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