]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/nfs/inode.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFS: use generic_write_checks() to sanity check direct writes
[net-next-2.6.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
4ce79717 42#include "nfs4_fs.h"
1da177e4
LT		 43#include "delegation.h"
44
45#define NFSDBG_FACILITY NFSDBG_VFS
46#define NFS_PARANOIA 1
47
48/* Maximum number of readahead requests
49 * FIXME: this should really be a sysctl so that users may tune it to suit
50 * their needs. People that do NFS over a slow network, might for
51 * instance want to reduce it to something closer to 1 for improved
52 * interactive response.
53 */
54#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
55
56static void nfs_invalidate_inode(struct inode *);
24aa1fe6 57static int nfs_update_inode(struct inode *, struct nfs_fattr *);
1da177e4
LT		 58
59static struct inode *nfs_alloc_inode(struct super_block *sb);
60static void nfs_destroy_inode(struct inode *);
61static int nfs_write_inode(struct inode *,int);
62static void nfs_delete_inode(struct inode *);
63static void nfs_clear_inode(struct inode *);
64static void nfs_umount_begin(struct super_block *);
65static int nfs_statfs(struct super_block *, struct kstatfs *);
66static int nfs_show_options(struct seq_file *, struct vfsmount *);
ada70d94 67static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT		 68
69static struct rpc_program nfs_program;
70
71static struct super_operations nfs_sops = {
72 .alloc_inode = nfs_alloc_inode,
73 .destroy_inode = nfs_destroy_inode,
74 .write_inode = nfs_write_inode,
75 .delete_inode = nfs_delete_inode,
76 .statfs = nfs_statfs,
77 .clear_inode = nfs_clear_inode,
78 .umount_begin = nfs_umount_begin,
79 .show_options = nfs_show_options,
80};
81
82/*
83 * RPC cruft for NFS
84 */
85static struct rpc_stat nfs_rpcstat = {
86 .program = &nfs_program
87};
88static struct rpc_version * nfs_version[] = {
89 NULL,
90 NULL,
91 &nfs_version2,
92#if defined(CONFIG_NFS_V3)
93 &nfs_version3,
94#elif defined(CONFIG_NFS_V4)
95 NULL,
96#endif
97#if defined(CONFIG_NFS_V4)
98 &nfs_version4,
99#endif
100};
101
102static struct rpc_program nfs_program = {
103 .name = "nfs",
104 .number = NFS_PROGRAM,
105 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
106 .version = nfs_version,
107 .stats = &nfs_rpcstat,
108 .pipe_dir_name = "/nfs",
109};
110
b7fa0554
AG		 111#ifdef CONFIG_NFS_V3_ACL
112static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
113static struct rpc_version * nfsacl_version[] = {
114 [3] = &nfsacl_version3,
115};
116
117struct rpc_program nfsacl_program = {
118 .name = "nfsacl",
119 .number = NFS_ACL_PROGRAM,
120 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
121 .version = nfsacl_version,
122 .stats = &nfsacl_rpcstat,
123};
124#endif /* CONFIG_NFS_V3_ACL */
125
1da177e4
LT		 126static inline unsigned long
127nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
128{
129 return nfs_fileid_to_ino_t(fattr->fileid);
130}
131
132static int
133nfs_write_inode(struct inode *inode, int sync)
134{
135 int flags = sync ? FLUSH_WAIT : 0;
136 int ret;
137
3da28eb1 138 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 139 if (ret < 0)
140 return ret;
141 return 0;
142}
143
144static void
145nfs_delete_inode(struct inode * inode)
146{
147 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
148
fef26658
MF		 149 truncate_inode_pages(&inode->i_data, 0);
150
1da177e4
LT		 151 nfs_wb_all(inode);
152 /*
153 * The following should never happen...
154 */
155 if (nfs_have_writebacks(inode)) {
156 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
157 }
158
159 clear_inode(inode);
160}
161
1da177e4
LT		 162static void
163nfs_clear_inode(struct inode *inode)
164{
165 struct nfs_inode *nfsi = NFS_I(inode);
166 struct rpc_cred *cred;
167
168 nfs_wb_all(inode);
169 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 170 nfs_zap_acl_cache(inode);
1da177e4
LT		 171 cred = nfsi->cache_access.cred;
172 if (cred)
173 put_rpccred(cred);
174 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
175}
176
177void
178nfs_umount_begin(struct super_block *sb)
179{
6a19275a 180 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT		 181
182 /* -EIO all pending I/O */
6a19275a 183 if (!IS_ERR(rpc))
1da177e4 184 rpc_killall_tasks(rpc);
b7fa0554
AG		 185 rpc = NFS_SB(sb)->client_acl;
186 if (!IS_ERR(rpc))
187 rpc_killall_tasks(rpc);
1da177e4
LT		 188}
189
190
191static inline unsigned long
192nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
193{
194 /* make sure blocksize is a power of two */
195 if ((bsize & (bsize - 1)) || nrbitsp) {
196 unsigned char nrbits;
197
198 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
199 ;
200 bsize = 1 << nrbits;
201 if (nrbitsp)
202 *nrbitsp = nrbits;
203 }
204
205 return bsize;
206}
207
208/*
209 * Calculate the number of 512byte blocks used.
210 */
211static inline unsigned long
212nfs_calc_block_size(u64 tsize)
213{
214 loff_t used = (tsize + 511) >> 9;
215 return (used > ULONG_MAX) ? ULONG_MAX : used;
216}
217
218/*
219 * Compute and set NFS server blocksize
220 */
221static inline unsigned long
222nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
223{
224 if (bsize < 1024)
225 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
226 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
227 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
228
229 return nfs_block_bits(bsize, nrbitsp);
230}
231
232/*
233 * Obtain the root inode of the file system.
234 */
235static struct inode *
236nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
237{
238 struct nfs_server *server = NFS_SB(sb);
239 struct inode *rooti;
240 int error;
241
242 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
243 if (error < 0) {
244 dprintk("nfs_get_root: getattr error = %d\n", -error);
245 return ERR_PTR(error);
246 }
247
248 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
249 if (!rooti)
250 return ERR_PTR(-ENOMEM);
251 return rooti;
252}
253
254/*
255 * Do NFS version-independent mount processing, and sanity checking
256 */
257static int
258nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
259{
260 struct nfs_server *server;
261 struct inode *root_inode;
262 struct nfs_fattr fattr;
263 struct nfs_fsinfo fsinfo = {
264 .fattr = &fattr,
265 };
266 struct nfs_pathconf pathinfo = {
267 .fattr = &fattr,
268 };
269 int no_root_error = 0;
270 unsigned long max_rpc_payload;
271
272 /* We probably want something more informative here */
273 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
274
275 server = NFS_SB(sb);
276
277 sb->s_magic = NFS_SUPER_MAGIC;
278
279 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
280 /* Did getting the root inode fail? */
281 if (IS_ERR(root_inode)) {
282 no_root_error = PTR_ERR(root_inode);
283 goto out_no_root;
284 }
285 sb->s_root = d_alloc_root(root_inode);
286 if (!sb->s_root) {
287 no_root_error = -ENOMEM;
288 goto out_no_root;
289 }
290 sb->s_root->d_op = server->rpc_ops->dentry_ops;
291
292 /* Get some general file system info */
293 if (server->namelen == 0 &&
294 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
295 server->namelen = pathinfo.max_namelen;
296 /* Work out a lot of parameters */
297 if (server->rsize == 0)
298 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
299 if (server->wsize == 0)
300 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
301
302 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
303 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
304 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
305 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
306
307 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
308 if (server->rsize > max_rpc_payload)
309 server->rsize = max_rpc_payload;
310 if (server->wsize > max_rpc_payload)
311 server->wsize = max_rpc_payload;
312
313 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
314 if (server->rpages > NFS_READ_MAXIOV) {
315 server->rpages = NFS_READ_MAXIOV;
316 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
317 }
318
319 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
320 if (server->wpages > NFS_WRITE_MAXIOV) {
321 server->wpages = NFS_WRITE_MAXIOV;
322 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
323 }
324
325 if (sb->s_blocksize == 0)
326 sb->s_blocksize = nfs_block_bits(server->wsize,
327 &sb->s_blocksize_bits);
328 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
329
330 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
331 if (server->dtsize > PAGE_CACHE_SIZE)
332 server->dtsize = PAGE_CACHE_SIZE;
333 if (server->dtsize > server->rsize)
334 server->dtsize = server->rsize;
335
336 if (server->flags & NFS_MOUNT_NOAC) {
337 server->acregmin = server->acregmax = 0;
338 server->acdirmin = server->acdirmax = 0;
339 sb->s_flags |= MS_SYNCHRONOUS;
340 }
341 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
342
343 sb->s_maxbytes = fsinfo.maxfilesize;
344 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
345 sb->s_maxbytes = MAX_LFS_FILESIZE;
346
347 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
348 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
349
350 /* We're airborne Set socket buffersize */
351 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
352 return 0;
353 /* Yargs. It didn't work out. */
354out_no_root:
355 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
356 if (!IS_ERR(root_inode))
357 iput(root_inode);
358 return no_root_error;
359}
360
eab5c084
CL		 361static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
362{
363 to->to_initval = timeo * HZ / 10;
364 to->to_retries = retrans;
365 if (!to->to_retries)
366 to->to_retries = 2;
367
368 switch (proto) {
369 case IPPROTO_TCP:
370 if (!to->to_initval)
371 to->to_initval = 60 * HZ;
03bf4b70
CL		 372 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
373 to->to_initval = NFS_MAX_TCP_TIMEOUT;
eab5c084
CL		 374 to->to_increment = to->to_initval;
375 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
376 to->to_exponential = 0;
377 break;
378 case IPPROTO_UDP:
379 default:
380 if (!to->to_initval)
381 to->to_initval = 11 * HZ / 10;
03bf4b70
CL		 382 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
383 to->to_initval = NFS_MAX_UDP_TIMEOUT;
384 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
eab5c084
CL		 385 to->to_exponential = 1;
386 break;
387 }
388}
389
1da177e4
LT		 390/*
391 * Create an RPC client handle.
392 */
393static struct rpc_clnt *
394nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
395{
396 struct rpc_timeout timeparms;
397 struct rpc_xprt *xprt = NULL;
398 struct rpc_clnt *clnt = NULL;
eab5c084 399 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
1da177e4 400
eab5c084 401 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
1da177e4
LT		 402
403 /* create transport and client */
eab5c084 404 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1da177e4 405 if (IS_ERR(xprt)) {
9085bbcb
TM		 406 dprintk("%s: cannot create RPC transport. Error = %ld\n",
407 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 408 return (struct rpc_clnt *)xprt;
409 }
410 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
411 server->rpc_ops->version, data->pseudoflavor);
412 if (IS_ERR(clnt)) {
9085bbcb
TM		 413 dprintk("%s: cannot create RPC client. Error = %ld\n",
414 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 415 goto out_fail;
416 }
417
418 clnt->cl_intr = 1;
419 clnt->cl_softrtry = 1;
420 clnt->cl_chatty = 1;
421
422 return clnt;
423
424out_fail:
1da177e4
LT		 425 return clnt;
426}
427
428/*
429 * The way this works is that the mount process passes a structure
430 * in the data argument which contains the server's IP address
431 * and the root file handle obtained from the server's mount
432 * daemon. We stash these away in the private superblock fields.
433 */
434static int
435nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
436{
437 struct nfs_server *server;
438 rpc_authflavor_t authflavor;
439
440 server = NFS_SB(sb);
441 sb->s_blocksize_bits = 0;
442 sb->s_blocksize = 0;
443 if (data->bsize)
444 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
445 if (data->rsize)
446 server->rsize = nfs_block_size(data->rsize, NULL);
447 if (data->wsize)
448 server->wsize = nfs_block_size(data->wsize, NULL);
449 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
450
451 server->acregmin = data->acregmin*HZ;
452 server->acregmax = data->acregmax*HZ;
453 server->acdirmin = data->acdirmin*HZ;
454 server->acdirmax = data->acdirmax*HZ;
455
456 /* Start lockd here, before we might error out */
457 if (!(server->flags & NFS_MOUNT_NONLM))
458 lockd_up();
459
460 server->namelen = data->namlen;
461 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
462 if (!server->hostname)
463 return -ENOMEM;
464 strcpy(server->hostname, data->hostname);
465
466 /* Check NFS protocol revision and initialize RPC op vector
467 * and file handle pool. */
1da177e4 468#ifdef CONFIG_NFS_V3
9085bbcb 469 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 470 server->rpc_ops = &nfs_v3_clientops;
471 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 472 } else {
473 server->rpc_ops = &nfs_v2_clientops;
474 }
9085bbcb
TM		 475#else
476 server->rpc_ops = &nfs_v2_clientops;
477#endif
1da177e4
LT		 478
479 /* Fill in pseudoflavor for mount version < 5 */
480 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
481 data->pseudoflavor = RPC_AUTH_UNIX;
482 authflavor = data->pseudoflavor; /* save for sb_init() */
483 /* XXX maybe we want to add a server->pseudoflavor field */
484
485 /* Create RPC client handles */
486 server->client = nfs_create_client(server, data);
487 if (IS_ERR(server->client))
488 return PTR_ERR(server->client);
489 /* RFC 2623, sec 2.3.2 */
490 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 491 struct rpc_auth *auth;
492
1da177e4
LT		 493 server->client_sys = rpc_clone_client(server->client);
494 if (IS_ERR(server->client_sys))
495 return PTR_ERR(server->client_sys);
6a19275a
BF		 496 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
497 if (IS_ERR(auth))
498 return PTR_ERR(auth);
1da177e4
LT		 499 } else {
500 atomic_inc(&server->client->cl_count);
501 server->client_sys = server->client;
502 }
1da177e4 503 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 504#ifdef CONFIG_NFS_V3_ACL
505 if (!(server->flags & NFS_MOUNT_NOACL)) {
506 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
507 /* No errors! Assume that Sun nfsacls are supported */
508 if (!IS_ERR(server->client_acl))
509 server->caps |= NFS_CAP_ACLS;
510 }
511#else
512 server->flags &= ~NFS_MOUNT_NOACL;
513#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 514 /*
515 * The VFS shouldn't apply the umask to mode bits. We will
516 * do so ourselves when necessary.
517 */
518 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 519 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
520 server->namelen = NFS3_MAXNAMLEN;
521 sb->s_time_gran = 1;
522 } else {
523 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
524 server->namelen = NFS2_MAXNAMLEN;
525 }
526
527 sb->s_op = &nfs_sops;
528 return nfs_sb_init(sb, authflavor);
529}
530
531static int
532nfs_statfs(struct super_block *sb, struct kstatfs *buf)
533{
534 struct nfs_server *server = NFS_SB(sb);
535 unsigned char blockbits;
536 unsigned long blockres;
537 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
538 struct nfs_fattr fattr;
539 struct nfs_fsstat res = {
540 .fattr = &fattr,
541 };
542 int error;
543
544 lock_kernel();
545
546 error = server->rpc_ops->statfs(server, rootfh, &res);
547 buf->f_type = NFS_SUPER_MAGIC;
548 if (error < 0)
549 goto out_err;
550
551 /*
552 * Current versions of glibc do not correctly handle the
553 * case where f_frsize != f_bsize. Eventually we want to
554 * report the value of wtmult in this field.
555 */
556 buf->f_frsize = sb->s_blocksize;
557
558 /*
559 * On most *nix systems, f_blocks, f_bfree, and f_bavail
560 * are reported in units of f_frsize. Linux hasn't had
561 * an f_frsize field in its statfs struct until recently,
562 * thus historically Linux's sys_statfs reports these
563 * fields in units of f_bsize.
564 */
565 buf->f_bsize = sb->s_blocksize;
566 blockbits = sb->s_blocksize_bits;
567 blockres = (1 << blockbits) - 1;
568 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
569 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
570 buf->f_bavail = (res.abytes + blockres) >> blockbits;
571
572 buf->f_files = res.tfiles;
573 buf->f_ffree = res.afiles;
574
575 buf->f_namelen = server->namelen;
576 out:
577 unlock_kernel();
578
579 return 0;
580
581 out_err:
582 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
583 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
584 goto out;
585
586}
587
588static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
589{
590 static struct proc_nfs_info {
591 int flag;
592 char *str;
593 char *nostr;
594 } nfs_info[] = {
595 { NFS_MOUNT_SOFT, ",soft", ",hard" },
596 { NFS_MOUNT_INTR, ",intr", "" },
597 { NFS_MOUNT_POSIX, ",posix", "" },
1da177e4
LT		 598 { NFS_MOUNT_NOCTO, ",nocto", "" },
599 { NFS_MOUNT_NOAC, ",noac", "" },
600 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
b7fa0554 601 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 602 { 0, NULL, NULL }
603 };
604 struct proc_nfs_info *nfs_infop;
605 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
3063d8a1
TM		 606 char buf[12];
607 char *proto;
1da177e4
LT		 608
609 seq_printf(m, ",v%d", nfss->rpc_ops->version);
610 seq_printf(m, ",rsize=%d", nfss->rsize);
611 seq_printf(m, ",wsize=%d", nfss->wsize);
612 if (nfss->acregmin != 3*HZ)
613 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
614 if (nfss->acregmax != 60*HZ)
615 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
616 if (nfss->acdirmin != 30*HZ)
617 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
618 if (nfss->acdirmax != 60*HZ)
619 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
620 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
621 if (nfss->flags & nfs_infop->flag)
622 seq_puts(m, nfs_infop->str);
623 else
624 seq_puts(m, nfs_infop->nostr);
625 }
3063d8a1
TM		 626 switch (nfss->client->cl_xprt->prot) {
627 case IPPROTO_TCP:
628 proto = "tcp";
629 break;
630 case IPPROTO_UDP:
631 proto = "udp";
632 break;
633 default:
634 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
635 proto = buf;
636 }
637 seq_printf(m, ",proto=%s", proto);
1da177e4
LT		 638 seq_puts(m, ",addr=");
639 seq_escape(m, nfss->hostname, " \t\n\\");
640 return 0;
641}
642
29884df0
TM		 643/**
644 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
645 */
646int nfs_sync_mapping(struct address_space *mapping)
647{
648 int ret;
649
650 if (mapping->nrpages == 0)
651 return 0;
652 unmap_mapping_range(mapping, 0, 0, 0);
653 ret = filemap_fdatawrite(mapping);
654 if (ret != 0)
655 goto out;
656 ret = filemap_fdatawait(mapping);
657 if (ret != 0)
658 goto out;
659 ret = nfs_wb_all(mapping->host);
660out:
661 return ret;
662}
663
1da177e4
LT		 664/*
665 * Invalidate the local caches
666 */
b37b03b7 667static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT		 668{
669 struct nfs_inode *nfsi = NFS_I(inode);
670 int mode = inode->i_mode;
671
672 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
673 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
674
675 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
676 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 677 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 678 else
55296809 679 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 680}
dc59250c 681
b37b03b7
TM		 682void nfs_zap_caches(struct inode *inode)
683{
684 spin_lock(&inode->i_lock);
685 nfs_zap_caches_locked(inode);
dc59250c 686 spin_unlock(&inode->i_lock);
ada70d94
TM		 687}
688
689static void nfs_zap_acl_cache(struct inode *inode)
690{
691 void (*clear_acl_cache)(struct inode *);
692
693 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
694 if (clear_acl_cache != NULL)
695 clear_acl_cache(inode);
dc59250c 696 spin_lock(&inode->i_lock);
55296809 697 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 698 spin_unlock(&inode->i_lock);
1da177e4
LT		 699}
700
701/*
b37b03b7
TM		 702 * Invalidate, but do not unhash, the inode.
703 * NB: must be called with inode->i_lock held!
1da177e4 704 */
b37b03b7 705static void nfs_invalidate_inode(struct inode *inode)
1da177e4 706{
b37b03b7
TM		 707 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
708 nfs_zap_caches_locked(inode);
1da177e4
LT		 709}
710
711struct nfs_find_desc {
712 struct nfs_fh *fh;
713 struct nfs_fattr *fattr;
714};
715
716/*
717 * In NFSv3 we can have 64bit inode numbers. In order to support
718 * this, and re-exported directories (also seen in NFSv2)
719 * we are forced to allow 2 different inodes to have the same
720 * i_ino.
721 */
722static int
723nfs_find_actor(struct inode *inode, void *opaque)
724{
725 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
726 struct nfs_fh *fh = desc->fh;
727 struct nfs_fattr *fattr = desc->fattr;
728
729 if (NFS_FILEID(inode) != fattr->fileid)
730 return 0;
731 if (nfs_compare_fh(NFS_FH(inode), fh))
732 return 0;
733 if (is_bad_inode(inode) || NFS_STALE(inode))
734 return 0;
735 return 1;
736}
737
738static int
739nfs_init_locked(struct inode *inode, void *opaque)
740{
741 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
742 struct nfs_fattr *fattr = desc->fattr;
743
744 NFS_FILEID(inode) = fattr->fileid;
745 nfs_copy_fh(NFS_FH(inode), desc->fh);
746 return 0;
747}
748
749/* Don't use READDIRPLUS on directories that we believe are too large */
750#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
751
752/*
753 * This is our front-end to iget that looks up inodes by file handle
754 * instead of inode number.
755 */
756struct inode *
757nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
758{
759 struct nfs_find_desc desc = {
760 .fh = fh,
761 .fattr = fattr
762 };
763 struct inode *inode = NULL;
764 unsigned long hash;
765
766 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
767 goto out_no_inode;
768
769 if (!fattr->nlink) {
770 printk("NFS: Buggy server - nlink == 0!\n");
771 goto out_no_inode;
772 }
773
774 hash = nfs_fattr_to_ino_t(fattr);
775
776 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
777 goto out_no_inode;
778
779 if (inode->i_state & I_NEW) {
780 struct nfs_inode *nfsi = NFS_I(inode);
781
782 /* We set i_ino for the few things that still rely on it,
783 * such as stat(2) */
784 inode->i_ino = hash;
785
786 /* We can't support update_atime(), since the server will reset it */
787 inode->i_flags |= S_NOATIME|S_NOCMTIME;
788 inode->i_mode = fattr->mode;
789 /* Why so? Because we want revalidate for devices/FIFOs, and
790 * that's precisely what we have in nfs_file_inode_operations.
791 */
92cfc62c 792 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 793 if (S_ISREG(inode->i_mode)) {
794 inode->i_fop = &nfs_file_operations;
795 inode->i_data.a_ops = &nfs_file_aops;
796 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
797 } else if (S_ISDIR(inode->i_mode)) {
798 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
799 inode->i_fop = &nfs_dir_operations;
800 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
801 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 802 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 803 } else if (S_ISLNK(inode->i_mode))
804 inode->i_op = &nfs_symlink_inode_operations;
805 else
806 init_special_inode(inode, inode->i_mode, fattr->rdev);
807
33801147
TM		 808 nfsi->read_cache_jiffies = fattr->time_start;
809 nfsi->last_updated = jiffies;
1da177e4
LT		 810 inode->i_atime = fattr->atime;
811 inode->i_mtime = fattr->mtime;
812 inode->i_ctime = fattr->ctime;
813 if (fattr->valid & NFS_ATTR_FATTR_V4)
814 nfsi->change_attr = fattr->change_attr;
815 inode->i_size = nfs_size_to_loff_t(fattr->size);
816 inode->i_nlink = fattr->nlink;
817 inode->i_uid = fattr->uid;
818 inode->i_gid = fattr->gid;
819 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
820 /*
821 * report the blocks in 512byte units
822 */
823 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
824 inode->i_blksize = inode->i_sb->s_blocksize;
825 } else {
826 inode->i_blocks = fattr->du.nfs2.blocks;
827 inode->i_blksize = fattr->du.nfs2.blocksize;
828 }
829 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
830 nfsi->attrtimeo_timestamp = jiffies;
831 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
832 nfsi->cache_access.cred = NULL;
833
834 unlock_new_inode(inode);
835 } else
836 nfs_refresh_inode(inode, fattr);
837 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
838 inode->i_sb->s_id,
839 (long long)NFS_FILEID(inode),
840 atomic_read(&inode->i_count));
841
842out:
843 return inode;
844
845out_no_inode:
846 printk("nfs_fhget: iget failed\n");
847 goto out;
848}
849
850#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
851
852int
853nfs_setattr(struct dentry *dentry, struct iattr *attr)
854{
855 struct inode *inode = dentry->d_inode;
856 struct nfs_fattr fattr;
857 int error;
858
859 if (attr->ia_valid & ATTR_SIZE) {
860 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
861 attr->ia_valid &= ~ATTR_SIZE;
862 }
863
864 /* Optimization: if the end result is no change, don't RPC */
865 attr->ia_valid &= NFS_VALID_ATTRS;
866 if (attr->ia_valid == 0)
867 return 0;
868
869 lock_kernel();
870 nfs_begin_data_update(inode);
871 /* Write all dirty data if we're changing file permissions or size */
872 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
873 if (filemap_fdatawrite(inode->i_mapping) == 0)
874 filemap_fdatawait(inode->i_mapping);
875 nfs_wb_all(inode);
876 }
642ac549
TM		 877 /*
878 * Return any delegations if we're going to change ACLs
879 */
880 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
881 nfs_inode_return_delegation(inode);
1da177e4 882 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 883 if (error == 0)
1da177e4 884 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 885 nfs_end_data_update(inode);
886 unlock_kernel();
887 return error;
888}
889
890/**
891 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
892 * @inode: pointer to struct inode
893 * @attr: pointer to struct iattr
894 *
895 * Note: we do this in the *proc.c in order to ensure that
896 * it works for things like exclusive creates too.
897 */
898void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
899{
900 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 901 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 902 int mode = attr->ia_mode & S_IALLUGO;
903 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 904 inode->i_mode = mode;
905 }
906 if ((attr->ia_valid & ATTR_UID) != 0)
907 inode->i_uid = attr->ia_uid;
908 if ((attr->ia_valid & ATTR_GID) != 0)
909 inode->i_gid = attr->ia_gid;
dc59250c 910 spin_lock(&inode->i_lock);
55296809 911 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 912 spin_unlock(&inode->i_lock);
65e4308d
TM		 913 }
914 if ((attr->ia_valid & ATTR_SIZE) != 0) {
915 inode->i_size = attr->ia_size;
916 vmtruncate(inode, attr->ia_size);
917 }
1da177e4
LT		 918}
919
412d582e
CL		 920static int nfs_wait_schedule(void *word)
921{
922 if (signal_pending(current))
923 return -ERESTARTSYS;
924 schedule();
925 return 0;
926}
927
1da177e4
LT		 928/*
929 * Wait for the inode to get unlocked.
1da177e4 930 */
412d582e 931static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 932{
933 struct rpc_clnt *clnt = NFS_CLIENT(inode);
934 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 935 sigset_t oldmask;
1da177e4 936 int error;
412d582e 937
412d582e
CL		 938 rpc_clnt_sigmask(clnt, &oldmask);
939 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
940 nfs_wait_schedule, TASK_INTERRUPTIBLE);
941 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 942
1da177e4
LT		 943 return error;
944}
945
412d582e
CL		 946static void nfs_wake_up_inode(struct inode *inode)
947{
948 struct nfs_inode *nfsi = NFS_I(inode);
949
950 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
951 smp_mb__after_clear_bit();
952 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
953}
954
1da177e4
LT		 955int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
956{
957 struct inode *inode = dentry->d_inode;
55296809 958 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 959 int err;
960
961 if (__IS_FLG(inode, MS_NOATIME))
962 need_atime = 0;
963 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
964 need_atime = 0;
965 /* We may force a getattr if the user cares about atime */
966 if (need_atime)
967 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
968 else
969 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
970 if (!err)
971 generic_fillattr(inode, stat);
972 return err;
973}
974
975struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
976{
977 struct nfs_open_context *ctx;
978
979 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
980 if (ctx != NULL) {
981 atomic_set(&ctx->count, 1);
982 ctx->dentry = dget(dentry);
983 ctx->cred = get_rpccred(cred);
984 ctx->state = NULL;
985 ctx->lockowner = current->files;
986 ctx->error = 0;
00a92642 987 ctx->dir_cookie = 0;
1da177e4
LT		 988 }
989 return ctx;
990}
991
992struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
993{
994 if (ctx != NULL)
995 atomic_inc(&ctx->count);
996 return ctx;
997}
998
999void put_nfs_open_context(struct nfs_open_context *ctx)
1000{
1001 if (atomic_dec_and_test(&ctx->count)) {
1002 if (!list_empty(&ctx->list)) {
1003 struct inode *inode = ctx->dentry->d_inode;
1004 spin_lock(&inode->i_lock);
1005 list_del(&ctx->list);
1006 spin_unlock(&inode->i_lock);
1007 }
1008 if (ctx->state != NULL)
1009 nfs4_close_state(ctx->state, ctx->mode);
1010 if (ctx->cred != NULL)
1011 put_rpccred(ctx->cred);
1012 dput(ctx->dentry);
1013 kfree(ctx);
1014 }
1015}
1016
1017/*
1018 * Ensure that mmap has a recent RPC credential for use when writing out
1019 * shared pages
1020 */
1021void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1022{
1023 struct inode *inode = filp->f_dentry->d_inode;
1024 struct nfs_inode *nfsi = NFS_I(inode);
1025
1026 filp->private_data = get_nfs_open_context(ctx);
1027 spin_lock(&inode->i_lock);
1028 list_add(&ctx->list, &nfsi->open_files);
1029 spin_unlock(&inode->i_lock);
1030}
1031
d530838b
TM		 1032/*
1033 * Given an inode, search for an open context with the desired characteristics
1034 */
1035struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1036{
1037 struct nfs_inode *nfsi = NFS_I(inode);
1038 struct nfs_open_context *pos, *ctx = NULL;
1039
1040 spin_lock(&inode->i_lock);
1041 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1042 if (cred != NULL && pos->cred != cred)
1043 continue;
1da177e4
LT		 1044 if ((pos->mode & mode) == mode) {
1045 ctx = get_nfs_open_context(pos);
1046 break;
1047 }
1048 }
1049 spin_unlock(&inode->i_lock);
1050 return ctx;
1051}
1052
1053void nfs_file_clear_open_context(struct file *filp)
1054{
1055 struct inode *inode = filp->f_dentry->d_inode;
1056 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1057
1058 if (ctx) {
1059 filp->private_data = NULL;
1060 spin_lock(&inode->i_lock);
1061 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1062 spin_unlock(&inode->i_lock);
1063 put_nfs_open_context(ctx);
1064 }
1065}
1066
1067/*
1068 * These allocate and release file read/write context information.
1069 */
1070int nfs_open(struct inode *inode, struct file *filp)
1071{
1072 struct nfs_open_context *ctx;
1073 struct rpc_cred *cred;
1074
1075 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1076 if (IS_ERR(cred))
1077 return PTR_ERR(cred);
1078 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
1079 put_rpccred(cred);
1080 if (ctx == NULL)
1081 return -ENOMEM;
1082 ctx->mode = filp->f_mode;
1083 nfs_file_set_open_context(filp, ctx);
1084 put_nfs_open_context(ctx);
1da177e4
LT		 1085 return 0;
1086}
1087
1088int nfs_release(struct inode *inode, struct file *filp)
1089{
1da177e4
LT		 1090 nfs_file_clear_open_context(filp);
1091 return 0;
1092}
1093
1094/*
1095 * This function is called whenever some part of NFS notices that
1096 * the cached attributes have to be refreshed.
1097 */
1098int
1099__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1100{
1101 int status = -ESTALE;
1102 struct nfs_fattr fattr;
1103 struct nfs_inode *nfsi = NFS_I(inode);
1da177e4
LT		 1104
1105 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1106 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1107
1108 lock_kernel();
1109 if (!inode || is_bad_inode(inode))
1110 goto out_nowait;
1111 if (NFS_STALE(inode))
1112 goto out_nowait;
1113
412d582e
CL		 1114 status = nfs_wait_on_inode(inode);
1115 if (status < 0)
1116 goto out;
1117 if (NFS_STALE(inode)) {
1118 status = -ESTALE;
1119 /* Do we trust the cached ESTALE? */
1120 if (NFS_ATTRTIMEO(inode) != 0) {
1121 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1122 /* no */
1123 } else
1124 goto out;
1125 }
1da177e4 1126 }
1da177e4 1127
1da177e4
LT		 1128 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1129 if (status != 0) {
1130 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1131 inode->i_sb->s_id,
1132 (long long)NFS_FILEID(inode), status);
1133 if (status == -ESTALE) {
1134 nfs_zap_caches(inode);
1135 if (!S_ISDIR(inode->i_mode))
412d582e 1136 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1137 }
1138 goto out;
1139 }
1140
33801147 1141 spin_lock(&inode->i_lock);
24aa1fe6 1142 status = nfs_update_inode(inode, &fattr);
1da177e4 1143 if (status) {
33801147 1144 spin_unlock(&inode->i_lock);
1da177e4
LT		 1145 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1146 inode->i_sb->s_id,
1147 (long long)NFS_FILEID(inode), status);
1148 goto out;
1149 }
dc59250c 1150 spin_unlock(&inode->i_lock);
55296809 1151
7d52e862 1152 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809 1153
24aa1fe6 1154 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1155 nfs_zap_acl_cache(inode);
55296809 1156
1da177e4
LT		 1157 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1158 inode->i_sb->s_id,
1159 (long long)NFS_FILEID(inode));
1160
412d582e
CL		 1161 out:
1162 nfs_wake_up_inode(inode);
1163
1da177e4
LT		 1164 out_nowait:
1165 unlock_kernel();
1166 return status;
1167}
1168
1169int nfs_attribute_timeout(struct inode *inode)
1170{
1171 struct nfs_inode *nfsi = NFS_I(inode);
1172
1173 if (nfs_have_delegation(inode, FMODE_READ))
1174 return 0;
1175 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1176}
1177
1178/**
1179 * nfs_revalidate_inode - Revalidate the inode attributes
1180 * @server - pointer to nfs_server struct
1181 * @inode - pointer to inode struct
1182 *
1183 * Updates inode attribute information by retrieving the data from the server.
1184 */
1185int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1186{
55296809 1187 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1188 && !nfs_attribute_timeout(inode))
1189 return NFS_STALE(inode) ? -ESTALE : 0;
1190 return __nfs_revalidate_inode(server, inode);
1191}
1192
7d52e862
TM		 1193/**
1194 * nfs_revalidate_mapping - Revalidate the pagecache
1195 * @inode - pointer to host inode
1196 * @mapping - pointer to mapping
1197 */
1198void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1199{
1200 struct nfs_inode *nfsi = NFS_I(inode);
1201
55296809 1202 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
29884df0
TM		 1203 if (S_ISREG(inode->i_mode))
1204 nfs_sync_mapping(mapping);
7d52e862 1205 invalidate_inode_pages2(mapping);
dc59250c
CL		 1206
1207 spin_lock(&inode->i_lock);
55296809 1208 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1209 if (S_ISDIR(inode->i_mode)) {
1210 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1211 /* This ensures we revalidate child dentries */
913a70fc 1212 nfsi->cache_change_attribute = jiffies;
7d52e862 1213 }
dc59250c
CL		 1214 spin_unlock(&inode->i_lock);
1215
7d52e862
TM		 1216 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1217 inode->i_sb->s_id,
1218 (long long)NFS_FILEID(inode));
1219 }
1220}
1221
1da177e4
LT		 1222/**
1223 * nfs_begin_data_update
1224 * @inode - pointer to inode
1225 * Declare that a set of operations will update file data on the server
1226 */
1227void nfs_begin_data_update(struct inode *inode)
1228{
1229 atomic_inc(&NFS_I(inode)->data_updates);
1230}
1231
1232/**
1233 * nfs_end_data_update
1234 * @inode - pointer to inode
1235 * Declare end of the operations that will update file data
1236 * This will mark the inode as immediately needing revalidation
1237 * of its attribute cache.
1238 */
1239void nfs_end_data_update(struct inode *inode)
1240{
1241 struct nfs_inode *nfsi = NFS_I(inode);
1242
1243 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1244 /* Directories and symlinks: invalidate page cache */
1245 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1246 spin_lock(&inode->i_lock);
55296809 1247 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1248 spin_unlock(&inode->i_lock);
1249 }
1da177e4 1250 }
913a70fc 1251 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1252 atomic_dec(&nfsi->data_updates);
1253}
1254
1da177e4 1255/**
33801147 1256 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1257 * @inode - pointer to inode
1258 * @fattr - updated attributes
1259 *
1260 * Verifies the attribute cache. If we have just changed the attributes,
1261 * so that fattr carries weak cache consistency data, then it may
1262 * also update the ctime/mtime/change_attribute.
1263 */
33801147 1264static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1265{
1266 struct nfs_inode *nfsi = NFS_I(inode);
1267 loff_t cur_size, new_isize;
1268 int data_unstable;
1269
dc59250c 1270
1da177e4
LT		 1271 /* Are we in the process of updating data on the server? */
1272 data_unstable = nfs_caches_unstable(inode);
1273
1274 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1275 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1276 && nfsi->change_attr == fattr->pre_change_attr)
1277 nfsi->change_attr = fattr->change_attr;
fe51beec 1278 if (nfsi->change_attr != fattr->change_attr) {
55296809 1279 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1280 if (!data_unstable)
55296809 1281 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1282 }
1da177e4
LT		 1283 }
1284
dc59250c 1285 if ((fattr->valid & NFS_ATTR_FATTR) == 0) {
1da177e4 1286 return 0;
dc59250c 1287 }
1da177e4
LT		 1288
1289 /* Has the inode gone and changed behind our back? */
1290 if (nfsi->fileid != fattr->fileid
dc59250c 1291 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1da177e4 1292 return -EIO;
dc59250c 1293 }
1da177e4
LT		 1294
1295 cur_size = i_size_read(inode);
1296 new_isize = nfs_size_to_loff_t(fattr->size);
1297
1298 /* If we have atomic WCC data, we may update some attributes */
1299 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1300 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1301 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1302 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1303 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1304 }
1305
1306 /* Verify a few of the more important attributes */
fe51beec 1307 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1308 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1309 if (!data_unstable)
55296809 1310 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec
TM		 1311 }
1312 if (cur_size != new_isize) {
55296809 1313 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1314 if (nfsi->npages == 0)
55296809 1315 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1316 }
1da177e4
LT		 1317
1318 /* Have any file permissions changed? */
1319 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1320 || inode->i_uid != fattr->uid
1321 || inode->i_gid != fattr->gid)
55296809 1322 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1323
1324 /* Has the link count changed? */
1325 if (inode->i_nlink != fattr->nlink)
55296809 1326 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1327
1328 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1329 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1330
33801147 1331 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1332 return 0;
1333}
1334
33801147
TM		 1335/**
1336 * nfs_refresh_inode - try to update the inode attribute cache
1337 * @inode - pointer to inode
1338 * @fattr - updated attributes
1339 *
1340 * Check that an RPC call that returned attributes has not overlapped with
1341 * other recent updates of the inode metadata, then decide whether it is
1342 * safe to do a full update of the inode attributes, or whether just to
1343 * call nfs_check_inode_attributes.
1344 */
1345int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1346{
1347 struct nfs_inode *nfsi = NFS_I(inode);
1348 int status;
1349
1350 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1351 return 0;
1352 spin_lock(&inode->i_lock);
1353 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
33801147 1354 if (time_after(fattr->time_start, nfsi->last_updated))
24aa1fe6 1355 status = nfs_update_inode(inode, fattr);
33801147
TM		 1356 else
1357 status = nfs_check_inode_attributes(inode, fattr);
1358
1359 spin_unlock(&inode->i_lock);
1360 return status;
1361}
1362
decf491f
TM		 1363/**
1364 * nfs_post_op_update_inode - try to update the inode attribute cache
1365 * @inode - pointer to inode
1366 * @fattr - updated attributes
1367 *
1368 * After an operation that has changed the inode metadata, mark the
1369 * attribute cache as being invalid, then try to update it.
1370 */
1371int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1372{
1373 struct nfs_inode *nfsi = NFS_I(inode);
1374 int status = 0;
1375
1376 spin_lock(&inode->i_lock);
1377 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1378 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1379 goto out;
1380 }
24aa1fe6 1381 status = nfs_update_inode(inode, fattr);
decf491f
TM		 1382out:
1383 spin_unlock(&inode->i_lock);
1384 return status;
1385}
1386
1da177e4
LT		 1387/*
1388 * Many nfs protocol calls return the new file attributes after
1389 * an operation. Here we update the inode to reflect the state
1390 * of the server's inode.
1391 *
1392 * This is a bit tricky because we have to make sure all dirty pages
1393 * have been sent off to the server before calling invalidate_inode_pages.
1394 * To make sure no other process adds more write requests while we try
1395 * our best to flush them, we make them sleep during the attribute refresh.
1396 *
1397 * A very similar scenario holds for the dir cache.
1398 */
24aa1fe6 1399static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1400{
1401 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1402 loff_t cur_isize, new_isize;
1da177e4 1403 unsigned int invalid = 0;
24aa1fe6 1404 int data_stable;
1da177e4
LT		 1405
1406 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1407 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1408 atomic_read(&inode->i_count), fattr->valid);
1409
1410 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1411 return 0;
1412
1413 if (nfsi->fileid != fattr->fileid) {
1414 printk(KERN_ERR "%s: inode number mismatch\n"
1415 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1416 __FUNCTION__,
1417 inode->i_sb->s_id, (long long)nfsi->fileid,
1418 inode->i_sb->s_id, (long long)fattr->fileid);
1419 goto out_err;
1420 }
1421
1422 /*
1423 * Make sure the inode's type hasn't changed.
1424 */
33801147 1425 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1da177e4
LT		 1426 goto out_changed;
1427
1428 /*
1429 * Update the read time so we don't revalidate too often.
1430 */
33801147
TM		 1431 nfsi->read_cache_jiffies = fattr->time_start;
1432 nfsi->last_updated = jiffies;
1da177e4
LT		 1433
1434 /* Are we racing with known updates of the metadata on the server? */
24aa1fe6
TM		 1435 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1436 if (data_stable)
1437 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1da177e4 1438
951a143b 1439 /* Check if our cached file size is stale */
1da177e4
LT		 1440 new_isize = nfs_size_to_loff_t(fattr->size);
1441 cur_isize = i_size_read(inode);
951a143b
TM		 1442 if (new_isize != cur_isize) {
1443 /* Do we perhaps have any outstanding writes? */
1444 if (nfsi->npages == 0) {
1445 /* No, but did we race with nfs_end_data_update()? */
24aa1fe6 1446 if (data_stable) {
1da177e4 1447 inode->i_size = new_isize;
951a143b 1448 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1449 }
951a143b
TM		 1450 invalid |= NFS_INO_INVALID_ATTR;
1451 } else if (new_isize > cur_isize) {
1da177e4
LT		 1452 inode->i_size = new_isize;
1453 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1454 }
24aa1fe6 1455 nfsi->cache_change_attribute = jiffies;
951a143b
TM		 1456 dprintk("NFS: isize change on server for file %s/%ld\n",
1457 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1458 }
1459
951a143b 1460 /* Check if the mtime agrees */
1da177e4
LT		 1461 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1462 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1463 dprintk("NFS: mtime change on server for file %s/%ld\n",
1464 inode->i_sb->s_id, inode->i_ino);
24aa1fe6
TM		 1465 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1466 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1467 }
1468
1469 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1470 && nfsi->change_attr != fattr->change_attr) {
951a143b 1471 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1da177e4 1472 inode->i_sb->s_id, inode->i_ino);
1da177e4 1473 nfsi->change_attr = fattr->change_attr;
24aa1fe6
TM		 1474 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1475 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1476 }
1477
ada70d94
TM		 1478 /* If ctime has changed we should definitely clear access+acl caches */
1479 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
24aa1fe6 1480 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
ada70d94 1481 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
24aa1fe6 1482 nfsi->cache_change_attribute = jiffies;
ada70d94 1483 }
1da177e4
LT		 1484 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1485
1486 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1487 inode->i_uid != fattr->uid ||
1488 inode->i_gid != fattr->gid)
ada70d94 1489 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1490
1491 inode->i_mode = fattr->mode;
1492 inode->i_nlink = fattr->nlink;
1493 inode->i_uid = fattr->uid;
1494 inode->i_gid = fattr->gid;
1495
1496 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1497 /*
1498 * report the blocks in 512byte units
1499 */
1500 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1501 inode->i_blksize = inode->i_sb->s_blocksize;
1502 } else {
1503 inode->i_blocks = fattr->du.nfs2.blocks;
1504 inode->i_blksize = fattr->du.nfs2.blocksize;
1505 }
1506
1507 /* Update attrtimeo value if we're out of the unstable period */
1508 if (invalid & NFS_INO_INVALID_ATTR) {
1509 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1510 nfsi->attrtimeo_timestamp = jiffies;
1511 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1512 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1513 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1514 nfsi->attrtimeo_timestamp = jiffies;
1515 }
1516 /* Don't invalidate the data if we were to blame */
1517 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1518 || S_ISLNK(inode->i_mode)))
1519 invalid &= ~NFS_INO_INVALID_DATA;
24aa1fe6
TM		 1520 if (data_stable)
1521 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1da177e4 1522 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1523 nfsi->cache_validity |= invalid;
1da177e4
LT		 1524
1525 return 0;
1526 out_changed:
1527 /*
1528 * Big trouble! The inode has become a different object.
1529 */
1530#ifdef NFS_PARANOIA
1531 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1532 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1533#endif
b37b03b7 1534 out_err:
1da177e4
LT		 1535 /*
1536 * No need to worry about unhashing the dentry, as the
1537 * lookup validation will know that the inode is bad.
1538 * (But we fall through to invalidate the caches.)
1539 */
1540 nfs_invalidate_inode(inode);
1da177e4
LT		 1541 return -ESTALE;
1542}
1543
1544/*
1545 * File system information
1546 */
1547
1548static int nfs_set_super(struct super_block *s, void *data)
1549{
1550 s->s_fs_info = data;
1551 return set_anon_super(s, data);
1552}
1553
1554static int nfs_compare_super(struct super_block *sb, void *data)
1555{
1556 struct nfs_server *server = data;
1557 struct nfs_server *old = NFS_SB(sb);
1558
1559 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1560 return 0;
1561 if (old->addr.sin_port != server->addr.sin_port)
1562 return 0;
1563 return !nfs_compare_fh(&old->fh, &server->fh);
1564}
1565
1566static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1567 int flags, const char *dev_name, void *raw_data)
1568{
1569 int error;
9085bbcb 1570 struct nfs_server *server = NULL;
1da177e4
LT		 1571 struct super_block *s;
1572 struct nfs_fh *root;
1573 struct nfs_mount_data *data = raw_data;
1574
9085bbcb
TM		 1575 s = ERR_PTR(-EINVAL);
1576 if (data == NULL) {
1577 dprintk("%s: missing data argument\n", __FUNCTION__);
1578 goto out_err;
1579 }
1580 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1581 dprintk("%s: bad mount version\n", __FUNCTION__);
1582 goto out_err;
1da177e4 1583 }
9085bbcb
TM		 1584 switch (data->version) {
1585 case 1:
1586 data->namlen = 0;
1587 case 2:
1588 data->bsize = 0;
1589 case 3:
1590 if (data->flags & NFS_MOUNT_VER3) {
1591 dprintk("%s: mount structure version %d does not support NFSv3\n",
1592 __FUNCTION__,
1593 data->version);
1594 goto out_err;
1595 }
1596 data->root.size = NFS2_FHSIZE;
1597 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1598 case 4:
1599 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1600 dprintk("%s: mount structure version %d does not support strong security\n",
1601 __FUNCTION__,
1602 data->version);
1603 goto out_err;
1604 }
1605 case 5:
1606 memset(data->context, 0, sizeof(data->context));
1607 }
1608#ifndef CONFIG_NFS_V3
1609 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1610 s = ERR_PTR(-EPROTONOSUPPORT);
1611 if (data->flags & NFS_MOUNT_VER3) {
1612 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1613 goto out_err;
1614 }
1615#endif /* CONFIG_NFS_V3 */
1da177e4 1616
9085bbcb 1617 s = ERR_PTR(-ENOMEM);
1da177e4
LT		 1618 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1619 if (!server)
9085bbcb 1620 goto out_err;
1da177e4
LT		 1621 memset(server, 0, sizeof(struct nfs_server));
1622 /* Zero out the NFS state stuff */
1623 init_nfsv4_state(server);
b7fa0554 1624 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1625
1da177e4
LT		 1626 root = &server->fh;
1627 if (data->flags & NFS_MOUNT_VER3)
1628 root->size = data->root.size;
1629 else
1630 root->size = NFS2_FHSIZE;
9085bbcb 1631 s = ERR_PTR(-EINVAL);
1da177e4 1632 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1633 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1634 goto out_err;
1da177e4
LT		 1635 }
1636 memcpy(root->data, data->root.data, root->size);
1637
1638 /* We now require that the mount process passes the remote address */
1639 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1640 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1641 dprintk("%s: mount program didn't pass remote address!\n",
1642 __FUNCTION__);
1643 goto out_err;
1da177e4
LT		 1644 }
1645
9085bbcb
TM		 1646 /* Fire up rpciod if not yet running */
1647 s = ERR_PTR(rpciod_up());
1648 if (IS_ERR(s)) {
1649 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1650 __FUNCTION__, PTR_ERR(s));
1651 goto out_err;
1da177e4
LT		 1652 }
1653
9085bbcb
TM		 1654 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1655 if (IS_ERR(s) || s->s_root)
1656 goto out_rpciod_down;
1da177e4 1657
9085bbcb 1658 s->s_flags = flags;
1da177e4
LT		 1659
1660 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1661 if (error) {
1662 up_write(&s->s_umount);
1663 deactivate_super(s);
1664 return ERR_PTR(error);
1665 }
1666 s->s_flags |= MS_ACTIVE;
1667 return s;
9085bbcb
TM		 1668out_rpciod_down:
1669 rpciod_down();
1670out_err:
1671 kfree(server);
1672 return s;
1da177e4
LT		 1673}
1674
1675static void nfs_kill_super(struct super_block *s)
1676{
1677 struct nfs_server *server = NFS_SB(s);
1678
1679 kill_anon_super(s);
1680
6a19275a 1681 if (!IS_ERR(server->client))
1da177e4 1682 rpc_shutdown_client(server->client);
6a19275a 1683 if (!IS_ERR(server->client_sys))
1da177e4 1684 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1685 if (!IS_ERR(server->client_acl))
1686 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1687
1688 if (!(server->flags & NFS_MOUNT_NONLM))
1689 lockd_down(); /* release rpc.lockd */
1690
1691 rpciod_down(); /* release rpciod */
1692
f99d49ad 1693 kfree(server->hostname);
1da177e4
LT		 1694 kfree(server);
1695}
1696
1697static struct file_system_type nfs_fs_type = {
1698 .owner = THIS_MODULE,
1699 .name = "nfs",
1700 .get_sb = nfs_get_sb,
1701 .kill_sb = nfs_kill_super,
1702 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1703};
1704
1705#ifdef CONFIG_NFS_V4
1706
1707static void nfs4_clear_inode(struct inode *);
1708
1709
1710static struct super_operations nfs4_sops = {
1711 .alloc_inode = nfs_alloc_inode,
1712 .destroy_inode = nfs_destroy_inode,
1713 .write_inode = nfs_write_inode,
1714 .delete_inode = nfs_delete_inode,
1715 .statfs = nfs_statfs,
1716 .clear_inode = nfs4_clear_inode,
1717 .umount_begin = nfs_umount_begin,
1718 .show_options = nfs_show_options,
1719};
1720
1721/*
1722 * Clean out any remaining NFSv4 state that might be left over due
1723 * to open() calls that passed nfs_atomic_lookup, but failed to call
1724 * nfs_open().
1725 */
1726static void nfs4_clear_inode(struct inode *inode)
1727{
1728 struct nfs_inode *nfsi = NFS_I(inode);
1729
1730 /* If we are holding a delegation, return it! */
cae7a073 1731 nfs_inode_return_delegation(inode);
1da177e4
LT		 1732 /* First call standard NFS clear_inode() code */
1733 nfs_clear_inode(inode);
1734 /* Now clear out any remaining state */
1735 while (!list_empty(&nfsi->open_states)) {
1736 struct nfs4_state *state;
1737
1738 state = list_entry(nfsi->open_states.next,
1739 struct nfs4_state,
1740 inode_states);
1741 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1742 __FUNCTION__,
1743 inode->i_sb->s_id,
1744 (long long)NFS_FILEID(inode),
1745 state);
1746 BUG_ON(atomic_read(&state->count) != 1);
1747 nfs4_close_state(state, state->state);
1748 }
1749}
1750
1751
1752static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1753{
1754 struct nfs_server *server;
1755 struct nfs4_client *clp = NULL;
1756 struct rpc_xprt *xprt = NULL;
1757 struct rpc_clnt *clnt = NULL;
1758 struct rpc_timeout timeparms;
1759 rpc_authflavor_t authflavour;
eab5c084 1760 int err = -EIO;
1da177e4
LT		 1761
1762 sb->s_blocksize_bits = 0;
1763 sb->s_blocksize = 0;
1764 server = NFS_SB(sb);
1765 if (data->rsize != 0)
1766 server->rsize = nfs_block_size(data->rsize, NULL);
1767 if (data->wsize != 0)
1768 server->wsize = nfs_block_size(data->wsize, NULL);
1769 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1770 server->caps = NFS_CAP_ATOMIC_OPEN;
1771
1772 server->acregmin = data->acregmin*HZ;
1773 server->acregmax = data->acregmax*HZ;
1774 server->acdirmin = data->acdirmin*HZ;
1775 server->acdirmax = data->acdirmax*HZ;
1776
1777 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1778
eab5c084 1779 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4
LT		 1780
1781 clp = nfs4_get_client(&server->addr.sin_addr);
1782 if (!clp) {
9085bbcb 1783 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1784 return -EIO;
1785 }
1786
1787 /* Now create transport and client */
1788 authflavour = RPC_AUTH_UNIX;
1789 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1790 if (data->auth_flavourlen != 1) {
1791 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1792 __FUNCTION__, data->auth_flavourlen);
1793 err = -EINVAL;
1794 goto out_fail;
1795 }
1da177e4
LT		 1796 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1797 err = -EFAULT;
1798 goto out_fail;
1799 }
1800 }
1801
1802 down_write(&clp->cl_sem);
6a19275a 1803 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1804 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT		 1805 if (IS_ERR(xprt)) {
1806 up_write(&clp->cl_sem);
1da177e4 1807 err = PTR_ERR(xprt);
9085bbcb
TM		 1808 dprintk("%s: cannot create RPC transport. Error = %d\n",
1809 __FUNCTION__, err);
1da177e4
LT		 1810 goto out_fail;
1811 }
1812 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1813 server->rpc_ops->version, authflavour);
1814 if (IS_ERR(clnt)) {
1815 up_write(&clp->cl_sem);
1da177e4 1816 err = PTR_ERR(clnt);
9085bbcb
TM		 1817 dprintk("%s: cannot create RPC client. Error = %d\n",
1818 __FUNCTION__, err);
1da177e4
LT		 1819 goto out_fail;
1820 }
1821 clnt->cl_intr = 1;
1822 clnt->cl_softrtry = 1;
1823 clnt->cl_chatty = 1;
1824 clp->cl_rpcclient = clnt;
1da177e4
LT		 1825 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1826 nfs_idmap_new(clp);
1827 }
1da177e4
LT		 1828 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1829 clnt = rpc_clone_client(clp->cl_rpcclient);
1830 if (!IS_ERR(clnt))
1831 server->nfs4_state = clp;
1832 up_write(&clp->cl_sem);
1833 clp = NULL;
1834
1835 if (IS_ERR(clnt)) {
9085bbcb
TM		 1836 err = PTR_ERR(clnt);
1837 dprintk("%s: cannot create RPC client. Error = %d\n",
1838 __FUNCTION__, err);
1839 return err;
1da177e4
LT		 1840 }
1841
1842 server->client = clnt;
1843
1844 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1845 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1846 return -ENOMEM;
1847 }
1848
1849 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1850 struct rpc_auth *auth;
1851
1852 auth = rpcauth_create(authflavour, clnt);
1853 if (IS_ERR(auth)) {
9085bbcb 1854 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1855 return PTR_ERR(auth);
1da177e4
LT		 1856 }
1857 }
1858
1859 sb->s_time_gran = 1;
1860
1861 sb->s_op = &nfs4_sops;
1862 err = nfs_sb_init(sb, authflavour);
1863 if (err == 0)
1864 return 0;
1865out_fail:
1866 if (clp)
1867 nfs4_put_client(clp);
1868 return err;
1869}
1870
1871static int nfs4_compare_super(struct super_block *sb, void *data)
1872{
1873 struct nfs_server *server = data;
1874 struct nfs_server *old = NFS_SB(sb);
1875
1876 if (strcmp(server->hostname, old->hostname) != 0)
1877 return 0;
1878 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1879 return 0;
1880 return 1;
1881}
1882
1883static void *
1884nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1885{
1886 void *p = NULL;
1887
1888 if (!src->len)
1889 return ERR_PTR(-EINVAL);
1890 if (src->len < maxlen)
1891 maxlen = src->len;
1892 if (dst == NULL) {
1893 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1894 if (p == NULL)
1895 return ERR_PTR(-ENOMEM);
1896 }
1897 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 1898 kfree(p);
1da177e4
LT		 1899 return ERR_PTR(-EFAULT);
1900 }
1901 dst[maxlen] = '\0';
1902 return dst;
1903}
1904
1905static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1906 int flags, const char *dev_name, void *raw_data)
1907{
1908 int error;
1909 struct nfs_server *server;
1910 struct super_block *s;
1911 struct nfs4_mount_data *data = raw_data;
1912 void *p;
1913
9085bbcb
TM		 1914 if (data == NULL) {
1915 dprintk("%s: missing data argument\n", __FUNCTION__);
1916 return ERR_PTR(-EINVAL);
1917 }
1918 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1919 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1920 return ERR_PTR(-EINVAL);
1921 }
1922
1923 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1924 if (!server)
1925 return ERR_PTR(-ENOMEM);
1926 memset(server, 0, sizeof(struct nfs_server));
1927 /* Zero out the NFS state stuff */
1928 init_nfsv4_state(server);
b7fa0554 1929 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1930
1da177e4
LT		 1931 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1932 if (IS_ERR(p))
1933 goto out_err;
1934 server->hostname = p;
1935
1936 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1937 if (IS_ERR(p))
1938 goto out_err;
1939 server->mnt_path = p;
1940
1941 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1942 sizeof(server->ip_addr) - 1);
1943 if (IS_ERR(p))
1944 goto out_err;
1945
1946 /* We now require that the mount process passes the remote address */
1947 if (data->host_addrlen != sizeof(server->addr)) {
1948 s = ERR_PTR(-EINVAL);
1949 goto out_free;
1950 }
1951 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1952 s = ERR_PTR(-EFAULT);
1953 goto out_free;
1954 }
1955 if (server->addr.sin_family != AF_INET ||
1956 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1957 dprintk("%s: mount program didn't pass remote IP address!\n",
1958 __FUNCTION__);
1da177e4
LT		 1959 s = ERR_PTR(-EINVAL);
1960 goto out_free;
1961 }
1962
9085bbcb
TM		 1963 /* Fire up rpciod if not yet running */
1964 s = ERR_PTR(rpciod_up());
1965 if (IS_ERR(s)) {
1966 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1967 __FUNCTION__, PTR_ERR(s));
1968 goto out_free;
1969 }
1970
1da177e4
LT		 1971 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1972
1973 if (IS_ERR(s) || s->s_root)
1974 goto out_free;
1975
1976 s->s_flags = flags;
1977
1da177e4
LT		 1978 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1979 if (error) {
1980 up_write(&s->s_umount);
1981 deactivate_super(s);
1982 return ERR_PTR(error);
1983 }
1984 s->s_flags |= MS_ACTIVE;
1985 return s;
1986out_err:
1987 s = (struct super_block *)p;
1988out_free:
f99d49ad
JJ		 1989 kfree(server->mnt_path);
1990 kfree(server->hostname);
1da177e4
LT		 1991 kfree(server);
1992 return s;
1993}
1994
1995static void nfs4_kill_super(struct super_block *sb)
1996{
1997 struct nfs_server *server = NFS_SB(sb);
1998
1999 nfs_return_all_delegations(sb);
2000 kill_anon_super(sb);
2001
2002 nfs4_renewd_prepare_shutdown(server);
2003
2004 if (server->client != NULL && !IS_ERR(server->client))
2005 rpc_shutdown_client(server->client);
2006 rpciod_down(); /* release rpciod */
2007
2008 destroy_nfsv4_state(server);
2009
f99d49ad 2010 kfree(server->hostname);
1da177e4
LT		 2011 kfree(server);
2012}
2013
2014static struct file_system_type nfs4_fs_type = {
2015 .owner = THIS_MODULE,
2016 .name = "nfs4",
2017 .get_sb = nfs4_get_sb,
2018 .kill_sb = nfs4_kill_super,
2019 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2020};
2021
2022#define nfs4_init_once(nfsi) \
2023 do { \
2024 INIT_LIST_HEAD(&(nfsi)->open_states); \
2025 nfsi->delegation = NULL; \
2026 nfsi->delegation_state = 0; \
2027 init_rwsem(&nfsi->rwsem); \
2028 } while(0)
2029#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
2030#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
2031#else
2032#define nfs4_init_once(nfsi) \
2033 do { } while (0)
2034#define register_nfs4fs() (0)
2035#define unregister_nfs4fs()
2036#endif
2037
2038extern int nfs_init_nfspagecache(void);
2039extern void nfs_destroy_nfspagecache(void);
2040extern int nfs_init_readpagecache(void);
2041extern void nfs_destroy_readpagecache(void);
2042extern int nfs_init_writepagecache(void);
2043extern void nfs_destroy_writepagecache(void);
2044#ifdef CONFIG_NFS_DIRECTIO
2045extern int nfs_init_directcache(void);
2046extern void nfs_destroy_directcache(void);
2047#endif
2048
2049static kmem_cache_t * nfs_inode_cachep;
2050
2051static struct inode *nfs_alloc_inode(struct super_block *sb)
2052{
2053 struct nfs_inode *nfsi;
2054 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2055 if (!nfsi)
2056 return NULL;
55296809
CL		 2057 nfsi->flags = 0UL;
2058 nfsi->cache_validity = 0UL;
223db122 2059 nfsi->cache_change_attribute = jiffies;
458818ed
TM		 2060#ifdef CONFIG_NFS_V3_ACL
2061 nfsi->acl_access = ERR_PTR(-EAGAIN);
2062 nfsi->acl_default = ERR_PTR(-EAGAIN);
2063#endif
e50a1c2e
BF		 2064#ifdef CONFIG_NFS_V4
2065 nfsi->nfs4_acl = NULL;
2066#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2067 return &nfsi->vfs_inode;
2068}
2069
2070static void nfs_destroy_inode(struct inode *inode)
2071{
2072 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2073}
2074
2075static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2076{
2077 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2078
2079 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2080 SLAB_CTOR_CONSTRUCTOR) {
2081 inode_init_once(&nfsi->vfs_inode);
2082 spin_lock_init(&nfsi->req_lock);
2083 INIT_LIST_HEAD(&nfsi->dirty);
2084 INIT_LIST_HEAD(&nfsi->commit);
2085 INIT_LIST_HEAD(&nfsi->open_files);
2086 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2087 atomic_set(&nfsi->data_updates, 0);
2088 nfsi->ndirty = 0;
2089 nfsi->ncommit = 0;
2090 nfsi->npages = 0;
1da177e4
LT		 2091 nfs4_init_once(nfsi);
2092 }
2093}
2094
75c96f85 2095static int nfs_init_inodecache(void)
1da177e4
LT		 2096{
2097 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2098 sizeof(struct nfs_inode),
2099 0, SLAB_RECLAIM_ACCOUNT,
2100 init_once, NULL);
2101 if (nfs_inode_cachep == NULL)
2102 return -ENOMEM;
2103
2104 return 0;
2105}
2106
75c96f85 2107static void nfs_destroy_inodecache(void)
1da177e4
LT		 2108{
2109 if (kmem_cache_destroy(nfs_inode_cachep))
2110 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2111}
2112
2113/*
2114 * Initialize NFS
2115 */
2116static int __init init_nfs_fs(void)
2117{
2118 int err;
2119
2120 err = nfs_init_nfspagecache();
2121 if (err)
2122 goto out4;
2123
2124 err = nfs_init_inodecache();
2125 if (err)
2126 goto out3;
2127
2128 err = nfs_init_readpagecache();
2129 if (err)
2130 goto out2;
2131
2132 err = nfs_init_writepagecache();
2133 if (err)
2134 goto out1;
2135
2136#ifdef CONFIG_NFS_DIRECTIO
2137 err = nfs_init_directcache();
2138 if (err)
2139 goto out0;
2140#endif
2141
2142#ifdef CONFIG_PROC_FS
2143 rpc_proc_register(&nfs_rpcstat);
2144#endif
2145 err = register_filesystem(&nfs_fs_type);
2146 if (err)
2147 goto out;
2148 if ((err = register_nfs4fs()) != 0)
2149 goto out;
2150 return 0;
2151out:
2152#ifdef CONFIG_PROC_FS
2153 rpc_proc_unregister("nfs");
2154#endif
2155 nfs_destroy_writepagecache();
2156#ifdef CONFIG_NFS_DIRECTIO
2157out0:
2158 nfs_destroy_directcache();
2159#endif
2160out1:
2161 nfs_destroy_readpagecache();
2162out2:
2163 nfs_destroy_inodecache();
2164out3:
2165 nfs_destroy_nfspagecache();
2166out4:
2167 return err;
2168}
2169
2170static void __exit exit_nfs_fs(void)
2171{
2172#ifdef CONFIG_NFS_DIRECTIO
2173 nfs_destroy_directcache();
2174#endif
2175 nfs_destroy_writepagecache();
2176 nfs_destroy_readpagecache();
2177 nfs_destroy_inodecache();
2178 nfs_destroy_nfspagecache();
2179#ifdef CONFIG_PROC_FS
2180 rpc_proc_unregister("nfs");
2181#endif
2182 unregister_filesystem(&nfs_fs_type);
2183 unregister_nfs4fs();
2184}
2185
2186/* Not quite true; I just maintain it */
2187MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2188MODULE_LICENSE("GPL");
2189
2190module_init(init_nfs_fs)
2191module_exit(exit_nfs_fs)
Clone of davem's net-next-2.6 tree