NFS: Store the file system "fsid" value in the NFS super block.
[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
8b4bdcf8 239 server->fsid = fsinfo->fattr->fsid;
03f28e3a 240 return nfs_fhget(sb, rootfh, fsinfo->fattr);
1da177e4
LT
241}
242
243/*
244 * Do NFS version-independent mount processing, and sanity checking
245 */
246static int
247nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
248{
249 struct nfs_server *server;
250 struct inode *root_inode;
251 struct nfs_fattr fattr;
252 struct nfs_fsinfo fsinfo = {
253 .fattr = &fattr,
254 };
255 struct nfs_pathconf pathinfo = {
256 .fattr = &fattr,
257 };
258 int no_root_error = 0;
259 unsigned long max_rpc_payload;
260
261 /* We probably want something more informative here */
262 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
263
264 server = NFS_SB(sb);
265
266 sb->s_magic = NFS_SUPER_MAGIC;
267
01d0ae8b
TM
268 server->io_stats = nfs_alloc_iostats();
269 if (server->io_stats == NULL)
270 return -ENOMEM;
271
1da177e4
LT
272 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
273 /* Did getting the root inode fail? */
274 if (IS_ERR(root_inode)) {
275 no_root_error = PTR_ERR(root_inode);
276 goto out_no_root;
277 }
278 sb->s_root = d_alloc_root(root_inode);
279 if (!sb->s_root) {
280 no_root_error = -ENOMEM;
281 goto out_no_root;
282 }
283 sb->s_root->d_op = server->rpc_ops->dentry_ops;
284
67ec9f46
CL
285 /* mount time stamp, in seconds */
286 server->mount_time = jiffies;
287
1da177e4
LT
288 /* Get some general file system info */
289 if (server->namelen == 0 &&
290 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
291 server->namelen = pathinfo.max_namelen;
292 /* Work out a lot of parameters */
293 if (server->rsize == 0)
294 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
295 if (server->wsize == 0)
296 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
297
298 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
299 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
300 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
301 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
302
303 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
304 if (server->rsize > max_rpc_payload)
305 server->rsize = max_rpc_payload;
40859d7e
CL
306 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
307 server->rsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 308 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4 309
40859d7e
CL
310 if (server->wsize > max_rpc_payload)
311 server->wsize = max_rpc_payload;
312 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
313 server->wsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 314 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4
LT
315
316 if (sb->s_blocksize == 0)
317 sb->s_blocksize = nfs_block_bits(server->wsize,
318 &sb->s_blocksize_bits);
319 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
320
321 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
322 if (server->dtsize > PAGE_CACHE_SIZE)
323 server->dtsize = PAGE_CACHE_SIZE;
324 if (server->dtsize > server->rsize)
325 server->dtsize = server->rsize;
326
327 if (server->flags & NFS_MOUNT_NOAC) {
328 server->acregmin = server->acregmax = 0;
329 server->acdirmin = server->acdirmax = 0;
330 sb->s_flags |= MS_SYNCHRONOUS;
331 }
332 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
333
334 sb->s_maxbytes = fsinfo.maxfilesize;
335 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
336 sb->s_maxbytes = MAX_LFS_FILESIZE;
337
338 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
339 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
340
341 /* We're airborne Set socket buffersize */
342 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
343 return 0;
344 /* Yargs. It didn't work out. */
345out_no_root:
346 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
347 if (!IS_ERR(root_inode))
348 iput(root_inode);
349 return no_root_error;
350}
351
eab5c084
CL
352static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
353{
354 to->to_initval = timeo * HZ / 10;
355 to->to_retries = retrans;
356 if (!to->to_retries)
357 to->to_retries = 2;
358
359 switch (proto) {
360 case IPPROTO_TCP:
361 if (!to->to_initval)
362 to->to_initval = 60 * HZ;
03bf4b70
CL
363 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
364 to->to_initval = NFS_MAX_TCP_TIMEOUT;
eab5c084
CL
365 to->to_increment = to->to_initval;
366 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
367 to->to_exponential = 0;
368 break;
369 case IPPROTO_UDP:
370 default:
371 if (!to->to_initval)
372 to->to_initval = 11 * HZ / 10;
03bf4b70
CL
373 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
374 to->to_initval = NFS_MAX_UDP_TIMEOUT;
375 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
eab5c084
CL
376 to->to_exponential = 1;
377 break;
378 }
379}
380
1da177e4
LT
381/*
382 * Create an RPC client handle.
383 */
384static struct rpc_clnt *
385nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
386{
387 struct rpc_timeout timeparms;
388 struct rpc_xprt *xprt = NULL;
389 struct rpc_clnt *clnt = NULL;
eab5c084 390 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
1da177e4 391
eab5c084 392 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
1da177e4 393
7a480e25
CL
394 server->retrans_timeo = timeparms.to_initval;
395 server->retrans_count = timeparms.to_retries;
396
1da177e4 397 /* create transport and client */
eab5c084 398 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 399 if (IS_ERR(xprt)) {
9085bbcb
TM
400 dprintk("%s: cannot create RPC transport. Error = %ld\n",
401 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT
402 return (struct rpc_clnt *)xprt;
403 }
404 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
405 server->rpc_ops->version, data->pseudoflavor);
406 if (IS_ERR(clnt)) {
9085bbcb
TM
407 dprintk("%s: cannot create RPC client. Error = %ld\n",
408 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT
409 goto out_fail;
410 }
411
412 clnt->cl_intr = 1;
413 clnt->cl_softrtry = 1;
1da177e4
LT
414
415 return clnt;
416
417out_fail:
1da177e4
LT
418 return clnt;
419}
420
421/*
422 * The way this works is that the mount process passes a structure
423 * in the data argument which contains the server's IP address
424 * and the root file handle obtained from the server's mount
425 * daemon. We stash these away in the private superblock fields.
426 */
427static int
428nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
429{
430 struct nfs_server *server;
431 rpc_authflavor_t authflavor;
432
433 server = NFS_SB(sb);
434 sb->s_blocksize_bits = 0;
435 sb->s_blocksize = 0;
436 if (data->bsize)
437 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
438 if (data->rsize)
439 server->rsize = nfs_block_size(data->rsize, NULL);
440 if (data->wsize)
441 server->wsize = nfs_block_size(data->wsize, NULL);
442 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
443
444 server->acregmin = data->acregmin*HZ;
445 server->acregmax = data->acregmax*HZ;
446 server->acdirmin = data->acdirmin*HZ;
447 server->acdirmax = data->acdirmax*HZ;
448
449 /* Start lockd here, before we might error out */
450 if (!(server->flags & NFS_MOUNT_NONLM))
451 lockd_up();
452
453 server->namelen = data->namlen;
454 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
455 if (!server->hostname)
456 return -ENOMEM;
457 strcpy(server->hostname, data->hostname);
458
459 /* Check NFS protocol revision and initialize RPC op vector
460 * and file handle pool. */
1da177e4 461#ifdef CONFIG_NFS_V3
9085bbcb 462 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT
463 server->rpc_ops = &nfs_v3_clientops;
464 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT
465 } else {
466 server->rpc_ops = &nfs_v2_clientops;
467 }
9085bbcb
TM
468#else
469 server->rpc_ops = &nfs_v2_clientops;
470#endif
1da177e4
LT
471
472 /* Fill in pseudoflavor for mount version < 5 */
473 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
474 data->pseudoflavor = RPC_AUTH_UNIX;
475 authflavor = data->pseudoflavor; /* save for sb_init() */
476 /* XXX maybe we want to add a server->pseudoflavor field */
477
478 /* Create RPC client handles */
479 server->client = nfs_create_client(server, data);
480 if (IS_ERR(server->client))
481 return PTR_ERR(server->client);
482 /* RFC 2623, sec 2.3.2 */
483 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF
484 struct rpc_auth *auth;
485
1da177e4
LT
486 server->client_sys = rpc_clone_client(server->client);
487 if (IS_ERR(server->client_sys))
488 return PTR_ERR(server->client_sys);
6a19275a
BF
489 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
490 if (IS_ERR(auth))
491 return PTR_ERR(auth);
1da177e4
LT
492 } else {
493 atomic_inc(&server->client->cl_count);
494 server->client_sys = server->client;
495 }
1da177e4 496 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG
497#ifdef CONFIG_NFS_V3_ACL
498 if (!(server->flags & NFS_MOUNT_NOACL)) {
499 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
500 /* No errors! Assume that Sun nfsacls are supported */
501 if (!IS_ERR(server->client_acl))
502 server->caps |= NFS_CAP_ACLS;
503 }
504#else
505 server->flags &= ~NFS_MOUNT_NOACL;
506#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG
507 /*
508 * The VFS shouldn't apply the umask to mode bits. We will
509 * do so ourselves when necessary.
510 */
511 sb->s_flags |= MS_POSIXACL;
1da177e4
LT
512 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
513 server->namelen = NFS3_MAXNAMLEN;
514 sb->s_time_gran = 1;
515 } else {
516 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
517 server->namelen = NFS2_MAXNAMLEN;
518 }
519
520 sb->s_op = &nfs_sops;
521 return nfs_sb_init(sb, authflavor);
522}
523
524static int
525nfs_statfs(struct super_block *sb, struct kstatfs *buf)
526{
527 struct nfs_server *server = NFS_SB(sb);
528 unsigned char blockbits;
529 unsigned long blockres;
530 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
531 struct nfs_fattr fattr;
532 struct nfs_fsstat res = {
533 .fattr = &fattr,
534 };
535 int error;
536
537 lock_kernel();
538
539 error = server->rpc_ops->statfs(server, rootfh, &res);
540 buf->f_type = NFS_SUPER_MAGIC;
541 if (error < 0)
542 goto out_err;
543
544 /*
545 * Current versions of glibc do not correctly handle the
546 * case where f_frsize != f_bsize. Eventually we want to
547 * report the value of wtmult in this field.
548 */
549 buf->f_frsize = sb->s_blocksize;
550
551 /*
552 * On most *nix systems, f_blocks, f_bfree, and f_bavail
553 * are reported in units of f_frsize. Linux hasn't had
554 * an f_frsize field in its statfs struct until recently,
555 * thus historically Linux's sys_statfs reports these
556 * fields in units of f_bsize.
557 */
558 buf->f_bsize = sb->s_blocksize;
559 blockbits = sb->s_blocksize_bits;
560 blockres = (1 << blockbits) - 1;
561 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
562 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
563 buf->f_bavail = (res.abytes + blockres) >> blockbits;
564
565 buf->f_files = res.tfiles;
566 buf->f_ffree = res.afiles;
567
568 buf->f_namelen = server->namelen;
569 out:
570 unlock_kernel();
1da177e4
LT
571 return 0;
572
573 out_err:
dc20f803 574 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
1da177e4
LT
575 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
576 goto out;
577
578}
579
d9ef5a8c 580static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
1da177e4
LT
581{
582 static struct proc_nfs_info {
583 int flag;
584 char *str;
585 char *nostr;
586 } nfs_info[] = {
587 { NFS_MOUNT_SOFT, ",soft", ",hard" },
588 { NFS_MOUNT_INTR, ",intr", "" },
1da177e4
LT
589 { NFS_MOUNT_NOCTO, ",nocto", "" },
590 { NFS_MOUNT_NOAC, ",noac", "" },
c8bded96 591 { NFS_MOUNT_NONLM, ",nolock", "" },
b7fa0554 592 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT
593 { 0, NULL, NULL }
594 };
595 struct proc_nfs_info *nfs_infop;
3063d8a1
TM
596 char buf[12];
597 char *proto;
1da177e4 598
c8bded96 599 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
1da177e4
LT
600 seq_printf(m, ",rsize=%d", nfss->rsize);
601 seq_printf(m, ",wsize=%d", nfss->wsize);
d9ef5a8c 602 if (nfss->acregmin != 3*HZ || showdefaults)
1da177e4 603 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
d9ef5a8c 604 if (nfss->acregmax != 60*HZ || showdefaults)
1da177e4 605 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
d9ef5a8c 606 if (nfss->acdirmin != 30*HZ || showdefaults)
1da177e4 607 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
d9ef5a8c 608 if (nfss->acdirmax != 60*HZ || showdefaults)
1da177e4
LT
609 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
610 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
611 if (nfss->flags & nfs_infop->flag)
612 seq_puts(m, nfs_infop->str);
613 else
614 seq_puts(m, nfs_infop->nostr);
615 }
3063d8a1
TM
616 switch (nfss->client->cl_xprt->prot) {
617 case IPPROTO_TCP:
618 proto = "tcp";
619 break;
620 case IPPROTO_UDP:
621 proto = "udp";
622 break;
623 default:
624 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
625 proto = buf;
626 }
627 seq_printf(m, ",proto=%s", proto);
7a480e25
CL
628 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
629 seq_printf(m, ",retrans=%u", nfss->retrans_count);
d9ef5a8c
CL
630}
631
632static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
633{
634 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
635
636 nfs_show_mount_options(m, nfss, 0);
637
1da177e4
LT
638 seq_puts(m, ",addr=");
639 seq_escape(m, nfss->hostname, " \t\n\\");
d9ef5a8c
CL
640
641 return 0;
642}
643
644static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
645{
646 int i, cpu;
647 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
648 struct rpc_auth *auth = nfss->client->cl_auth;
649 struct nfs_iostats totals = { };
650
651 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
652
653 /*
654 * Display all mount option settings
655 */
656 seq_printf(m, "\n\topts:\t");
657 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
658 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
659 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
660 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
661 nfs_show_mount_options(m, nfss, 1);
662
67ec9f46
CL
663 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
664
d9ef5a8c
CL
665 seq_printf(m, "\n\tcaps:\t");
666 seq_printf(m, "caps=0x%x", nfss->caps);
667 seq_printf(m, ",wtmult=%d", nfss->wtmult);
668 seq_printf(m, ",dtsize=%d", nfss->dtsize);
669 seq_printf(m, ",bsize=%d", nfss->bsize);
670 seq_printf(m, ",namelen=%d", nfss->namelen);
671
672#ifdef CONFIG_NFS_V4
673 if (nfss->rpc_ops->version == 4) {
674 seq_printf(m, "\n\tnfsv4:\t");
675 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
676 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
677 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
678 }
679#endif
680
681 /*
682 * Display security flavor in effect for this mount
683 */
684 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
685 if (auth->au_flavor)
686 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
687
688 /*
689 * Display superblock I/O counters
690 */
b9d9506d 691 for_each_possible_cpu(cpu) {
d9ef5a8c
CL
692 struct nfs_iostats *stats;
693
d9ef5a8c
CL
694 preempt_disable();
695 stats = per_cpu_ptr(nfss->io_stats, cpu);
696
697 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
698 totals.events[i] += stats->events[i];
699 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
700 totals.bytes[i] += stats->bytes[i];
701
702 preempt_enable();
703 }
704
705 seq_printf(m, "\n\tevents:\t");
706 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
707 seq_printf(m, "%lu ", totals.events[i]);
708 seq_printf(m, "\n\tbytes:\t");
709 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
710 seq_printf(m, "%Lu ", totals.bytes[i]);
4ece3a2d
CL
711 seq_printf(m, "\n");
712
713 rpc_print_iostats(m, nfss->client);
d9ef5a8c 714
1da177e4
LT
715 return 0;
716}
717
29884df0
TM
718/**
719 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
720 */
721int nfs_sync_mapping(struct address_space *mapping)
722{
723 int ret;
724
725 if (mapping->nrpages == 0)
726 return 0;
727 unmap_mapping_range(mapping, 0, 0, 0);
28fd1298 728 ret = filemap_write_and_wait(mapping);
29884df0
TM
729 if (ret != 0)
730 goto out;
731 ret = nfs_wb_all(mapping->host);
732out:
733 return ret;
734}
735
1da177e4
LT
736/*
737 * Invalidate the local caches
738 */
b37b03b7 739static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT
740{
741 struct nfs_inode *nfsi = NFS_I(inode);
742 int mode = inode->i_mode;
743
91d5b470
CL
744 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
745
1da177e4
LT
746 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
747 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
748
749 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
750 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 751 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 752 else
55296809 753 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 754}
dc59250c 755
b37b03b7
TM
756void nfs_zap_caches(struct inode *inode)
757{
758 spin_lock(&inode->i_lock);
759 nfs_zap_caches_locked(inode);
dc59250c 760 spin_unlock(&inode->i_lock);
ada70d94
TM
761}
762
763static void nfs_zap_acl_cache(struct inode *inode)
764{
765 void (*clear_acl_cache)(struct inode *);
766
767 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
768 if (clear_acl_cache != NULL)
769 clear_acl_cache(inode);
dc59250c 770 spin_lock(&inode->i_lock);
55296809 771 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 772 spin_unlock(&inode->i_lock);
1da177e4
LT
773}
774
775/*
b37b03b7
TM
776 * Invalidate, but do not unhash, the inode.
777 * NB: must be called with inode->i_lock held!
1da177e4 778 */
b37b03b7 779static void nfs_invalidate_inode(struct inode *inode)
1da177e4 780{
b37b03b7
TM
781 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
782 nfs_zap_caches_locked(inode);
1da177e4
LT
783}
784
785struct nfs_find_desc {
786 struct nfs_fh *fh;
787 struct nfs_fattr *fattr;
788};
789
790/*
791 * In NFSv3 we can have 64bit inode numbers. In order to support
792 * this, and re-exported directories (also seen in NFSv2)
793 * we are forced to allow 2 different inodes to have the same
794 * i_ino.
795 */
796static int
797nfs_find_actor(struct inode *inode, void *opaque)
798{
799 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
800 struct nfs_fh *fh = desc->fh;
801 struct nfs_fattr *fattr = desc->fattr;
802
803 if (NFS_FILEID(inode) != fattr->fileid)
804 return 0;
805 if (nfs_compare_fh(NFS_FH(inode), fh))
806 return 0;
807 if (is_bad_inode(inode) || NFS_STALE(inode))
808 return 0;
809 return 1;
810}
811
812static int
813nfs_init_locked(struct inode *inode, void *opaque)
814{
815 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
816 struct nfs_fattr *fattr = desc->fattr;
817
818 NFS_FILEID(inode) = fattr->fileid;
819 nfs_copy_fh(NFS_FH(inode), desc->fh);
820 return 0;
821}
822
823/* Don't use READDIRPLUS on directories that we believe are too large */
824#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
825
826/*
827 * This is our front-end to iget that looks up inodes by file handle
828 * instead of inode number.
829 */
830struct inode *
831nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
832{
833 struct nfs_find_desc desc = {
834 .fh = fh,
835 .fattr = fattr
836 };
03f28e3a 837 struct inode *inode = ERR_PTR(-ENOENT);
1da177e4
LT
838 unsigned long hash;
839
840 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
841 goto out_no_inode;
842
843 if (!fattr->nlink) {
844 printk("NFS: Buggy server - nlink == 0!\n");
845 goto out_no_inode;
846 }
847
848 hash = nfs_fattr_to_ino_t(fattr);
849
03f28e3a
TM
850 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
851 if (inode == NULL) {
852 inode = ERR_PTR(-ENOMEM);
1da177e4 853 goto out_no_inode;
03f28e3a 854 }
1da177e4
LT
855
856 if (inode->i_state & I_NEW) {
857 struct nfs_inode *nfsi = NFS_I(inode);
858
859 /* We set i_ino for the few things that still rely on it,
860 * such as stat(2) */
861 inode->i_ino = hash;
862
863 /* We can't support update_atime(), since the server will reset it */
864 inode->i_flags |= S_NOATIME|S_NOCMTIME;
865 inode->i_mode = fattr->mode;
866 /* Why so? Because we want revalidate for devices/FIFOs, and
867 * that's precisely what we have in nfs_file_inode_operations.
868 */
92cfc62c 869 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT
870 if (S_ISREG(inode->i_mode)) {
871 inode->i_fop = &nfs_file_operations;
872 inode->i_data.a_ops = &nfs_file_aops;
873 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
874 } else if (S_ISDIR(inode->i_mode)) {
875 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
876 inode->i_fop = &nfs_dir_operations;
877 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
878 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 879 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT
880 } else if (S_ISLNK(inode->i_mode))
881 inode->i_op = &nfs_symlink_inode_operations;
882 else
883 init_special_inode(inode, inode->i_mode, fattr->rdev);
884
33801147
TM
885 nfsi->read_cache_jiffies = fattr->time_start;
886 nfsi->last_updated = jiffies;
1da177e4
LT
887 inode->i_atime = fattr->atime;
888 inode->i_mtime = fattr->mtime;
889 inode->i_ctime = fattr->ctime;
890 if (fattr->valid & NFS_ATTR_FATTR_V4)
891 nfsi->change_attr = fattr->change_attr;
892 inode->i_size = nfs_size_to_loff_t(fattr->size);
893 inode->i_nlink = fattr->nlink;
894 inode->i_uid = fattr->uid;
895 inode->i_gid = fattr->gid;
896 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
897 /*
898 * report the blocks in 512byte units
899 */
900 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
901 inode->i_blksize = inode->i_sb->s_blocksize;
902 } else {
903 inode->i_blocks = fattr->du.nfs2.blocks;
904 inode->i_blksize = fattr->du.nfs2.blocksize;
905 }
906 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
907 nfsi->attrtimeo_timestamp = jiffies;
908 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
909 nfsi->cache_access.cred = NULL;
910
911 unlock_new_inode(inode);
912 } else
913 nfs_refresh_inode(inode, fattr);
914 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
915 inode->i_sb->s_id,
916 (long long)NFS_FILEID(inode),
917 atomic_read(&inode->i_count));
918
919out:
920 return inode;
921
922out_no_inode:
03f28e3a 923 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
1da177e4
LT
924 goto out;
925}
926
927#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
928
929int
930nfs_setattr(struct dentry *dentry, struct iattr *attr)
931{
932 struct inode *inode = dentry->d_inode;
933 struct nfs_fattr fattr;
934 int error;
935
91d5b470
CL
936 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
937
1da177e4
LT
938 if (attr->ia_valid & ATTR_SIZE) {
939 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
940 attr->ia_valid &= ~ATTR_SIZE;
941 }
942
943 /* Optimization: if the end result is no change, don't RPC */
944 attr->ia_valid &= NFS_VALID_ATTRS;
945 if (attr->ia_valid == 0)
946 return 0;
947
948 lock_kernel();
949 nfs_begin_data_update(inode);
755c1e20
TM
950 /* Write all dirty data */
951 filemap_write_and_wait(inode->i_mapping);
952 nfs_wb_all(inode);
642ac549
TM
953 /*
954 * Return any delegations if we're going to change ACLs
955 */
956 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
957 nfs_inode_return_delegation(inode);
1da177e4 958 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 959 if (error == 0)
1da177e4 960 nfs_refresh_inode(inode, &fattr);
65e4308d
TM
961 nfs_end_data_update(inode);
962 unlock_kernel();
963 return error;
964}
965
966/**
967 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
968 * @inode: pointer to struct inode
969 * @attr: pointer to struct iattr
970 *
971 * Note: we do this in the *proc.c in order to ensure that
972 * it works for things like exclusive creates too.
973 */
974void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
975{
976 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 977 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM
978 int mode = attr->ia_mode & S_IALLUGO;
979 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT
980 inode->i_mode = mode;
981 }
982 if ((attr->ia_valid & ATTR_UID) != 0)
983 inode->i_uid = attr->ia_uid;
984 if ((attr->ia_valid & ATTR_GID) != 0)
985 inode->i_gid = attr->ia_gid;
dc59250c 986 spin_lock(&inode->i_lock);
55296809 987 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 988 spin_unlock(&inode->i_lock);
65e4308d
TM
989 }
990 if ((attr->ia_valid & ATTR_SIZE) != 0) {
91d5b470 991 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
65e4308d
TM
992 inode->i_size = attr->ia_size;
993 vmtruncate(inode, attr->ia_size);
994 }
1da177e4
LT
995}
996
412d582e
CL
997static int nfs_wait_schedule(void *word)
998{
999 if (signal_pending(current))
1000 return -ERESTARTSYS;
1001 schedule();
1002 return 0;
1003}
1004
1da177e4
LT
1005/*
1006 * Wait for the inode to get unlocked.
1da177e4 1007 */
412d582e 1008static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT
1009{
1010 struct rpc_clnt *clnt = NFS_CLIENT(inode);
1011 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 1012 sigset_t oldmask;
1da177e4 1013 int error;
412d582e 1014
412d582e
CL
1015 rpc_clnt_sigmask(clnt, &oldmask);
1016 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
1017 nfs_wait_schedule, TASK_INTERRUPTIBLE);
1018 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 1019
1da177e4
LT
1020 return error;
1021}
1022
412d582e
CL
1023static void nfs_wake_up_inode(struct inode *inode)
1024{
1025 struct nfs_inode *nfsi = NFS_I(inode);
1026
1027 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
1028 smp_mb__after_clear_bit();
1029 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
1030}
1031
1da177e4
LT
1032int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1033{
1034 struct inode *inode = dentry->d_inode;
55296809 1035 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT
1036 int err;
1037
70b9ecbd 1038 /* Flush out writes to the server in order to update c/mtime */
c42de9dd 1039 nfs_sync_inode_wait(inode, 0, 0, FLUSH_NOCOMMIT);
fc33a7bb
CH
1040
1041 /*
1042 * We may force a getattr if the user cares about atime.
1043 *
1044 * Note that we only have to check the vfsmount flags here:
1045 * - NFS always sets S_NOATIME by so checking it would give a
1046 * bogus result
1047 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1048 * no point in checking those.
1049 */
1050 if ((mnt->mnt_flags & MNT_NOATIME) ||
1051 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
1da177e4 1052 need_atime = 0;
fc33a7bb 1053
1da177e4
LT
1054 if (need_atime)
1055 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
1056 else
1057 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
1058 if (!err)
1059 generic_fillattr(inode, stat);
1060 return err;
1061}
1062
b92dccf6 1063static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
1da177e4
LT
1064{
1065 struct nfs_open_context *ctx;
1066
1067 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
1068 if (ctx != NULL) {
1069 atomic_set(&ctx->count, 1);
1070 ctx->dentry = dget(dentry);
b92dccf6 1071 ctx->vfsmnt = mntget(mnt);
1da177e4
LT
1072 ctx->cred = get_rpccred(cred);
1073 ctx->state = NULL;
1074 ctx->lockowner = current->files;
1075 ctx->error = 0;
00a92642 1076 ctx->dir_cookie = 0;
1da177e4
LT
1077 }
1078 return ctx;
1079}
1080
1081struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
1082{
1083 if (ctx != NULL)
1084 atomic_inc(&ctx->count);
1085 return ctx;
1086}
1087
1088void put_nfs_open_context(struct nfs_open_context *ctx)
1089{
1090 if (atomic_dec_and_test(&ctx->count)) {
1091 if (!list_empty(&ctx->list)) {
1092 struct inode *inode = ctx->dentry->d_inode;
1093 spin_lock(&inode->i_lock);
1094 list_del(&ctx->list);
1095 spin_unlock(&inode->i_lock);
1096 }
1097 if (ctx->state != NULL)
1098 nfs4_close_state(ctx->state, ctx->mode);
1099 if (ctx->cred != NULL)
1100 put_rpccred(ctx->cred);
1101 dput(ctx->dentry);
b92dccf6 1102 mntput(ctx->vfsmnt);
1da177e4
LT
1103 kfree(ctx);
1104 }
1105}
1106
1107/*
1108 * Ensure that mmap has a recent RPC credential for use when writing out
1109 * shared pages
1110 */
b92dccf6 1111static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1da177e4
LT
1112{
1113 struct inode *inode = filp->f_dentry->d_inode;
1114 struct nfs_inode *nfsi = NFS_I(inode);
1115
1116 filp->private_data = get_nfs_open_context(ctx);
1117 spin_lock(&inode->i_lock);
1118 list_add(&ctx->list, &nfsi->open_files);
1119 spin_unlock(&inode->i_lock);
1120}
1121
d530838b
TM
1122/*
1123 * Given an inode, search for an open context with the desired characteristics
1124 */
1125struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT
1126{
1127 struct nfs_inode *nfsi = NFS_I(inode);
1128 struct nfs_open_context *pos, *ctx = NULL;
1129
1130 spin_lock(&inode->i_lock);
1131 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM
1132 if (cred != NULL && pos->cred != cred)
1133 continue;
1da177e4
LT
1134 if ((pos->mode & mode) == mode) {
1135 ctx = get_nfs_open_context(pos);
1136 break;
1137 }
1138 }
1139 spin_unlock(&inode->i_lock);
1140 return ctx;
1141}
1142
b92dccf6 1143static void nfs_file_clear_open_context(struct file *filp)
1da177e4
LT
1144{
1145 struct inode *inode = filp->f_dentry->d_inode;
1146 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1147
1148 if (ctx) {
1149 filp->private_data = NULL;
1150 spin_lock(&inode->i_lock);
1151 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1152 spin_unlock(&inode->i_lock);
1153 put_nfs_open_context(ctx);
1154 }
1155}
1156
1157/*
1158 * These allocate and release file read/write context information.
1159 */
1160int nfs_open(struct inode *inode, struct file *filp)
1161{
1162 struct nfs_open_context *ctx;
1163 struct rpc_cred *cred;
1164
1165 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1166 if (IS_ERR(cred))
1167 return PTR_ERR(cred);
b92dccf6 1168 ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
1da177e4
LT
1169 put_rpccred(cred);
1170 if (ctx == NULL)
1171 return -ENOMEM;
1172 ctx->mode = filp->f_mode;
1173 nfs_file_set_open_context(filp, ctx);
1174 put_nfs_open_context(ctx);
1da177e4
LT
1175 return 0;
1176}
1177
1178int nfs_release(struct inode *inode, struct file *filp)
1179{
1da177e4
LT
1180 nfs_file_clear_open_context(filp);
1181 return 0;
1182}
1183
1184/*
1185 * This function is called whenever some part of NFS notices that
1186 * the cached attributes have to be refreshed.
1187 */
1188int
1189__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1190{
1191 int status = -ESTALE;
1192 struct nfs_fattr fattr;
1193 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4
LT
1194
1195 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1196 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1197
1842bfb4 1198 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
1da177e4
LT
1199 lock_kernel();
1200 if (!inode || is_bad_inode(inode))
1201 goto out_nowait;
1202 if (NFS_STALE(inode))
1203 goto out_nowait;
1204
412d582e
CL
1205 status = nfs_wait_on_inode(inode);
1206 if (status < 0)
1207 goto out;
1208 if (NFS_STALE(inode)) {
1209 status = -ESTALE;
1210 /* Do we trust the cached ESTALE? */
1211 if (NFS_ATTRTIMEO(inode) != 0) {
44b11874 1212 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME)) {
412d582e
CL
1213 /* no */
1214 } else
1215 goto out;
1216 }
1da177e4 1217 }
1da177e4 1218
1da177e4
LT
1219 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1220 if (status != 0) {
1221 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1222 inode->i_sb->s_id,
1223 (long long)NFS_FILEID(inode), status);
1224 if (status == -ESTALE) {
1225 nfs_zap_caches(inode);
1226 if (!S_ISDIR(inode->i_mode))
412d582e 1227 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT
1228 }
1229 goto out;
1230 }
1231
33801147 1232 spin_lock(&inode->i_lock);
24aa1fe6 1233 status = nfs_update_inode(inode, &fattr);
1da177e4 1234 if (status) {
33801147 1235 spin_unlock(&inode->i_lock);
1da177e4
LT
1236 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1237 inode->i_sb->s_id,
1238 (long long)NFS_FILEID(inode), status);
1239 goto out;
1240 }
dc59250c 1241 spin_unlock(&inode->i_lock);
55296809 1242
24aa1fe6 1243 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1244 nfs_zap_acl_cache(inode);
55296809 1245
1da177e4
LT
1246 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1247 inode->i_sb->s_id,
1248 (long long)NFS_FILEID(inode));
1249
412d582e
CL
1250 out:
1251 nfs_wake_up_inode(inode);
1252
1da177e4
LT
1253 out_nowait:
1254 unlock_kernel();
1255 return status;
1256}
1257
1258int nfs_attribute_timeout(struct inode *inode)
1259{
1260 struct nfs_inode *nfsi = NFS_I(inode);
1261
1262 if (nfs_have_delegation(inode, FMODE_READ))
1263 return 0;
1264 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1265}
1266
1267/**
1268 * nfs_revalidate_inode - Revalidate the inode attributes
1269 * @server - pointer to nfs_server struct
1270 * @inode - pointer to inode struct
1271 *
1272 * Updates inode attribute information by retrieving the data from the server.
1273 */
1274int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1275{
44b11874 1276 if (!(NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATTR)
1da177e4
LT
1277 && !nfs_attribute_timeout(inode))
1278 return NFS_STALE(inode) ? -ESTALE : 0;
1279 return __nfs_revalidate_inode(server, inode);
1280}
1281
7d52e862
TM
1282/**
1283 * nfs_revalidate_mapping - Revalidate the pagecache
1284 * @inode - pointer to host inode
1285 * @mapping - pointer to mapping
1286 */
44b11874 1287int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
7d52e862
TM
1288{
1289 struct nfs_inode *nfsi = NFS_I(inode);
44b11874
TM
1290 int ret = 0;
1291
1292 if (NFS_STALE(inode))
1293 ret = -ESTALE;
1294 if ((nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
1295 || nfs_attribute_timeout(inode))
1296 ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
7d52e862 1297
55296809 1298 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
91d5b470 1299 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
29884df0
TM
1300 if (S_ISREG(inode->i_mode))
1301 nfs_sync_mapping(mapping);
7d52e862 1302 invalidate_inode_pages2(mapping);
dc59250c
CL
1303
1304 spin_lock(&inode->i_lock);
55296809 1305 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM
1306 if (S_ISDIR(inode->i_mode)) {
1307 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1308 /* This ensures we revalidate child dentries */
913a70fc 1309 nfsi->cache_change_attribute = jiffies;
7d52e862 1310 }
dc59250c
CL
1311 spin_unlock(&inode->i_lock);
1312
7d52e862
TM
1313 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1314 inode->i_sb->s_id,
1315 (long long)NFS_FILEID(inode));
1316 }
44b11874 1317 return ret;
7d52e862
TM
1318}
1319
1da177e4
LT
1320/**
1321 * nfs_begin_data_update
1322 * @inode - pointer to inode
1323 * Declare that a set of operations will update file data on the server
1324 */
1325void nfs_begin_data_update(struct inode *inode)
1326{
1327 atomic_inc(&NFS_I(inode)->data_updates);
1328}
1329
1330/**
1331 * nfs_end_data_update
1332 * @inode - pointer to inode
1333 * Declare end of the operations that will update file data
1334 * This will mark the inode as immediately needing revalidation
1335 * of its attribute cache.
1336 */
1337void nfs_end_data_update(struct inode *inode)
1338{
1339 struct nfs_inode *nfsi = NFS_I(inode);
1340
1341 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM
1342 /* Directories and symlinks: invalidate page cache */
1343 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1344 spin_lock(&inode->i_lock);
55296809 1345 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM
1346 spin_unlock(&inode->i_lock);
1347 }
1da177e4 1348 }
913a70fc 1349 nfsi->cache_change_attribute = jiffies;
1da177e4
LT
1350 atomic_dec(&nfsi->data_updates);
1351}
1352
a895b4a1
TM
1353static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1354{
1355 struct nfs_inode *nfsi = NFS_I(inode);
1356
a895b4a1
TM
1357 /* If we have atomic WCC data, we may update some attributes */
1358 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1359 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
1360 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1361 nfsi->cache_change_attribute = jiffies;
1362 }
1363 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
1364 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1365 nfsi->cache_change_attribute = jiffies;
1366 }
1367 if (inode->i_size == fattr->pre_size && nfsi->npages == 0) {
1368 inode->i_size = fattr->size;
1369 nfsi->cache_change_attribute = jiffies;
1370 }
1371 }
1372}
1373
1da177e4 1374/**
33801147 1375 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT
1376 * @inode - pointer to inode
1377 * @fattr - updated attributes
1378 *
1379 * Verifies the attribute cache. If we have just changed the attributes,
1380 * so that fattr carries weak cache consistency data, then it may
1381 * also update the ctime/mtime/change_attribute.
1382 */
33801147 1383static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT
1384{
1385 struct nfs_inode *nfsi = NFS_I(inode);
1386 loff_t cur_size, new_isize;
1387 int data_unstable;
1388
dc59250c 1389
ca62b9c3
TM
1390 /* Has the inode gone and changed behind our back? */
1391 if (nfsi->fileid != fattr->fileid
1392 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1393 return -EIO;
1394 }
1395
1da177e4
LT
1396 /* Are we in the process of updating data on the server? */
1397 data_unstable = nfs_caches_unstable(inode);
1398
a895b4a1
TM
1399 /* Do atomic weak cache consistency updates */
1400 nfs_wcc_update_inode(inode, fattr);
1da177e4 1401
9d1e9232 1402 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 &&
f1bb0b92
TM
1403 nfsi->change_attr != fattr->change_attr)
1404 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4 1405
1da177e4 1406 /* Verify a few of the more important attributes */
f1bb0b92
TM
1407 if (!timespec_equal(&inode->i_mtime, &fattr->mtime))
1408 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
ca62b9c3
TM
1409
1410 cur_size = i_size_read(inode);
1411 new_isize = nfs_size_to_loff_t(fattr->size);
fb374d24
TM
1412 if (cur_size != new_isize && nfsi->npages == 0)
1413 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4
LT
1414
1415 /* Have any file permissions changed? */
1416 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1417 || inode->i_uid != fattr->uid
1418 || inode->i_gid != fattr->gid)
55296809 1419 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT
1420
1421 /* Has the link count changed? */
1422 if (inode->i_nlink != fattr->nlink)
55296809 1423 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT
1424
1425 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1426 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1427
33801147 1428 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT
1429 return 0;
1430}
1431
33801147
TM
1432/**
1433 * nfs_refresh_inode - try to update the inode attribute cache
1434 * @inode - pointer to inode
1435 * @fattr - updated attributes
1436 *
1437 * Check that an RPC call that returned attributes has not overlapped with
1438 * other recent updates of the inode metadata, then decide whether it is
1439 * safe to do a full update of the inode attributes, or whether just to
1440 * call nfs_check_inode_attributes.
1441 */
1442int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1443{
1444 struct nfs_inode *nfsi = NFS_I(inode);
1445 int status;
1446
1447 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1448 return 0;
1449 spin_lock(&inode->i_lock);
33801147 1450 if (time_after(fattr->time_start, nfsi->last_updated))
24aa1fe6 1451 status = nfs_update_inode(inode, fattr);
33801147
TM
1452 else
1453 status = nfs_check_inode_attributes(inode, fattr);
1454
1455 spin_unlock(&inode->i_lock);
1456 return status;
1457}
1458
decf491f
TM
1459/**
1460 * nfs_post_op_update_inode - try to update the inode attribute cache
1461 * @inode - pointer to inode
1462 * @fattr - updated attributes
1463 *
1464 * After an operation that has changed the inode metadata, mark the
1465 * attribute cache as being invalid, then try to update it.
1466 */
1467int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1468{
1469 struct nfs_inode *nfsi = NFS_I(inode);
1470 int status = 0;
1471
1472 spin_lock(&inode->i_lock);
1473 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
f1bb0b92 1474 nfsi->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
decf491f
TM
1475 goto out;
1476 }
24aa1fe6 1477 status = nfs_update_inode(inode, fattr);
decf491f
TM
1478out:
1479 spin_unlock(&inode->i_lock);
1480 return status;
1481}
1482
1da177e4
LT
1483/*
1484 * Many nfs protocol calls return the new file attributes after
1485 * an operation. Here we update the inode to reflect the state
1486 * of the server's inode.
1487 *
1488 * This is a bit tricky because we have to make sure all dirty pages
1489 * have been sent off to the server before calling invalidate_inode_pages.
1490 * To make sure no other process adds more write requests while we try
1491 * our best to flush them, we make them sleep during the attribute refresh.
1492 *
1493 * A very similar scenario holds for the dir cache.
1494 */
24aa1fe6 1495static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1da177e4 1496{
8b4bdcf8 1497 struct nfs_server *server;
1da177e4 1498 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1499 loff_t cur_isize, new_isize;
1da177e4 1500 unsigned int invalid = 0;
24aa1fe6 1501 int data_stable;
1da177e4
LT
1502
1503 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1504 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1505 atomic_read(&inode->i_count), fattr->valid);
1506
325cfed9
CL
1507 if (nfsi->fileid != fattr->fileid)
1508 goto out_fileid;
1da177e4
LT
1509
1510 /*
1511 * Make sure the inode's type hasn't changed.
1512 */
33801147 1513 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT
1514 goto out_changed;
1515
8b4bdcf8
TM
1516 server = NFS_SERVER(inode);
1517 /* Update the fsid if and only if this is the root directory */
1518 if (inode == inode->i_sb->s_root->d_inode
1519 && !nfs_fsid_equal(&server->fsid, &fattr->fsid))
1520 server->fsid = fattr->fsid;
1521
1da177e4
LT
1522 /*
1523 * Update the read time so we don't revalidate too often.
1524 */
33801147
TM
1525 nfsi->read_cache_jiffies = fattr->time_start;
1526 nfsi->last_updated = jiffies;
1da177e4
LT
1527
1528 /* Are we racing with known updates of the metadata on the server? */
24aa1fe6
TM
1529 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1530 if (data_stable)
f1bb0b92 1531 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE|NFS_INO_INVALID_ATIME);
1da177e4 1532
a895b4a1
TM
1533 /* Do atomic weak cache consistency updates */
1534 nfs_wcc_update_inode(inode, fattr);
1535
951a143b 1536 /* Check if our cached file size is stale */
1da177e4
LT
1537 new_isize = nfs_size_to_loff_t(fattr->size);
1538 cur_isize = i_size_read(inode);
951a143b
TM
1539 if (new_isize != cur_isize) {
1540 /* Do we perhaps have any outstanding writes? */
1541 if (nfsi->npages == 0) {
1542 /* No, but did we race with nfs_end_data_update()? */
24aa1fe6 1543 if (data_stable) {
1da177e4 1544 inode->i_size = new_isize;
951a143b 1545 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1546 }
951a143b
TM
1547 invalid |= NFS_INO_INVALID_ATTR;
1548 } else if (new_isize > cur_isize) {
1da177e4
LT
1549 inode->i_size = new_isize;
1550 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1551 }
24aa1fe6 1552 nfsi->cache_change_attribute = jiffies;
951a143b
TM
1553 dprintk("NFS: isize change on server for file %s/%ld\n",
1554 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT
1555 }
1556
951a143b 1557 /* Check if the mtime agrees */
1da177e4
LT
1558 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1559 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM
1560 dprintk("NFS: mtime change on server for file %s/%ld\n",
1561 inode->i_sb->s_id, inode->i_ino);
24aa1fe6
TM
1562 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1563 nfsi->cache_change_attribute = jiffies;
1da177e4
LT
1564 }
1565
ada70d94
TM
1566 /* If ctime has changed we should definitely clear access+acl caches */
1567 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
24aa1fe6 1568 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
ada70d94 1569 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
24aa1fe6 1570 nfsi->cache_change_attribute = jiffies;
ada70d94 1571 }
1da177e4
LT
1572 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1573
1574 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1575 inode->i_uid != fattr->uid ||
1576 inode->i_gid != fattr->gid)
ada70d94 1577 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT
1578
1579 inode->i_mode = fattr->mode;
1580 inode->i_nlink = fattr->nlink;
1581 inode->i_uid = fattr->uid;
1582 inode->i_gid = fattr->gid;
1583
1584 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1585 /*
1586 * report the blocks in 512byte units
1587 */
1588 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1589 inode->i_blksize = inode->i_sb->s_blocksize;
1590 } else {
1591 inode->i_blocks = fattr->du.nfs2.blocks;
1592 inode->i_blksize = fattr->du.nfs2.blocksize;
1593 }
1594
9d1e9232
TM
1595 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0 &&
1596 nfsi->change_attr != fattr->change_attr) {
1597 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1598 inode->i_sb->s_id, inode->i_ino);
1599 nfsi->change_attr = fattr->change_attr;
1600 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1601 nfsi->cache_change_attribute = jiffies;
ca62b9c3
TM
1602 }
1603
1da177e4
LT
1604 /* Update attrtimeo value if we're out of the unstable period */
1605 if (invalid & NFS_INO_INVALID_ATTR) {
91d5b470 1606 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
1da177e4
LT
1607 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1608 nfsi->attrtimeo_timestamp = jiffies;
1609 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1610 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1611 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1612 nfsi->attrtimeo_timestamp = jiffies;
1613 }
1614 /* Don't invalidate the data if we were to blame */
1615 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1616 || S_ISLNK(inode->i_mode)))
1617 invalid &= ~NFS_INO_INVALID_DATA;
24aa1fe6
TM
1618 if (data_stable)
1619 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1da177e4 1620 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1621 nfsi->cache_validity |= invalid;
1da177e4
LT
1622
1623 return 0;
1624 out_changed:
1625 /*
1626 * Big trouble! The inode has become a different object.
1627 */
1628#ifdef NFS_PARANOIA
1629 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1630 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1631#endif
b37b03b7 1632 out_err:
1da177e4
LT
1633 /*
1634 * No need to worry about unhashing the dentry, as the
1635 * lookup validation will know that the inode is bad.
1636 * (But we fall through to invalidate the caches.)
1637 */
1638 nfs_invalidate_inode(inode);
1da177e4 1639 return -ESTALE;
325cfed9
CL
1640
1641 out_fileid:
1642 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1643 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1644 NFS_SERVER(inode)->hostname, inode->i_sb->s_id,
1645 (long long)nfsi->fileid, (long long)fattr->fileid);
1646 goto out_err;
1da177e4
LT
1647}
1648
1649/*
1650 * File system information
1651 */
1652
1653static int nfs_set_super(struct super_block *s, void *data)
1654{
1655 s->s_fs_info = data;
1656 return set_anon_super(s, data);
1657}
1658
1659static int nfs_compare_super(struct super_block *sb, void *data)
1660{
1661 struct nfs_server *server = data;
1662 struct nfs_server *old = NFS_SB(sb);
1663
1664 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1665 return 0;
1666 if (old->addr.sin_port != server->addr.sin_port)
1667 return 0;
1668 return !nfs_compare_fh(&old->fh, &server->fh);
1669}
1670
1671static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1672 int flags, const char *dev_name, void *raw_data)
1673{
1674 int error;
9085bbcb 1675 struct nfs_server *server = NULL;
1da177e4
LT
1676 struct super_block *s;
1677 struct nfs_fh *root;
1678 struct nfs_mount_data *data = raw_data;
1679
9085bbcb
TM
1680 s = ERR_PTR(-EINVAL);
1681 if (data == NULL) {
1682 dprintk("%s: missing data argument\n", __FUNCTION__);
1683 goto out_err;
1684 }
1685 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1686 dprintk("%s: bad mount version\n", __FUNCTION__);
1687 goto out_err;
1da177e4 1688 }
9085bbcb
TM
1689 switch (data->version) {
1690 case 1:
1691 data->namlen = 0;
1692 case 2:
1693 data->bsize = 0;
1694 case 3:
1695 if (data->flags & NFS_MOUNT_VER3) {
1696 dprintk("%s: mount structure version %d does not support NFSv3\n",
1697 __FUNCTION__,
1698 data->version);
1699 goto out_err;
1700 }
1701 data->root.size = NFS2_FHSIZE;
1702 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1703 case 4:
1704 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1705 dprintk("%s: mount structure version %d does not support strong security\n",
1706 __FUNCTION__,
1707 data->version);
1708 goto out_err;
1709 }
1710 case 5:
1711 memset(data->context, 0, sizeof(data->context));
1712 }
1713#ifndef CONFIG_NFS_V3
1714 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1715 s = ERR_PTR(-EPROTONOSUPPORT);
1716 if (data->flags & NFS_MOUNT_VER3) {
1717 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1718 goto out_err;
1719 }
1720#endif /* CONFIG_NFS_V3 */
1da177e4 1721
9085bbcb 1722 s = ERR_PTR(-ENOMEM);
bd647545 1723 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4 1724 if (!server)
9085bbcb 1725 goto out_err;
1da177e4
LT
1726 /* Zero out the NFS state stuff */
1727 init_nfsv4_state(server);
b7fa0554 1728 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1729
1da177e4
LT
1730 root = &server->fh;
1731 if (data->flags & NFS_MOUNT_VER3)
1732 root->size = data->root.size;
1733 else
1734 root->size = NFS2_FHSIZE;
9085bbcb 1735 s = ERR_PTR(-EINVAL);
1da177e4 1736 if (root->size > sizeof(root->data)) {
9085bbcb
TM
1737 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1738 goto out_err;
1da177e4
LT
1739 }
1740 memcpy(root->data, data->root.data, root->size);
1741
1742 /* We now require that the mount process passes the remote address */
1743 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1744 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM
1745 dprintk("%s: mount program didn't pass remote address!\n",
1746 __FUNCTION__);
1747 goto out_err;
1da177e4
LT
1748 }
1749
9085bbcb
TM
1750 /* Fire up rpciod if not yet running */
1751 s = ERR_PTR(rpciod_up());
1752 if (IS_ERR(s)) {
1753 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1754 __FUNCTION__, PTR_ERR(s));
1755 goto out_err;
1da177e4
LT
1756 }
1757
9085bbcb
TM
1758 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1759 if (IS_ERR(s) || s->s_root)
1760 goto out_rpciod_down;
1da177e4 1761
9085bbcb 1762 s->s_flags = flags;
1da177e4 1763
9b04c997 1764 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT
1765 if (error) {
1766 up_write(&s->s_umount);
1767 deactivate_super(s);
1768 return ERR_PTR(error);
1769 }
1770 s->s_flags |= MS_ACTIVE;
1771 return s;
9085bbcb
TM
1772out_rpciod_down:
1773 rpciod_down();
1774out_err:
1775 kfree(server);
1776 return s;
1da177e4
LT
1777}
1778
1779static void nfs_kill_super(struct super_block *s)
1780{
1781 struct nfs_server *server = NFS_SB(s);
1782
1783 kill_anon_super(s);
1784
6a19275a 1785 if (!IS_ERR(server->client))
1da177e4 1786 rpc_shutdown_client(server->client);
6a19275a 1787 if (!IS_ERR(server->client_sys))
1da177e4 1788 rpc_shutdown_client(server->client_sys);
b7fa0554
AG
1789 if (!IS_ERR(server->client_acl))
1790 rpc_shutdown_client(server->client_acl);
1da177e4
LT
1791
1792 if (!(server->flags & NFS_MOUNT_NONLM))
1793 lockd_down(); /* release rpc.lockd */
1794
1795 rpciod_down(); /* release rpciod */
1796
01d0ae8b 1797 nfs_free_iostats(server->io_stats);
f99d49ad 1798 kfree(server->hostname);
1da177e4
LT
1799 kfree(server);
1800}
1801
1802static struct file_system_type nfs_fs_type = {
1803 .owner = THIS_MODULE,
1804 .name = "nfs",
1805 .get_sb = nfs_get_sb,
1806 .kill_sb = nfs_kill_super,
1807 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1808};
1809
1810#ifdef CONFIG_NFS_V4
1811
1812static void nfs4_clear_inode(struct inode *);
1813
1814
1815static struct super_operations nfs4_sops = {
1816 .alloc_inode = nfs_alloc_inode,
1817 .destroy_inode = nfs_destroy_inode,
1818 .write_inode = nfs_write_inode,
1da177e4
LT
1819 .statfs = nfs_statfs,
1820 .clear_inode = nfs4_clear_inode,
1821 .umount_begin = nfs_umount_begin,
1822 .show_options = nfs_show_options,
d9ef5a8c 1823 .show_stats = nfs_show_stats,
1da177e4
LT
1824};
1825
1826/*
1827 * Clean out any remaining NFSv4 state that might be left over due
1828 * to open() calls that passed nfs_atomic_lookup, but failed to call
1829 * nfs_open().
1830 */
1831static void nfs4_clear_inode(struct inode *inode)
1832{
1833 struct nfs_inode *nfsi = NFS_I(inode);
1834
1835 /* If we are holding a delegation, return it! */
cae7a073 1836 nfs_inode_return_delegation(inode);
1da177e4
LT
1837 /* First call standard NFS clear_inode() code */
1838 nfs_clear_inode(inode);
1839 /* Now clear out any remaining state */
1840 while (!list_empty(&nfsi->open_states)) {
1841 struct nfs4_state *state;
1842
1843 state = list_entry(nfsi->open_states.next,
1844 struct nfs4_state,
1845 inode_states);
1846 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1847 __FUNCTION__,
1848 inode->i_sb->s_id,
1849 (long long)NFS_FILEID(inode),
1850 state);
1851 BUG_ON(atomic_read(&state->count) != 1);
1852 nfs4_close_state(state, state->state);
1853 }
1854}
1855
1856
1857static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1858{
1859 struct nfs_server *server;
1860 struct nfs4_client *clp = NULL;
1861 struct rpc_xprt *xprt = NULL;
1862 struct rpc_clnt *clnt = NULL;
1863 struct rpc_timeout timeparms;
1864 rpc_authflavor_t authflavour;
eab5c084 1865 int err = -EIO;
1da177e4
LT
1866
1867 sb->s_blocksize_bits = 0;
1868 sb->s_blocksize = 0;
1869 server = NFS_SB(sb);
1870 if (data->rsize != 0)
1871 server->rsize = nfs_block_size(data->rsize, NULL);
1872 if (data->wsize != 0)
1873 server->wsize = nfs_block_size(data->wsize, NULL);
1874 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1875 server->caps = NFS_CAP_ATOMIC_OPEN;
1876
1877 server->acregmin = data->acregmin*HZ;
1878 server->acregmax = data->acregmax*HZ;
1879 server->acdirmin = data->acdirmin*HZ;
1880 server->acdirmax = data->acdirmax*HZ;
1881
1882 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1883
eab5c084 1884 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4 1885
7a480e25
CL
1886 server->retrans_timeo = timeparms.to_initval;
1887 server->retrans_count = timeparms.to_retries;
1888
1da177e4
LT
1889 clp = nfs4_get_client(&server->addr.sin_addr);
1890 if (!clp) {
9085bbcb 1891 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT
1892 return -EIO;
1893 }
1894
1895 /* Now create transport and client */
1896 authflavour = RPC_AUTH_UNIX;
1897 if (data->auth_flavourlen != 0) {
9085bbcb
TM
1898 if (data->auth_flavourlen != 1) {
1899 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1900 __FUNCTION__, data->auth_flavourlen);
1901 err = -EINVAL;
1902 goto out_fail;
1903 }
1da177e4
LT
1904 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1905 err = -EFAULT;
1906 goto out_fail;
1907 }
1908 }
1909
1910 down_write(&clp->cl_sem);
6a19275a 1911 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1912 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT
1913 if (IS_ERR(xprt)) {
1914 up_write(&clp->cl_sem);
1da177e4 1915 err = PTR_ERR(xprt);
9085bbcb
TM
1916 dprintk("%s: cannot create RPC transport. Error = %d\n",
1917 __FUNCTION__, err);
1da177e4
LT
1918 goto out_fail;
1919 }
1920 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1921 server->rpc_ops->version, authflavour);
1922 if (IS_ERR(clnt)) {
1923 up_write(&clp->cl_sem);
1da177e4 1924 err = PTR_ERR(clnt);
9085bbcb
TM
1925 dprintk("%s: cannot create RPC client. Error = %d\n",
1926 __FUNCTION__, err);
1da177e4
LT
1927 goto out_fail;
1928 }
1929 clnt->cl_intr = 1;
1930 clnt->cl_softrtry = 1;
1da177e4 1931 clp->cl_rpcclient = clnt;
1da177e4
LT
1932 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1933 nfs_idmap_new(clp);
1934 }
1da177e4
LT
1935 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1936 clnt = rpc_clone_client(clp->cl_rpcclient);
1937 if (!IS_ERR(clnt))
1938 server->nfs4_state = clp;
1939 up_write(&clp->cl_sem);
1940 clp = NULL;
1941
1942 if (IS_ERR(clnt)) {
9085bbcb
TM
1943 err = PTR_ERR(clnt);
1944 dprintk("%s: cannot create RPC client. Error = %d\n",
1945 __FUNCTION__, err);
1946 return err;
1da177e4
LT
1947 }
1948
1949 server->client = clnt;
1950
1951 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1952 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT
1953 return -ENOMEM;
1954 }
1955
1956 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF
1957 struct rpc_auth *auth;
1958
1959 auth = rpcauth_create(authflavour, clnt);
1960 if (IS_ERR(auth)) {
9085bbcb 1961 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1962 return PTR_ERR(auth);
1da177e4
LT
1963 }
1964 }
1965
1966 sb->s_time_gran = 1;
1967
1968 sb->s_op = &nfs4_sops;
1969 err = nfs_sb_init(sb, authflavour);
1970 if (err == 0)
1971 return 0;
1972out_fail:
1973 if (clp)
1974 nfs4_put_client(clp);
1975 return err;
1976}
1977
1978static int nfs4_compare_super(struct super_block *sb, void *data)
1979{
1980 struct nfs_server *server = data;
1981 struct nfs_server *old = NFS_SB(sb);
1982
1983 if (strcmp(server->hostname, old->hostname) != 0)
1984 return 0;
1985 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1986 return 0;
1987 return 1;
1988}
1989
1990static void *
1991nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1992{
1993 void *p = NULL;
1994
1995 if (!src->len)
1996 return ERR_PTR(-EINVAL);
1997 if (src->len < maxlen)
1998 maxlen = src->len;
1999 if (dst == NULL) {
2000 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
2001 if (p == NULL)
2002 return ERR_PTR(-ENOMEM);
2003 }
2004 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 2005 kfree(p);
1da177e4
LT
2006 return ERR_PTR(-EFAULT);
2007 }
2008 dst[maxlen] = '\0';
2009 return dst;
2010}
2011
2012static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
2013 int flags, const char *dev_name, void *raw_data)
2014{
2015 int error;
2016 struct nfs_server *server;
2017 struct super_block *s;
2018 struct nfs4_mount_data *data = raw_data;
2019 void *p;
2020
9085bbcb
TM
2021 if (data == NULL) {
2022 dprintk("%s: missing data argument\n", __FUNCTION__);
2023 return ERR_PTR(-EINVAL);
2024 }
2025 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
2026 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT
2027 return ERR_PTR(-EINVAL);
2028 }
2029
bd647545 2030 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4
LT
2031 if (!server)
2032 return ERR_PTR(-ENOMEM);
1da177e4
LT
2033 /* Zero out the NFS state stuff */
2034 init_nfsv4_state(server);
b7fa0554 2035 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 2036
1da177e4
LT
2037 p = nfs_copy_user_string(NULL, &data->hostname, 256);
2038 if (IS_ERR(p))
2039 goto out_err;
2040 server->hostname = p;
2041
2042 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
2043 if (IS_ERR(p))
2044 goto out_err;
2045 server->mnt_path = p;
2046
2047 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
2048 sizeof(server->ip_addr) - 1);
2049 if (IS_ERR(p))
2050 goto out_err;
2051
2052 /* We now require that the mount process passes the remote address */
2053 if (data->host_addrlen != sizeof(server->addr)) {
2054 s = ERR_PTR(-EINVAL);
2055 goto out_free;
2056 }
2057 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
2058 s = ERR_PTR(-EFAULT);
2059 goto out_free;
2060 }
2061 if (server->addr.sin_family != AF_INET ||
2062 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM
2063 dprintk("%s: mount program didn't pass remote IP address!\n",
2064 __FUNCTION__);
1da177e4
LT
2065 s = ERR_PTR(-EINVAL);
2066 goto out_free;
2067 }
2068
9085bbcb
TM
2069 /* Fire up rpciod if not yet running */
2070 s = ERR_PTR(rpciod_up());
2071 if (IS_ERR(s)) {
2072 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2073 __FUNCTION__, PTR_ERR(s));
2074 goto out_free;
2075 }
2076
1da177e4
LT
2077 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
2078
2079 if (IS_ERR(s) || s->s_root)
2080 goto out_free;
2081
2082 s->s_flags = flags;
2083
9b04c997 2084 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1da177e4
LT
2085 if (error) {
2086 up_write(&s->s_umount);
2087 deactivate_super(s);
2088 return ERR_PTR(error);
2089 }
2090 s->s_flags |= MS_ACTIVE;
2091 return s;
2092out_err:
2093 s = (struct super_block *)p;
2094out_free:
f99d49ad
JJ
2095 kfree(server->mnt_path);
2096 kfree(server->hostname);
1da177e4
LT
2097 kfree(server);
2098 return s;
2099}
2100
2101static void nfs4_kill_super(struct super_block *sb)
2102{
2103 struct nfs_server *server = NFS_SB(sb);
2104
2105 nfs_return_all_delegations(sb);
2106 kill_anon_super(sb);
2107
2108 nfs4_renewd_prepare_shutdown(server);
2109
2110 if (server->client != NULL && !IS_ERR(server->client))
2111 rpc_shutdown_client(server->client);
1da177e4
LT
2112
2113 destroy_nfsv4_state(server);
2114
967b9281
TM
2115 rpciod_down();
2116
01d0ae8b 2117 nfs_free_iostats(server->io_stats);
f99d49ad 2118 kfree(server->hostname);
1da177e4
LT
2119 kfree(server);
2120}
2121
2122static struct file_system_type nfs4_fs_type = {
2123 .owner = THIS_MODULE,
2124 .name = "nfs4",
2125 .get_sb = nfs4_get_sb,
2126 .kill_sb = nfs4_kill_super,
2127 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2128};
2129
a72b4422
TM
2130static const int nfs_set_port_min = 0;
2131static const int nfs_set_port_max = 65535;
2132static int param_set_port(const char *val, struct kernel_param *kp)
2133{
2134 char *endp;
2135 int num = simple_strtol(val, &endp, 0);
2136 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
2137 return -EINVAL;
2138 *((int *)kp->arg) = num;
2139 return 0;
2140}
2141
2142module_param_call(callback_tcpport, param_set_port, param_get_int,
2143 &nfs_callback_set_tcpport, 0644);
2144
58df095b
TM
2145static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
2146{
2147 char *endp;
2148 int num = simple_strtol(val, &endp, 0);
2149 int jif = num * HZ;
2150 if (endp == val || *endp || num < 0 || jif < num)
2151 return -EINVAL;
2152 *((int *)kp->arg) = jif;
2153 return 0;
2154}
2155
2156module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
2157 &nfs_idmap_cache_timeout, 0644);
2158
1da177e4
LT
2159#define nfs4_init_once(nfsi) \
2160 do { \
2161 INIT_LIST_HEAD(&(nfsi)->open_states); \
2162 nfsi->delegation = NULL; \
2163 nfsi->delegation_state = 0; \
2164 init_rwsem(&nfsi->rwsem); \
2165 } while(0)
a72b4422
TM
2166
2167static inline int register_nfs4fs(void)
2168{
2169 int ret;
2170
2171 ret = nfs_register_sysctl();
2172 if (ret != 0)
2173 return ret;
2174 ret = register_filesystem(&nfs4_fs_type);
2175 if (ret != 0)
2176 nfs_unregister_sysctl();
2177 return ret;
2178}
2179
2180static inline void unregister_nfs4fs(void)
2181{
2182 unregister_filesystem(&nfs4_fs_type);
2183 nfs_unregister_sysctl();
2184}
1da177e4
LT
2185#else
2186#define nfs4_init_once(nfsi) \
2187 do { } while (0)
2188#define register_nfs4fs() (0)
2189#define unregister_nfs4fs()
2190#endif
2191
2192extern int nfs_init_nfspagecache(void);
2193extern void nfs_destroy_nfspagecache(void);
2194extern int nfs_init_readpagecache(void);
2195extern void nfs_destroy_readpagecache(void);
2196extern int nfs_init_writepagecache(void);
2197extern void nfs_destroy_writepagecache(void);
2198#ifdef CONFIG_NFS_DIRECTIO
2199extern int nfs_init_directcache(void);
2200extern void nfs_destroy_directcache(void);
2201#endif
2202
2203static kmem_cache_t * nfs_inode_cachep;
2204
2205static struct inode *nfs_alloc_inode(struct super_block *sb)
2206{
2207 struct nfs_inode *nfsi;
2208 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2209 if (!nfsi)
2210 return NULL;
55296809
CL
2211 nfsi->flags = 0UL;
2212 nfsi->cache_validity = 0UL;
223db122 2213 nfsi->cache_change_attribute = jiffies;
458818ed
TM
2214#ifdef CONFIG_NFS_V3_ACL
2215 nfsi->acl_access = ERR_PTR(-EAGAIN);
2216 nfsi->acl_default = ERR_PTR(-EAGAIN);
2217#endif
e50a1c2e
BF
2218#ifdef CONFIG_NFS_V4
2219 nfsi->nfs4_acl = NULL;
2220#endif /* CONFIG_NFS_V4 */
1da177e4
LT
2221 return &nfsi->vfs_inode;
2222}
2223
2224static void nfs_destroy_inode(struct inode *inode)
2225{
2226 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2227}
2228
2229static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2230{
2231 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2232
2233 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2234 SLAB_CTOR_CONSTRUCTOR) {
2235 inode_init_once(&nfsi->vfs_inode);
2236 spin_lock_init(&nfsi->req_lock);
2237 INIT_LIST_HEAD(&nfsi->dirty);
2238 INIT_LIST_HEAD(&nfsi->commit);
2239 INIT_LIST_HEAD(&nfsi->open_files);
2240 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2241 atomic_set(&nfsi->data_updates, 0);
2242 nfsi->ndirty = 0;
2243 nfsi->ncommit = 0;
2244 nfsi->npages = 0;
1da177e4
LT
2245 nfs4_init_once(nfsi);
2246 }
2247}
2248
75c96f85 2249static int nfs_init_inodecache(void)
1da177e4
LT
2250{
2251 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2252 sizeof(struct nfs_inode),
fffb60f9
PJ
2253 0, (SLAB_RECLAIM_ACCOUNT|
2254 SLAB_MEM_SPREAD),
1da177e4
LT
2255 init_once, NULL);
2256 if (nfs_inode_cachep == NULL)
2257 return -ENOMEM;
2258
2259 return 0;
2260}
2261
75c96f85 2262static void nfs_destroy_inodecache(void)
1da177e4
LT
2263{
2264 if (kmem_cache_destroy(nfs_inode_cachep))
2265 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2266}
2267
2268/*
2269 * Initialize NFS
2270 */
2271static int __init init_nfs_fs(void)
2272{
2273 int err;
2274
2275 err = nfs_init_nfspagecache();
2276 if (err)
2277 goto out4;
2278
2279 err = nfs_init_inodecache();
2280 if (err)
2281 goto out3;
2282
2283 err = nfs_init_readpagecache();
2284 if (err)
2285 goto out2;
2286
2287 err = nfs_init_writepagecache();
2288 if (err)
2289 goto out1;
2290
2291#ifdef CONFIG_NFS_DIRECTIO
2292 err = nfs_init_directcache();
2293 if (err)
2294 goto out0;
2295#endif
2296
2297#ifdef CONFIG_PROC_FS
2298 rpc_proc_register(&nfs_rpcstat);
2299#endif
2300 err = register_filesystem(&nfs_fs_type);
2301 if (err)
2302 goto out;
2303 if ((err = register_nfs4fs()) != 0)
2304 goto out;
2305 return 0;
2306out:
2307#ifdef CONFIG_PROC_FS
2308 rpc_proc_unregister("nfs");
2309#endif
1da177e4 2310#ifdef CONFIG_NFS_DIRECTIO
1da177e4 2311 nfs_destroy_directcache();
6b59a754 2312out0:
1da177e4 2313#endif
6b59a754 2314 nfs_destroy_writepagecache();
1da177e4
LT
2315out1:
2316 nfs_destroy_readpagecache();
2317out2:
2318 nfs_destroy_inodecache();
2319out3:
2320 nfs_destroy_nfspagecache();
2321out4:
2322 return err;
2323}
2324
2325static void __exit exit_nfs_fs(void)
2326{
2327#ifdef CONFIG_NFS_DIRECTIO
2328 nfs_destroy_directcache();
2329#endif
2330 nfs_destroy_writepagecache();
2331 nfs_destroy_readpagecache();
2332 nfs_destroy_inodecache();
2333 nfs_destroy_nfspagecache();
2334#ifdef CONFIG_PROC_FS
2335 rpc_proc_unregister("nfs");
2336#endif
2337 unregister_filesystem(&nfs_fs_type);
2338 unregister_nfs4fs();
2339}
2340
2341/* Not quite true; I just maintain it */
2342MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2343MODULE_LICENSE("GPL");
2344
2345module_init(init_nfs_fs)
2346module_exit(exit_nfs_fs)