projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFS: clean up some mount options
[GitHub/mt8127/android_kernel_alcatel_ttab.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
LT		 44#include "delegation.h"
45
46#define NFSDBG_FACILITY NFSDBG_VFS
47#define NFS_PARANOIA 1
48
49/* Maximum number of readahead requests
50 * FIXME: this should really be a sysctl so that users may tune it to suit
51 * their needs. People that do NFS over a slow network, might for
52 * instance want to reduce it to something closer to 1 for improved
53 * interactive response.
54 */
55#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
56
57static void nfs_invalidate_inode(struct inode *);
24aa1fe6 58static int nfs_update_inode(struct inode *, struct nfs_fattr *);
1da177e4
LT		 59
60static struct inode *nfs_alloc_inode(struct super_block *sb);
61static void nfs_destroy_inode(struct inode *);
62static int nfs_write_inode(struct inode *,int);
63static void nfs_delete_inode(struct inode *);
64static void nfs_clear_inode(struct inode *);
65static void nfs_umount_begin(struct super_block *);
66static int nfs_statfs(struct super_block *, struct kstatfs *);
67static int nfs_show_options(struct seq_file *, struct vfsmount *);
ada70d94 68static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT		 69
70static struct rpc_program nfs_program;
71
72static struct super_operations nfs_sops = {
73 .alloc_inode = nfs_alloc_inode,
74 .destroy_inode = nfs_destroy_inode,
75 .write_inode = nfs_write_inode,
76 .delete_inode = nfs_delete_inode,
77 .statfs = nfs_statfs,
78 .clear_inode = nfs_clear_inode,
79 .umount_begin = nfs_umount_begin,
80 .show_options = nfs_show_options,
81};
82
83/*
84 * RPC cruft for NFS
85 */
86static struct rpc_stat nfs_rpcstat = {
87 .program = &nfs_program
88};
89static struct rpc_version * nfs_version[] = {
90 NULL,
91 NULL,
92 &nfs_version2,
93#if defined(CONFIG_NFS_V3)
94 &nfs_version3,
95#elif defined(CONFIG_NFS_V4)
96 NULL,
97#endif
98#if defined(CONFIG_NFS_V4)
99 &nfs_version4,
100#endif
101};
102
103static struct rpc_program nfs_program = {
104 .name = "nfs",
105 .number = NFS_PROGRAM,
106 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
107 .version = nfs_version,
108 .stats = &nfs_rpcstat,
109 .pipe_dir_name = "/nfs",
110};
111
b7fa0554
AG		 112#ifdef CONFIG_NFS_V3_ACL
113static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
114static struct rpc_version * nfsacl_version[] = {
115 [3] = &nfsacl_version3,
116};
117
118struct rpc_program nfsacl_program = {
119 .name = "nfsacl",
120 .number = NFS_ACL_PROGRAM,
121 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
122 .version = nfsacl_version,
123 .stats = &nfsacl_rpcstat,
124};
125#endif /* CONFIG_NFS_V3_ACL */
126
1da177e4
LT		 127static inline unsigned long
128nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
129{
130 return nfs_fileid_to_ino_t(fattr->fileid);
131}
132
133static int
134nfs_write_inode(struct inode *inode, int sync)
135{
136 int flags = sync ? FLUSH_WAIT : 0;
137 int ret;
138
3da28eb1 139 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 140 if (ret < 0)
141 return ret;
142 return 0;
143}
144
145static void
146nfs_delete_inode(struct inode * inode)
147{
148 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
149
fef26658
MF		 150 truncate_inode_pages(&inode->i_data, 0);
151
1da177e4
LT		 152 nfs_wb_all(inode);
153 /*
154 * The following should never happen...
155 */
156 if (nfs_have_writebacks(inode)) {
157 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
158 }
159
160 clear_inode(inode);
161}
162
1da177e4
LT		 163static void
164nfs_clear_inode(struct inode *inode)
165{
166 struct nfs_inode *nfsi = NFS_I(inode);
167 struct rpc_cred *cred;
168
169 nfs_wb_all(inode);
170 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 171 nfs_zap_acl_cache(inode);
1da177e4
LT		 172 cred = nfsi->cache_access.cred;
173 if (cred)
174 put_rpccred(cred);
175 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
176}
177
178void
179nfs_umount_begin(struct super_block *sb)
180{
6a19275a 181 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT		 182
183 /* -EIO all pending I/O */
6a19275a 184 if (!IS_ERR(rpc))
1da177e4 185 rpc_killall_tasks(rpc);
b7fa0554
AG		 186 rpc = NFS_SB(sb)->client_acl;
187 if (!IS_ERR(rpc))
188 rpc_killall_tasks(rpc);
1da177e4
LT		 189}
190
191
192static inline unsigned long
193nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
194{
195 /* make sure blocksize is a power of two */
196 if ((bsize & (bsize - 1)) || nrbitsp) {
197 unsigned char nrbits;
198
199 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
200 ;
201 bsize = 1 << nrbits;
202 if (nrbitsp)
203 *nrbitsp = nrbits;
204 }
205
206 return bsize;
207}
208
209/*
210 * Calculate the number of 512byte blocks used.
211 */
212static inline unsigned long
213nfs_calc_block_size(u64 tsize)
214{
215 loff_t used = (tsize + 511) >> 9;
216 return (used > ULONG_MAX) ? ULONG_MAX : used;
217}
218
219/*
220 * Compute and set NFS server blocksize
221 */
222static inline unsigned long
223nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
224{
40859d7e
CL		 225 if (bsize < NFS_MIN_FILE_IO_SIZE)
226 bsize = NFS_DEF_FILE_IO_SIZE;
227 else if (bsize >= NFS_MAX_FILE_IO_SIZE)
228 bsize = NFS_MAX_FILE_IO_SIZE;
1da177e4
LT		 229
230 return nfs_block_bits(bsize, nrbitsp);
231}
232
233/*
234 * Obtain the root inode of the file system.
235 */
236static struct inode *
237nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
238{
239 struct nfs_server *server = NFS_SB(sb);
240 struct inode *rooti;
241 int error;
242
243 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
244 if (error < 0) {
245 dprintk("nfs_get_root: getattr error = %d\n", -error);
246 return ERR_PTR(error);
247 }
248
249 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
250 if (!rooti)
251 return ERR_PTR(-ENOMEM);
252 return rooti;
253}
254
255/*
256 * Do NFS version-independent mount processing, and sanity checking
257 */
258static int
259nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
260{
261 struct nfs_server *server;
262 struct inode *root_inode;
263 struct nfs_fattr fattr;
264 struct nfs_fsinfo fsinfo = {
265 .fattr = &fattr,
266 };
267 struct nfs_pathconf pathinfo = {
268 .fattr = &fattr,
269 };
270 int no_root_error = 0;
271 unsigned long max_rpc_payload;
272
273 /* We probably want something more informative here */
274 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
275
276 server = NFS_SB(sb);
277
278 sb->s_magic = NFS_SUPER_MAGIC;
279
280 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
281 /* Did getting the root inode fail? */
282 if (IS_ERR(root_inode)) {
283 no_root_error = PTR_ERR(root_inode);
284 goto out_no_root;
285 }
286 sb->s_root = d_alloc_root(root_inode);
287 if (!sb->s_root) {
288 no_root_error = -ENOMEM;
289 goto out_no_root;
290 }
291 sb->s_root->d_op = server->rpc_ops->dentry_ops;
292
293 /* Get some general file system info */
294 if (server->namelen == 0 &&
295 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
296 server->namelen = pathinfo.max_namelen;
297 /* Work out a lot of parameters */
298 if (server->rsize == 0)
299 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
300 if (server->wsize == 0)
301 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
302
303 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
304 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
305 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
306 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
307
308 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
309 if (server->rsize > max_rpc_payload)
310 server->rsize = max_rpc_payload;
40859d7e
CL		 311 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
312 server->rsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 313 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4 314
40859d7e
CL		 315 if (server->wsize > max_rpc_payload)
316 server->wsize = max_rpc_payload;
317 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
318 server->wsize = NFS_MAX_FILE_IO_SIZE;
1da177e4 319 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1da177e4
LT		 320
321 if (sb->s_blocksize == 0)
322 sb->s_blocksize = nfs_block_bits(server->wsize,
323 &sb->s_blocksize_bits);
324 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
325
326 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
327 if (server->dtsize > PAGE_CACHE_SIZE)
328 server->dtsize = PAGE_CACHE_SIZE;
329 if (server->dtsize > server->rsize)
330 server->dtsize = server->rsize;
331
332 if (server->flags & NFS_MOUNT_NOAC) {
333 server->acregmin = server->acregmax = 0;
334 server->acdirmin = server->acdirmax = 0;
335 sb->s_flags |= MS_SYNCHRONOUS;
336 }
337 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
338
339 sb->s_maxbytes = fsinfo.maxfilesize;
340 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
341 sb->s_maxbytes = MAX_LFS_FILESIZE;
342
343 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
344 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
345
346 /* We're airborne Set socket buffersize */
347 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
348 return 0;
349 /* Yargs. It didn't work out. */
350out_no_root:
351 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
352 if (!IS_ERR(root_inode))
353 iput(root_inode);
354 return no_root_error;
355}
356
eab5c084
CL		 357static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
358{
359 to->to_initval = timeo * HZ / 10;
360 to->to_retries = retrans;
361 if (!to->to_retries)
362 to->to_retries = 2;
363
364 switch (proto) {
365 case IPPROTO_TCP:
366 if (!to->to_initval)
367 to->to_initval = 60 * HZ;
03bf4b70
CL		 368 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
369 to->to_initval = NFS_MAX_TCP_TIMEOUT;
eab5c084
CL		 370 to->to_increment = to->to_initval;
371 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
372 to->to_exponential = 0;
373 break;
374 case IPPROTO_UDP:
375 default:
376 if (!to->to_initval)
377 to->to_initval = 11 * HZ / 10;
03bf4b70
CL		 378 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
379 to->to_initval = NFS_MAX_UDP_TIMEOUT;
380 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
eab5c084
CL		 381 to->to_exponential = 1;
382 break;
383 }
384}
385
1da177e4
LT		 386/*
387 * Create an RPC client handle.
388 */
389static struct rpc_clnt *
390nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
391{
392 struct rpc_timeout timeparms;
393 struct rpc_xprt *xprt = NULL;
394 struct rpc_clnt *clnt = NULL;
eab5c084 395 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
1da177e4 396
eab5c084 397 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
1da177e4 398
7a480e25
CL		 399 server->retrans_timeo = timeparms.to_initval;
400 server->retrans_count = timeparms.to_retries;
401
1da177e4 402 /* create transport and client */
eab5c084 403 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 404 if (IS_ERR(xprt)) {
9085bbcb
TM		 405 dprintk("%s: cannot create RPC transport. Error = %ld\n",
406 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 407 return (struct rpc_clnt *)xprt;
408 }
409 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
410 server->rpc_ops->version, data->pseudoflavor);
411 if (IS_ERR(clnt)) {
9085bbcb
TM		 412 dprintk("%s: cannot create RPC client. Error = %ld\n",
413 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 414 goto out_fail;
415 }
416
417 clnt->cl_intr = 1;
418 clnt->cl_softrtry = 1;
1da177e4
LT		 419
420 return clnt;
421
422out_fail:
1da177e4
LT		 423 return clnt;
424}
425
426/*
427 * The way this works is that the mount process passes a structure
428 * in the data argument which contains the server's IP address
429 * and the root file handle obtained from the server's mount
430 * daemon. We stash these away in the private superblock fields.
431 */
432static int
433nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
434{
435 struct nfs_server *server;
436 rpc_authflavor_t authflavor;
437
438 server = NFS_SB(sb);
439 sb->s_blocksize_bits = 0;
440 sb->s_blocksize = 0;
441 if (data->bsize)
442 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
443 if (data->rsize)
444 server->rsize = nfs_block_size(data->rsize, NULL);
445 if (data->wsize)
446 server->wsize = nfs_block_size(data->wsize, NULL);
447 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
448
449 server->acregmin = data->acregmin*HZ;
450 server->acregmax = data->acregmax*HZ;
451 server->acdirmin = data->acdirmin*HZ;
452 server->acdirmax = data->acdirmax*HZ;
453
454 /* Start lockd here, before we might error out */
455 if (!(server->flags & NFS_MOUNT_NONLM))
456 lockd_up();
457
458 server->namelen = data->namlen;
459 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
460 if (!server->hostname)
461 return -ENOMEM;
462 strcpy(server->hostname, data->hostname);
463
464 /* Check NFS protocol revision and initialize RPC op vector
465 * and file handle pool. */
1da177e4 466#ifdef CONFIG_NFS_V3
9085bbcb 467 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 468 server->rpc_ops = &nfs_v3_clientops;
469 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 470 } else {
471 server->rpc_ops = &nfs_v2_clientops;
472 }
9085bbcb
TM		 473#else
474 server->rpc_ops = &nfs_v2_clientops;
475#endif
1da177e4
LT		 476
477 /* Fill in pseudoflavor for mount version < 5 */
478 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
479 data->pseudoflavor = RPC_AUTH_UNIX;
480 authflavor = data->pseudoflavor; /* save for sb_init() */
481 /* XXX maybe we want to add a server->pseudoflavor field */
482
483 /* Create RPC client handles */
484 server->client = nfs_create_client(server, data);
485 if (IS_ERR(server->client))
486 return PTR_ERR(server->client);
487 /* RFC 2623, sec 2.3.2 */
488 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 489 struct rpc_auth *auth;
490
1da177e4
LT		 491 server->client_sys = rpc_clone_client(server->client);
492 if (IS_ERR(server->client_sys))
493 return PTR_ERR(server->client_sys);
6a19275a
BF		 494 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
495 if (IS_ERR(auth))
496 return PTR_ERR(auth);
1da177e4
LT		 497 } else {
498 atomic_inc(&server->client->cl_count);
499 server->client_sys = server->client;
500 }
1da177e4 501 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 502#ifdef CONFIG_NFS_V3_ACL
503 if (!(server->flags & NFS_MOUNT_NOACL)) {
504 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
505 /* No errors! Assume that Sun nfsacls are supported */
506 if (!IS_ERR(server->client_acl))
507 server->caps |= NFS_CAP_ACLS;
508 }
509#else
510 server->flags &= ~NFS_MOUNT_NOACL;
511#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 512 /*
513 * The VFS shouldn't apply the umask to mode bits. We will
514 * do so ourselves when necessary.
515 */
516 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 517 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
518 server->namelen = NFS3_MAXNAMLEN;
519 sb->s_time_gran = 1;
520 } else {
521 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
522 server->namelen = NFS2_MAXNAMLEN;
523 }
524
525 sb->s_op = &nfs_sops;
526 return nfs_sb_init(sb, authflavor);
527}
528
529static int
530nfs_statfs(struct super_block *sb, struct kstatfs *buf)
531{
532 struct nfs_server *server = NFS_SB(sb);
533 unsigned char blockbits;
534 unsigned long blockres;
535 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
536 struct nfs_fattr fattr;
537 struct nfs_fsstat res = {
538 .fattr = &fattr,
539 };
540 int error;
541
542 lock_kernel();
543
544 error = server->rpc_ops->statfs(server, rootfh, &res);
545 buf->f_type = NFS_SUPER_MAGIC;
546 if (error < 0)
547 goto out_err;
548
549 /*
550 * Current versions of glibc do not correctly handle the
551 * case where f_frsize != f_bsize. Eventually we want to
552 * report the value of wtmult in this field.
553 */
554 buf->f_frsize = sb->s_blocksize;
555
556 /*
557 * On most *nix systems, f_blocks, f_bfree, and f_bavail
558 * are reported in units of f_frsize. Linux hasn't had
559 * an f_frsize field in its statfs struct until recently,
560 * thus historically Linux's sys_statfs reports these
561 * fields in units of f_bsize.
562 */
563 buf->f_bsize = sb->s_blocksize;
564 blockbits = sb->s_blocksize_bits;
565 blockres = (1 << blockbits) - 1;
566 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
567 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
568 buf->f_bavail = (res.abytes + blockres) >> blockbits;
569
570 buf->f_files = res.tfiles;
571 buf->f_ffree = res.afiles;
572
573 buf->f_namelen = server->namelen;
574 out:
575 unlock_kernel();
1da177e4
LT		 576 return 0;
577
578 out_err:
dc20f803 579 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
1da177e4
LT		 580 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
581 goto out;
582
583}
584
585static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
586{
587 static struct proc_nfs_info {
588 int flag;
589 char *str;
590 char *nostr;
591 } nfs_info[] = {
592 { NFS_MOUNT_SOFT, ",soft", ",hard" },
593 { NFS_MOUNT_INTR, ",intr", "" },
1da177e4
LT		 594 { NFS_MOUNT_NOCTO, ",nocto", "" },
595 { NFS_MOUNT_NOAC, ",noac", "" },
c8bded96 596 { NFS_MOUNT_NONLM, ",nolock", "" },
b7fa0554 597 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 598 { 0, NULL, NULL }
599 };
600 struct proc_nfs_info *nfs_infop;
601 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
3063d8a1
TM		 602 char buf[12];
603 char *proto;
1da177e4 604
c8bded96 605 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
1da177e4
LT		 606 seq_printf(m, ",rsize=%d", nfss->rsize);
607 seq_printf(m, ",wsize=%d", nfss->wsize);
608 if (nfss->acregmin != 3*HZ)
609 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
610 if (nfss->acregmax != 60*HZ)
611 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
612 if (nfss->acdirmin != 30*HZ)
613 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
614 if (nfss->acdirmax != 60*HZ)
615 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
616 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
617 if (nfss->flags & nfs_infop->flag)
618 seq_puts(m, nfs_infop->str);
619 else
620 seq_puts(m, nfs_infop->nostr);
621 }
3063d8a1
TM		 622 switch (nfss->client->cl_xprt->prot) {
623 case IPPROTO_TCP:
624 proto = "tcp";
625 break;
626 case IPPROTO_UDP:
627 proto = "udp";
628 break;
629 default:
630 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
631 proto = buf;
632 }
633 seq_printf(m, ",proto=%s", proto);
7a480e25
CL		 634 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
635 seq_printf(m, ",retrans=%u", nfss->retrans_count);
1da177e4
LT		 636 seq_puts(m, ",addr=");
637 seq_escape(m, nfss->hostname, " \t\n\\");
638 return 0;
639}
640
29884df0
TM		 641/**
642 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
643 */
644int nfs_sync_mapping(struct address_space *mapping)
645{
646 int ret;
647
648 if (mapping->nrpages == 0)
649 return 0;
650 unmap_mapping_range(mapping, 0, 0, 0);
28fd1298 651 ret = filemap_write_and_wait(mapping);
29884df0
TM		 652 if (ret != 0)
653 goto out;
654 ret = nfs_wb_all(mapping->host);
655out:
656 return ret;
657}
658
1da177e4
LT		 659/*
660 * Invalidate the local caches
661 */
b37b03b7 662static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT		 663{
664 struct nfs_inode *nfsi = NFS_I(inode);
665 int mode = inode->i_mode;
666
667 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
668 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
669
670 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
671 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 672 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 673 else
55296809 674 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 675}
dc59250c 676
b37b03b7
TM		 677void nfs_zap_caches(struct inode *inode)
678{
679 spin_lock(&inode->i_lock);
680 nfs_zap_caches_locked(inode);
dc59250c 681 spin_unlock(&inode->i_lock);
ada70d94
TM		 682}
683
684static void nfs_zap_acl_cache(struct inode *inode)
685{
686 void (*clear_acl_cache)(struct inode *);
687
688 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
689 if (clear_acl_cache != NULL)
690 clear_acl_cache(inode);
dc59250c 691 spin_lock(&inode->i_lock);
55296809 692 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 693 spin_unlock(&inode->i_lock);
1da177e4
LT		 694}
695
696/*
b37b03b7
TM		 697 * Invalidate, but do not unhash, the inode.
698 * NB: must be called with inode->i_lock held!
1da177e4 699 */
b37b03b7 700static void nfs_invalidate_inode(struct inode *inode)
1da177e4 701{
b37b03b7
TM		 702 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
703 nfs_zap_caches_locked(inode);
1da177e4
LT		 704}
705
706struct nfs_find_desc {
707 struct nfs_fh *fh;
708 struct nfs_fattr *fattr;
709};
710
711/*
712 * In NFSv3 we can have 64bit inode numbers. In order to support
713 * this, and re-exported directories (also seen in NFSv2)
714 * we are forced to allow 2 different inodes to have the same
715 * i_ino.
716 */
717static int
718nfs_find_actor(struct inode *inode, void *opaque)
719{
720 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
721 struct nfs_fh *fh = desc->fh;
722 struct nfs_fattr *fattr = desc->fattr;
723
724 if (NFS_FILEID(inode) != fattr->fileid)
725 return 0;
726 if (nfs_compare_fh(NFS_FH(inode), fh))
727 return 0;
728 if (is_bad_inode(inode) || NFS_STALE(inode))
729 return 0;
730 return 1;
731}
732
733static int
734nfs_init_locked(struct inode *inode, void *opaque)
735{
736 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
737 struct nfs_fattr *fattr = desc->fattr;
738
739 NFS_FILEID(inode) = fattr->fileid;
740 nfs_copy_fh(NFS_FH(inode), desc->fh);
741 return 0;
742}
743
744/* Don't use READDIRPLUS on directories that we believe are too large */
745#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
746
747/*
748 * This is our front-end to iget that looks up inodes by file handle
749 * instead of inode number.
750 */
751struct inode *
752nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
753{
754 struct nfs_find_desc desc = {
755 .fh = fh,
756 .fattr = fattr
757 };
758 struct inode *inode = NULL;
759 unsigned long hash;
760
761 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
762 goto out_no_inode;
763
764 if (!fattr->nlink) {
765 printk("NFS: Buggy server - nlink == 0!\n");
766 goto out_no_inode;
767 }
768
769 hash = nfs_fattr_to_ino_t(fattr);
770
771 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
772 goto out_no_inode;
773
774 if (inode->i_state & I_NEW) {
775 struct nfs_inode *nfsi = NFS_I(inode);
776
777 /* We set i_ino for the few things that still rely on it,
778 * such as stat(2) */
779 inode->i_ino = hash;
780
781 /* We can't support update_atime(), since the server will reset it */
782 inode->i_flags |= S_NOATIME|S_NOCMTIME;
783 inode->i_mode = fattr->mode;
784 /* Why so? Because we want revalidate for devices/FIFOs, and
785 * that's precisely what we have in nfs_file_inode_operations.
786 */
92cfc62c 787 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 788 if (S_ISREG(inode->i_mode)) {
789 inode->i_fop = &nfs_file_operations;
790 inode->i_data.a_ops = &nfs_file_aops;
791 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
792 } else if (S_ISDIR(inode->i_mode)) {
793 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
794 inode->i_fop = &nfs_dir_operations;
795 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
796 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 797 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 798 } else if (S_ISLNK(inode->i_mode))
799 inode->i_op = &nfs_symlink_inode_operations;
800 else
801 init_special_inode(inode, inode->i_mode, fattr->rdev);
802
33801147
TM		 803 nfsi->read_cache_jiffies = fattr->time_start;
804 nfsi->last_updated = jiffies;
1da177e4
LT		 805 inode->i_atime = fattr->atime;
806 inode->i_mtime = fattr->mtime;
807 inode->i_ctime = fattr->ctime;
808 if (fattr->valid & NFS_ATTR_FATTR_V4)
809 nfsi->change_attr = fattr->change_attr;
810 inode->i_size = nfs_size_to_loff_t(fattr->size);
811 inode->i_nlink = fattr->nlink;
812 inode->i_uid = fattr->uid;
813 inode->i_gid = fattr->gid;
814 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
815 /*
816 * report the blocks in 512byte units
817 */
818 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
819 inode->i_blksize = inode->i_sb->s_blocksize;
820 } else {
821 inode->i_blocks = fattr->du.nfs2.blocks;
822 inode->i_blksize = fattr->du.nfs2.blocksize;
823 }
824 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
825 nfsi->attrtimeo_timestamp = jiffies;
826 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
827 nfsi->cache_access.cred = NULL;
828
829 unlock_new_inode(inode);
830 } else
831 nfs_refresh_inode(inode, fattr);
832 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
833 inode->i_sb->s_id,
834 (long long)NFS_FILEID(inode),
835 atomic_read(&inode->i_count));
836
837out:
838 return inode;
839
840out_no_inode:
841 printk("nfs_fhget: iget failed\n");
842 goto out;
843}
844
845#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
846
847int
848nfs_setattr(struct dentry *dentry, struct iattr *attr)
849{
850 struct inode *inode = dentry->d_inode;
851 struct nfs_fattr fattr;
852 int error;
853
854 if (attr->ia_valid & ATTR_SIZE) {
855 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
856 attr->ia_valid &= ~ATTR_SIZE;
857 }
858
859 /* Optimization: if the end result is no change, don't RPC */
860 attr->ia_valid &= NFS_VALID_ATTRS;
861 if (attr->ia_valid == 0)
862 return 0;
863
864 lock_kernel();
865 nfs_begin_data_update(inode);
755c1e20
TM		 866 /* Write all dirty data */
867 filemap_write_and_wait(inode->i_mapping);
868 nfs_wb_all(inode);
642ac549
TM		 869 /*
870 * Return any delegations if we're going to change ACLs
871 */
872 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
873 nfs_inode_return_delegation(inode);
1da177e4 874 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 875 if (error == 0)
1da177e4 876 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 877 nfs_end_data_update(inode);
878 unlock_kernel();
879 return error;
880}
881
882/**
883 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
884 * @inode: pointer to struct inode
885 * @attr: pointer to struct iattr
886 *
887 * Note: we do this in the *proc.c in order to ensure that
888 * it works for things like exclusive creates too.
889 */
890void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
891{
892 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 893 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 894 int mode = attr->ia_mode & S_IALLUGO;
895 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 896 inode->i_mode = mode;
897 }
898 if ((attr->ia_valid & ATTR_UID) != 0)
899 inode->i_uid = attr->ia_uid;
900 if ((attr->ia_valid & ATTR_GID) != 0)
901 inode->i_gid = attr->ia_gid;
dc59250c 902 spin_lock(&inode->i_lock);
55296809 903 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 904 spin_unlock(&inode->i_lock);
65e4308d
TM		 905 }
906 if ((attr->ia_valid & ATTR_SIZE) != 0) {
907 inode->i_size = attr->ia_size;
908 vmtruncate(inode, attr->ia_size);
909 }
1da177e4
LT		 910}
911
412d582e
CL		 912static int nfs_wait_schedule(void *word)
913{
914 if (signal_pending(current))
915 return -ERESTARTSYS;
916 schedule();
917 return 0;
918}
919
1da177e4
LT		 920/*
921 * Wait for the inode to get unlocked.
1da177e4 922 */
412d582e 923static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 924{
925 struct rpc_clnt *clnt = NFS_CLIENT(inode);
926 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 927 sigset_t oldmask;
1da177e4 928 int error;
412d582e 929
412d582e
CL		 930 rpc_clnt_sigmask(clnt, &oldmask);
931 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
932 nfs_wait_schedule, TASK_INTERRUPTIBLE);
933 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 934
1da177e4
LT		 935 return error;
936}
937
412d582e
CL		 938static void nfs_wake_up_inode(struct inode *inode)
939{
940 struct nfs_inode *nfsi = NFS_I(inode);
941
942 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
943 smp_mb__after_clear_bit();
944 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
945}
946
1da177e4
LT		 947int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
948{
949 struct inode *inode = dentry->d_inode;
55296809 950 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 951 int err;
952
70b9ecbd
TM		 953 /* Flush out writes to the server in order to update c/mtime */
954 nfs_sync_inode(inode, 0, 0, FLUSH_WAIT|FLUSH_NOCOMMIT);
fc33a7bb
CH		 955
956 /*
957 * We may force a getattr if the user cares about atime.
958 *
959 * Note that we only have to check the vfsmount flags here:
960 * - NFS always sets S_NOATIME by so checking it would give a
961 * bogus result
962 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
963 * no point in checking those.
964 */
965 if ((mnt->mnt_flags & MNT_NOATIME) ||
966 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
1da177e4 967 need_atime = 0;
fc33a7bb 968
1da177e4
LT		 969 if (need_atime)
970 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
971 else
972 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
973 if (!err)
974 generic_fillattr(inode, stat);
975 return err;
976}
977
b92dccf6 978static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
1da177e4
LT		 979{
980 struct nfs_open_context *ctx;
981
982 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
983 if (ctx != NULL) {
984 atomic_set(&ctx->count, 1);
985 ctx->dentry = dget(dentry);
b92dccf6 986 ctx->vfsmnt = mntget(mnt);
1da177e4
LT		 987 ctx->cred = get_rpccred(cred);
988 ctx->state = NULL;
989 ctx->lockowner = current->files;
990 ctx->error = 0;
00a92642 991 ctx->dir_cookie = 0;
1da177e4
LT		 992 }
993 return ctx;
994}
995
996struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
997{
998 if (ctx != NULL)
999 atomic_inc(&ctx->count);
1000 return ctx;
1001}
1002
1003void put_nfs_open_context(struct nfs_open_context *ctx)
1004{
1005 if (atomic_dec_and_test(&ctx->count)) {
1006 if (!list_empty(&ctx->list)) {
1007 struct inode *inode = ctx->dentry->d_inode;
1008 spin_lock(&inode->i_lock);
1009 list_del(&ctx->list);
1010 spin_unlock(&inode->i_lock);
1011 }
1012 if (ctx->state != NULL)
1013 nfs4_close_state(ctx->state, ctx->mode);
1014 if (ctx->cred != NULL)
1015 put_rpccred(ctx->cred);
1016 dput(ctx->dentry);
b92dccf6 1017 mntput(ctx->vfsmnt);
1da177e4
LT		 1018 kfree(ctx);
1019 }
1020}
1021
1022/*
1023 * Ensure that mmap has a recent RPC credential for use when writing out
1024 * shared pages
1025 */
b92dccf6 1026static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1da177e4
LT		 1027{
1028 struct inode *inode = filp->f_dentry->d_inode;
1029 struct nfs_inode *nfsi = NFS_I(inode);
1030
1031 filp->private_data = get_nfs_open_context(ctx);
1032 spin_lock(&inode->i_lock);
1033 list_add(&ctx->list, &nfsi->open_files);
1034 spin_unlock(&inode->i_lock);
1035}
1036
d530838b
TM		 1037/*
1038 * Given an inode, search for an open context with the desired characteristics
1039 */
1040struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1041{
1042 struct nfs_inode *nfsi = NFS_I(inode);
1043 struct nfs_open_context *pos, *ctx = NULL;
1044
1045 spin_lock(&inode->i_lock);
1046 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1047 if (cred != NULL && pos->cred != cred)
1048 continue;
1da177e4
LT		 1049 if ((pos->mode & mode) == mode) {
1050 ctx = get_nfs_open_context(pos);
1051 break;
1052 }
1053 }
1054 spin_unlock(&inode->i_lock);
1055 return ctx;
1056}
1057
b92dccf6 1058static void nfs_file_clear_open_context(struct file *filp)
1da177e4
LT		 1059{
1060 struct inode *inode = filp->f_dentry->d_inode;
1061 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1062
1063 if (ctx) {
1064 filp->private_data = NULL;
1065 spin_lock(&inode->i_lock);
1066 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1067 spin_unlock(&inode->i_lock);
1068 put_nfs_open_context(ctx);
1069 }
1070}
1071
1072/*
1073 * These allocate and release file read/write context information.
1074 */
1075int nfs_open(struct inode *inode, struct file *filp)
1076{
1077 struct nfs_open_context *ctx;
1078 struct rpc_cred *cred;
1079
1080 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1081 if (IS_ERR(cred))
1082 return PTR_ERR(cred);
b92dccf6 1083 ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
1da177e4
LT		 1084 put_rpccred(cred);
1085 if (ctx == NULL)
1086 return -ENOMEM;
1087 ctx->mode = filp->f_mode;
1088 nfs_file_set_open_context(filp, ctx);
1089 put_nfs_open_context(ctx);
1da177e4
LT		 1090 return 0;
1091}
1092
1093int nfs_release(struct inode *inode, struct file *filp)
1094{
1da177e4
LT		 1095 nfs_file_clear_open_context(filp);
1096 return 0;
1097}
1098
1099/*
1100 * This function is called whenever some part of NFS notices that
1101 * the cached attributes have to be refreshed.
1102 */
1103int
1104__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1105{
1106 int status = -ESTALE;
1107 struct nfs_fattr fattr;
1108 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4
LT		 1109
1110 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1111 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1112
1113 lock_kernel();
1114 if (!inode || is_bad_inode(inode))
1115 goto out_nowait;
1116 if (NFS_STALE(inode))
1117 goto out_nowait;
1118
412d582e
CL		 1119 status = nfs_wait_on_inode(inode);
1120 if (status < 0)
1121 goto out;
1122 if (NFS_STALE(inode)) {
1123 status = -ESTALE;
1124 /* Do we trust the cached ESTALE? */
1125 if (NFS_ATTRTIMEO(inode) != 0) {
1126 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1127 /* no */
1128 } else
1129 goto out;
1130 }
1da177e4 1131 }
1da177e4 1132
1da177e4
LT		 1133 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1134 if (status != 0) {
1135 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1136 inode->i_sb->s_id,
1137 (long long)NFS_FILEID(inode), status);
1138 if (status == -ESTALE) {
1139 nfs_zap_caches(inode);
1140 if (!S_ISDIR(inode->i_mode))
412d582e 1141 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1142 }
1143 goto out;
1144 }
1145
33801147 1146 spin_lock(&inode->i_lock);
24aa1fe6 1147 status = nfs_update_inode(inode, &fattr);
1da177e4 1148 if (status) {
33801147 1149 spin_unlock(&inode->i_lock);
1da177e4
LT		 1150 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1151 inode->i_sb->s_id,
1152 (long long)NFS_FILEID(inode), status);
1153 goto out;
1154 }
dc59250c 1155 spin_unlock(&inode->i_lock);
55296809 1156
7d52e862 1157 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809 1158
24aa1fe6 1159 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1160 nfs_zap_acl_cache(inode);
55296809 1161
1da177e4
LT		 1162 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1163 inode->i_sb->s_id,
1164 (long long)NFS_FILEID(inode));
1165
412d582e
CL		 1166 out:
1167 nfs_wake_up_inode(inode);
1168
1da177e4
LT		 1169 out_nowait:
1170 unlock_kernel();
1171 return status;
1172}
1173
1174int nfs_attribute_timeout(struct inode *inode)
1175{
1176 struct nfs_inode *nfsi = NFS_I(inode);
1177
1178 if (nfs_have_delegation(inode, FMODE_READ))
1179 return 0;
1180 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1181}
1182
1183/**
1184 * nfs_revalidate_inode - Revalidate the inode attributes
1185 * @server - pointer to nfs_server struct
1186 * @inode - pointer to inode struct
1187 *
1188 * Updates inode attribute information by retrieving the data from the server.
1189 */
1190int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1191{
55296809 1192 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1193 && !nfs_attribute_timeout(inode))
1194 return NFS_STALE(inode) ? -ESTALE : 0;
1195 return __nfs_revalidate_inode(server, inode);
1196}
1197
7d52e862
TM		 1198/**
1199 * nfs_revalidate_mapping - Revalidate the pagecache
1200 * @inode - pointer to host inode
1201 * @mapping - pointer to mapping
1202 */
1203void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1204{
1205 struct nfs_inode *nfsi = NFS_I(inode);
1206
55296809 1207 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
29884df0
TM		 1208 if (S_ISREG(inode->i_mode))
1209 nfs_sync_mapping(mapping);
7d52e862 1210 invalidate_inode_pages2(mapping);
dc59250c
CL		 1211
1212 spin_lock(&inode->i_lock);
55296809 1213 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1214 if (S_ISDIR(inode->i_mode)) {
1215 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1216 /* This ensures we revalidate child dentries */
913a70fc 1217 nfsi->cache_change_attribute = jiffies;
7d52e862 1218 }
dc59250c
CL		 1219 spin_unlock(&inode->i_lock);
1220
7d52e862
TM		 1221 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1222 inode->i_sb->s_id,
1223 (long long)NFS_FILEID(inode));
1224 }
1225}
1226
1da177e4
LT		 1227/**
1228 * nfs_begin_data_update
1229 * @inode - pointer to inode
1230 * Declare that a set of operations will update file data on the server
1231 */
1232void nfs_begin_data_update(struct inode *inode)
1233{
1234 atomic_inc(&NFS_I(inode)->data_updates);
1235}
1236
1237/**
1238 * nfs_end_data_update
1239 * @inode - pointer to inode
1240 * Declare end of the operations that will update file data
1241 * This will mark the inode as immediately needing revalidation
1242 * of its attribute cache.
1243 */
1244void nfs_end_data_update(struct inode *inode)
1245{
1246 struct nfs_inode *nfsi = NFS_I(inode);
1247
1248 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1249 /* Directories and symlinks: invalidate page cache */
1250 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1251 spin_lock(&inode->i_lock);
55296809 1252 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1253 spin_unlock(&inode->i_lock);
1254 }
1da177e4 1255 }
913a70fc 1256 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1257 atomic_dec(&nfsi->data_updates);
1258}
1259
a895b4a1
TM		 1260static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1261{
1262 struct nfs_inode *nfsi = NFS_I(inode);
1263
1264 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1265 && nfsi->change_attr == fattr->pre_change_attr) {
1266 nfsi->change_attr = fattr->change_attr;
1267 nfsi->cache_change_attribute = jiffies;
1268 }
1269
1270 /* If we have atomic WCC data, we may update some attributes */
1271 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1272 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
1273 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1274 nfsi->cache_change_attribute = jiffies;
1275 }
1276 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
1277 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1278 nfsi->cache_change_attribute = jiffies;
1279 }
1280 if (inode->i_size == fattr->pre_size && nfsi->npages == 0) {
1281 inode->i_size = fattr->size;
1282 nfsi->cache_change_attribute = jiffies;
1283 }
1284 }
1285}
1286
1da177e4 1287/**
33801147 1288 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1289 * @inode - pointer to inode
1290 * @fattr - updated attributes
1291 *
1292 * Verifies the attribute cache. If we have just changed the attributes,
1293 * so that fattr carries weak cache consistency data, then it may
1294 * also update the ctime/mtime/change_attribute.
1295 */
33801147 1296static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1297{
1298 struct nfs_inode *nfsi = NFS_I(inode);
1299 loff_t cur_size, new_isize;
1300 int data_unstable;
1301
dc59250c 1302
a895b4a1
TM		 1303 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1304 return 0;
1305
ca62b9c3
TM		 1306 /* Has the inode gone and changed behind our back? */
1307 if (nfsi->fileid != fattr->fileid
1308 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1309 return -EIO;
1310 }
1311
1da177e4
LT		 1312 /* Are we in the process of updating data on the server? */
1313 data_unstable = nfs_caches_unstable(inode);
1314
a895b4a1
TM		 1315 /* Do atomic weak cache consistency updates */
1316 nfs_wcc_update_inode(inode, fattr);
1da177e4 1317
ca62b9c3
TM		 1318 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0) {
1319 if (nfsi->change_attr == fattr->change_attr)
1320 goto out;
a895b4a1
TM		 1321 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1322 if (!data_unstable)
1323 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
dc59250c 1324 }
1da177e4 1325
1da177e4 1326 /* Verify a few of the more important attributes */
fe51beec 1327 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1328 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1329 if (!data_unstable)
55296809 1330 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1331 }
ca62b9c3
TM		 1332
1333 cur_size = i_size_read(inode);
1334 new_isize = nfs_size_to_loff_t(fattr->size);
fb374d24
TM		 1335 if (cur_size != new_isize && nfsi->npages == 0)
1336 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1da177e4
LT		 1337
1338 /* Have any file permissions changed? */
1339 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1340 || inode->i_uid != fattr->uid
1341 || inode->i_gid != fattr->gid)
55296809 1342 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1343
1344 /* Has the link count changed? */
1345 if (inode->i_nlink != fattr->nlink)
55296809 1346 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4 1347
ca62b9c3 1348out:
1da177e4 1349 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1350 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1351
33801147 1352 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1353 return 0;
1354}
1355
33801147
TM		 1356/**
1357 * nfs_refresh_inode - try to update the inode attribute cache
1358 * @inode - pointer to inode
1359 * @fattr - updated attributes
1360 *
1361 * Check that an RPC call that returned attributes has not overlapped with
1362 * other recent updates of the inode metadata, then decide whether it is
1363 * safe to do a full update of the inode attributes, or whether just to
1364 * call nfs_check_inode_attributes.
1365 */
1366int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1367{
1368 struct nfs_inode *nfsi = NFS_I(inode);
1369 int status;
1370
1371 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1372 return 0;
1373 spin_lock(&inode->i_lock);
1374 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
33801147 1375 if (time_after(fattr->time_start, nfsi->last_updated))
24aa1fe6 1376 status = nfs_update_inode(inode, fattr);
33801147
TM		 1377 else
1378 status = nfs_check_inode_attributes(inode, fattr);
1379
1380 spin_unlock(&inode->i_lock);
1381 return status;
1382}
1383
decf491f
TM		 1384/**
1385 * nfs_post_op_update_inode - try to update the inode attribute cache
1386 * @inode - pointer to inode
1387 * @fattr - updated attributes
1388 *
1389 * After an operation that has changed the inode metadata, mark the
1390 * attribute cache as being invalid, then try to update it.
1391 */
1392int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1393{
1394 struct nfs_inode *nfsi = NFS_I(inode);
1395 int status = 0;
1396
1397 spin_lock(&inode->i_lock);
1398 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1399 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1400 goto out;
1401 }
24aa1fe6 1402 status = nfs_update_inode(inode, fattr);
decf491f
TM		 1403out:
1404 spin_unlock(&inode->i_lock);
1405 return status;
1406}
1407
1da177e4
LT		 1408/*
1409 * Many nfs protocol calls return the new file attributes after
1410 * an operation. Here we update the inode to reflect the state
1411 * of the server's inode.
1412 *
1413 * This is a bit tricky because we have to make sure all dirty pages
1414 * have been sent off to the server before calling invalidate_inode_pages.
1415 * To make sure no other process adds more write requests while we try
1416 * our best to flush them, we make them sleep during the attribute refresh.
1417 *
1418 * A very similar scenario holds for the dir cache.
1419 */
24aa1fe6 1420static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1421{
1422 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1423 loff_t cur_isize, new_isize;
1da177e4 1424 unsigned int invalid = 0;
24aa1fe6 1425 int data_stable;
1da177e4
LT		 1426
1427 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1428 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1429 atomic_read(&inode->i_count), fattr->valid);
1430
1431 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1432 return 0;
1433
325cfed9
CL		 1434 if (nfsi->fileid != fattr->fileid)
1435 goto out_fileid;
1da177e4
LT		 1436
1437 /*
1438 * Make sure the inode's type hasn't changed.
1439 */
33801147 1440 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT		 1441 goto out_changed;
1442
1443 /*
1444 * Update the read time so we don't revalidate too often.
1445 */
33801147
TM		 1446 nfsi->read_cache_jiffies = fattr->time_start;
1447 nfsi->last_updated = jiffies;
1da177e4
LT		 1448
1449 /* Are we racing with known updates of the metadata on the server? */
24aa1fe6
TM		 1450 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1451 if (data_stable)
1452 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1da177e4 1453
a895b4a1
TM		 1454 /* Do atomic weak cache consistency updates */
1455 nfs_wcc_update_inode(inode, fattr);
1456
951a143b 1457 /* Check if our cached file size is stale */
1da177e4
LT		 1458 new_isize = nfs_size_to_loff_t(fattr->size);
1459 cur_isize = i_size_read(inode);
951a143b
TM		 1460 if (new_isize != cur_isize) {
1461 /* Do we perhaps have any outstanding writes? */
1462 if (nfsi->npages == 0) {
1463 /* No, but did we race with nfs_end_data_update()? */
24aa1fe6 1464 if (data_stable) {
1da177e4 1465 inode->i_size = new_isize;
951a143b 1466 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1467 }
951a143b
TM		 1468 invalid |= NFS_INO_INVALID_ATTR;
1469 } else if (new_isize > cur_isize) {
1da177e4
LT		 1470 inode->i_size = new_isize;
1471 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1472 }
24aa1fe6 1473 nfsi->cache_change_attribute = jiffies;
951a143b
TM		 1474 dprintk("NFS: isize change on server for file %s/%ld\n",
1475 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1476 }
1477
951a143b 1478 /* Check if the mtime agrees */
1da177e4
LT		 1479 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1480 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1481 dprintk("NFS: mtime change on server for file %s/%ld\n",
1482 inode->i_sb->s_id, inode->i_ino);
24aa1fe6
TM		 1483 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1484 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1485 }
1486
ada70d94
TM		 1487 /* If ctime has changed we should definitely clear access+acl caches */
1488 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
24aa1fe6 1489 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
ada70d94 1490 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
24aa1fe6 1491 nfsi->cache_change_attribute = jiffies;
ada70d94 1492 }
1da177e4
LT		 1493 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1494
1495 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1496 inode->i_uid != fattr->uid ||
1497 inode->i_gid != fattr->gid)
ada70d94 1498 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1499
1500 inode->i_mode = fattr->mode;
1501 inode->i_nlink = fattr->nlink;
1502 inode->i_uid = fattr->uid;
1503 inode->i_gid = fattr->gid;
1504
1505 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1506 /*
1507 * report the blocks in 512byte units
1508 */
1509 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1510 inode->i_blksize = inode->i_sb->s_blocksize;
1511 } else {
1512 inode->i_blocks = fattr->du.nfs2.blocks;
1513 inode->i_blksize = fattr->du.nfs2.blocksize;
1514 }
1515
ca62b9c3
TM		 1516 if ((fattr->valid & NFS_ATTR_FATTR_V4)) {
1517 if (nfsi->change_attr != fattr->change_attr) {
1518 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1519 inode->i_sb->s_id, inode->i_ino);
1520 nfsi->change_attr = fattr->change_attr;
1521 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1522 nfsi->cache_change_attribute = jiffies;
1523 } else
1524 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA);
1525 }
1526
1da177e4
LT		 1527 /* Update attrtimeo value if we're out of the unstable period */
1528 if (invalid & NFS_INO_INVALID_ATTR) {
1529 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1530 nfsi->attrtimeo_timestamp = jiffies;
1531 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1532 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1533 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1534 nfsi->attrtimeo_timestamp = jiffies;
1535 }
1536 /* Don't invalidate the data if we were to blame */
1537 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1538 || S_ISLNK(inode->i_mode)))
1539 invalid &= ~NFS_INO_INVALID_DATA;
24aa1fe6
TM		 1540 if (data_stable)
1541 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1da177e4 1542 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1543 nfsi->cache_validity |= invalid;
1da177e4
LT		 1544
1545 return 0;
1546 out_changed:
1547 /*
1548 * Big trouble! The inode has become a different object.
1549 */
1550#ifdef NFS_PARANOIA
1551 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1552 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1553#endif
b37b03b7 1554 out_err:
1da177e4
LT		 1555 /*
1556 * No need to worry about unhashing the dentry, as the
1557 * lookup validation will know that the inode is bad.
1558 * (But we fall through to invalidate the caches.)
1559 */
1560 nfs_invalidate_inode(inode);
1da177e4 1561 return -ESTALE;
325cfed9
CL		 1562
1563 out_fileid:
1564 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1565 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1566 NFS_SERVER(inode)->hostname, inode->i_sb->s_id,
1567 (long long)nfsi->fileid, (long long)fattr->fileid);
1568 goto out_err;
1da177e4
LT		 1569}
1570
1571/*
1572 * File system information
1573 */
1574
1575static int nfs_set_super(struct super_block *s, void *data)
1576{
1577 s->s_fs_info = data;
1578 return set_anon_super(s, data);
1579}
1580
1581static int nfs_compare_super(struct super_block *sb, void *data)
1582{
1583 struct nfs_server *server = data;
1584 struct nfs_server *old = NFS_SB(sb);
1585
1586 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1587 return 0;
1588 if (old->addr.sin_port != server->addr.sin_port)
1589 return 0;
1590 return !nfs_compare_fh(&old->fh, &server->fh);
1591}
1592
1593static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1594 int flags, const char *dev_name, void *raw_data)
1595{
1596 int error;
9085bbcb 1597 struct nfs_server *server = NULL;
1da177e4
LT		 1598 struct super_block *s;
1599 struct nfs_fh *root;
1600 struct nfs_mount_data *data = raw_data;
1601
9085bbcb
TM		 1602 s = ERR_PTR(-EINVAL);
1603 if (data == NULL) {
1604 dprintk("%s: missing data argument\n", __FUNCTION__);
1605 goto out_err;
1606 }
1607 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1608 dprintk("%s: bad mount version\n", __FUNCTION__);
1609 goto out_err;
1da177e4 1610 }
9085bbcb
TM		 1611 switch (data->version) {
1612 case 1:
1613 data->namlen = 0;
1614 case 2:
1615 data->bsize = 0;
1616 case 3:
1617 if (data->flags & NFS_MOUNT_VER3) {
1618 dprintk("%s: mount structure version %d does not support NFSv3\n",
1619 __FUNCTION__,
1620 data->version);
1621 goto out_err;
1622 }
1623 data->root.size = NFS2_FHSIZE;
1624 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1625 case 4:
1626 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1627 dprintk("%s: mount structure version %d does not support strong security\n",
1628 __FUNCTION__,
1629 data->version);
1630 goto out_err;
1631 }
1632 case 5:
1633 memset(data->context, 0, sizeof(data->context));
1634 }
1635#ifndef CONFIG_NFS_V3
1636 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1637 s = ERR_PTR(-EPROTONOSUPPORT);
1638 if (data->flags & NFS_MOUNT_VER3) {
1639 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1640 goto out_err;
1641 }
1642#endif /* CONFIG_NFS_V3 */
1da177e4 1643
9085bbcb 1644 s = ERR_PTR(-ENOMEM);
bd647545 1645 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4 1646 if (!server)
9085bbcb 1647 goto out_err;
1da177e4
LT		 1648 /* Zero out the NFS state stuff */
1649 init_nfsv4_state(server);
b7fa0554 1650 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1651
1da177e4
LT		 1652 root = &server->fh;
1653 if (data->flags & NFS_MOUNT_VER3)
1654 root->size = data->root.size;
1655 else
1656 root->size = NFS2_FHSIZE;
9085bbcb 1657 s = ERR_PTR(-EINVAL);
1da177e4 1658 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1659 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1660 goto out_err;
1da177e4
LT		 1661 }
1662 memcpy(root->data, data->root.data, root->size);
1663
1664 /* We now require that the mount process passes the remote address */
1665 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1666 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1667 dprintk("%s: mount program didn't pass remote address!\n",
1668 __FUNCTION__);
1669 goto out_err;
1da177e4
LT		 1670 }
1671
9085bbcb
TM		 1672 /* Fire up rpciod if not yet running */
1673 s = ERR_PTR(rpciod_up());
1674 if (IS_ERR(s)) {
1675 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1676 __FUNCTION__, PTR_ERR(s));
1677 goto out_err;
1da177e4
LT		 1678 }
1679
9085bbcb
TM		 1680 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1681 if (IS_ERR(s) || s->s_root)
1682 goto out_rpciod_down;
1da177e4 1683
9085bbcb 1684 s->s_flags = flags;
1da177e4
LT		 1685
1686 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1687 if (error) {
1688 up_write(&s->s_umount);
1689 deactivate_super(s);
1690 return ERR_PTR(error);
1691 }
1692 s->s_flags |= MS_ACTIVE;
1693 return s;
9085bbcb
TM		 1694out_rpciod_down:
1695 rpciod_down();
1696out_err:
1697 kfree(server);
1698 return s;
1da177e4
LT		 1699}
1700
1701static void nfs_kill_super(struct super_block *s)
1702{
1703 struct nfs_server *server = NFS_SB(s);
1704
1705 kill_anon_super(s);
1706
6a19275a 1707 if (!IS_ERR(server->client))
1da177e4 1708 rpc_shutdown_client(server->client);
6a19275a 1709 if (!IS_ERR(server->client_sys))
1da177e4 1710 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1711 if (!IS_ERR(server->client_acl))
1712 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1713
1714 if (!(server->flags & NFS_MOUNT_NONLM))
1715 lockd_down(); /* release rpc.lockd */
1716
1717 rpciod_down(); /* release rpciod */
1718
f99d49ad 1719 kfree(server->hostname);
1da177e4
LT		 1720 kfree(server);
1721}
1722
1723static struct file_system_type nfs_fs_type = {
1724 .owner = THIS_MODULE,
1725 .name = "nfs",
1726 .get_sb = nfs_get_sb,
1727 .kill_sb = nfs_kill_super,
1728 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1729};
1730
1731#ifdef CONFIG_NFS_V4
1732
1733static void nfs4_clear_inode(struct inode *);
1734
1735
1736static struct super_operations nfs4_sops = {
1737 .alloc_inode = nfs_alloc_inode,
1738 .destroy_inode = nfs_destroy_inode,
1739 .write_inode = nfs_write_inode,
1740 .delete_inode = nfs_delete_inode,
1741 .statfs = nfs_statfs,
1742 .clear_inode = nfs4_clear_inode,
1743 .umount_begin = nfs_umount_begin,
1744 .show_options = nfs_show_options,
1745};
1746
1747/*
1748 * Clean out any remaining NFSv4 state that might be left over due
1749 * to open() calls that passed nfs_atomic_lookup, but failed to call
1750 * nfs_open().
1751 */
1752static void nfs4_clear_inode(struct inode *inode)
1753{
1754 struct nfs_inode *nfsi = NFS_I(inode);
1755
1756 /* If we are holding a delegation, return it! */
cae7a073 1757 nfs_inode_return_delegation(inode);
1da177e4
LT		 1758 /* First call standard NFS clear_inode() code */
1759 nfs_clear_inode(inode);
1760 /* Now clear out any remaining state */
1761 while (!list_empty(&nfsi->open_states)) {
1762 struct nfs4_state *state;
1763
1764 state = list_entry(nfsi->open_states.next,
1765 struct nfs4_state,
1766 inode_states);
1767 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1768 __FUNCTION__,
1769 inode->i_sb->s_id,
1770 (long long)NFS_FILEID(inode),
1771 state);
1772 BUG_ON(atomic_read(&state->count) != 1);
1773 nfs4_close_state(state, state->state);
1774 }
1775}
1776
1777
1778static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1779{
1780 struct nfs_server *server;
1781 struct nfs4_client *clp = NULL;
1782 struct rpc_xprt *xprt = NULL;
1783 struct rpc_clnt *clnt = NULL;
1784 struct rpc_timeout timeparms;
1785 rpc_authflavor_t authflavour;
eab5c084 1786 int err = -EIO;
1da177e4
LT		 1787
1788 sb->s_blocksize_bits = 0;
1789 sb->s_blocksize = 0;
1790 server = NFS_SB(sb);
1791 if (data->rsize != 0)
1792 server->rsize = nfs_block_size(data->rsize, NULL);
1793 if (data->wsize != 0)
1794 server->wsize = nfs_block_size(data->wsize, NULL);
1795 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1796 server->caps = NFS_CAP_ATOMIC_OPEN;
1797
1798 server->acregmin = data->acregmin*HZ;
1799 server->acregmax = data->acregmax*HZ;
1800 server->acdirmin = data->acdirmin*HZ;
1801 server->acdirmax = data->acdirmax*HZ;
1802
1803 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1804
eab5c084 1805 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4 1806
7a480e25
CL		 1807 server->retrans_timeo = timeparms.to_initval;
1808 server->retrans_count = timeparms.to_retries;
1809
1da177e4
LT		 1810 clp = nfs4_get_client(&server->addr.sin_addr);
1811 if (!clp) {
9085bbcb 1812 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1813 return -EIO;
1814 }
1815
1816 /* Now create transport and client */
1817 authflavour = RPC_AUTH_UNIX;
1818 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1819 if (data->auth_flavourlen != 1) {
1820 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1821 __FUNCTION__, data->auth_flavourlen);
1822 err = -EINVAL;
1823 goto out_fail;
1824 }
1da177e4
LT		 1825 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1826 err = -EFAULT;
1827 goto out_fail;
1828 }
1829 }
1830
1831 down_write(&clp->cl_sem);
6a19275a 1832 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1833 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT		 1834 if (IS_ERR(xprt)) {
1835 up_write(&clp->cl_sem);
1da177e4 1836 err = PTR_ERR(xprt);
9085bbcb
TM		 1837 dprintk("%s: cannot create RPC transport. Error = %d\n",
1838 __FUNCTION__, err);
1da177e4
LT		 1839 goto out_fail;
1840 }
1841 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1842 server->rpc_ops->version, authflavour);
1843 if (IS_ERR(clnt)) {
1844 up_write(&clp->cl_sem);
1da177e4 1845 err = PTR_ERR(clnt);
9085bbcb
TM		 1846 dprintk("%s: cannot create RPC client. Error = %d\n",
1847 __FUNCTION__, err);
1da177e4
LT		 1848 goto out_fail;
1849 }
1850 clnt->cl_intr = 1;
1851 clnt->cl_softrtry = 1;
1da177e4 1852 clp->cl_rpcclient = clnt;
1da177e4
LT		 1853 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1854 nfs_idmap_new(clp);
1855 }
1da177e4
LT		 1856 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1857 clnt = rpc_clone_client(clp->cl_rpcclient);
1858 if (!IS_ERR(clnt))
1859 server->nfs4_state = clp;
1860 up_write(&clp->cl_sem);
1861 clp = NULL;
1862
1863 if (IS_ERR(clnt)) {
9085bbcb
TM		 1864 err = PTR_ERR(clnt);
1865 dprintk("%s: cannot create RPC client. Error = %d\n",
1866 __FUNCTION__, err);
1867 return err;
1da177e4
LT		 1868 }
1869
1870 server->client = clnt;
1871
1872 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1873 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1874 return -ENOMEM;
1875 }
1876
1877 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1878 struct rpc_auth *auth;
1879
1880 auth = rpcauth_create(authflavour, clnt);
1881 if (IS_ERR(auth)) {
9085bbcb 1882 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1883 return PTR_ERR(auth);
1da177e4
LT		 1884 }
1885 }
1886
1887 sb->s_time_gran = 1;
1888
1889 sb->s_op = &nfs4_sops;
1890 err = nfs_sb_init(sb, authflavour);
1891 if (err == 0)
1892 return 0;
1893out_fail:
1894 if (clp)
1895 nfs4_put_client(clp);
1896 return err;
1897}
1898
1899static int nfs4_compare_super(struct super_block *sb, void *data)
1900{
1901 struct nfs_server *server = data;
1902 struct nfs_server *old = NFS_SB(sb);
1903
1904 if (strcmp(server->hostname, old->hostname) != 0)
1905 return 0;
1906 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1907 return 0;
1908 return 1;
1909}
1910
1911static void *
1912nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1913{
1914 void *p = NULL;
1915
1916 if (!src->len)
1917 return ERR_PTR(-EINVAL);
1918 if (src->len < maxlen)
1919 maxlen = src->len;
1920 if (dst == NULL) {
1921 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1922 if (p == NULL)
1923 return ERR_PTR(-ENOMEM);
1924 }
1925 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 1926 kfree(p);
1da177e4
LT		 1927 return ERR_PTR(-EFAULT);
1928 }
1929 dst[maxlen] = '\0';
1930 return dst;
1931}
1932
1933static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1934 int flags, const char *dev_name, void *raw_data)
1935{
1936 int error;
1937 struct nfs_server *server;
1938 struct super_block *s;
1939 struct nfs4_mount_data *data = raw_data;
1940 void *p;
1941
9085bbcb
TM		 1942 if (data == NULL) {
1943 dprintk("%s: missing data argument\n", __FUNCTION__);
1944 return ERR_PTR(-EINVAL);
1945 }
1946 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1947 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1948 return ERR_PTR(-EINVAL);
1949 }
1950
bd647545 1951 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1da177e4
LT		 1952 if (!server)
1953 return ERR_PTR(-ENOMEM);
1da177e4
LT		 1954 /* Zero out the NFS state stuff */
1955 init_nfsv4_state(server);
b7fa0554 1956 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1957
1da177e4
LT		 1958 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1959 if (IS_ERR(p))
1960 goto out_err;
1961 server->hostname = p;
1962
1963 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1964 if (IS_ERR(p))
1965 goto out_err;
1966 server->mnt_path = p;
1967
1968 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1969 sizeof(server->ip_addr) - 1);
1970 if (IS_ERR(p))
1971 goto out_err;
1972
1973 /* We now require that the mount process passes the remote address */
1974 if (data->host_addrlen != sizeof(server->addr)) {
1975 s = ERR_PTR(-EINVAL);
1976 goto out_free;
1977 }
1978 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1979 s = ERR_PTR(-EFAULT);
1980 goto out_free;
1981 }
1982 if (server->addr.sin_family != AF_INET ||
1983 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1984 dprintk("%s: mount program didn't pass remote IP address!\n",
1985 __FUNCTION__);
1da177e4
LT		 1986 s = ERR_PTR(-EINVAL);
1987 goto out_free;
1988 }
1989
9085bbcb
TM		 1990 /* Fire up rpciod if not yet running */
1991 s = ERR_PTR(rpciod_up());
1992 if (IS_ERR(s)) {
1993 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1994 __FUNCTION__, PTR_ERR(s));
1995 goto out_free;
1996 }
1997
1da177e4
LT		 1998 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1999
2000 if (IS_ERR(s) || s->s_root)
2001 goto out_free;
2002
2003 s->s_flags = flags;
2004
1da177e4
LT		 2005 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
2006 if (error) {
2007 up_write(&s->s_umount);
2008 deactivate_super(s);
2009 return ERR_PTR(error);
2010 }
2011 s->s_flags |= MS_ACTIVE;
2012 return s;
2013out_err:
2014 s = (struct super_block *)p;
2015out_free:
f99d49ad
JJ		 2016 kfree(server->mnt_path);
2017 kfree(server->hostname);
1da177e4
LT		 2018 kfree(server);
2019 return s;
2020}
2021
2022static void nfs4_kill_super(struct super_block *sb)
2023{
2024 struct nfs_server *server = NFS_SB(sb);
2025
2026 nfs_return_all_delegations(sb);
2027 kill_anon_super(sb);
2028
2029 nfs4_renewd_prepare_shutdown(server);
2030
2031 if (server->client != NULL && !IS_ERR(server->client))
2032 rpc_shutdown_client(server->client);
1da177e4
LT		 2033
2034 destroy_nfsv4_state(server);
2035
967b9281
TM		 2036 rpciod_down();
2037
f99d49ad 2038 kfree(server->hostname);
1da177e4
LT		 2039 kfree(server);
2040}
2041
2042static struct file_system_type nfs4_fs_type = {
2043 .owner = THIS_MODULE,
2044 .name = "nfs4",
2045 .get_sb = nfs4_get_sb,
2046 .kill_sb = nfs4_kill_super,
2047 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2048};
2049
a72b4422
TM		 2050static const int nfs_set_port_min = 0;
2051static const int nfs_set_port_max = 65535;
2052static int param_set_port(const char *val, struct kernel_param *kp)
2053{
2054 char *endp;
2055 int num = simple_strtol(val, &endp, 0);
2056 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
2057 return -EINVAL;
2058 *((int *)kp->arg) = num;
2059 return 0;
2060}
2061
2062module_param_call(callback_tcpport, param_set_port, param_get_int,
2063 &nfs_callback_set_tcpport, 0644);
2064
58df095b
TM		 2065static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
2066{
2067 char *endp;
2068 int num = simple_strtol(val, &endp, 0);
2069 int jif = num * HZ;
2070 if (endp == val || *endp || num < 0 || jif < num)
2071 return -EINVAL;
2072 *((int *)kp->arg) = jif;
2073 return 0;
2074}
2075
2076module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
2077 &nfs_idmap_cache_timeout, 0644);
2078
1da177e4
LT		 2079#define nfs4_init_once(nfsi) \
2080 do { \
2081 INIT_LIST_HEAD(&(nfsi)->open_states); \
2082 nfsi->delegation = NULL; \
2083 nfsi->delegation_state = 0; \
2084 init_rwsem(&nfsi->rwsem); \
2085 } while(0)
a72b4422
TM		 2086
2087static inline int register_nfs4fs(void)
2088{
2089 int ret;
2090
2091 ret = nfs_register_sysctl();
2092 if (ret != 0)
2093 return ret;
2094 ret = register_filesystem(&nfs4_fs_type);
2095 if (ret != 0)
2096 nfs_unregister_sysctl();
2097 return ret;
2098}
2099
2100static inline void unregister_nfs4fs(void)
2101{
2102 unregister_filesystem(&nfs4_fs_type);
2103 nfs_unregister_sysctl();
2104}
1da177e4
LT		 2105#else
2106#define nfs4_init_once(nfsi) \
2107 do { } while (0)
2108#define register_nfs4fs() (0)
2109#define unregister_nfs4fs()
2110#endif
2111
2112extern int nfs_init_nfspagecache(void);
2113extern void nfs_destroy_nfspagecache(void);
2114extern int nfs_init_readpagecache(void);
2115extern void nfs_destroy_readpagecache(void);
2116extern int nfs_init_writepagecache(void);
2117extern void nfs_destroy_writepagecache(void);
2118#ifdef CONFIG_NFS_DIRECTIO
2119extern int nfs_init_directcache(void);
2120extern void nfs_destroy_directcache(void);
2121#endif
2122
2123static kmem_cache_t * nfs_inode_cachep;
2124
2125static struct inode *nfs_alloc_inode(struct super_block *sb)
2126{
2127 struct nfs_inode *nfsi;
2128 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2129 if (!nfsi)
2130 return NULL;
55296809
CL		 2131 nfsi->flags = 0UL;
2132 nfsi->cache_validity = 0UL;
223db122 2133 nfsi->cache_change_attribute = jiffies;
458818ed
TM		 2134#ifdef CONFIG_NFS_V3_ACL
2135 nfsi->acl_access = ERR_PTR(-EAGAIN);
2136 nfsi->acl_default = ERR_PTR(-EAGAIN);
2137#endif
e50a1c2e
BF		 2138#ifdef CONFIG_NFS_V4
2139 nfsi->nfs4_acl = NULL;
2140#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2141 return &nfsi->vfs_inode;
2142}
2143
2144static void nfs_destroy_inode(struct inode *inode)
2145{
2146 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2147}
2148
2149static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2150{
2151 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2152
2153 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2154 SLAB_CTOR_CONSTRUCTOR) {
2155 inode_init_once(&nfsi->vfs_inode);
2156 spin_lock_init(&nfsi->req_lock);
2157 INIT_LIST_HEAD(&nfsi->dirty);
2158 INIT_LIST_HEAD(&nfsi->commit);
2159 INIT_LIST_HEAD(&nfsi->open_files);
2160 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2161 atomic_set(&nfsi->data_updates, 0);
2162 nfsi->ndirty = 0;
2163 nfsi->ncommit = 0;
2164 nfsi->npages = 0;
1da177e4
LT		 2165 nfs4_init_once(nfsi);
2166 }
2167}
2168
75c96f85 2169static int nfs_init_inodecache(void)
1da177e4
LT		 2170{
2171 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2172 sizeof(struct nfs_inode),
2173 0, SLAB_RECLAIM_ACCOUNT,
2174 init_once, NULL);
2175 if (nfs_inode_cachep == NULL)
2176 return -ENOMEM;
2177
2178 return 0;
2179}
2180
75c96f85 2181static void nfs_destroy_inodecache(void)
1da177e4
LT		 2182{
2183 if (kmem_cache_destroy(nfs_inode_cachep))
2184 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2185}
2186
2187/*
2188 * Initialize NFS
2189 */
2190static int __init init_nfs_fs(void)
2191{
2192 int err;
2193
2194 err = nfs_init_nfspagecache();
2195 if (err)
2196 goto out4;
2197
2198 err = nfs_init_inodecache();
2199 if (err)
2200 goto out3;
2201
2202 err = nfs_init_readpagecache();
2203 if (err)
2204 goto out2;
2205
2206 err = nfs_init_writepagecache();
2207 if (err)
2208 goto out1;
2209
2210#ifdef CONFIG_NFS_DIRECTIO
2211 err = nfs_init_directcache();
2212 if (err)
2213 goto out0;
2214#endif
2215
2216#ifdef CONFIG_PROC_FS
2217 rpc_proc_register(&nfs_rpcstat);
2218#endif
2219 err = register_filesystem(&nfs_fs_type);
2220 if (err)
2221 goto out;
2222 if ((err = register_nfs4fs()) != 0)
2223 goto out;
2224 return 0;
2225out:
2226#ifdef CONFIG_PROC_FS
2227 rpc_proc_unregister("nfs");
2228#endif
1da177e4 2229#ifdef CONFIG_NFS_DIRECTIO
1da177e4 2230 nfs_destroy_directcache();
6b59a754 2231out0:
1da177e4 2232#endif
6b59a754 2233 nfs_destroy_writepagecache();
1da177e4
LT		 2234out1:
2235 nfs_destroy_readpagecache();
2236out2:
2237 nfs_destroy_inodecache();
2238out3:
2239 nfs_destroy_nfspagecache();
2240out4:
2241 return err;
2242}
2243
2244static void __exit exit_nfs_fs(void)
2245{
2246#ifdef CONFIG_NFS_DIRECTIO
2247 nfs_destroy_directcache();
2248#endif
2249 nfs_destroy_writepagecache();
2250 nfs_destroy_readpagecache();
2251 nfs_destroy_inodecache();
2252 nfs_destroy_nfspagecache();
2253#ifdef CONFIG_PROC_FS
2254 rpc_proc_unregister("nfs");
2255#endif
2256 unregister_filesystem(&nfs_fs_type);
2257 unregister_nfs4fs();
2258}
2259
2260/* Not quite true; I just maintain it */
2261MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2262MODULE_LICENSE("GPL");
2263
2264module_init(init_nfs_fs)
2265module_exit(exit_nfs_fs)
kernel source for telekom puls (alcatel/mediatek)