projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFS: introduce mechanism for tracking NFS client metrics
[linux-2.6-block.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
4ce79717 42#include "nfs4_fs.h"
a72b4422 43#include "callback.h"
1da177e4 44#include "delegation.h"
d9ef5a8c 45#include "iostat.h"
1da177e4
LT		 46
47#define NFSDBG_FACILITY NFSDBG_VFS
48#define NFS_PARANOIA 1
49
50/* Maximum number of readahead requests
51 * FIXME: this should really be a sysctl so that users may tune it to suit
52 * their needs. People that do NFS over a slow network, might for
53 * instance want to reduce it to something closer to 1 for improved
54 * interactive response.
55 */
56#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
57
58static void nfs_invalidate_inode(struct inode *);
24aa1fe6 59static int nfs_update_inode(struct inode *, struct nfs_fattr *);
1da177e4
LT		 60
61static struct inode *nfs_alloc_inode(struct super_block *sb);
62static void nfs_destroy_inode(struct inode *);
63static int nfs_write_inode(struct inode *,int);
64static void nfs_delete_inode(struct inode *);
65static void nfs_clear_inode(struct inode *);
66static void nfs_umount_begin(struct super_block *);
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,
78 .delete_inode = nfs_delete_inode,
79 .statfs = nfs_statfs,
80 .clear_inode = nfs_clear_inode,
81 .umount_begin = nfs_umount_begin,
82 .show_options = nfs_show_options,
d9ef5a8c 83 .show_stats = nfs_show_stats,
1da177e4
LT		 84};
85
86/*
87 * RPC cruft for NFS
88 */
89static struct rpc_stat nfs_rpcstat = {
90 .program = &nfs_program
91};
92static struct rpc_version * nfs_version[] = {
93 NULL,
94 NULL,
95 &nfs_version2,
96#if defined(CONFIG_NFS_V3)
97 &nfs_version3,
98#elif defined(CONFIG_NFS_V4)
99 NULL,
100#endif
101#if defined(CONFIG_NFS_V4)
102 &nfs_version4,
103#endif
104};
105
106static struct rpc_program nfs_program = {
107 .name = "nfs",
108 .number = NFS_PROGRAM,
109 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
110 .version = nfs_version,
111 .stats = &nfs_rpcstat,
112 .pipe_dir_name = "/nfs",
113};
114
b7fa0554
AG		 115#ifdef CONFIG_NFS_V3_ACL
116static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
117static struct rpc_version * nfsacl_version[] = {
118 [3] = &nfsacl_version3,
119};
120
121struct rpc_program nfsacl_program = {
122 .name = "nfsacl",
123 .number = NFS_ACL_PROGRAM,
124 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
125 .version = nfsacl_version,
126 .stats = &nfsacl_rpcstat,
127};
128#endif /* CONFIG_NFS_V3_ACL */
129
1da177e4
LT		 130static inline unsigned long
131nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
132{
133 return nfs_fileid_to_ino_t(fattr->fileid);
134}
135
136static int
137nfs_write_inode(struct inode *inode, int sync)
138{
139 int flags = sync ? FLUSH_WAIT : 0;
140 int ret;
141
3da28eb1 142 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 143 if (ret < 0)
144 return ret;
145 return 0;
146}
147
148static void
149nfs_delete_inode(struct inode * inode)
150{
151 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
152
fef26658
MF		 153 truncate_inode_pages(&inode->i_data, 0);
154
1da177e4
LT		 155 nfs_wb_all(inode);
156 /*
157 * The following should never happen...
158 */
159 if (nfs_have_writebacks(inode)) {
160 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
161 }
162
163 clear_inode(inode);
164}
165
1da177e4
LT		 166static void
167nfs_clear_inode(struct inode *inode)
168{
169 struct nfs_inode *nfsi = NFS_I(inode);
170 struct rpc_cred *cred;
171
172 nfs_wb_all(inode);
173 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 174 nfs_zap_acl_cache(inode);
1da177e4
LT		 175 cred = nfsi->cache_access.cred;
176 if (cred)
177 put_rpccred(cred);
178 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
179}
180
181void
182nfs_umount_begin(struct super_block *sb)
183{
6a19275a 184 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT		 185
186 /* -EIO all pending I/O */
6a19275a 187 if (!IS_ERR(rpc))
1da177e4 188 rpc_killall_tasks(rpc);
b7fa0554
AG		 189 rpc = NFS_SB(sb)->client_acl;
190 if (!IS_ERR(rpc))
191 rpc_killall_tasks(rpc);
1da177e4
LT		 192}
193
194
195static inline unsigned long
196nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
197{
198 /* make sure blocksize is a power of two */
199 if ((bsize & (bsize - 1)) || nrbitsp) {
200 unsigned char nrbits;
201
202 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
203 ;
204 bsize = 1 << nrbits;
205 if (nrbitsp)
206 *nrbitsp = nrbits;
207 }
208
209 return bsize;
210}
211
212/*
213 * Calculate the number of 512byte blocks used.
214 */
215static inline unsigned long
216nfs_calc_block_size(u64 tsize)
217{
218 loff_t used = (tsize + 511) >> 9;
219 return (used > ULONG_MAX) ? ULONG_MAX : used;
220}
221
222/*
223 * Compute and set NFS server blocksize
224 */
225static inline unsigned long
226nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
227{
40859d7e
CL		 228 if (bsize < NFS_MIN_FILE_IO_SIZE)
229 bsize = NFS_DEF_FILE_IO_SIZE;
230 else if (bsize >= NFS_MAX_FILE_IO_SIZE)
231 bsize = NFS_MAX_FILE_IO_SIZE;
1da177e4
LT		 232
233 return nfs_block_bits(bsize, nrbitsp);
234}
235
236/*
237 * Obtain the root inode of the file system.
238 */
239static struct inode *
240nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
241{
242 struct nfs_server *server = NFS_SB(sb);
243 struct inode *rooti;
244 int error;
245
246 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
247 if (error < 0) {
248 dprintk("nfs_get_root: getattr error = %d\n", -error);
249 return ERR_PTR(error);
250 }
251
252 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
253 if (!rooti)
254 return ERR_PTR(-ENOMEM);
255 return rooti;
256}
257
258/*
259 * Do NFS version-independent mount processing, and sanity checking
260 */
261static int
262nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
263{
264 struct nfs_server *server;
265 struct inode *root_inode;
266 struct nfs_fattr fattr;
267 struct nfs_fsinfo fsinfo = {
268 .fattr = &fattr,
269 };
270 struct nfs_pathconf pathinfo = {
271 .fattr = &fattr,
272 };
273 int no_root_error = 0;
274 unsigned long max_rpc_payload;
275
276 /* We probably want something more informative here */
277 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
278
279 server = NFS_SB(sb);
280
281 sb->s_magic = NFS_SUPER_MAGIC;
282
283 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
284 /* Did getting the root inode fail? */
285 if (IS_ERR(root_inode)) {
286 no_root_error = PTR_ERR(root_inode);
287 goto out_no_root;
288 }
289 sb->s_root = d_alloc_root(root_inode);
290 if (!sb->s_root) {
291 no_root_error = -ENOMEM;
292 goto out_no_root;
293 }
294 sb->s_root->d_op = server->rpc_ops->dentry_ops;
295
d9ef5a8c
CL		 296 server->io_stats = nfs_alloc_iostats();
297 if (!server->io_stats) {
298 no_root_error = -ENOMEM;
299 goto out_no_root;
300 }
301
1da177e4
LT		 302 /* Get some general file system info */
303 if (server->namelen == 0 &&
304 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
305 server->namelen = pathinfo.max_namelen;
306 /* Work out a lot of parameters */
307 if (server->rsize == 0)
308 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
309 if (server->wsize == 0)
310 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
311
312 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
313 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
314 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
315 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
316
317 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
318 if (server->rsize > max_rpc_payload)
319 server->rsize = max_rpc_payload;
40859d7e
CL		 320 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
321 server->rsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 322 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4 323
40859d7e
CL		 324 if (server->wsize > max_rpc_payload)
325 server->wsize = max_rpc_payload;
326 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
327 server->wsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 328 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4
LT		 329
330 if (sb->s_blocksize == 0)
331 sb->s_blocksize = nfs_block_bits(server->wsize,
332 &sb->s_blocksize_bits);
333 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
334
335 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
336 if (server->dtsize > PAGE_CACHE_SIZE)
337 server->dtsize = PAGE_CACHE_SIZE;
338 if (server->dtsize > server->rsize)
339 server->dtsize = server->rsize;
340
341 if (server->flags & NFS_MOUNT_NOAC) {
342 server->acregmin = server->acregmax = 0;
343 server->acdirmin = server->acdirmax = 0;
344 sb->s_flags |= MS_SYNCHRONOUS;
345 }
346 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
347
348 sb->s_maxbytes = fsinfo.maxfilesize;
349 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
350 sb->s_maxbytes = MAX_LFS_FILESIZE;
351
352 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
353 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
354
355 /* We're airborne Set socket buffersize */
356 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
357 return 0;
358 /* Yargs. It didn't work out. */
359out_no_root:
360 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
361 if (!IS_ERR(root_inode))
362 iput(root_inode);
363 return no_root_error;
364}
365
eab5c084
CL		 366static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
367{
368 to->to_initval = timeo * HZ / 10;
369 to->to_retries = retrans;
370 if (!to->to_retries)
371 to->to_retries = 2;
372
373 switch (proto) {
374 case IPPROTO_TCP:
375 if (!to->to_initval)
376 to->to_initval = 60 * HZ;
03bf4b70
CL		 377 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
378 to->to_initval = NFS_MAX_TCP_TIMEOUT;
eab5c084
CL		 379 to->to_increment = to->to_initval;
380 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
381 to->to_exponential = 0;
382 break;
383 case IPPROTO_UDP:
384 default:
385 if (!to->to_initval)
386 to->to_initval = 11 * HZ / 10;
03bf4b70
CL		 387 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
388 to->to_initval = NFS_MAX_UDP_TIMEOUT;
389 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
eab5c084
CL		 390 to->to_exponential = 1;
391 break;
392 }
393}
394
1da177e4
LT		 395/*
396 * Create an RPC client handle.
397 */
398static struct rpc_clnt *
399nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
400{
401 struct rpc_timeout timeparms;
402 struct rpc_xprt *xprt = NULL;
403 struct rpc_clnt *clnt = NULL;
eab5c084 404 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
1da177e4 405
eab5c084 406 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
1da177e4 407
7a480e25
CL		 408 server->retrans_timeo = timeparms.to_initval;
409 server->retrans_count = timeparms.to_retries;
410
1da177e4 411 /* create transport and client */
eab5c084 412 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 413 if (IS_ERR(xprt)) {
9085bbcb
TM		 414 dprintk("%s: cannot create RPC transport. Error = %ld\n",
415 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 416 return (struct rpc_clnt *)xprt;
417 }
418 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
419 server->rpc_ops->version, data->pseudoflavor);
420 if (IS_ERR(clnt)) {
9085bbcb
TM		 421 dprintk("%s: cannot create RPC client. Error = %ld\n",
422 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 423 goto out_fail;
424 }
425
426 clnt->cl_intr = 1;
427 clnt->cl_softrtry = 1;
1da177e4
LT		 428
429 return clnt;
430
431out_fail:
1da177e4
LT		 432 return clnt;
433}
434
435/*
436 * The way this works is that the mount process passes a structure
437 * in the data argument which contains the server's IP address
438 * and the root file handle obtained from the server's mount
439 * daemon. We stash these away in the private superblock fields.
440 */
441static int
442nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
443{
444 struct nfs_server *server;
445 rpc_authflavor_t authflavor;
446
447 server = NFS_SB(sb);
448 sb->s_blocksize_bits = 0;
449 sb->s_blocksize = 0;
450 if (data->bsize)
451 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
452 if (data->rsize)
453 server->rsize = nfs_block_size(data->rsize, NULL);
454 if (data->wsize)
455 server->wsize = nfs_block_size(data->wsize, NULL);
456 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
457
458 server->acregmin = data->acregmin*HZ;
459 server->acregmax = data->acregmax*HZ;
460 server->acdirmin = data->acdirmin*HZ;
461 server->acdirmax = data->acdirmax*HZ;
462
463 /* Start lockd here, before we might error out */
464 if (!(server->flags & NFS_MOUNT_NONLM))
465 lockd_up();
466
467 server->namelen = data->namlen;
468 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
469 if (!server->hostname)
470 return -ENOMEM;
471 strcpy(server->hostname, data->hostname);
472
473 /* Check NFS protocol revision and initialize RPC op vector
474 * and file handle pool. */
1da177e4 475#ifdef CONFIG_NFS_V3
9085bbcb 476 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 477 server->rpc_ops = &nfs_v3_clientops;
478 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 479 } else {
480 server->rpc_ops = &nfs_v2_clientops;
481 }
9085bbcb
TM		 482#else
483 server->rpc_ops = &nfs_v2_clientops;
484#endif
1da177e4
LT		 485
486 /* Fill in pseudoflavor for mount version < 5 */
487 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
488 data->pseudoflavor = RPC_AUTH_UNIX;
489 authflavor = data->pseudoflavor; /* save for sb_init() */
490 /* XXX maybe we want to add a server->pseudoflavor field */
491
492 /* Create RPC client handles */
493 server->client = nfs_create_client(server, data);
494 if (IS_ERR(server->client))
495 return PTR_ERR(server->client);
496 /* RFC 2623, sec 2.3.2 */
497 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 498 struct rpc_auth *auth;
499
1da177e4
LT		 500 server->client_sys = rpc_clone_client(server->client);
501 if (IS_ERR(server->client_sys))
502 return PTR_ERR(server->client_sys);
6a19275a
BF		 503 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
504 if (IS_ERR(auth))
505 return PTR_ERR(auth);
1da177e4
LT		 506 } else {
507 atomic_inc(&server->client->cl_count);
508 server->client_sys = server->client;
509 }
1da177e4 510 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 511#ifdef CONFIG_NFS_V3_ACL
512 if (!(server->flags & NFS_MOUNT_NOACL)) {
513 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
514 /* No errors! Assume that Sun nfsacls are supported */
515 if (!IS_ERR(server->client_acl))
516 server->caps |= NFS_CAP_ACLS;
517 }
518#else
519 server->flags &= ~NFS_MOUNT_NOACL;
520#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 521 /*
522 * The VFS shouldn't apply the umask to mode bits. We will
523 * do so ourselves when necessary.
524 */
525 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 526 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
527 server->namelen = NFS3_MAXNAMLEN;
528 sb->s_time_gran = 1;
529 } else {
530 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
531 server->namelen = NFS2_MAXNAMLEN;
532 }
533
534 sb->s_op = &nfs_sops;
535 return nfs_sb_init(sb, authflavor);
536}
537
538static int
539nfs_statfs(struct super_block *sb, struct kstatfs *buf)
540{
541 struct nfs_server *server = NFS_SB(sb);
542 unsigned char blockbits;
543 unsigned long blockres;
544 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
545 struct nfs_fattr fattr;
546 struct nfs_fsstat res = {
547 .fattr = &fattr,
548 };
549 int error;
550
551 lock_kernel();
552
553 error = server->rpc_ops->statfs(server, rootfh, &res);
554 buf->f_type = NFS_SUPER_MAGIC;
555 if (error < 0)
556 goto out_err;
557
558 /*
559 * Current versions of glibc do not correctly handle the
560 * case where f_frsize != f_bsize. Eventually we want to
561 * report the value of wtmult in this field.
562 */
563 buf->f_frsize = sb->s_blocksize;
564
565 /*
566 * On most *nix systems, f_blocks, f_bfree, and f_bavail
567 * are reported in units of f_frsize. Linux hasn't had
568 * an f_frsize field in its statfs struct until recently,
569 * thus historically Linux's sys_statfs reports these
570 * fields in units of f_bsize.
571 */
572 buf->f_bsize = sb->s_blocksize;
573 blockbits = sb->s_blocksize_bits;
574 blockres = (1 << blockbits) - 1;
575 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
576 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
577 buf->f_bavail = (res.abytes + blockres) >> blockbits;
578
579 buf->f_files = res.tfiles;
580 buf->f_ffree = res.afiles;
581
582 buf->f_namelen = server->namelen;
583 out:
584 unlock_kernel();
1da177e4
LT		 585 return 0;
586
587 out_err:
dc20f803 588 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
1da177e4
LT		 589 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
590 goto out;
591
592}
593
d9ef5a8c 594static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
1da177e4
LT		 595{
596 static struct proc_nfs_info {
597 int flag;
598 char *str;
599 char *nostr;
600 } nfs_info[] = {
601 { NFS_MOUNT_SOFT, ",soft", ",hard" },
602 { NFS_MOUNT_INTR, ",intr", "" },
1da177e4
LT		 603 { NFS_MOUNT_NOCTO, ",nocto", "" },
604 { NFS_MOUNT_NOAC, ",noac", "" },
c8bded96 605 { NFS_MOUNT_NONLM, ",nolock", "" },
b7fa0554 606 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 607 { 0, NULL, NULL }
608 };
609 struct proc_nfs_info *nfs_infop;
3063d8a1
TM		 610 char buf[12];
611 char *proto;
1da177e4 612
c8bded96 613 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
1da177e4
LT		 614 seq_printf(m, ",rsize=%d", nfss->rsize);
615 seq_printf(m, ",wsize=%d", nfss->wsize);
d9ef5a8c 616 if (nfss->acregmin != 3*HZ || showdefaults)
1da177e4 617 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
d9ef5a8c 618 if (nfss->acregmax != 60*HZ || showdefaults)
1da177e4 619 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
d9ef5a8c 620 if (nfss->acdirmin != 30*HZ || showdefaults)
1da177e4 621 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
d9ef5a8c 622 if (nfss->acdirmax != 60*HZ || showdefaults)
1da177e4
LT		 623 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
624 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
625 if (nfss->flags & nfs_infop->flag)
626 seq_puts(m, nfs_infop->str);
627 else
628 seq_puts(m, nfs_infop->nostr);
629 }
3063d8a1
TM		 630 switch (nfss->client->cl_xprt->prot) {
631 case IPPROTO_TCP:
632 proto = "tcp";
633 break;
634 case IPPROTO_UDP:
635 proto = "udp";
636 break;
637 default:
638 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
639 proto = buf;
640 }
641 seq_printf(m, ",proto=%s", proto);
7a480e25
CL		 642 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
643 seq_printf(m, ",retrans=%u", nfss->retrans_count);
d9ef5a8c
CL		 644}
645
646static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
647{
648 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
649
650 nfs_show_mount_options(m, nfss, 0);
651
1da177e4
LT		 652 seq_puts(m, ",addr=");
653 seq_escape(m, nfss->hostname, " \t\n\\");
d9ef5a8c
CL		 654
655 return 0;
656}
657
658static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
659{
660 int i, cpu;
661 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
662 struct rpc_auth *auth = nfss->client->cl_auth;
663 struct nfs_iostats totals = { };
664
665 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
666
667 /*
668 * Display all mount option settings
669 */
670 seq_printf(m, "\n\topts:\t");
671 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
672 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
673 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
674 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
675 nfs_show_mount_options(m, nfss, 1);
676
677 seq_printf(m, "\n\tcaps:\t");
678 seq_printf(m, "caps=0x%x", nfss->caps);
679 seq_printf(m, ",wtmult=%d", nfss->wtmult);
680 seq_printf(m, ",dtsize=%d", nfss->dtsize);
681 seq_printf(m, ",bsize=%d", nfss->bsize);
682 seq_printf(m, ",namelen=%d", nfss->namelen);
683
684#ifdef CONFIG_NFS_V4
685 if (nfss->rpc_ops->version == 4) {
686 seq_printf(m, "\n\tnfsv4:\t");
687 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
688 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
689 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
690 }
691#endif
692
693 /*
694 * Display security flavor in effect for this mount
695 */
696 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
697 if (auth->au_flavor)
698 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
699
700 /*
701 * Display superblock I/O counters
702 */
703 for (cpu = 0; cpu < NR_CPUS; cpu++) {
704 struct nfs_iostats *stats;
705
706 if (!cpu_possible(cpu))
707 continue;
708
709 preempt_disable();
710 stats = per_cpu_ptr(nfss->io_stats, cpu);
711
712 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
713 totals.events[i] += stats->events[i];
714 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
715 totals.bytes[i] += stats->bytes[i];
716
717 preempt_enable();
718 }
719
720 seq_printf(m, "\n\tevents:\t");
721 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
722 seq_printf(m, "%lu ", totals.events[i]);
723 seq_printf(m, "\n\tbytes:\t");
724 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
725 seq_printf(m, "%Lu ", totals.bytes[i]);
726
1da177e4
LT		 727 return 0;
728}
729
29884df0
TM		 730/**
731 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
732 */
733int nfs_sync_mapping(struct address_space *mapping)
734{
735 int ret;
736
737 if (mapping->nrpages == 0)
738 return 0;
739 unmap_mapping_range(mapping, 0, 0, 0);
28fd1298 740 ret = filemap_write_and_wait(mapping);
29884df0
TM		 741 if (ret != 0)
742 goto out;
743 ret = nfs_wb_all(mapping->host);
744out:
745 return ret;
746}
747
1da177e4
LT		 748/*
749 * Invalidate the local caches
750 */
b37b03b7 751static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT		 752{
753 struct nfs_inode *nfsi = NFS_I(inode);
754 int mode = inode->i_mode;
755
756 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
757 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
758
759 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
760 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 761 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 762 else
55296809 763 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 764}
dc59250c 765
b37b03b7
TM		 766void nfs_zap_caches(struct inode *inode)
767{
768 spin_lock(&inode->i_lock);
769 nfs_zap_caches_locked(inode);
dc59250c 770 spin_unlock(&inode->i_lock);
ada70d94
TM		 771}
772
773static void nfs_zap_acl_cache(struct inode *inode)
774{
775 void (*clear_acl_cache)(struct inode *);
776
777 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
778 if (clear_acl_cache != NULL)
779 clear_acl_cache(inode);
dc59250c 780 spin_lock(&inode->i_lock);
55296809 781 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 782 spin_unlock(&inode->i_lock);
1da177e4
LT		 783}
784
785/*
b37b03b7
TM		 786 * Invalidate, but do not unhash, the inode.
787 * NB: must be called with inode->i_lock held!
1da177e4 788 */
b37b03b7 789static void nfs_invalidate_inode(struct inode *inode)
1da177e4 790{
b37b03b7
TM		 791 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
792 nfs_zap_caches_locked(inode);
1da177e4
LT		 793}
794
795struct nfs_find_desc {
796 struct nfs_fh *fh;
797 struct nfs_fattr *fattr;
798};
799
800/*
801 * In NFSv3 we can have 64bit inode numbers. In order to support
802 * this, and re-exported directories (also seen in NFSv2)
803 * we are forced to allow 2 different inodes to have the same
804 * i_ino.
805 */
806static int
807nfs_find_actor(struct inode *inode, void *opaque)
808{
809 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
810 struct nfs_fh *fh = desc->fh;
811 struct nfs_fattr *fattr = desc->fattr;
812
813 if (NFS_FILEID(inode) != fattr->fileid)
814 return 0;
815 if (nfs_compare_fh(NFS_FH(inode), fh))
816 return 0;
817 if (is_bad_inode(inode) || NFS_STALE(inode))
818 return 0;
819 return 1;
820}
821
822static int
823nfs_init_locked(struct inode *inode, void *opaque)
824{
825 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
826 struct nfs_fattr *fattr = desc->fattr;
827
828 NFS_FILEID(inode) = fattr->fileid;
829 nfs_copy_fh(NFS_FH(inode), desc->fh);
830 return 0;
831}
832
833/* Don't use READDIRPLUS on directories that we believe are too large */
834#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
835
836/*
837 * This is our front-end to iget that looks up inodes by file handle
838 * instead of inode number.
839 */
840struct inode *
841nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
842{
843 struct nfs_find_desc desc = {
844 .fh = fh,
845 .fattr = fattr
846 };
847 struct inode *inode = NULL;
848 unsigned long hash;
849
850 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
851 goto out_no_inode;
852
853 if (!fattr->nlink) {
854 printk("NFS: Buggy server - nlink == 0!\n");
855 goto out_no_inode;
856 }
857
858 hash = nfs_fattr_to_ino_t(fattr);
859
860 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
861 goto out_no_inode;
862
863 if (inode->i_state & I_NEW) {
864 struct nfs_inode *nfsi = NFS_I(inode);
865
866 /* We set i_ino for the few things that still rely on it,
867 * such as stat(2) */
868 inode->i_ino = hash;
869
870 /* We can't support update_atime(), since the server will reset it */
871 inode->i_flags |= S_NOATIME|S_NOCMTIME;
872 inode->i_mode = fattr->mode;
873 /* Why so? Because we want revalidate for devices/FIFOs, and
874 * that's precisely what we have in nfs_file_inode_operations.
875 */
92cfc62c 876 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 877 if (S_ISREG(inode->i_mode)) {
878 inode->i_fop = &nfs_file_operations;
879 inode->i_data.a_ops = &nfs_file_aops;
880 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
881 } else if (S_ISDIR(inode->i_mode)) {
882 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
883 inode->i_fop = &nfs_dir_operations;
884 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
885 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 886 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 887 } else if (S_ISLNK(inode->i_mode))
888 inode->i_op = &nfs_symlink_inode_operations;
889 else
890 init_special_inode(inode, inode->i_mode, fattr->rdev);
891
33801147
TM		 892 nfsi->read_cache_jiffies = fattr->time_start;
893 nfsi->last_updated = jiffies;
1da177e4
LT		 894 inode->i_atime = fattr->atime;
895 inode->i_mtime = fattr->mtime;
896 inode->i_ctime = fattr->ctime;
897 if (fattr->valid & NFS_ATTR_FATTR_V4)
898 nfsi->change_attr = fattr->change_attr;
899 inode->i_size = nfs_size_to_loff_t(fattr->size);
900 inode->i_nlink = fattr->nlink;
901 inode->i_uid = fattr->uid;
902 inode->i_gid = fattr->gid;
903 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
904 /*
905 * report the blocks in 512byte units
906 */
907 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
908 inode->i_blksize = inode->i_sb->s_blocksize;
909 } else {
910 inode->i_blocks = fattr->du.nfs2.blocks;
911 inode->i_blksize = fattr->du.nfs2.blocksize;
912 }
913 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
914 nfsi->attrtimeo_timestamp = jiffies;
915 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
916 nfsi->cache_access.cred = NULL;
917
918 unlock_new_inode(inode);
919 } else
920 nfs_refresh_inode(inode, fattr);
921 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
922 inode->i_sb->s_id,
923 (long long)NFS_FILEID(inode),
924 atomic_read(&inode->i_count));
925
926out:
927 return inode;
928
929out_no_inode:
930 printk("nfs_fhget: iget failed\n");
931 goto out;
932}
933
934#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
935
936int
937nfs_setattr(struct dentry *dentry, struct iattr *attr)
938{
939 struct inode *inode = dentry->d_inode;
940 struct nfs_fattr fattr;
941 int error;
942
943 if (attr->ia_valid & ATTR_SIZE) {
944 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
945 attr->ia_valid &= ~ATTR_SIZE;
946 }
947
948 /* Optimization: if the end result is no change, don't RPC */
949 attr->ia_valid &= NFS_VALID_ATTRS;
950 if (attr->ia_valid == 0)
951 return 0;
952
953 lock_kernel();
954 nfs_begin_data_update(inode);
755c1e20
TM		 955 /* Write all dirty data */
956 filemap_write_and_wait(inode->i_mapping);
957 nfs_wb_all(inode);
642ac549
TM		 958 /*
959 * Return any delegations if we're going to change ACLs
960 */
961 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
962 nfs_inode_return_delegation(inode);
1da177e4 963 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 964 if (error == 0)
1da177e4 965 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 966 nfs_end_data_update(inode);
967 unlock_kernel();
968 return error;
969}
970
971/**
972 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
973 * @inode: pointer to struct inode
974 * @attr: pointer to struct iattr
975 *
976 * Note: we do this in the *proc.c in order to ensure that
977 * it works for things like exclusive creates too.
978 */
979void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
980{
981 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 982 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 983 int mode = attr->ia_mode & S_IALLUGO;
984 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 985 inode->i_mode = mode;
986 }
987 if ((attr->ia_valid & ATTR_UID) != 0)
988 inode->i_uid = attr->ia_uid;
989 if ((attr->ia_valid & ATTR_GID) != 0)
990 inode->i_gid = attr->ia_gid;
dc59250c 991 spin_lock(&inode->i_lock);
55296809 992 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 993 spin_unlock(&inode->i_lock);
65e4308d
TM		 994 }
995 if ((attr->ia_valid & ATTR_SIZE) != 0) {
996 inode->i_size = attr->ia_size;
997 vmtruncate(inode, attr->ia_size);
998 }
1da177e4
LT		 999}
1000
412d582e
CL		 1001static int nfs_wait_schedule(void *word)
1002{
1003 if (signal_pending(current))
1004 return -ERESTARTSYS;
1005 schedule();
1006 return 0;
1007}
1008
1da177e4
LT		 1009/*
1010 * Wait for the inode to get unlocked.
1da177e4 1011 */
412d582e 1012static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 1013{
1014 struct rpc_clnt *clnt = NFS_CLIENT(inode);
1015 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 1016 sigset_t oldmask;
1da177e4 1017 int error;
412d582e 1018
412d582e
CL		 1019 rpc_clnt_sigmask(clnt, &oldmask);
1020 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
1021 nfs_wait_schedule, TASK_INTERRUPTIBLE);
1022 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 1023
1da177e4
LT		 1024 return error;
1025}
1026
412d582e
CL		 1027static void nfs_wake_up_inode(struct inode *inode)
1028{
1029 struct nfs_inode *nfsi = NFS_I(inode);
1030
1031 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
1032 smp_mb__after_clear_bit();
1033 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
1034}
1035
1da177e4
LT		 1036int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1037{
1038 struct inode *inode = dentry->d_inode;
55296809 1039 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 1040 int err;
1041
70b9ecbd
TM		 1042 /* Flush out writes to the server in order to update c/mtime */
1043 nfs_sync_inode(inode, 0, 0, FLUSH_WAIT|FLUSH_NOCOMMIT);
fc33a7bb
CH		 1044
1045 /*
1046 * We may force a getattr if the user cares about atime.
1047 *
1048 * Note that we only have to check the vfsmount flags here:
1049 * - NFS always sets S_NOATIME by so checking it would give a
1050 * bogus result
1051 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1052 * no point in checking those.
1053 */
1054 if ((mnt->mnt_flags & MNT_NOATIME) ||
1055 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
1da177e4 1056 need_atime = 0;
fc33a7bb 1057
1da177e4
LT		 1058 if (need_atime)
1059 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
1060 else
1061 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
1062 if (!err)
1063 generic_fillattr(inode, stat);
1064 return err;
1065}
1066
b92dccf6 1067static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
1da177e4
LT		 1068{
1069 struct nfs_open_context *ctx;
1070
1071 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
1072 if (ctx != NULL) {
1073 atomic_set(&ctx->count, 1);
1074 ctx->dentry = dget(dentry);
b92dccf6 1075 ctx->vfsmnt = mntget(mnt);
1da177e4
LT		 1076 ctx->cred = get_rpccred(cred);
1077 ctx->state = NULL;
1078 ctx->lockowner = current->files;
1079 ctx->error = 0;
00a92642 1080 ctx->dir_cookie = 0;
1da177e4
LT		 1081 }
1082 return ctx;
1083}
1084
1085struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
1086{
1087 if (ctx != NULL)
1088 atomic_inc(&ctx->count);
1089 return ctx;
1090}
1091
1092void put_nfs_open_context(struct nfs_open_context *ctx)
1093{
1094 if (atomic_dec_and_test(&ctx->count)) {
1095 if (!list_empty(&ctx->list)) {
1096 struct inode *inode = ctx->dentry->d_inode;
1097 spin_lock(&inode->i_lock);
1098 list_del(&ctx->list);
1099 spin_unlock(&inode->i_lock);
1100 }
1101 if (ctx->state != NULL)
1102 nfs4_close_state(ctx->state, ctx->mode);
1103 if (ctx->cred != NULL)
1104 put_rpccred(ctx->cred);
1105 dput(ctx->dentry);
b92dccf6 1106 mntput(ctx->vfsmnt);
1da177e4
LT		 1107 kfree(ctx);
1108 }
1109}
1110
1111/*
1112 * Ensure that mmap has a recent RPC credential for use when writing out
1113 * shared pages
1114 */
b92dccf6 1115static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1da177e4
LT		 1116{
1117 struct inode *inode = filp->f_dentry->d_inode;
1118 struct nfs_inode *nfsi = NFS_I(inode);
1119
1120 filp->private_data = get_nfs_open_context(ctx);
1121 spin_lock(&inode->i_lock);
1122 list_add(&ctx->list, &nfsi->open_files);
1123 spin_unlock(&inode->i_lock);
1124}
1125
d530838b
TM		 1126/*
1127 * Given an inode, search for an open context with the desired characteristics
1128 */
1129struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1130{
1131 struct nfs_inode *nfsi = NFS_I(inode);
1132 struct nfs_open_context *pos, *ctx = NULL;
1133
1134 spin_lock(&inode->i_lock);
1135 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1136 if (cred != NULL && pos->cred != cred)
1137 continue;
1da177e4
LT		 1138 if ((pos->mode & mode) == mode) {
1139 ctx = get_nfs_open_context(pos);
1140 break;
1141 }
1142 }
1143 spin_unlock(&inode->i_lock);
1144 return ctx;
1145}
1146
b92dccf6 1147static void nfs_file_clear_open_context(struct file *filp)
1da177e4
LT		 1148{
1149 struct inode *inode = filp->f_dentry->d_inode;
1150 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1151
1152 if (ctx) {
1153 filp->private_data = NULL;
1154 spin_lock(&inode->i_lock);
1155 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1156 spin_unlock(&inode->i_lock);
1157 put_nfs_open_context(ctx);
1158 }
1159}
1160
1161/*
1162 * These allocate and release file read/write context information.
1163 */
1164int nfs_open(struct inode *inode, struct file *filp)
1165{
1166 struct nfs_open_context *ctx;
1167 struct rpc_cred *cred;
1168
1169 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1170 if (IS_ERR(cred))
1171 return PTR_ERR(cred);
b92dccf6 1172 ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
1da177e4
LT		 1173 put_rpccred(cred);
1174 if (ctx == NULL)
1175 return -ENOMEM;
1176 ctx->mode = filp->f_mode;
1177 nfs_file_set_open_context(filp, ctx);
1178 put_nfs_open_context(ctx);
1da177e4
LT		 1179 return 0;
1180}
1181
1182int nfs_release(struct inode *inode, struct file *filp)
1183{
1da177e4
LT		 1184 nfs_file_clear_open_context(filp);
1185 return 0;
1186}
1187
1188/*
1189 * This function is called whenever some part of NFS notices that
1190 * the cached attributes have to be refreshed.
1191 */
1192int
1193__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1194{
1195 int status = -ESTALE;
1196 struct nfs_fattr fattr;
1197 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4
LT		 1198
1199 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1200 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1201
1202 lock_kernel();
1203 if (!inode || is_bad_inode(inode))
1204 goto out_nowait;
1205 if (NFS_STALE(inode))
1206 goto out_nowait;
1207
412d582e
CL		 1208 status = nfs_wait_on_inode(inode);
1209 if (status < 0)
1210 goto out;
1211 if (NFS_STALE(inode)) {
1212 status = -ESTALE;
1213 /* Do we trust the cached ESTALE? */
1214 if (NFS_ATTRTIMEO(inode) != 0) {
1215 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1216 /* no */
1217 } else
1218 goto out;
1219 }
1da177e4 1220 }
1da177e4 1221
1da177e4
LT		 1222 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1223 if (status != 0) {
1224 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1225 inode->i_sb->s_id,
1226 (long long)NFS_FILEID(inode), status);
1227 if (status == -ESTALE) {
1228 nfs_zap_caches(inode);
1229 if (!S_ISDIR(inode->i_mode))
412d582e 1230 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1231 }
1232 goto out;
1233 }
1234
33801147 1235 spin_lock(&inode->i_lock);
24aa1fe6 1236 status = nfs_update_inode(inode, &fattr);
1da177e4 1237 if (status) {
33801147 1238 spin_unlock(&inode->i_lock);
1da177e4
LT		 1239 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1240 inode->i_sb->s_id,
1241 (long long)NFS_FILEID(inode), status);
1242 goto out;
1243 }
dc59250c 1244 spin_unlock(&inode->i_lock);
55296809 1245
7d52e862 1246 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809 1247
24aa1fe6 1248 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1249 nfs_zap_acl_cache(inode);
55296809 1250
1da177e4
LT		 1251 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1252 inode->i_sb->s_id,
1253 (long long)NFS_FILEID(inode));
1254
412d582e
CL		 1255 out:
1256 nfs_wake_up_inode(inode);
1257
1da177e4
LT		 1258 out_nowait:
1259 unlock_kernel();
1260 return status;
1261}
1262
1263int nfs_attribute_timeout(struct inode *inode)
1264{
1265 struct nfs_inode *nfsi = NFS_I(inode);
1266
1267 if (nfs_have_delegation(inode, FMODE_READ))
1268 return 0;
1269 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1270}
1271
1272/**
1273 * nfs_revalidate_inode - Revalidate the inode attributes
1274 * @server - pointer to nfs_server struct
1275 * @inode - pointer to inode struct
1276 *
1277 * Updates inode attribute information by retrieving the data from the server.
1278 */
1279int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1280{
55296809 1281 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1282 && !nfs_attribute_timeout(inode))
1283 return NFS_STALE(inode) ? -ESTALE : 0;
1284 return __nfs_revalidate_inode(server, inode);
1285}
1286
7d52e862
TM		 1287/**
1288 * nfs_revalidate_mapping - Revalidate the pagecache
1289 * @inode - pointer to host inode
1290 * @mapping - pointer to mapping
1291 */
1292void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1293{
1294 struct nfs_inode *nfsi = NFS_I(inode);
1295
55296809 1296 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
29884df0
TM		 1297 if (S_ISREG(inode->i_mode))
1298 nfs_sync_mapping(mapping);
7d52e862 1299 invalidate_inode_pages2(mapping);
dc59250c
CL		 1300
1301 spin_lock(&inode->i_lock);
55296809 1302 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1303 if (S_ISDIR(inode->i_mode)) {
1304 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1305 /* This ensures we revalidate child dentries */
913a70fc 1306 nfsi->cache_change_attribute = jiffies;
7d52e862 1307 }
dc59250c
CL		 1308 spin_unlock(&inode->i_lock);
1309
7d52e862
TM		 1310 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1311 inode->i_sb->s_id,
1312 (long long)NFS_FILEID(inode));
1313 }
1314}
1315
1da177e4
LT		 1316/**
1317 * nfs_begin_data_update
1318 * @inode - pointer to inode
1319 * Declare that a set of operations will update file data on the server
1320 */
1321void nfs_begin_data_update(struct inode *inode)
1322{
1323 atomic_inc(&NFS_I(inode)->data_updates);
1324}
1325
1326/**
1327 * nfs_end_data_update
1328 * @inode - pointer to inode
1329 * Declare end of the operations that will update file data
1330 * This will mark the inode as immediately needing revalidation
1331 * of its attribute cache.
1332 */
1333void nfs_end_data_update(struct inode *inode)
1334{
1335 struct nfs_inode *nfsi = NFS_I(inode);
1336
1337 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1338 /* Directories and symlinks: invalidate page cache */
1339 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1340 spin_lock(&inode->i_lock);
55296809 1341 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1342 spin_unlock(&inode->i_lock);
1343 }
1da177e4 1344 }
913a70fc 1345 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1346 atomic_dec(&nfsi->data_updates);
1347}
1348
a895b4a1
TM		 1349static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1350{
1351 struct nfs_inode *nfsi = NFS_I(inode);
1352
1353 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1354 && nfsi->change_attr == fattr->pre_change_attr) {
1355 nfsi->change_attr = fattr->change_attr;
1356 nfsi->cache_change_attribute = jiffies;
1357 }
1358
1359 /* If we have atomic WCC data, we may update some attributes */
1360 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1361 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
1362 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1363 nfsi->cache_change_attribute = jiffies;
1364 }
1365 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
1366 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1367 nfsi->cache_change_attribute = jiffies;
1368 }
1369 if (inode->i_size == fattr->pre_size && nfsi->npages == 0) {
1370 inode->i_size = fattr->size;
1371 nfsi->cache_change_attribute = jiffies;
1372 }
1373 }
1374}
1375
1da177e4 1376/**
33801147 1377 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1378 * @inode - pointer to inode
1379 * @fattr - updated attributes
1380 *
1381 * Verifies the attribute cache. If we have just changed the attributes,
1382 * so that fattr carries weak cache consistency data, then it may
1383 * also update the ctime/mtime/change_attribute.
1384 */
33801147 1385static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1386{
1387 struct nfs_inode *nfsi = NFS_I(inode);
1388 loff_t cur_size, new_isize;
1389 int data_unstable;
1390
dc59250c 1391
a895b4a1
TM		 1392 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1393 return 0;
1394
ca62b9c3
TM		 1395 /* Has the inode gone and changed behind our back? */
1396 if (nfsi->fileid != fattr->fileid
1397 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1398 return -EIO;
1399 }
1400
1da177e4
LT		 1401 /* Are we in the process of updating data on the server? */
1402 data_unstable = nfs_caches_unstable(inode);
1403
a895b4a1
TM		 1404 /* Do atomic weak cache consistency updates */
1405 nfs_wcc_update_inode(inode, fattr);
1da177e4 1406
ca62b9c3
TM		 1407 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0) {
1408 if (nfsi->change_attr == fattr->change_attr)
1409 goto out;
a895b4a1
TM		 1410 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1411 if (!data_unstable)
1412 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
dc59250c 1413 }
1da177e4 1414
1da177e4 1415 /* Verify a few of the more important attributes */
fe51beec 1416 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1417 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1418 if (!data_unstable)
55296809 1419 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1420 }
ca62b9c3
TM		 1421
1422 cur_size = i_size_read(inode);
1423 new_isize = nfs_size_to_loff_t(fattr->size);
fb374d24
TM		 1424 if (cur_size != new_isize && nfsi->npages == 0)
1425 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4
LT		 1426
1427 /* Have any file permissions changed? */
1428 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1429 || inode->i_uid != fattr->uid
1430 || inode->i_gid != fattr->gid)
55296809 1431 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1432
1433 /* Has the link count changed? */
1434 if (inode->i_nlink != fattr->nlink)
55296809 1435 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4 1436
ca62b9c3 1437out:
1da177e4 1438 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1439 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1440
33801147 1441 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1442 return 0;
1443}
1444
33801147
TM		 1445/**
1446 * nfs_refresh_inode - try to update the inode attribute cache
1447 * @inode - pointer to inode
1448 * @fattr - updated attributes
1449 *
1450 * Check that an RPC call that returned attributes has not overlapped with
1451 * other recent updates of the inode metadata, then decide whether it is
1452 * safe to do a full update of the inode attributes, or whether just to
1453 * call nfs_check_inode_attributes.
1454 */
1455int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1456{
1457 struct nfs_inode *nfsi = NFS_I(inode);
1458 int status;
1459
1460 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1461 return 0;
1462 spin_lock(&inode->i_lock);
1463 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
33801147 1464 if (time_after(fattr->time_start, nfsi->last_updated))
24aa1fe6 1465 status = nfs_update_inode(inode, fattr);
33801147
TM		 1466 else
1467 status = nfs_check_inode_attributes(inode, fattr);
1468
1469 spin_unlock(&inode->i_lock);
1470 return status;
1471}
1472
decf491f
TM		 1473/**
1474 * nfs_post_op_update_inode - try to update the inode attribute cache
1475 * @inode - pointer to inode
1476 * @fattr - updated attributes
1477 *
1478 * After an operation that has changed the inode metadata, mark the
1479 * attribute cache as being invalid, then try to update it.
1480 */
1481int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1482{
1483 struct nfs_inode *nfsi = NFS_I(inode);
1484 int status = 0;
1485
1486 spin_lock(&inode->i_lock);
1487 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1488 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1489 goto out;
1490 }
24aa1fe6 1491 status = nfs_update_inode(inode, fattr);
decf491f
TM		 1492out:
1493 spin_unlock(&inode->i_lock);
1494 return status;
1495}
1496
1da177e4
LT		 1497/*
1498 * Many nfs protocol calls return the new file attributes after
1499 * an operation. Here we update the inode to reflect the state
1500 * of the server's inode.
1501 *
1502 * This is a bit tricky because we have to make sure all dirty pages
1503 * have been sent off to the server before calling invalidate_inode_pages.
1504 * To make sure no other process adds more write requests while we try
1505 * our best to flush them, we make them sleep during the attribute refresh.
1506 *
1507 * A very similar scenario holds for the dir cache.
1508 */
24aa1fe6 1509static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1510{
1511 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1512 loff_t cur_isize, new_isize;
1da177e4 1513 unsigned int invalid = 0;
24aa1fe6 1514 int data_stable;
1da177e4
LT		 1515
1516 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1517 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1518 atomic_read(&inode->i_count), fattr->valid);
1519
1520 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1521 return 0;
1522
325cfed9
CL		 1523 if (nfsi->fileid != fattr->fileid)
1524 goto out_fileid;
1da177e4
LT		 1525
1526 /*
1527 * Make sure the inode's type hasn't changed.
1528 */
33801147 1529 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT		 1530 goto out_changed;
1531
1532 /*
1533 * Update the read time so we don't revalidate too often.
1534 */
33801147
TM		 1535 nfsi->read_cache_jiffies = fattr->time_start;
1536 nfsi->last_updated = jiffies;
1da177e4
LT		 1537
1538 /* Are we racing with known updates of the metadata on the server? */
24aa1fe6
TM		 1539 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1540 if (data_stable)
1541 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1da177e4 1542
a895b4a1
TM		 1543 /* Do atomic weak cache consistency updates */
1544 nfs_wcc_update_inode(inode, fattr);
1545
951a143b 1546 /* Check if our cached file size is stale */
1da177e4
LT		 1547 new_isize = nfs_size_to_loff_t(fattr->size);
1548 cur_isize = i_size_read(inode);
951a143b
TM		 1549 if (new_isize != cur_isize) {
1550 /* Do we perhaps have any outstanding writes? */
1551 if (nfsi->npages == 0) {
1552 /* No, but did we race with nfs_end_data_update()? */
24aa1fe6 1553 if (data_stable) {
1da177e4 1554 inode->i_size = new_isize;
951a143b 1555 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1556 }
951a143b
TM		 1557 invalid |= NFS_INO_INVALID_ATTR;
1558 } else if (new_isize > cur_isize) {
1da177e4
LT		 1559 inode->i_size = new_isize;
1560 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1561 }
24aa1fe6 1562 nfsi->cache_change_attribute = jiffies;
951a143b
TM		 1563 dprintk("NFS: isize change on server for file %s/%ld\n",
1564 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1565 }
1566
951a143b 1567 /* Check if the mtime agrees */
1da177e4
LT		 1568 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1569 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1570 dprintk("NFS: mtime change on server for file %s/%ld\n",
1571 inode->i_sb->s_id, inode->i_ino);
24aa1fe6
TM		 1572 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1573 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1574 }
1575
ada70d94
TM		 1576 /* If ctime has changed we should definitely clear access+acl caches */
1577 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
24aa1fe6 1578 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
ada70d94 1579 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
24aa1fe6 1580 nfsi->cache_change_attribute = jiffies;
ada70d94 1581 }
1da177e4
LT		 1582 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1583
1584 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1585 inode->i_uid != fattr->uid ||
1586 inode->i_gid != fattr->gid)
ada70d94 1587 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1588
1589 inode->i_mode = fattr->mode;
1590 inode->i_nlink = fattr->nlink;
1591 inode->i_uid = fattr->uid;
1592 inode->i_gid = fattr->gid;
1593
1594 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1595 /*
1596 * report the blocks in 512byte units
1597 */
1598 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1599 inode->i_blksize = inode->i_sb->s_blocksize;
1600 } else {
1601 inode->i_blocks = fattr->du.nfs2.blocks;
1602 inode->i_blksize = fattr->du.nfs2.blocksize;
1603 }
1604
ca62b9c3
TM		 1605 if ((fattr->valid & NFS_ATTR_FATTR_V4)) {
1606 if (nfsi->change_attr != fattr->change_attr) {
1607 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1608 inode->i_sb->s_id, inode->i_ino);
1609 nfsi->change_attr = fattr->change_attr;
1610 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1611 nfsi->cache_change_attribute = jiffies;
1612 } else
1613 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA);
1614 }
1615
1da177e4
LT		 1616 /* Update attrtimeo value if we're out of the unstable period */
1617 if (invalid & NFS_INO_INVALID_ATTR) {
1618 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1619 nfsi->attrtimeo_timestamp = jiffies;
1620 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1621 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1622 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1623 nfsi->attrtimeo_timestamp = jiffies;
1624 }
1625 /* Don't invalidate the data if we were to blame */
1626 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1627 || S_ISLNK(inode->i_mode)))
1628 invalid &= ~NFS_INO_INVALID_DATA;
24aa1fe6
TM		 1629 if (data_stable)
1630 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1da177e4 1631 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1632 nfsi->cache_validity |= invalid;
1da177e4
LT		 1633
1634 return 0;
1635 out_changed:
1636 /*
1637 * Big trouble! The inode has become a different object.
1638 */
1639#ifdef NFS_PARANOIA
1640 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1641 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1642#endif
b37b03b7 1643 out_err:
1da177e4
LT		 1644 /*
1645 * No need to worry about unhashing the dentry, as the
1646 * lookup validation will know that the inode is bad.
1647 * (But we fall through to invalidate the caches.)
1648 */
1649 nfs_invalidate_inode(inode);
1da177e4 1650 return -ESTALE;
325cfed9
CL		 1651
1652 out_fileid:
1653 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1654 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1655 NFS_SERVER(inode)->hostname, inode->i_sb->s_id,
1656 (long long)nfsi->fileid, (long long)fattr->fileid);
1657 goto out_err;
1da177e4
LT		 1658}
1659
1660/*
1661 * File system information
1662 */
1663
1664static int nfs_set_super(struct super_block *s, void *data)
1665{
1666 s->s_fs_info = data;
1667 return set_anon_super(s, data);
1668}
1669
1670static int nfs_compare_super(struct super_block *sb, void *data)
1671{
1672 struct nfs_server *server = data;
1673 struct nfs_server *old = NFS_SB(sb);
1674
1675 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1676 return 0;
1677 if (old->addr.sin_port != server->addr.sin_port)
1678 return 0;
1679 return !nfs_compare_fh(&old->fh, &server->fh);
1680}
1681
1682static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1683 int flags, const char *dev_name, void *raw_data)
1684{
1685 int error;
9085bbcb 1686 struct nfs_server *server = NULL;
1da177e4
LT		 1687 struct super_block *s;
1688 struct nfs_fh *root;
1689 struct nfs_mount_data *data = raw_data;
1690
9085bbcb
TM		 1691 s = ERR_PTR(-EINVAL);
1692 if (data == NULL) {
1693 dprintk("%s: missing data argument\n", __FUNCTION__);
1694 goto out_err;
1695 }
1696 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1697 dprintk("%s: bad mount version\n", __FUNCTION__);
1698 goto out_err;
1da177e4 1699 }
9085bbcb
TM		 1700 switch (data->version) {
1701 case 1:
1702 data->namlen = 0;
1703 case 2:
1704 data->bsize = 0;
1705 case 3:
1706 if (data->flags & NFS_MOUNT_VER3) {
1707 dprintk("%s: mount structure version %d does not support NFSv3\n",
1708 __FUNCTION__,
1709 data->version);
1710 goto out_err;
1711 }
1712 data->root.size = NFS2_FHSIZE;
1713 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1714 case 4:
1715 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1716 dprintk("%s: mount structure version %d does not support strong security\n",
1717 __FUNCTION__,
1718 data->version);
1719 goto out_err;
1720 }
1721 case 5:
1722 memset(data->context, 0, sizeof(data->context));
1723 }
1724#ifndef CONFIG_NFS_V3
1725 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1726 s = ERR_PTR(-EPROTONOSUPPORT);
1727 if (data->flags & NFS_MOUNT_VER3) {
1728 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1729 goto out_err;
1730 }
1731#endif /* CONFIG_NFS_V3 */
1da177e4 1732
9085bbcb 1733 s = ERR_PTR(-ENOMEM);
bd647545 1734 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4 1735 if (!server)
9085bbcb 1736 goto out_err;
1da177e4
LT		 1737 /* Zero out the NFS state stuff */
1738 init_nfsv4_state(server);
b7fa0554 1739 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1740
1da177e4
LT		 1741 root = &server->fh;
1742 if (data->flags & NFS_MOUNT_VER3)
1743 root->size = data->root.size;
1744 else
1745 root->size = NFS2_FHSIZE;
9085bbcb 1746 s = ERR_PTR(-EINVAL);
1da177e4 1747 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1748 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1749 goto out_err;
1da177e4
LT		 1750 }
1751 memcpy(root->data, data->root.data, root->size);
1752
1753 /* We now require that the mount process passes the remote address */
1754 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1755 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1756 dprintk("%s: mount program didn't pass remote address!\n",
1757 __FUNCTION__);
1758 goto out_err;
1da177e4
LT		 1759 }
1760
9085bbcb
TM		 1761 /* Fire up rpciod if not yet running */
1762 s = ERR_PTR(rpciod_up());
1763 if (IS_ERR(s)) {
1764 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1765 __FUNCTION__, PTR_ERR(s));
1766 goto out_err;
1da177e4
LT		 1767 }
1768
9085bbcb
TM		 1769 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1770 if (IS_ERR(s) || s->s_root)
1771 goto out_rpciod_down;
1da177e4 1772
9085bbcb 1773 s->s_flags = flags;
1da177e4
LT		 1774
1775 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1776 if (error) {
1777 up_write(&s->s_umount);
1778 deactivate_super(s);
1779 return ERR_PTR(error);
1780 }
1781 s->s_flags |= MS_ACTIVE;
1782 return s;
9085bbcb
TM		 1783out_rpciod_down:
1784 rpciod_down();
1785out_err:
1786 kfree(server);
1787 return s;
1da177e4
LT		 1788}
1789
1790static void nfs_kill_super(struct super_block *s)
1791{
1792 struct nfs_server *server = NFS_SB(s);
1793
1794 kill_anon_super(s);
1795
6a19275a 1796 if (!IS_ERR(server->client))
1da177e4 1797 rpc_shutdown_client(server->client);
6a19275a 1798 if (!IS_ERR(server->client_sys))
1da177e4 1799 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1800 if (!IS_ERR(server->client_acl))
1801 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1802
1803 if (!(server->flags & NFS_MOUNT_NONLM))
1804 lockd_down(); /* release rpc.lockd */
1805
1806 rpciod_down(); /* release rpciod */
1807
f99d49ad 1808 kfree(server->hostname);
1da177e4
LT		 1809 kfree(server);
1810}
1811
1812static struct file_system_type nfs_fs_type = {
1813 .owner = THIS_MODULE,
1814 .name = "nfs",
1815 .get_sb = nfs_get_sb,
1816 .kill_sb = nfs_kill_super,
1817 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1818};
1819
1820#ifdef CONFIG_NFS_V4
1821
1822static void nfs4_clear_inode(struct inode *);
1823
1824
1825static struct super_operations nfs4_sops = {
1826 .alloc_inode = nfs_alloc_inode,
1827 .destroy_inode = nfs_destroy_inode,
1828 .write_inode = nfs_write_inode,
1829 .delete_inode = nfs_delete_inode,
1830 .statfs = nfs_statfs,
1831 .clear_inode = nfs4_clear_inode,
1832 .umount_begin = nfs_umount_begin,
1833 .show_options = nfs_show_options,
d9ef5a8c 1834 .show_stats = nfs_show_stats,
1da177e4
LT		 1835};
1836
1837/*
1838 * Clean out any remaining NFSv4 state that might be left over due
1839 * to open() calls that passed nfs_atomic_lookup, but failed to call
1840 * nfs_open().
1841 */
1842static void nfs4_clear_inode(struct inode *inode)
1843{
1844 struct nfs_inode *nfsi = NFS_I(inode);
1845
1846 /* If we are holding a delegation, return it! */
cae7a073 1847 nfs_inode_return_delegation(inode);
1da177e4
LT		 1848 /* First call standard NFS clear_inode() code */
1849 nfs_clear_inode(inode);
1850 /* Now clear out any remaining state */
1851 while (!list_empty(&nfsi->open_states)) {
1852 struct nfs4_state *state;
1853
1854 state = list_entry(nfsi->open_states.next,
1855 struct nfs4_state,
1856 inode_states);
1857 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1858 __FUNCTION__,
1859 inode->i_sb->s_id,
1860 (long long)NFS_FILEID(inode),
1861 state);
1862 BUG_ON(atomic_read(&state->count) != 1);
1863 nfs4_close_state(state, state->state);
1864 }
1865}
1866
1867
1868static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1869{
1870 struct nfs_server *server;
1871 struct nfs4_client *clp = NULL;
1872 struct rpc_xprt *xprt = NULL;
1873 struct rpc_clnt *clnt = NULL;
1874 struct rpc_timeout timeparms;
1875 rpc_authflavor_t authflavour;
eab5c084 1876 int err = -EIO;
1da177e4
LT		 1877
1878 sb->s_blocksize_bits = 0;
1879 sb->s_blocksize = 0;
1880 server = NFS_SB(sb);
1881 if (data->rsize != 0)
1882 server->rsize = nfs_block_size(data->rsize, NULL);
1883 if (data->wsize != 0)
1884 server->wsize = nfs_block_size(data->wsize, NULL);
1885 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1886 server->caps = NFS_CAP_ATOMIC_OPEN;
1887
1888 server->acregmin = data->acregmin*HZ;
1889 server->acregmax = data->acregmax*HZ;
1890 server->acdirmin = data->acdirmin*HZ;
1891 server->acdirmax = data->acdirmax*HZ;
1892
1893 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1894
eab5c084 1895 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4 1896
7a480e25
CL		 1897 server->retrans_timeo = timeparms.to_initval;
1898 server->retrans_count = timeparms.to_retries;
1899
1da177e4
LT		 1900 clp = nfs4_get_client(&server->addr.sin_addr);
1901 if (!clp) {
9085bbcb 1902 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1903 return -EIO;
1904 }
1905
1906 /* Now create transport and client */
1907 authflavour = RPC_AUTH_UNIX;
1908 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1909 if (data->auth_flavourlen != 1) {
1910 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1911 __FUNCTION__, data->auth_flavourlen);
1912 err = -EINVAL;
1913 goto out_fail;
1914 }
1da177e4
LT		 1915 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1916 err = -EFAULT;
1917 goto out_fail;
1918 }
1919 }
1920
1921 down_write(&clp->cl_sem);
6a19275a 1922 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1923 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT		 1924 if (IS_ERR(xprt)) {
1925 up_write(&clp->cl_sem);
1da177e4 1926 err = PTR_ERR(xprt);
9085bbcb
TM		 1927 dprintk("%s: cannot create RPC transport. Error = %d\n",
1928 __FUNCTION__, err);
1da177e4
LT		 1929 goto out_fail;
1930 }
1931 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1932 server->rpc_ops->version, authflavour);
1933 if (IS_ERR(clnt)) {
1934 up_write(&clp->cl_sem);
1da177e4 1935 err = PTR_ERR(clnt);
9085bbcb
TM		 1936 dprintk("%s: cannot create RPC client. Error = %d\n",
1937 __FUNCTION__, err);
1da177e4
LT		 1938 goto out_fail;
1939 }
1940 clnt->cl_intr = 1;
1941 clnt->cl_softrtry = 1;
1da177e4 1942 clp->cl_rpcclient = clnt;
1da177e4
LT		 1943 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1944 nfs_idmap_new(clp);
1945 }
1da177e4
LT		 1946 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1947 clnt = rpc_clone_client(clp->cl_rpcclient);
1948 if (!IS_ERR(clnt))
1949 server->nfs4_state = clp;
1950 up_write(&clp->cl_sem);
1951 clp = NULL;
1952
1953 if (IS_ERR(clnt)) {
9085bbcb
TM		 1954 err = PTR_ERR(clnt);
1955 dprintk("%s: cannot create RPC client. Error = %d\n",
1956 __FUNCTION__, err);
1957 return err;
1da177e4
LT		 1958 }
1959
1960 server->client = clnt;
1961
1962 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1963 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1964 return -ENOMEM;
1965 }
1966
1967 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1968 struct rpc_auth *auth;
1969
1970 auth = rpcauth_create(authflavour, clnt);
1971 if (IS_ERR(auth)) {
9085bbcb 1972 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1973 return PTR_ERR(auth);
1da177e4
LT		 1974 }
1975 }
1976
1977 sb->s_time_gran = 1;
1978
1979 sb->s_op = &nfs4_sops;
1980 err = nfs_sb_init(sb, authflavour);
1981 if (err == 0)
1982 return 0;
1983out_fail:
1984 if (clp)
1985 nfs4_put_client(clp);
1986 return err;
1987}
1988
1989static int nfs4_compare_super(struct super_block *sb, void *data)
1990{
1991 struct nfs_server *server = data;
1992 struct nfs_server *old = NFS_SB(sb);
1993
1994 if (strcmp(server->hostname, old->hostname) != 0)
1995 return 0;
1996 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1997 return 0;
1998 return 1;
1999}
2000
2001static void *
2002nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
2003{
2004 void *p = NULL;
2005
2006 if (!src->len)
2007 return ERR_PTR(-EINVAL);
2008 if (src->len < maxlen)
2009 maxlen = src->len;
2010 if (dst == NULL) {
2011 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
2012 if (p == NULL)
2013 return ERR_PTR(-ENOMEM);
2014 }
2015 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 2016 kfree(p);
1da177e4
LT		 2017 return ERR_PTR(-EFAULT);
2018 }
2019 dst[maxlen] = '\0';
2020 return dst;
2021}
2022
2023static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
2024 int flags, const char *dev_name, void *raw_data)
2025{
2026 int error;
2027 struct nfs_server *server;
2028 struct super_block *s;
2029 struct nfs4_mount_data *data = raw_data;
2030 void *p;
2031
9085bbcb
TM		 2032 if (data == NULL) {
2033 dprintk("%s: missing data argument\n", __FUNCTION__);
2034 return ERR_PTR(-EINVAL);
2035 }
2036 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
2037 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 2038 return ERR_PTR(-EINVAL);
2039 }
2040
bd647545 2041 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4
LT		 2042 if (!server)
2043 return ERR_PTR(-ENOMEM);
1da177e4
LT		 2044 /* Zero out the NFS state stuff */
2045 init_nfsv4_state(server);
b7fa0554 2046 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 2047
1da177e4
LT		 2048 p = nfs_copy_user_string(NULL, &data->hostname, 256);
2049 if (IS_ERR(p))
2050 goto out_err;
2051 server->hostname = p;
2052
2053 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
2054 if (IS_ERR(p))
2055 goto out_err;
2056 server->mnt_path = p;
2057
2058 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
2059 sizeof(server->ip_addr) - 1);
2060 if (IS_ERR(p))
2061 goto out_err;
2062
2063 /* We now require that the mount process passes the remote address */
2064 if (data->host_addrlen != sizeof(server->addr)) {
2065 s = ERR_PTR(-EINVAL);
2066 goto out_free;
2067 }
2068 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
2069 s = ERR_PTR(-EFAULT);
2070 goto out_free;
2071 }
2072 if (server->addr.sin_family != AF_INET ||
2073 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 2074 dprintk("%s: mount program didn't pass remote IP address!\n",
2075 __FUNCTION__);
1da177e4
LT		 2076 s = ERR_PTR(-EINVAL);
2077 goto out_free;
2078 }
2079
9085bbcb
TM		 2080 /* Fire up rpciod if not yet running */
2081 s = ERR_PTR(rpciod_up());
2082 if (IS_ERR(s)) {
2083 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2084 __FUNCTION__, PTR_ERR(s));
2085 goto out_free;
2086 }
2087
1da177e4
LT		 2088 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
2089
2090 if (IS_ERR(s) || s->s_root)
2091 goto out_free;
2092
2093 s->s_flags = flags;
2094
1da177e4
LT		 2095 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
2096 if (error) {
2097 up_write(&s->s_umount);
2098 deactivate_super(s);
2099 return ERR_PTR(error);
2100 }
2101 s->s_flags |= MS_ACTIVE;
2102 return s;
2103out_err:
2104 s = (struct super_block *)p;
2105out_free:
f99d49ad
JJ		 2106 kfree(server->mnt_path);
2107 kfree(server->hostname);
d9ef5a8c 2108 nfs_free_iostats(server->io_stats);
1da177e4
LT		 2109 kfree(server);
2110 return s;
2111}
2112
2113static void nfs4_kill_super(struct super_block *sb)
2114{
2115 struct nfs_server *server = NFS_SB(sb);
2116
2117 nfs_return_all_delegations(sb);
2118 kill_anon_super(sb);
2119
2120 nfs4_renewd_prepare_shutdown(server);
2121
2122 if (server->client != NULL && !IS_ERR(server->client))
2123 rpc_shutdown_client(server->client);
1da177e4
LT		 2124
2125 destroy_nfsv4_state(server);
2126
967b9281
TM		 2127 rpciod_down();
2128
f99d49ad 2129 kfree(server->hostname);
1da177e4
LT		 2130 kfree(server);
2131}
2132
2133static struct file_system_type nfs4_fs_type = {
2134 .owner = THIS_MODULE,
2135 .name = "nfs4",
2136 .get_sb = nfs4_get_sb,
2137 .kill_sb = nfs4_kill_super,
2138 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2139};
2140
a72b4422
TM		 2141static const int nfs_set_port_min = 0;
2142static const int nfs_set_port_max = 65535;
2143static int param_set_port(const char *val, struct kernel_param *kp)
2144{
2145 char *endp;
2146 int num = simple_strtol(val, &endp, 0);
2147 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
2148 return -EINVAL;
2149 *((int *)kp->arg) = num;
2150 return 0;
2151}
2152
2153module_param_call(callback_tcpport, param_set_port, param_get_int,
2154 &nfs_callback_set_tcpport, 0644);
2155
58df095b
TM		 2156static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
2157{
2158 char *endp;
2159 int num = simple_strtol(val, &endp, 0);
2160 int jif = num * HZ;
2161 if (endp == val || *endp || num < 0 || jif < num)
2162 return -EINVAL;
2163 *((int *)kp->arg) = jif;
2164 return 0;
2165}
2166
2167module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
2168 &nfs_idmap_cache_timeout, 0644);
2169
1da177e4
LT		 2170#define nfs4_init_once(nfsi) \
2171 do { \
2172 INIT_LIST_HEAD(&(nfsi)->open_states); \
2173 nfsi->delegation = NULL; \
2174 nfsi->delegation_state = 0; \
2175 init_rwsem(&nfsi->rwsem); \
2176 } while(0)
a72b4422
TM		 2177
2178static inline int register_nfs4fs(void)
2179{
2180 int ret;
2181
2182 ret = nfs_register_sysctl();
2183 if (ret != 0)
2184 return ret;
2185 ret = register_filesystem(&nfs4_fs_type);
2186 if (ret != 0)
2187 nfs_unregister_sysctl();
2188 return ret;
2189}
2190
2191static inline void unregister_nfs4fs(void)
2192{
2193 unregister_filesystem(&nfs4_fs_type);
2194 nfs_unregister_sysctl();
2195}
1da177e4
LT		 2196#else
2197#define nfs4_init_once(nfsi) \
2198 do { } while (0)
2199#define register_nfs4fs() (0)
2200#define unregister_nfs4fs()
2201#endif
2202
2203extern int nfs_init_nfspagecache(void);
2204extern void nfs_destroy_nfspagecache(void);
2205extern int nfs_init_readpagecache(void);
2206extern void nfs_destroy_readpagecache(void);
2207extern int nfs_init_writepagecache(void);
2208extern void nfs_destroy_writepagecache(void);
2209#ifdef CONFIG_NFS_DIRECTIO
2210extern int nfs_init_directcache(void);
2211extern void nfs_destroy_directcache(void);
2212#endif
2213
2214static kmem_cache_t * nfs_inode_cachep;
2215
2216static struct inode *nfs_alloc_inode(struct super_block *sb)
2217{
2218 struct nfs_inode *nfsi;
2219 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2220 if (!nfsi)
2221 return NULL;
55296809
CL		 2222 nfsi->flags = 0UL;
2223 nfsi->cache_validity = 0UL;
223db122 2224 nfsi->cache_change_attribute = jiffies;
458818ed
TM		 2225#ifdef CONFIG_NFS_V3_ACL
2226 nfsi->acl_access = ERR_PTR(-EAGAIN);
2227 nfsi->acl_default = ERR_PTR(-EAGAIN);
2228#endif
e50a1c2e
BF		 2229#ifdef CONFIG_NFS_V4
2230 nfsi->nfs4_acl = NULL;
2231#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2232 return &nfsi->vfs_inode;
2233}
2234
2235static void nfs_destroy_inode(struct inode *inode)
2236{
2237 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2238}
2239
2240static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2241{
2242 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2243
2244 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2245 SLAB_CTOR_CONSTRUCTOR) {
2246 inode_init_once(&nfsi->vfs_inode);
2247 spin_lock_init(&nfsi->req_lock);
2248 INIT_LIST_HEAD(&nfsi->dirty);
2249 INIT_LIST_HEAD(&nfsi->commit);
2250 INIT_LIST_HEAD(&nfsi->open_files);
2251 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2252 atomic_set(&nfsi->data_updates, 0);
2253 nfsi->ndirty = 0;
2254 nfsi->ncommit = 0;
2255 nfsi->npages = 0;
1da177e4
LT		 2256 nfs4_init_once(nfsi);
2257 }
2258}
2259
75c96f85 2260static int nfs_init_inodecache(void)
1da177e4
LT		 2261{
2262 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2263 sizeof(struct nfs_inode),
2264 0, SLAB_RECLAIM_ACCOUNT,
2265 init_once, NULL);
2266 if (nfs_inode_cachep == NULL)
2267 return -ENOMEM;
2268
2269 return 0;
2270}
2271
75c96f85 2272static void nfs_destroy_inodecache(void)
1da177e4
LT		 2273{
2274 if (kmem_cache_destroy(nfs_inode_cachep))
2275 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2276}
2277
2278/*
2279 * Initialize NFS
2280 */
2281static int __init init_nfs_fs(void)
2282{
2283 int err;
2284
2285 err = nfs_init_nfspagecache();
2286 if (err)
2287 goto out4;
2288
2289 err = nfs_init_inodecache();
2290 if (err)
2291 goto out3;
2292
2293 err = nfs_init_readpagecache();
2294 if (err)
2295 goto out2;
2296
2297 err = nfs_init_writepagecache();
2298 if (err)
2299 goto out1;
2300
2301#ifdef CONFIG_NFS_DIRECTIO
2302 err = nfs_init_directcache();
2303 if (err)
2304 goto out0;
2305#endif
2306
2307#ifdef CONFIG_PROC_FS
2308 rpc_proc_register(&nfs_rpcstat);
2309#endif
2310 err = register_filesystem(&nfs_fs_type);
2311 if (err)
2312 goto out;
2313 if ((err = register_nfs4fs()) != 0)
2314 goto out;
2315 return 0;
2316out:
2317#ifdef CONFIG_PROC_FS
2318 rpc_proc_unregister("nfs");
2319#endif
1da177e4 2320#ifdef CONFIG_NFS_DIRECTIO
1da177e4 2321 nfs_destroy_directcache();
6b59a754 2322out0:
1da177e4 2323#endif
6b59a754 2324 nfs_destroy_writepagecache();
1da177e4
LT		 2325out1:
2326 nfs_destroy_readpagecache();
2327out2:
2328 nfs_destroy_inodecache();
2329out3:
2330 nfs_destroy_nfspagecache();
2331out4:
2332 return err;
2333}
2334
2335static void __exit exit_nfs_fs(void)
2336{
2337#ifdef CONFIG_NFS_DIRECTIO
2338 nfs_destroy_directcache();
2339#endif
2340 nfs_destroy_writepagecache();
2341 nfs_destroy_readpagecache();
2342 nfs_destroy_inodecache();
2343 nfs_destroy_nfspagecache();
2344#ifdef CONFIG_PROC_FS
2345 rpc_proc_unregister("nfs");
2346#endif
2347 unregister_filesystem(&nfs_fs_type);
2348 unregister_nfs4fs();
2349}
2350
2351/* Not quite true; I just maintain it */
2352MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2353MODULE_LICENSE("GPL");
2354
2355module_init(init_nfs_fs)
2356module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)