpccred(cred);
174 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
175}
176
177void
178nfs_umount_begin(struct super_block *sb)
179{
6a19275a 180 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT		 181
182 /* -EIO all pending I/O */
6a19275a 183 if (!IS_ERR(rpc))
1da177e4 184 rpc_killall_tasks(rpc);
b7fa0554
AG		 185 rpc = NFS_SB(sb)->client_acl;
186 if (!IS_ERR(rpc))
187 rpc_killall_tasks(rpc);
1da177e4
LT		 188}
189
190
191static inline unsigned long
192nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
193{
194 /* make sure blocksize is a power of two */
195 if ((bsize & (bsize - 1)) || nrbitsp) {
196 unsigned char nrbits;
197
198 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
199 ;
200 bsize = 1 << nrbits;
201 if (nrbitsp)
202 *nrbitsp = nrbits;
203 }
204
205 return bsize;
206}
207
208/*
209 * Calculate the number of 512byte blocks used.
210 */
211static inline unsigned long
212nfs_calc_block_size(u64 tsize)
213{
214 loff_t used = (tsize + 511) >> 9;
215 return (used > ULONG_MAX) ? ULONG_MAX : used;
216}
217
218/*
219 * Compute and set NFS server blocksize
220 */
221static inline unsigned long
222nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
223{
224 if (bsize < 1024)
225 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
226 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
227 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
228
229 return nfs_block_bits(bsize, nrbitsp);
230}
231
232/*
233 * Obtain the root inode of the file system.
234 */
235static struct inode *
236nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
237{
238 struct nfs_server *server = NFS_SB(sb);
239 struct inode *rooti;
240 int error;
241
242 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
243 if (error < 0) {
244 dprintk("nfs_get_root: getattr error = %d\n", -error);
245 return ERR_PTR(error);
246 }
247
248 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
249 if (!rooti)
250 return ERR_PTR(-ENOMEM);
251 return rooti;
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
279 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
280 /* Did getting the root inode fail? */
281 if (IS_ERR(root_inode)) {
282 no_root_error = PTR_ERR(root_inode);
283 goto out_no_root;
284 }
285 sb->s_root = d_alloc_root(root_inode);
286 if (!sb->s_root) {
287 no_root_error = -ENOMEM;
288 goto out_no_root;
289 }
290 sb->s_root->d_op = server->rpc_ops->dentry_ops;
291
292 /* Get some general file system info */
293 if (server->namelen == 0 &&
294 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
295 server->namelen = pathinfo.max_namelen;
296 /* Work out a lot of parameters */
297 if (server->rsize == 0)
298 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
299 if (server->wsize == 0)
300 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
301
302 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
303 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
304 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
305 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
306
307 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
308 if (server->rsize > max_rpc_payload)
309 server->rsize = max_rpc_payload;
310 if (server->wsize > max_rpc_payload)
311 server->wsize = max_rpc_payload;
312
313 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
314 if (server->rpages > NFS_READ_MAXIOV) {
315 server->rpages = NFS_READ_MAXIOV;
316 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
317 }
318
319 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
320 if (server->wpages > NFS_WRITE_MAXIOV) {
321 server->wpages = NFS_WRITE_MAXIOV;
322 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
323 }
324
325 if (sb->s_blocksize == 0)
326 sb->s_blocksize = nfs_block_bits(server->wsize,
327 &sb->s_blocksize_bits);
328 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
329
330 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
331 if (server->dtsize > PAGE_CACHE_SIZE)
332 server->dtsize = PAGE_CACHE_SIZE;
333 if (server->dtsize > server->rsize)
334 server->dtsize = server->rsize;
335
336 if (server->flags & NFS_MOUNT_NOAC) {
337 server->acregmin = server->acregmax = 0;
338 server->acdirmin = server->acdirmax = 0;
339 sb->s_flags |= MS_SYNCHRONOUS;
340 }
341 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
342
343 sb->s_maxbytes = fsinfo.maxfilesize;
344 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
345 sb->s_maxbytes = MAX_LFS_FILESIZE;
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
LT		 402
403 /* create transport and client */
eab5c084 404 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 405 if (IS_ERR(xprt)) {
9085bbcb
TM		 406 dprintk("%s: cannot create RPC transport. Error = %ld\n",
407 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 408 return (struct rpc_clnt *)xprt;
409 }
410 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
411 server->rpc_ops->version, data->pseudoflavor);
412 if (IS_ERR(clnt)) {
9085bbcb
TM		 413 dprintk("%s: cannot create RPC client. Error = %ld\n",
414 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 415 goto out_fail;
416 }
417
418 clnt->cl_intr = 1;
419 clnt->cl_softrtry = 1;
420 clnt->cl_chatty = 1;
421
422 return clnt;
423
424out_fail:
1da177e4
LT		 425 return clnt;
426}
427
428/*
429 * The way this works is that the mount process passes a structure
430 * in the data argument which contains the server's IP address
431 * and the root file handle obtained from the server's mount
432 * daemon. We stash these away in the private superblock fields.
433 */
434static int
435nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
436{
437 struct nfs_server *server;
438 rpc_authflavor_t authflavor;
439
440 server = NFS_SB(sb);
441 sb->s_blocksize_bits = 0;
442 sb->s_blocksize = 0;
443 if (data->bsize)
444 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
445 if (data->rsize)
446 server->rsize = nfs_block_size(data->rsize, NULL);
447 if (data->wsize)
448 server->wsize = nfs_block_size(data->wsize, NULL);
449 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
450
451 server->acregmin = data->acregmin*HZ;
452 server->acregmax = data->acregmax*HZ;
453 server->acdirmin = data->acdirmin*HZ;
454 server->acdirmax = data->acdirmax*HZ;
455
456 /* Start lockd here, before we might error out */
457 if (!(server->flags & NFS_MOUNT_NONLM))
458 lockd_up();
459
460 server->namelen = data->namlen;
461 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
462 if (!server->hostname)
463 return -ENOMEM;
464 strcpy(server->hostname, data->hostname);
465
466 /* Check NFS protocol revision and initialize RPC op vector
467 * and file handle pool. */
1da177e4 468#ifdef CONFIG_NFS_V3
9085bbcb 469 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 470 server->rpc_ops = &nfs_v3_clientops;
471 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 472 } else {
473 server->rpc_ops = &nfs_v2_clientops;
474 }
9085bbcb
TM		 475#else
476 server->rpc_ops = &nfs_v2_clientops;
477#endif
1da177e4
LT		 478
479 /* Fill in pseudoflavor for mount version < 5 */
480 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
481 data->pseudoflavor = RPC_AUTH_UNIX;
482 authflavor = data->pseudoflavor; /* save for sb_init() */
483 /* XXX maybe we want to add a server->pseudoflavor field */
484
485 /* Create RPC client handles */
486 server->client = nfs_create_client(server, data);
487 if (IS_ERR(server->client))
488 return PTR_ERR(server->client);
489 /* RFC 2623, sec 2.3.2 */
490 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 491 struct rpc_auth *auth;
492
1da177e4
LT		 493 server->client_sys = rpc_clone_client(server->client);
494 if (IS_ERR(server->client_sys))
495 return PTR_ERR(server->client_sys);
6a19275a
BF		 496 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
497 if (IS_ERR(auth))
498 return PTR_ERR(auth);
1da177e4
LT		 499 } else {
500 atomic_inc(&server->client->cl_count);
501 server->client_sys = server->client;
502 }
1da177e4 503 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 504#ifdef CONFIG_NFS_V3_ACL
505 if (!(server->flags & NFS_MOUNT_NOACL)) {
506 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
507 /* No errors! Assume that Sun nfsacls are supported */
508 if (!IS_ERR(server->client_acl))
509 server->caps |= NFS_CAP_ACLS;
510 }
511#else
512 server->flags &= ~NFS_MOUNT_NOACL;
513#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 514 /*
515 * The VFS shouldn't apply the umask to mode bits. We will
516 * do so ourselves when necessary.
517 */
518 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 519 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
520 server->namelen = NFS3_MAXNAMLEN;
521 sb->s_time_gran = 1;
522 } else {
523 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
524 server->namelen = NFS2_MAXNAMLEN;
525 }
526
527 sb->s_op = &nfs_sops;
528 return nfs_sb_init(sb, authflavor);
529}
530
531static int
532nfs_statfs(struct super_block *sb, struct kstatfs *buf)
533{
534 struct nfs_server *server = NFS_SB(sb);
535 unsigned char blockbits;
536 unsigned long blockres;
537 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
538 struct nfs_fattr fattr;
539 struct nfs_fsstat res = {
540 .fattr = &fattr,
541 };
542 int error;
543
544 lock_kernel();
545
546 error = server->rpc_ops->statfs(server, rootfh, &res);
547 buf->f_type = NFS_SUPER_MAGIC;
548 if (error < 0)
549 goto out_err;
550
551 /*
552 * Current versions of glibc do not correctly handle the
553 * case where f_frsize != f_bsize. Eventually we want to
554 * report the value of wtmult in this field.
555 */
556 buf->f_frsize = sb->s_blocksize;
557
558 /*
559 * On most *nix systems, f_blocks, f_bfree, and f_bavail
560 * are reported in units of f_frsize. Linux hasn't had
561 * an f_frsize field in its statfs struct until recently,
562 * thus historically Linux's sys_statfs reports these
563 * fields in units of f_bsize.
564 */
565 buf->f_bsize = sb->s_blocksize;
566 blockbits = sb->s_blocksize_bits;
567 blockres = (1 << blockbits) - 1;
568 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
569 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
570 buf->f_bavail = (res.abytes + blockres) >> blockbits;
571
572 buf->f_files = res.tfiles;
573 buf->f_ffree = res.afiles;
574
575 buf->f_namelen = server->namelen;
576 out:
577 unlock_kernel();
578
579 return 0;
580
581 out_err:
582 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
583 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
584 goto out;
585
586}
587
588static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
589{
590 static struct proc_nfs_info {
591 int flag;
592 char *str;
593 char *nostr;
594 } nfs_info[] = {
595 { NFS_MOUNT_SOFT, ",soft", ",hard" },
596 { NFS_MOUNT_INTR, ",intr", "" },
597 { NFS_MOUNT_POSIX, ",posix", "" },
1da177e4
LT		 598 { NFS_MOUNT_NOCTO, ",nocto", "" },
599 { NFS_MOUNT_NOAC, ",noac", "" },
600 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
b7fa0554 601 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 602 { 0, NULL, NULL }
603 };
604 struct proc_nfs_info *nfs_infop;
605 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
3063d8a1
TM		 606 char buf[12];
607 char *proto;
1da177e4
LT		 608
609 seq_printf(m, ",v%d", nfss->rpc_ops->version);
610 seq_printf(m, ",rsize=%d", nfss->rsize);
611 seq_printf(m, ",wsize=%d", nfss->wsize);
612 if (nfss->acregmin != 3*HZ)
613 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
614 if (nfss->acregmax != 60*HZ)
615 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
616 if (nfss->acdirmin != 30*HZ)
617 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
618 if (nfss->acdirmax != 60*HZ)
619 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
620 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
621 if (nfss->flags & nfs_infop->flag)
622 seq_puts(m, nfs_infop->str);
623 else
624 seq_puts(m, nfs_infop->nostr);
625 }
3063d8a1
TM		 626 switch (nfss->client->cl_xprt->prot) {
627 case IPPROTO_TCP:
628 proto = "tcp";
629 break;
630 case IPPROTO_UDP:
631 proto = "udp";
632 break;
633 default:
634 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
635 proto = buf;
636 }
637 seq_printf(m, ",proto=%s", proto);
1da177e4
LT		 638 seq_puts(m, ",addr=");
639 seq_escape(m, nfss->hostname, " \t\n\\");
640 return 0;
641}
642
643/*
644 * Invalidate the local caches
645 */
646void
647nfs_zap_caches(struct inode *inode)
648{
649 struct nfs_inode *nfsi = NFS_I(inode);
650 int mode = inode->i_mode;
651
dc59250c
CL		 652 spin_lock(&inode->i_lock);
653
1da177e4
LT		 654 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
655 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
656
657 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
658 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 659 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 660 else
55296809 661 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
dc59250c
CL		 662
663 spin_unlock(&inode->i_lock);
ada70d94
TM		 664}
665
666static void nfs_zap_acl_cache(struct inode *inode)
667{
668 void (*clear_acl_cache)(struct inode *);
669
670 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
671 if (clear_acl_cache != NULL)
672 clear_acl_cache(inode);
dc59250c 673 spin_lock(&inode->i_lock);
55296809 674 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 675 spin_unlock(&inode->i_lock);
1da177e4
LT		 676}
677
678/*
679 * Invalidate, but do not unhash, the inode
680 */
681static void
682nfs_invalidate_inode(struct inode *inode)
683{
684 umode_t save_mode = inode->i_mode;
685
686 make_bad_inode(inode);
687 inode->i_mode = save_mode;
688 nfs_zap_caches(inode);
689}
690
691struct nfs_find_desc {
692 struct nfs_fh *fh;
693 struct nfs_fattr *fattr;
694};
695
696/*
697 * In NFSv3 we can have 64bit inode numbers. In order to support
698 * this, and re-exported directories (also seen in NFSv2)
699 * we are forced to allow 2 different inodes to have the same
700 * i_ino.
701 */
702static int
703nfs_find_actor(struct inode *inode, void *opaque)
704{
705 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
706 struct nfs_fh *fh = desc->fh;
707 struct nfs_fattr *fattr = desc->fattr;
708
709 if (NFS_FILEID(inode) != fattr->fileid)
710 return 0;
711 if (nfs_compare_fh(NFS_FH(inode), fh))
712 return 0;
713 if (is_bad_inode(inode) || NFS_STALE(inode))
714 return 0;
715 return 1;
716}
717
718static int
719nfs_init_locked(struct inode *inode, void *opaque)
720{
721 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
722 struct nfs_fattr *fattr = desc->fattr;
723
724 NFS_FILEID(inode) = fattr->fileid;
725 nfs_copy_fh(NFS_FH(inode), desc->fh);
726 return 0;
727}
728
729/* Don't use READDIRPLUS on directories that we believe are too large */
730#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
731
732/*
733 * This is our front-end to iget that looks up inodes by file handle
734 * instead of inode number.
735 */
736struct inode *
737nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
738{
739 struct nfs_find_desc desc = {
740 .fh = fh,
741 .fattr = fattr
742 };
743 struct inode *inode = NULL;
744 unsigned long hash;
745
746 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
747 goto out_no_inode;
748
749 if (!fattr->nlink) {
750 printk("NFS: Buggy server - nlink == 0!\n");
751 goto out_no_inode;
752 }
753
754 hash = nfs_fattr_to_ino_t(fattr);
755
756 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
757 goto out_no_inode;
758
759 if (inode->i_state & I_NEW) {
760 struct nfs_inode *nfsi = NFS_I(inode);
761
762 /* We set i_ino for the few things that still rely on it,
763 * such as stat(2) */
764 inode->i_ino = hash;
765
766 /* We can't support update_atime(), since the server will reset it */
767 inode->i_flags |= S_NOATIME|S_NOCMTIME;
768 inode->i_mode = fattr->mode;
769 /* Why so? Because we want revalidate for devices/FIFOs, and
770 * that's precisely what we have in nfs_file_inode_operations.
771 */
92cfc62c 772 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 773 if (S_ISREG(inode->i_mode)) {
774 inode->i_fop = &nfs_file_operations;
775 inode->i_data.a_ops = &nfs_file_aops;
776 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
777 } else if (S_ISDIR(inode->i_mode)) {
778 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
779 inode->i_fop = &nfs_dir_operations;
780 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
781 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 782 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 783 } else if (S_ISLNK(inode->i_mode))
784 inode->i_op = &nfs_symlink_inode_operations;
785 else
786 init_special_inode(inode, inode->i_mode, fattr->rdev);
787
33801147
TM		 788 nfsi->read_cache_jiffies = fattr->time_start;
789 nfsi->last_updated = jiffies;
1da177e4
LT		 790 inode->i_atime = fattr->atime;
791 inode->i_mtime = fattr->mtime;
792 inode->i_ctime = fattr->ctime;
793 if (fattr->valid & NFS_ATTR_FATTR_V4)
794 nfsi->change_attr = fattr->change_attr;
795 inode->i_size = nfs_size_to_loff_t(fattr->size);
796 inode->i_nlink = fattr->nlink;
797 inode->i_uid = fattr->uid;
798 inode->i_gid = fattr->gid;
799 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
800 /*
801 * report the blocks in 512byte units
802 */
803 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
804 inode->i_blksize = inode->i_sb->s_blocksize;
805 } else {
806 inode->i_blocks = fattr->du.nfs2.blocks;
807 inode->i_blksize = fattr->du.nfs2.blocksize;
808 }
809 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
810 nfsi->attrtimeo_timestamp = jiffies;
811 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
812 nfsi->cache_access.cred = NULL;
813
814 unlock_new_inode(inode);
815 } else
816 nfs_refresh_inode(inode, fattr);
817 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
818 inode->i_sb->s_id,
819 (long long)NFS_FILEID(inode),
820 atomic_read(&inode->i_count));
821
822out:
823 return inode;
824
825out_no_inode:
826 printk("nfs_fhget: iget failed\n");
827 goto out;
828}
829
830#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
831
832int
833nfs_setattr(struct dentry *dentry, struct iattr *attr)
834{
835 struct inode *inode = dentry->d_inode;
836 struct nfs_fattr fattr;
837 int error;
838
839 if (attr->ia_valid & ATTR_SIZE) {
840 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
841 attr->ia_valid &= ~ATTR_SIZE;
842 }
843
844 /* Optimization: if the end result is no change, don't RPC */
845 attr->ia_valid &= NFS_VALID_ATTRS;
846 if (attr->ia_valid == 0)
847 return 0;
848
849 lock_kernel();
850 nfs_begin_data_update(inode);
851 /* Write all dirty data if we're changing file permissions or size */
852 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
853 if (filemap_fdatawrite(inode->i_mapping) == 0)
854 filemap_fdatawait(inode->i_mapping);
855 nfs_wb_all(inode);
856 }
642ac549
TM		 857 /*
858 * Return any delegations if we're going to change ACLs
859 */
860 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
861 nfs_inode_return_delegation(inode);
1da177e4 862 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 863 if (error == 0)
1da177e4 864 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 865 nfs_end_data_update(inode);
866 unlock_kernel();
867 return error;
868}
869
870/**
871 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
872 * @inode: pointer to struct inode
873 * @attr: pointer to struct iattr
874 *
875 * Note: we do this in the *proc.c in order to ensure that
876 * it works for things like exclusive creates too.
877 */
878void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
879{
880 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 881 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 882 int mode = attr->ia_mode & S_IALLUGO;
883 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 884 inode->i_mode = mode;
885 }
886 if ((attr->ia_valid & ATTR_UID) != 0)
887 inode->i_uid = attr->ia_uid;
888 if ((attr->ia_valid & ATTR_GID) != 0)
889 inode->i_gid = attr->ia_gid;
dc59250c 890 spin_lock(&inode->i_lock);
55296809 891 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 892 spin_unlock(&inode->i_lock);
65e4308d
TM		 893 }
894 if ((attr->ia_valid & ATTR_SIZE) != 0) {
895 inode->i_size = attr->ia_size;
896 vmtruncate(inode, attr->ia_size);
897 }
1da177e4
LT		 898}
899
412d582e
CL		 900static int nfs_wait_schedule(void *word)
901{
902 if (signal_pending(current))
903 return -ERESTARTSYS;
904 schedule();
905 return 0;
906}
907
1da177e4
LT		 908/*
909 * Wait for the inode to get unlocked.
1da177e4 910 */
412d582e 911static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 912{
913 struct rpc_clnt *clnt = NFS_CLIENT(inode);
914 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 915 sigset_t oldmask;
1da177e4 916 int error;
412d582e 917
412d582e
CL		 918 rpc_clnt_sigmask(clnt, &oldmask);
919 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
920 nfs_wait_schedule, TASK_INTERRUPTIBLE);
921 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 922
1da177e4
LT		 923 return error;
924}
925
412d582e
CL		 926static void nfs_wake_up_inode(struct inode *inode)
927{
928 struct nfs_inode *nfsi = NFS_I(inode);
929
930 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
931 smp_mb__after_clear_bit();
932 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
933}
934
1da177e4
LT		 935int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
936{
937 struct inode *inode = dentry->d_inode;
55296809 938 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 939 int err;
940
941 if (__IS_FLG(inode, MS_NOATIME))
942 need_atime = 0;
943 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
944 need_atime = 0;
945 /* We may force a getattr if the user cares about atime */
946 if (need_atime)
947 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
948 else
949 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
950 if (!err)
951 generic_fillattr(inode, stat);
952 return err;
953}
954
955struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
956{
957 struct nfs_open_context *ctx;
958
959 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
960 if (ctx != NULL) {
961 atomic_set(&ctx->count, 1);
962 ctx->dentry = dget(dentry);
963 ctx->cred = get_rpccred(cred);
964 ctx->state = NULL;
965 ctx->lockowner = current->files;
966 ctx->error = 0;
00a92642 967 ctx->dir_cookie = 0;
1da177e4
LT		 968 }
969 return ctx;
970}
971
972struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
973{
974 if (ctx != NULL)
975 atomic_inc(&ctx->count);
976 return ctx;
977}
978
979void put_nfs_open_context(struct nfs_open_context *ctx)
980{
981 if (atomic_dec_and_test(&ctx->count)) {
982 if (!list_empty(&ctx->list)) {
983 struct inode *inode = ctx->dentry->d_inode;
984 spin_lock(&inode->i_lock);
985 list_del(&ctx->list);
986 spin_unlock(&inode->i_lock);
987 }
988 if (ctx->state != NULL)
989 nfs4_close_state(ctx->state, ctx->mode);
990 if (ctx->cred != NULL)
991 put_rpccred(ctx->cred);
992 dput(ctx->dentry);
993 kfree(ctx);
994 }
995}
996
997/*
998 * Ensure that mmap has a recent RPC credential for use when writing out
999 * shared pages
1000 */
1001void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1002{
1003 struct inode *inode = filp->f_dentry->d_inode;
1004 struct nfs_inode *nfsi = NFS_I(inode);
1005
1006 filp->private_data = get_nfs_open_context(ctx);
1007 spin_lock(&inode->i_lock);
1008 list_add(&ctx->list, &nfsi->open_files);
1009 spin_unlock(&inode->i_lock);
1010}
1011
d530838b
TM		 1012/*
1013 * Given an inode, search for an open context with the desired characteristics
1014 */
1015struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1016{
1017 struct nfs_inode *nfsi = NFS_I(inode);
1018 struct nfs_open_context *pos, *ctx = NULL;
1019
1020 spin_lock(&inode->i_lock);
1021 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1022 if (cred != NULL && pos->cred != cred)
1023 continue;
1da177e4
LT		 1024 if ((pos->mode & mode) == mode) {
1025 ctx = get_nfs_open_context(pos);
1026 break;
1027 }
1028 }
1029 spin_unlock(&inode->i_lock);
1030 return ctx;
1031}
1032
1033void nfs_file_clear_open_context(struct file *filp)
1034{
1035 struct inode *inode = filp->f_dentry->d_inode;
1036 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1037
1038 if (ctx) {
1039 filp->private_data = NULL;
1040 spin_lock(&inode->i_lock);
1041 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1042 spin_unlock(&inode->i_lock);
1043 put_nfs_open_context(ctx);
1044 }
1045}
1046
1047/*
1048 * These allocate and release file read/write context information.
1049 */
1050int nfs_open(struct inode *inode, struct file *filp)
1051{
1052 struct nfs_open_context *ctx;
1053 struct rpc_cred *cred;
1054
1055 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1056 if (IS_ERR(cred))
1057 return PTR_ERR(cred);
1058 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
1059 put_rpccred(cred);
1060 if (ctx == NULL)
1061 return -ENOMEM;
1062 ctx->mode = filp->f_mode;
1063 nfs_file_set_open_context(filp, ctx);
1064 put_nfs_open_context(ctx);
1da177e4
LT		 1065 return 0;
1066}
1067
1068int nfs_release(struct inode *inode, struct file *filp)
1069{
1da177e4
LT		 1070 nfs_file_clear_open_context(filp);
1071 return 0;
1072}
1073
1074/*
1075 * This function is called whenever some part of NFS notices that
1076 * the cached attributes have to be refreshed.
1077 */
1078int
1079__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1080{
1081 int status = -ESTALE;
1082 struct nfs_fattr fattr;
1083 struct nfs_inode *nfsi = NFS_I(inode);
1084 unsigned long verifier;
55296809 1085 unsigned long cache_validity;
1da177e4
LT		 1086
1087 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1088 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1089
1090 lock_kernel();
1091 if (!inode || is_bad_inode(inode))
1092 goto out_nowait;
1093 if (NFS_STALE(inode))
1094 goto out_nowait;
1095
412d582e
CL		 1096 status = nfs_wait_on_inode(inode);
1097 if (status < 0)
1098 goto out;
1099 if (NFS_STALE(inode)) {
1100 status = -ESTALE;
1101 /* Do we trust the cached ESTALE? */
1102 if (NFS_ATTRTIMEO(inode) != 0) {
1103 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1104 /* no */
1105 } else
1106 goto out;
1107 }
1da177e4 1108 }
1da177e4
LT		 1109
1110 /* Protect against RPC races by saving the change attribute */
1111 verifier = nfs_save_change_attribute(inode);
1112 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1113 if (status != 0) {
1114 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1115 inode->i_sb->s_id,
1116 (long long)NFS_FILEID(inode), status);
1117 if (status == -ESTALE) {
1118 nfs_zap_caches(inode);
1119 if (!S_ISDIR(inode->i_mode))
412d582e 1120 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1121 }
1122 goto out;
1123 }
1124
33801147 1125 spin_lock(&inode->i_lock);
1da177e4
LT		 1126 status = nfs_update_inode(inode, &fattr, verifier);
1127 if (status) {
33801147 1128 spin_unlock(&inode->i_lock);
1da177e4
LT		 1129 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1130 inode->i_sb->s_id,
1131 (long long)NFS_FILEID(inode), status);
1132 goto out;
1133 }
55296809
CL		 1134 cache_validity = nfsi->cache_validity;
1135 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1136
1da177e4
LT		 1137 /*
1138 * We may need to keep the attributes marked as invalid if
1139 * we raced with nfs_end_attr_update().
1140 */
913a70fc 1141 if (time_after_eq(verifier, nfsi->cache_change_attribute))
55296809 1142 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
dc59250c 1143 spin_unlock(&inode->i_lock);
55296809 1144
7d52e862 1145 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809
CL		 1146
1147 if (cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1148 nfs_zap_acl_cache(inode);
55296809 1149
1da177e4
LT		 1150 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1151 inode->i_sb->s_id,
1152 (long long)NFS_FILEID(inode));
1153
412d582e
CL		 1154 out:
1155 nfs_wake_up_inode(inode);
1156
1da177e4
LT		 1157 out_nowait:
1158 unlock_kernel();
1159 return status;
1160}
1161
1162int nfs_attribute_timeout(struct inode *inode)
1163{
1164 struct nfs_inode *nfsi = NFS_I(inode);
1165
1166 if (nfs_have_delegation(inode, FMODE_READ))
1167 return 0;
1168 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1169}
1170
1171/**
1172 * nfs_revalidate_inode - Revalidate the inode attributes
1173 * @server - pointer to nfs_server struct
1174 * @inode - pointer to inode struct
1175 *
1176 * Updates inode attribute information by retrieving the data from the server.
1177 */
1178int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1179{
55296809 1180 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1181 && !nfs_attribute_timeout(inode))
1182 return NFS_STALE(inode) ? -ESTALE : 0;
1183 return __nfs_revalidate_inode(server, inode);
1184}
1185
7d52e862
TM		 1186/**
1187 * nfs_revalidate_mapping - Revalidate the pagecache
1188 * @inode - pointer to host inode
1189 * @mapping - pointer to mapping
1190 */
1191void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1192{
1193 struct nfs_inode *nfsi = NFS_I(inode);
1194
55296809 1195 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
7d52e862
TM		 1196 if (S_ISREG(inode->i_mode)) {
1197 if (filemap_fdatawrite(mapping) == 0)
1198 filemap_fdatawait(mapping);
1199 nfs_wb_all(inode);
1200 }
1201 invalidate_inode_pages2(mapping);
dc59250c
CL		 1202
1203 spin_lock(&inode->i_lock);
55296809 1204 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1205 if (S_ISDIR(inode->i_mode)) {
1206 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1207 /* This ensures we revalidate child dentries */
913a70fc 1208 nfsi->cache_change_attribute = jiffies;
7d52e862 1209 }
dc59250c
CL		 1210 spin_unlock(&inode->i_lock);
1211
7d52e862
TM		 1212 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1213 inode->i_sb->s_id,
1214 (long long)NFS_FILEID(inode));
1215 }
1216}
1217
1da177e4
LT		 1218/**
1219 * nfs_begin_data_update
1220 * @inode - pointer to inode
1221 * Declare that a set of operations will update file data on the server
1222 */
1223void nfs_begin_data_update(struct inode *inode)
1224{
1225 atomic_inc(&NFS_I(inode)->data_updates);
1226}
1227
1228/**
1229 * nfs_end_data_update
1230 * @inode - pointer to inode
1231 * Declare end of the operations that will update file data
1232 * This will mark the inode as immediately needing revalidation
1233 * of its attribute cache.
1234 */
1235void nfs_end_data_update(struct inode *inode)
1236{
1237 struct nfs_inode *nfsi = NFS_I(inode);
1238
1239 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1240 /* Directories and symlinks: invalidate page cache */
1241 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1242 spin_lock(&inode->i_lock);
55296809 1243 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1244 spin_unlock(&inode->i_lock);
1245 }
1da177e4 1246 }
913a70fc 1247 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1248 atomic_dec(&nfsi->data_updates);
1249}
1250
1da177e4 1251/**
33801147 1252 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1253 * @inode - pointer to inode
1254 * @fattr - updated attributes
1255 *
1256 * Verifies the attribute cache. If we have just changed the attributes,
1257 * so that fattr carries weak cache consistency data, then it may
1258 * also update the ctime/mtime/change_attribute.
1259 */
33801147 1260static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1261{
1262 struct nfs_inode *nfsi = NFS_I(inode);
1263 loff_t cur_size, new_isize;
1264 int data_unstable;
1265
dc59250c 1266
1da177e4
LT		 1267 /* Are we in the process of updating data on the server? */
1268 data_unstable = nfs_caches_unstable(inode);
1269
1270 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1271 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1272 && nfsi->change_attr == fattr->pre_change_attr)
1273 nfsi->change_attr = fattr->change_attr;
fe51beec 1274 if (nfsi->change_attr != fattr->change_attr) {
55296809 1275 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1276 if (!data_unstable)
55296809 1277 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1278 }
1da177e4
LT		 1279 }
1280
dc59250c 1281 if ((fattr->valid & NFS_ATTR_FATTR) == 0) {
1da177e4 1282 return 0;
dc59250c 1283 }
1da177e4
LT		 1284
1285 /* Has the inode gone and changed behind our back? */
1286 if (nfsi->fileid != fattr->fileid
dc59250c 1287 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1da177e4 1288 return -EIO;
dc59250c 1289 }
1da177e4
LT		 1290
1291 cur_size = i_size_read(inode);
1292 new_isize = nfs_size_to_loff_t(fattr->size);
1293
1294 /* If we have atomic WCC data, we may update some attributes */
1295 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1296 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1297 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1298 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1299 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1300 }
1301
1302 /* Verify a few of the more important attributes */
fe51beec 1303 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1304 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1305 if (!data_unstable)
55296809 1306 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec
TM		 1307 }
1308 if (cur_size != new_isize) {
55296809 1309 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1310 if (nfsi->npages == 0)
55296809 1311 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1312 }
1da177e4
LT		 1313
1314 /* Have any file permissions changed? */
1315 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1316 || inode->i_uid != fattr->uid
1317 || inode->i_gid != fattr->gid)
55296809 1318 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1319
1320 /* Has the link count changed? */
1321 if (inode->i_nlink != fattr->nlink)
55296809 1322 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1323
1324 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1325 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1326
33801147 1327 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1328 return 0;
1329}
1330
33801147
TM		 1331/**
1332 * nfs_refresh_inode - try to update the inode attribute cache
1333 * @inode - pointer to inode
1334 * @fattr - updated attributes
1335 *
1336 * Check that an RPC call that returned attributes has not overlapped with
1337 * other recent updates of the inode metadata, then decide whether it is
1338 * safe to do a full update of the inode attributes, or whether just to
1339 * call nfs_check_inode_attributes.
1340 */
1341int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1342{
1343 struct nfs_inode *nfsi = NFS_I(inode);
1344 int status;
1345
1346 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1347 return 0;
1348 spin_lock(&inode->i_lock);
1349 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1350 if (nfs_verify_change_attribute(inode, fattr->time_start))
1351 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1352 if (time_after(fattr->time_start, nfsi->last_updated))
1353 status = nfs_update_inode(inode, fattr, fattr->time_start);
1354 else
1355 status = nfs_check_inode_attributes(inode, fattr);
1356
1357 spin_unlock(&inode->i_lock);
1358 return status;
1359}
1360
decf491f
TM		 1361/**
1362 * nfs_post_op_update_inode - try to update the inode attribute cache
1363 * @inode - pointer to inode
1364 * @fattr - updated attributes
1365 *
1366 * After an operation that has changed the inode metadata, mark the
1367 * attribute cache as being invalid, then try to update it.
1368 */
1369int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1370{
1371 struct nfs_inode *nfsi = NFS_I(inode);
1372 int status = 0;
1373
1374 spin_lock(&inode->i_lock);
1375 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1376 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1377 goto out;
1378 }
1379 status = nfs_update_inode(inode, fattr, fattr->time_start);
1380 if (time_after_eq(fattr->time_start, nfsi->cache_change_attribute))
1381 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1382 nfsi->cache_change_attribute = jiffies;
1383out:
1384 spin_unlock(&inode->i_lock);
1385 return status;
1386}
1387
1da177e4
LT		 1388/*
1389 * Many nfs protocol calls return the new file attributes after
1390 * an operation. Here we update the inode to reflect the state
1391 * of the server's inode.
1392 *
1393 * This is a bit tricky because we have to make sure all dirty pages
1394 * have been sent off to the server before calling invalidate_inode_pages.
1395 * To make sure no other process adds more write requests while we try
1396 * our best to flush them, we make them sleep during the attribute refresh.
1397 *
1398 * A very similar scenario holds for the dir cache.
1399 */
1400static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1401{
1402 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1403 loff_t cur_isize, new_isize;
1da177e4 1404 unsigned int invalid = 0;
1da177e4
LT		 1405 int data_unstable;
1406
1407 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1408 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1409 atomic_read(&inode->i_count), fattr->valid);
1410
1411 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1412 return 0;
1413
1414 if (nfsi->fileid != fattr->fileid) {
1415 printk(KERN_ERR "%s: inode number mismatch\n"
1416 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1417 __FUNCTION__,
1418 inode->i_sb->s_id, (long long)nfsi->fileid,
1419 inode->i_sb->s_id, (long long)fattr->fileid);
1420 goto out_err;
1421 }
1422
1423 /*
1424 * Make sure the inode's type hasn't changed.
1425 */
33801147 1426 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT		 1427 goto out_changed;
1428
1429 /*
1430 * Update the read time so we don't revalidate too often.
1431 */
33801147
TM		 1432 nfsi->read_cache_jiffies = fattr->time_start;
1433 nfsi->last_updated = jiffies;
1da177e4
LT		 1434
1435 /* Are we racing with known updates of the metadata on the server? */
b3c52da3
TM		 1436 data_unstable = ! (nfs_verify_change_attribute(inode, verifier) ||
1437 (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE));
1da177e4 1438
951a143b 1439 /* Check if our cached file size is stale */
1da177e4
LT		 1440 new_isize = nfs_size_to_loff_t(fattr->size);
1441 cur_isize = i_size_read(inode);
951a143b
TM		 1442 if (new_isize != cur_isize) {
1443 /* Do we perhaps have any outstanding writes? */
1444 if (nfsi->npages == 0) {
1445 /* No, but did we race with nfs_end_data_update()? */
913a70fc 1446 if (time_after_eq(verifier, nfsi->cache_change_attribute)) {
1da177e4 1447 inode->i_size = new_isize;
951a143b 1448 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1449 }
951a143b
TM		 1450 invalid |= NFS_INO_INVALID_ATTR;
1451 } else if (new_isize > cur_isize) {
1da177e4
LT		 1452 inode->i_size = new_isize;
1453 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1454 }
951a143b
TM		 1455 dprintk("NFS: isize change on server for file %s/%ld\n",
1456 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1457 }
1458
951a143b 1459 /* Check if the mtime agrees */
1da177e4
LT		 1460 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1461 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1462 dprintk("NFS: mtime change on server for file %s/%ld\n",
1463 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1464 if (!data_unstable)
1465 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1466 }
1467
1468 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1469 && nfsi->change_attr != fattr->change_attr) {
951a143b 1470 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1da177e4 1471 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1472 nfsi->change_attr = fattr->change_attr;
1473 if (!data_unstable)
ada70d94 1474 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1475 }
1476
ada70d94
TM		 1477 /* If ctime has changed we should definitely clear access+acl caches */
1478 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1479 if (!data_unstable)
1480 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1481 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1482 }
1da177e4
LT		 1483 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1484
1485 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1486 inode->i_uid != fattr->uid ||
1487 inode->i_gid != fattr->gid)
ada70d94 1488 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1489
1490 inode->i_mode = fattr->mode;
1491 inode->i_nlink = fattr->nlink;
1492 inode->i_uid = fattr->uid;
1493 inode->i_gid = fattr->gid;
1494
1495 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1496 /*
1497 * report the blocks in 512byte units
1498 */
1499 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1500 inode->i_blksize = inode->i_sb->s_blocksize;
1501 } else {
1502 inode->i_blocks = fattr->du.nfs2.blocks;
1503 inode->i_blksize = fattr->du.nfs2.blocksize;
1504 }
1505
1506 /* Update attrtimeo value if we're out of the unstable period */
1507 if (invalid & NFS_INO_INVALID_ATTR) {
1508 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1509 nfsi->attrtimeo_timestamp = jiffies;
1510 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1511 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1512 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1513 nfsi->attrtimeo_timestamp = jiffies;
1514 }
1515 /* Don't invalidate the data if we were to blame */
1516 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1517 || S_ISLNK(inode->i_mode)))
1518 invalid &= ~NFS_INO_INVALID_DATA;
1519 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1520 nfsi->cache_validity |= invalid;
1da177e4
LT		 1521
1522 return 0;
1523 out_changed:
1524 /*
1525 * Big trouble! The inode has become a different object.
1526 */
1527#ifdef NFS_PARANOIA
1528 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1529 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1530#endif
1531 /*
1532 * No need to worry about unhashing the dentry, as the
1533 * lookup validation will know that the inode is bad.
1534 * (But we fall through to invalidate the caches.)
1535 */
1536 nfs_invalidate_inode(inode);
1537 out_err:
412d582e 1538 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1539 return -ESTALE;
1540}
1541
1542/*
1543 * File system information
1544 */
1545
1546static int nfs_set_super(struct super_block *s, void *data)
1547{
1548 s->s_fs_info = data;
1549 return set_anon_super(s, data);
1550}
1551
1552static int nfs_compare_super(struct super_block *sb, void *data)
1553{
1554 struct nfs_server *server = data;
1555 struct nfs_server *old = NFS_SB(sb);
1556
1557 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1558 return 0;
1559 if (old->addr.sin_port != server->addr.sin_port)
1560 return 0;
1561 return !nfs_compare_fh(&old->fh, &server->fh);
1562}
1563
1564static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1565 int flags, const char *dev_name, void *raw_data)
1566{
1567 int error;
9085bbcb 1568 struct nfs_server *server = NULL;
1da177e4
LT		 1569 struct super_block *s;
1570 struct nfs_fh *root;
1571 struct nfs_mount_data *data = raw_data;
1572
9085bbcb
TM		 1573 s = ERR_PTR(-EINVAL);
1574 if (data == NULL) {
1575 dprintk("%s: missing data argument\n", __FUNCTION__);
1576 goto out_err;
1577 }
1578 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1579 dprintk("%s: bad mount version\n", __FUNCTION__);
1580 goto out_err;
1da177e4 1581 }
9085bbcb
TM		 1582 switch (data->version) {
1583 case 1:
1584 data->namlen = 0;
1585 case 2:
1586 data->bsize = 0;
1587 case 3:
1588 if (data->flags & NFS_MOUNT_VER3) {
1589 dprintk("%s: mount structure version %d does not support NFSv3\n",
1590 __FUNCTION__,
1591 data->version);
1592 goto out_err;
1593 }
1594 data->root.size = NFS2_FHSIZE;
1595 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1596 case 4:
1597 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1598 dprintk("%s: mount structure version %d does not support strong security\n",
1599 __FUNCTION__,
1600 data->version);
1601 goto out_err;
1602 }
1603 case 5:
1604 memset(data->context, 0, sizeof(data->context));
1605 }
1606#ifndef CONFIG_NFS_V3
1607 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1608 s = ERR_PTR(-EPROTONOSUPPORT);
1609 if (data->flags & NFS_MOUNT_VER3) {
1610 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1611 goto out_err;
1612 }
1613#endif /* CONFIG_NFS_V3 */
1da177e4 1614
9085bbcb 1615 s = ERR_PTR(-ENOMEM);
1da177e4
LT		 1616 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1617 if (!server)
9085bbcb 1618 goto out_err;
1da177e4
LT		 1619 memset(server, 0, sizeof(struct nfs_server));
1620 /* Zero out the NFS state stuff */
1621 init_nfsv4_state(server);
b7fa0554 1622 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1623
1da177e4
LT		 1624 root = &server->fh;
1625 if (data->flags & NFS_MOUNT_VER3)
1626 root->size = data->root.size;
1627 else
1628 root->size = NFS2_FHSIZE;
9085bbcb 1629 s = ERR_PTR(-EINVAL);
1da177e4 1630 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1631 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1632 goto out_err;
1da177e4
LT		 1633 }
1634 memcpy(root->data, data->root.data, root->size);
1635
1636 /* We now require that the mount process passes the remote address */
1637 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1638 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1639 dprintk("%s: mount program didn't pass remote address!\n",
1640 __FUNCTION__);
1641 goto out_err;
1da177e4
LT		 1642 }
1643
9085bbcb
TM		 1644 /* Fire up rpciod if not yet running */
1645 s = ERR_PTR(rpciod_up());
1646 if (IS_ERR(s)) {
1647 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1648 __FUNCTION__, PTR_ERR(s));
1649 goto out_err;
1da177e4
LT		 1650 }
1651
9085bbcb
TM		 1652 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1653 if (IS_ERR(s) || s->s_root)
1654 goto out_rpciod_down;
1da177e4 1655
9085bbcb 1656 s->s_flags = flags;
1da177e4
LT		 1657
1658 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1659 if (error) {
1660 up_write(&s->s_umount);
1661 deactivate_super(s);
1662 return ERR_PTR(error);
1663 }
1664 s->s_flags |= MS_ACTIVE;
1665 return s;
9085bbcb
TM		 1666out_rpciod_down:
1667 rpciod_down();
1668out_err:
1669 kfree(server);
1670 return s;
1da177e4
LT		 1671}
1672
1673static void nfs_kill_super(struct super_block *s)
1674{
1675 struct nfs_server *server = NFS_SB(s);
1676
1677 kill_anon_super(s);
1678
6a19275a 1679 if (!IS_ERR(server->client))
1da177e4 1680 rpc_shutdown_client(server->client);
6a19275a 1681 if (!IS_ERR(server->client_sys))
1da177e4 1682 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1683 if (!IS_ERR(server->client_acl))
1684 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1685
1686 if (!(server->flags & NFS_MOUNT_NONLM))
1687 lockd_down(); /* release rpc.lockd */
1688
1689 rpciod_down(); /* release rpciod */
1690
f99d49ad 1691 kfree(server->hostname);
1da177e4
LT		 1692 kfree(server);
1693}
1694
1695static struct file_system_type nfs_fs_type = {
1696 .owner = THIS_MODULE,
1697 .name = "nfs",
1698 .get_sb = nfs_get_sb,
1699 .kill_sb = nfs_kill_super,
1700 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1701};
1702
1703#ifdef CONFIG_NFS_V4
1704
1705static void nfs4_clear_inode(struct inode *);
1706
1707
1708static struct super_operations nfs4_sops = {
1709 .alloc_inode = nfs_alloc_inode,
1710 .destroy_inode = nfs_destroy_inode,
1711 .write_inode = nfs_write_inode,
1712 .delete_inode = nfs_delete_inode,
1713 .statfs = nfs_statfs,
1714 .clear_inode = nfs4_clear_inode,
1715 .umount_begin = nfs_umount_begin,
1716 .show_options = nfs_show_options,
1717};
1718
1719/*
1720 * Clean out any remaining NFSv4 state that might be left over due
1721 * to open() calls that passed nfs_atomic_lookup, but failed to call
1722 * nfs_open().
1723 */
1724static void nfs4_clear_inode(struct inode *inode)
1725{
1726 struct nfs_inode *nfsi = NFS_I(inode);
1727
1728 /* If we are holding a delegation, return it! */
cae7a073 1729 nfs_inode_return_delegation(inode);
1da177e4
LT		 1730 /* First call standard NFS clear_inode() code */
1731 nfs_clear_inode(inode);
1732 /* Now clear out any remaining state */
1733 while (!list_empty(&nfsi->open_states)) {
1734 struct nfs4_state *state;
1735
1736 state = list_entry(nfsi->open_states.next,
1737 struct nfs4_state,
1738 inode_states);
1739 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1740 __FUNCTION__,
1741 inode->i_sb->s_id,
1742 (long long)NFS_FILEID(inode),
1743 state);
1744 BUG_ON(atomic_read(&state->count) != 1);
1745 nfs4_close_state(state, state->state);
1746 }
1747}
1748
1749
1750static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1751{
1752 struct nfs_server *server;
1753 struct nfs4_client *clp = NULL;
1754 struct rpc_xprt *xprt = NULL;
1755 struct rpc_clnt *clnt = NULL;
1756 struct rpc_timeout timeparms;
1757 rpc_authflavor_t authflavour;
eab5c084 1758 int err = -EIO;
1da177e4
LT		 1759
1760 sb->s_blocksize_bits = 0;
1761 sb->s_blocksize = 0;
1762 server = NFS_SB(sb);
1763 if (data->rsize != 0)
1764 server->rsize = nfs_block_size(data->rsize, NULL);
1765 if (data->wsize != 0)
1766 server->wsize = nfs_block_size(data->wsize, NULL);
1767 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1768 server->caps = NFS_CAP_ATOMIC_OPEN;
1769
1770 server->acregmin = data->acregmin*HZ;
1771 server->acregmax = data->acregmax*HZ;
1772 server->acdirmin = data->acdirmin*HZ;
1773 server->acdirmax = data->acdirmax*HZ;
1774
1775 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1776
eab5c084 1777 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4
LT		 1778
1779 clp = nfs4_get_client(&server->addr.sin_addr);
1780 if (!clp) {
9085bbcb 1781 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1782 return -EIO;
1783 }
1784
1785 /* Now create transport and client */
1786 authflavour = RPC_AUTH_UNIX;
1787 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1788 if (data->auth_flavourlen != 1) {
1789 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1790 __FUNCTION__, data->auth_flavourlen);
1791 err = -EINVAL;
1792 goto out_fail;
1793 }
1da177e4
LT		 1794 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1795 err = -EFAULT;
1796 goto out_fail;
1797 }
1798 }
1799
1800 down_write(&clp->cl_sem);
6a19275a 1801 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1802 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT		 1803 if (IS_ERR(xprt)) {
1804 up_write(&clp->cl_sem);
1da177e4 1805 err = PTR_ERR(xprt);
9085bbcb
TM		 1806 dprintk("%s: cannot create RPC transport. Error = %d\n",
1807 __FUNCTION__, err);
1da177e4
LT		 1808 goto out_fail;
1809 }
1810 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1811 server->rpc_ops->version, authflavour);
1812 if (IS_ERR(clnt)) {
1813 up_write(&clp->cl_sem);
1da177e4 1814 err = PTR_ERR(clnt);
9085bbcb
TM		 1815 dprintk("%s: cannot create RPC client. Error = %d\n",
1816 __FUNCTION__, err);
1da177e4
LT		 1817 goto out_fail;
1818 }
1819 clnt->cl_intr = 1;
1820 clnt->cl_softrtry = 1;
1821 clnt->cl_chatty = 1;
1822 clp->cl_rpcclient = clnt;
1823 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1824 if (IS_ERR(clp->cl_cred)) {
1825 up_write(&clp->cl_sem);
1826 err = PTR_ERR(clp->cl_cred);
1827 clp->cl_cred = NULL;
1828 goto out_fail;
1829 }
1830 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1831 nfs_idmap_new(clp);
1832 }
1833 if (list_empty(&clp->cl_superblocks)) {
1834 err = nfs4_init_client(clp);
1835 if (err != 0) {
1836 up_write(&clp->cl_sem);
1837 goto out_fail;
1838 }
1839 }
1840 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1841 clnt = rpc_clone_client(clp->cl_rpcclient);
1842 if (!IS_ERR(clnt))
1843 server->nfs4_state = clp;
1844 up_write(&clp->cl_sem);
1845 clp = NULL;
1846
1847 if (IS_ERR(clnt)) {
9085bbcb
TM		 1848 err = PTR_ERR(clnt);
1849 dprintk("%s: cannot create RPC client. Error = %d\n",
1850 __FUNCTION__, err);
1851 return err;
1da177e4
LT		 1852 }
1853
1854 server->client = clnt;
1855
1856 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1857 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1858 return -ENOMEM;
1859 }
1860
1861 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1862 struct rpc_auth *auth;
1863
1864 auth = rpcauth_create(authflavour, clnt);
1865 if (IS_ERR(auth)) {
9085bbcb 1866 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1867 return PTR_ERR(auth);
1da177e4
LT		 1868 }
1869 }
1870
1871 sb->s_time_gran = 1;
1872
1873 sb->s_op = &nfs4_sops;
1874 err = nfs_sb_init(sb, authflavour);
1875 if (err == 0)
1876 return 0;
1877out_fail:
1878 if (clp)
1879 nfs4_put_client(clp);
1880 return err;
1881}
1882
1883static int nfs4_compare_super(struct super_block *sb, void *data)
1884{
1885 struct nfs_server *server = data;
1886 struct nfs_server *old = NFS_SB(sb);
1887
1888 if (strcmp(server->hostname, old->hostname) != 0)
1889 return 0;
1890 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1891 return 0;
1892 return 1;
1893}
1894
1895static void *
1896nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1897{
1898 void *p = NULL;
1899
1900 if (!src->len)
1901 return ERR_PTR(-EINVAL);
1902 if (src->len < maxlen)
1903 maxlen = src->len;
1904 if (dst == NULL) {
1905 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1906 if (p == NULL)
1907 return ERR_PTR(-ENOMEM);
1908 }
1909 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 1910 kfree(p);
1da177e4
LT		 1911 return ERR_PTR(-EFAULT);
1912 }
1913 dst[maxlen] = '\0';
1914 return dst;
1915}
1916
1917static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1918 int flags, const char *dev_name, void *raw_data)
1919{
1920 int error;
1921 struct nfs_server *server;
1922 struct super_block *s;
1923 struct nfs4_mount_data *data = raw_data;
1924 void *p;
1925
9085bbcb
TM		 1926 if (data == NULL) {
1927 dprintk("%s: missing data argument\n", __FUNCTION__);
1928 return ERR_PTR(-EINVAL);
1929 }
1930 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1931 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1932 return ERR_PTR(-EINVAL);
1933 }
1934
1935 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1936 if (!server)
1937 return ERR_PTR(-ENOMEM);
1938 memset(server, 0, sizeof(struct nfs_server));
1939 /* Zero out the NFS state stuff */
1940 init_nfsv4_state(server);
b7fa0554 1941 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1942
1da177e4
LT		 1943 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1944 if (IS_ERR(p))
1945 goto out_err;
1946 server->hostname = p;
1947
1948 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1949 if (IS_ERR(p))
1950 goto out_err;
1951 server->mnt_path = p;
1952
1953 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1954 sizeof(server->ip_addr) - 1);
1955 if (IS_ERR(p))
1956 goto out_err;
1957
1958 /* We now require that the mount process passes the remote address */
1959 if (data->host_addrlen != sizeof(server->addr)) {
1960 s = ERR_PTR(-EINVAL);
1961 goto out_free;
1962 }
1963 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1964 s = ERR_PTR(-EFAULT);
1965 goto out_free;
1966 }
1967 if (server->addr.sin_family != AF_INET ||
1968 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1969 dprintk("%s: mount program didn't pass remote IP address!\n",
1970 __FUNCTION__);
1da177e4
LT		 1971 s = ERR_PTR(-EINVAL);
1972 goto out_free;
1973 }
1974
9085bbcb
TM		 1975 /* Fire up rpciod if not yet running */
1976 s = ERR_PTR(rpciod_up());
1977 if (IS_ERR(s)) {
1978 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1979 __FUNCTION__, PTR_ERR(s));
1980 goto out_free;
1981 }
1982
1da177e4
LT		 1983 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1984
1985 if (IS_ERR(s) || s->s_root)
1986 goto out_free;
1987
1988 s->s_flags = flags;
1989
1da177e4
LT		 1990 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1991 if (error) {
1992 up_write(&s->s_umount);
1993 deactivate_super(s);
1994 return ERR_PTR(error);
1995 }
1996 s->s_flags |= MS_ACTIVE;
1997 return s;
1998out_err:
1999 s = (struct super_block *)p;
2000out_free:
f99d49ad
JJ		 2001 kfree(server->mnt_path);
2002 kfree(server->hostname);
1da177e4
LT		 2003 kfree(server);
2004 return s;
2005}
2006
2007static void nfs4_kill_super(struct super_block *sb)
2008{
2009 struct nfs_server *server = NFS_SB(sb);
2010
2011 nfs_return_all_delegations(sb);
2012 kill_anon_super(sb);
2013
2014 nfs4_renewd_prepare_shutdown(server);
2015
2016 if (server->client != NULL && !IS_ERR(server->client))
2017 rpc_shutdown_client(server->client);
2018 rpciod_down(); /* release rpciod */
2019
2020 destroy_nfsv4_state(server);
2021
f99d49ad 2022 kfree(server->hostname);
1da177e4
LT		 2023 kfree(server);
2024}
2025
2026static struct file_system_type nfs4_fs_type = {
2027 .owner = THIS_MODULE,
2028 .name = "nfs4",
2029 .get_sb = nfs4_get_sb,
2030 .kill_sb = nfs4_kill_super,
2031 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2032};
2033
2034#define nfs4_init_once(nfsi) \
2035 do { \
2036 INIT_LIST_HEAD(&(nfsi)->open_states); \
2037 nfsi->delegation = NULL; \
2038 nfsi->delegation_state = 0; \
2039 init_rwsem(&nfsi->rwsem); \
2040 } while(0)
2041#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
2042#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
2043#else
2044#define nfs4_init_once(nfsi) \
2045 do { } while (0)
2046#define register_nfs4fs() (0)
2047#define unregister_nfs4fs()
2048#endif
2049
2050extern int nfs_init_nfspagecache(void);
2051extern void nfs_destroy_nfspagecache(void);
2052extern int nfs_init_readpagecache(void);
2053extern void nfs_destroy_readpagecache(void);
2054extern int nfs_init_writepagecache(void);
2055extern void nfs_destroy_writepagecache(void);
2056#ifdef CONFIG_NFS_DIRECTIO
2057extern int nfs_init_directcache(void);
2058extern void nfs_destroy_directcache(void);
2059#endif
2060
2061static kmem_cache_t * nfs_inode_cachep;
2062
2063static struct inode *nfs_alloc_inode(struct super_block *sb)
2064{
2065 struct nfs_inode *nfsi;
2066 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2067 if (!nfsi)
2068 return NULL;
55296809
CL		 2069 nfsi->flags = 0UL;
2070 nfsi->cache_validity = 0UL;
458818ed
TM		 2071#ifdef CONFIG_NFS_V3_ACL
2072 nfsi->acl_access = ERR_PTR(-EAGAIN);
2073 nfsi->acl_default = ERR_PTR(-EAGAIN);
2074#endif
e50a1c2e
BF		 2075#ifdef CONFIG_NFS_V4
2076 nfsi->nfs4_acl = NULL;
2077#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2078 return &nfsi->vfs_inode;
2079}
2080
2081static void nfs_destroy_inode(struct inode *inode)
2082{
2083 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2084}
2085
2086static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2087{
2088 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2089
2090 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2091 SLAB_CTOR_CONSTRUCTOR) {
2092 inode_init_once(&nfsi->vfs_inode);
2093 spin_lock_init(&nfsi->req_lock);
2094 INIT_LIST_HEAD(&nfsi->dirty);
2095 INIT_LIST_HEAD(&nfsi->commit);
2096 INIT_LIST_HEAD(&nfsi->open_files);
2097 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2098 atomic_set(&nfsi->data_updates, 0);
2099 nfsi->ndirty = 0;
2100 nfsi->ncommit = 0;
2101 nfsi->npages = 0;
1da177e4
LT		 2102 nfs4_init_once(nfsi);
2103 }
2104}
2105
75c96f85 2106static int nfs_init_inodecache(void)
1da177e4
LT		 2107{
2108 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2109 sizeof(struct nfs_inode),
2110 0, SLAB_RECLAIM_ACCOUNT,
2111 init_once, NULL);
2112 if (nfs_inode_cachep == NULL)
2113 return -ENOMEM;
2114
2115 return 0;
2116}
2117
75c96f85 2118static void nfs_destroy_inodecache(void)
1da177e4
LT		 2119{
2120 if (kmem_cache_destroy(nfs_inode_cachep))
2121 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2122}
2123
2124/*
2125 * Initialize NFS
2126 */
2127static int __init init_nfs_fs(void)
2128{
2129 int err;
2130
2131 err = nfs_init_nfspagecache();
2132 if (err)
2133 goto out4;
2134
2135 err = nfs_init_inodecache();
2136 if (err)
2137 goto out3;
2138
2139 err = nfs_init_readpagecache();
2140 if (err)
2141 goto out2;
2142
2143 err = nfs_init_writepagecache();
2144 if (err)
2145 goto out1;
2146
2147#ifdef CONFIG_NFS_DIRECTIO
2148 err = nfs_init_directcache();
2149 if (err)
2150 goto out0;
2151#endif
2152
2153#ifdef CONFIG_PROC_FS
2154 rpc_proc_register(&nfs_rpcstat);
2155#endif
2156 err = register_filesystem(&nfs_fs_type);
2157 if (err)
2158 goto out;
2159 if ((err = register_nfs4fs()) != 0)
2160 goto out;
2161 return 0;
2162out:
2163#ifdef CONFIG_PROC_FS
2164 rpc_proc_unregister("nfs");
2165#endif
2166 nfs_destroy_writepagecache();
2167#ifdef CONFIG_NFS_DIRECTIO
2168out0:
2169 nfs_destroy_directcache();
2170#endif
2171out1:
2172 nfs_destroy_readpagecache();
2173out2:
2174 nfs_destroy_inodecache();
2175out3:
2176 nfs_destroy_nfspagecache();
2177out4:
2178 return err;
2179}
2180
2181static void __exit exit_nfs_fs(void)
2182{
2183#ifdef CONFIG_NFS_DIRECTIO
2184 nfs_destroy_directcache();
2185#endif
2186 nfs_destroy_writepagecache();
2187 nfs_destroy_readpagecache();
2188 nfs_destroy_inodecache();
2189 nfs_destroy_nfspagecache();
2190#ifdef CONFIG_PROC_FS
2191 rpc_proc_unregister("nfs");
2192#endif
2193 unregister_filesystem(&nfs_fs_type);
2194 unregister_nfs4fs();
2195}
2196
2197/* Not quite true; I just maintain it */
2198MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2199MODULE_LICENSE("GPL");
2200
2201module_init(init_nfs_fs)
2202module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)