projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
NFS: Fix cache consistency regression
[linux-2.6-block.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
4ce79717 42#include "nfs4_fs.h"
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 *);
57static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long);
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
643/*
644 * Invalidate the local caches
645 */
b37b03b7 646static void nfs_zap_caches_locked(struct inode *inode)
1da177e4
LT		 647{
648 struct nfs_inode *nfsi = NFS_I(inode);
649 int mode = inode->i_mode;
650
651 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
652 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
653
654 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
655 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 656 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 657 else
55296809 658 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
b37b03b7 659}
dc59250c 660
b37b03b7
TM		 661void nfs_zap_caches(struct inode *inode)
662{
663 spin_lock(&inode->i_lock);
664 nfs_zap_caches_locked(inode);
dc59250c 665 spin_unlock(&inode->i_lock);
ada70d94
TM		 666}
667
668static void nfs_zap_acl_cache(struct inode *inode)
669{
670 void (*clear_acl_cache)(struct inode *);
671
672 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
673 if (clear_acl_cache != NULL)
674 clear_acl_cache(inode);
dc59250c 675 spin_lock(&inode->i_lock);
55296809 676 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 677 spin_unlock(&inode->i_lock);
1da177e4
LT		 678}
679
680/*
b37b03b7
TM		 681 * Invalidate, but do not unhash, the inode.
682 * NB: must be called with inode->i_lock held!
1da177e4 683 */
b37b03b7 684static void nfs_invalidate_inode(struct inode *inode)
1da177e4 685{
b37b03b7
TM		 686 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
687 nfs_zap_caches_locked(inode);
1da177e4
LT		 688}
689
690struct nfs_find_desc {
691 struct nfs_fh *fh;
692 struct nfs_fattr *fattr;
693};
694
695/*
696 * In NFSv3 we can have 64bit inode numbers. In order to support
697 * this, and re-exported directories (also seen in NFSv2)
698 * we are forced to allow 2 different inodes to have the same
699 * i_ino.
700 */
701static int
702nfs_find_actor(struct inode *inode, void *opaque)
703{
704 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
705 struct nfs_fh *fh = desc->fh;
706 struct nfs_fattr *fattr = desc->fattr;
707
708 if (NFS_FILEID(inode) != fattr->fileid)
709 return 0;
710 if (nfs_compare_fh(NFS_FH(inode), fh))
711 return 0;
712 if (is_bad_inode(inode) || NFS_STALE(inode))
713 return 0;
714 return 1;
715}
716
717static int
718nfs_init_locked(struct inode *inode, void *opaque)
719{
720 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
721 struct nfs_fattr *fattr = desc->fattr;
722
723 NFS_FILEID(inode) = fattr->fileid;
724 nfs_copy_fh(NFS_FH(inode), desc->fh);
725 return 0;
726}
727
728/* Don't use READDIRPLUS on directories that we believe are too large */
729#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
730
731/*
732 * This is our front-end to iget that looks up inodes by file handle
733 * instead of inode number.
734 */
735struct inode *
736nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
737{
738 struct nfs_find_desc desc = {
739 .fh = fh,
740 .fattr = fattr
741 };
742 struct inode *inode = NULL;
743 unsigned long hash;
744
745 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
746 goto out_no_inode;
747
748 if (!fattr->nlink) {
749 printk("NFS: Buggy server - nlink == 0!\n");
750 goto out_no_inode;
751 }
752
753 hash = nfs_fattr_to_ino_t(fattr);
754
755 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
756 goto out_no_inode;
757
758 if (inode->i_state & I_NEW) {
759 struct nfs_inode *nfsi = NFS_I(inode);
760
761 /* We set i_ino for the few things that still rely on it,
762 * such as stat(2) */
763 inode->i_ino = hash;
764
765 /* We can't support update_atime(), since the server will reset it */
766 inode->i_flags |= S_NOATIME|S_NOCMTIME;
767 inode->i_mode = fattr->mode;
768 /* Why so? Because we want revalidate for devices/FIFOs, and
769 * that's precisely what we have in nfs_file_inode_operations.
770 */
92cfc62c 771 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 772 if (S_ISREG(inode->i_mode)) {
773 inode->i_fop = &nfs_file_operations;
774 inode->i_data.a_ops = &nfs_file_aops;
775 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
776 } else if (S_ISDIR(inode->i_mode)) {
777 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
778 inode->i_fop = &nfs_dir_operations;
779 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
780 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 781 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 782 } else if (S_ISLNK(inode->i_mode))
783 inode->i_op = &nfs_symlink_inode_operations;
784 else
785 init_special_inode(inode, inode->i_mode, fattr->rdev);
786
33801147
TM		 787 nfsi->read_cache_jiffies = fattr->time_start;
788 nfsi->last_updated = jiffies;
1da177e4
LT		 789 inode->i_atime = fattr->atime;
790 inode->i_mtime = fattr->mtime;
791 inode->i_ctime = fattr->ctime;
792 if (fattr->valid & NFS_ATTR_FATTR_V4)
793 nfsi->change_attr = fattr->change_attr;
794 inode->i_size = nfs_size_to_loff_t(fattr->size);
795 inode->i_nlink = fattr->nlink;
796 inode->i_uid = fattr->uid;
797 inode->i_gid = fattr->gid;
798 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
799 /*
800 * report the blocks in 512byte units
801 */
802 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
803 inode->i_blksize = inode->i_sb->s_blocksize;
804 } else {
805 inode->i_blocks = fattr->du.nfs2.blocks;
806 inode->i_blksize = fattr->du.nfs2.blocksize;
807 }
808 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
809 nfsi->attrtimeo_timestamp = jiffies;
810 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
811 nfsi->cache_access.cred = NULL;
812
813 unlock_new_inode(inode);
814 } else
815 nfs_refresh_inode(inode, fattr);
816 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
817 inode->i_sb->s_id,
818 (long long)NFS_FILEID(inode),
819 atomic_read(&inode->i_count));
820
821out:
822 return inode;
823
824out_no_inode:
825 printk("nfs_fhget: iget failed\n");
826 goto out;
827}
828
829#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
830
831int
832nfs_setattr(struct dentry *dentry, struct iattr *attr)
833{
834 struct inode *inode = dentry->d_inode;
835 struct nfs_fattr fattr;
836 int error;
837
838 if (attr->ia_valid & ATTR_SIZE) {
839 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
840 attr->ia_valid &= ~ATTR_SIZE;
841 }
842
843 /* Optimization: if the end result is no change, don't RPC */
844 attr->ia_valid &= NFS_VALID_ATTRS;
845 if (attr->ia_valid == 0)
846 return 0;
847
848 lock_kernel();
849 nfs_begin_data_update(inode);
850 /* Write all dirty data if we're changing file permissions or size */
851 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
852 if (filemap_fdatawrite(inode->i_mapping) == 0)
853 filemap_fdatawait(inode->i_mapping);
854 nfs_wb_all(inode);
855 }
642ac549
TM		 856 /*
857 * Return any delegations if we're going to change ACLs
858 */
859 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
860 nfs_inode_return_delegation(inode);
1da177e4 861 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 862 if (error == 0)
1da177e4 863 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 864 nfs_end_data_update(inode);
865 unlock_kernel();
866 return error;
867}
868
869/**
870 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
871 * @inode: pointer to struct inode
872 * @attr: pointer to struct iattr
873 *
874 * Note: we do this in the *proc.c in order to ensure that
875 * it works for things like exclusive creates too.
876 */
877void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
878{
879 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 880 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 881 int mode = attr->ia_mode & S_IALLUGO;
882 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 883 inode->i_mode = mode;
884 }
885 if ((attr->ia_valid & ATTR_UID) != 0)
886 inode->i_uid = attr->ia_uid;
887 if ((attr->ia_valid & ATTR_GID) != 0)
888 inode->i_gid = attr->ia_gid;
dc59250c 889 spin_lock(&inode->i_lock);
55296809 890 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 891 spin_unlock(&inode->i_lock);
65e4308d
TM		 892 }
893 if ((attr->ia_valid & ATTR_SIZE) != 0) {
894 inode->i_size = attr->ia_size;
895 vmtruncate(inode, attr->ia_size);
896 }
1da177e4
LT		 897}
898
412d582e
CL		 899static int nfs_wait_schedule(void *word)
900{
901 if (signal_pending(current))
902 return -ERESTARTSYS;
903 schedule();
904 return 0;
905}
906
1da177e4
LT		 907/*
908 * Wait for the inode to get unlocked.
1da177e4 909 */
412d582e 910static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 911{
912 struct rpc_clnt *clnt = NFS_CLIENT(inode);
913 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 914 sigset_t oldmask;
1da177e4 915 int error;
412d582e 916
412d582e
CL		 917 rpc_clnt_sigmask(clnt, &oldmask);
918 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
919 nfs_wait_schedule, TASK_INTERRUPTIBLE);
920 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 921
1da177e4
LT		 922 return error;
923}
924
412d582e
CL		 925static void nfs_wake_up_inode(struct inode *inode)
926{
927 struct nfs_inode *nfsi = NFS_I(inode);
928
929 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
930 smp_mb__after_clear_bit();
931 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
932}
933
1da177e4
LT		 934int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
935{
936 struct inode *inode = dentry->d_inode;
55296809 937 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 938 int err;
939
940 if (__IS_FLG(inode, MS_NOATIME))
941 need_atime = 0;
942 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
943 need_atime = 0;
944 /* We may force a getattr if the user cares about atime */
945 if (need_atime)
946 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
947 else
948 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
949 if (!err)
950 generic_fillattr(inode, stat);
951 return err;
952}
953
954struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
955{
956 struct nfs_open_context *ctx;
957
958 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
959 if (ctx != NULL) {
960 atomic_set(&ctx->count, 1);
961 ctx->dentry = dget(dentry);
962 ctx->cred = get_rpccred(cred);
963 ctx->state = NULL;
964 ctx->lockowner = current->files;
965 ctx->error = 0;
00a92642 966 ctx->dir_cookie = 0;
1da177e4
LT		 967 }
968 return ctx;
969}
970
971struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
972{
973 if (ctx != NULL)
974 atomic_inc(&ctx->count);
975 return ctx;
976}
977
978void put_nfs_open_context(struct nfs_open_context *ctx)
979{
980 if (atomic_dec_and_test(&ctx->count)) {
981 if (!list_empty(&ctx->list)) {
982 struct inode *inode = ctx->dentry->d_inode;
983 spin_lock(&inode->i_lock);
984 list_del(&ctx->list);
985 spin_unlock(&inode->i_lock);
986 }
987 if (ctx->state != NULL)
988 nfs4_close_state(ctx->state, ctx->mode);
989 if (ctx->cred != NULL)
990 put_rpccred(ctx->cred);
991 dput(ctx->dentry);
992 kfree(ctx);
993 }
994}
995
996/*
997 * Ensure that mmap has a recent RPC credential for use when writing out
998 * shared pages
999 */
1000void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1001{
1002 struct inode *inode = filp->f_dentry->d_inode;
1003 struct nfs_inode *nfsi = NFS_I(inode);
1004
1005 filp->private_data = get_nfs_open_context(ctx);
1006 spin_lock(&inode->i_lock);
1007 list_add(&ctx->list, &nfsi->open_files);
1008 spin_unlock(&inode->i_lock);
1009}
1010
d530838b
TM		 1011/*
1012 * Given an inode, search for an open context with the desired characteristics
1013 */
1014struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1da177e4
LT		 1015{
1016 struct nfs_inode *nfsi = NFS_I(inode);
1017 struct nfs_open_context *pos, *ctx = NULL;
1018
1019 spin_lock(&inode->i_lock);
1020 list_for_each_entry(pos, &nfsi->open_files, list) {
d530838b
TM		 1021 if (cred != NULL && pos->cred != cred)
1022 continue;
1da177e4
LT		 1023 if ((pos->mode & mode) == mode) {
1024 ctx = get_nfs_open_context(pos);
1025 break;
1026 }
1027 }
1028 spin_unlock(&inode->i_lock);
1029 return ctx;
1030}
1031
1032void nfs_file_clear_open_context(struct file *filp)
1033{
1034 struct inode *inode = filp->f_dentry->d_inode;
1035 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1036
1037 if (ctx) {
1038 filp->private_data = NULL;
1039 spin_lock(&inode->i_lock);
1040 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1041 spin_unlock(&inode->i_lock);
1042 put_nfs_open_context(ctx);
1043 }
1044}
1045
1046/*
1047 * These allocate and release file read/write context information.
1048 */
1049int nfs_open(struct inode *inode, struct file *filp)
1050{
1051 struct nfs_open_context *ctx;
1052 struct rpc_cred *cred;
1053
1054 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1055 if (IS_ERR(cred))
1056 return PTR_ERR(cred);
1057 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
1058 put_rpccred(cred);
1059 if (ctx == NULL)
1060 return -ENOMEM;
1061 ctx->mode = filp->f_mode;
1062 nfs_file_set_open_context(filp, ctx);
1063 put_nfs_open_context(ctx);
1da177e4
LT		 1064 return 0;
1065}
1066
1067int nfs_release(struct inode *inode, struct file *filp)
1068{
1da177e4
LT		 1069 nfs_file_clear_open_context(filp);
1070 return 0;
1071}
1072
1073/*
1074 * This function is called whenever some part of NFS notices that
1075 * the cached attributes have to be refreshed.
1076 */
1077int
1078__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1079{
1080 int status = -ESTALE;
1081 struct nfs_fattr fattr;
1082 struct nfs_inode *nfsi = NFS_I(inode);
1083 unsigned long verifier;
55296809 1084 unsigned long cache_validity;
1da177e4
LT		 1085
1086 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1087 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1088
1089 lock_kernel();
1090 if (!inode || is_bad_inode(inode))
1091 goto out_nowait;
1092 if (NFS_STALE(inode))
1093 goto out_nowait;
1094
412d582e
CL		 1095 status = nfs_wait_on_inode(inode);
1096 if (status < 0)
1097 goto out;
1098 if (NFS_STALE(inode)) {
1099 status = -ESTALE;
1100 /* Do we trust the cached ESTALE? */
1101 if (NFS_ATTRTIMEO(inode) != 0) {
1102 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1103 /* no */
1104 } else
1105 goto out;
1106 }
1da177e4 1107 }
1da177e4
LT		 1108
1109 /* Protect against RPC races by saving the change attribute */
1110 verifier = nfs_save_change_attribute(inode);
1111 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1112 if (status != 0) {
1113 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1114 inode->i_sb->s_id,
1115 (long long)NFS_FILEID(inode), status);
1116 if (status == -ESTALE) {
1117 nfs_zap_caches(inode);
1118 if (!S_ISDIR(inode->i_mode))
412d582e 1119 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1120 }
1121 goto out;
1122 }
1123
33801147 1124 spin_lock(&inode->i_lock);
1da177e4
LT		 1125 status = nfs_update_inode(inode, &fattr, verifier);
1126 if (status) {
33801147 1127 spin_unlock(&inode->i_lock);
1da177e4
LT		 1128 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1129 inode->i_sb->s_id,
1130 (long long)NFS_FILEID(inode), status);
1131 goto out;
1132 }
55296809
CL		 1133 cache_validity = nfsi->cache_validity;
1134 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1135
1da177e4
LT		 1136 /*
1137 * We may need to keep the attributes marked as invalid if
1138 * we raced with nfs_end_attr_update().
1139 */
913a70fc 1140 if (time_after_eq(verifier, nfsi->cache_change_attribute))
55296809 1141 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
dc59250c 1142 spin_unlock(&inode->i_lock);
55296809 1143
7d52e862 1144 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809
CL		 1145
1146 if (cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1147 nfs_zap_acl_cache(inode);
55296809 1148
1da177e4
LT		 1149 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1150 inode->i_sb->s_id,
1151 (long long)NFS_FILEID(inode));
1152
412d582e
CL		 1153 out:
1154 nfs_wake_up_inode(inode);
1155
1da177e4
LT		 1156 out_nowait:
1157 unlock_kernel();
1158 return status;
1159}
1160
1161int nfs_attribute_timeout(struct inode *inode)
1162{
1163 struct nfs_inode *nfsi = NFS_I(inode);
1164
1165 if (nfs_have_delegation(inode, FMODE_READ))
1166 return 0;
1167 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1168}
1169
1170/**
1171 * nfs_revalidate_inode - Revalidate the inode attributes
1172 * @server - pointer to nfs_server struct
1173 * @inode - pointer to inode struct
1174 *
1175 * Updates inode attribute information by retrieving the data from the server.
1176 */
1177int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1178{
55296809 1179 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1180 && !nfs_attribute_timeout(inode))
1181 return NFS_STALE(inode) ? -ESTALE : 0;
1182 return __nfs_revalidate_inode(server, inode);
1183}
1184
7d52e862
TM		 1185/**
1186 * nfs_revalidate_mapping - Revalidate the pagecache
1187 * @inode - pointer to host inode
1188 * @mapping - pointer to mapping
1189 */
1190void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1191{
1192 struct nfs_inode *nfsi = NFS_I(inode);
1193
55296809 1194 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
7d52e862
TM		 1195 if (S_ISREG(inode->i_mode)) {
1196 if (filemap_fdatawrite(mapping) == 0)
1197 filemap_fdatawait(mapping);
1198 nfs_wb_all(inode);
1199 }
1200 invalidate_inode_pages2(mapping);
dc59250c
CL		 1201
1202 spin_lock(&inode->i_lock);
55296809 1203 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1204 if (S_ISDIR(inode->i_mode)) {
1205 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1206 /* This ensures we revalidate child dentries */
913a70fc 1207 nfsi->cache_change_attribute = jiffies;
7d52e862 1208 }
dc59250c
CL		 1209 spin_unlock(&inode->i_lock);
1210
7d52e862
TM		 1211 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1212 inode->i_sb->s_id,
1213 (long long)NFS_FILEID(inode));
1214 }
1215}
1216
1da177e4
LT		 1217/**
1218 * nfs_begin_data_update
1219 * @inode - pointer to inode
1220 * Declare that a set of operations will update file data on the server
1221 */
1222void nfs_begin_data_update(struct inode *inode)
1223{
1224 atomic_inc(&NFS_I(inode)->data_updates);
1225}
1226
1227/**
1228 * nfs_end_data_update
1229 * @inode - pointer to inode
1230 * Declare end of the operations that will update file data
1231 * This will mark the inode as immediately needing revalidation
1232 * of its attribute cache.
1233 */
1234void nfs_end_data_update(struct inode *inode)
1235{
1236 struct nfs_inode *nfsi = NFS_I(inode);
1237
1238 if (!nfs_have_delegation(inode, FMODE_READ)) {
decf491f
TM		 1239 /* Directories and symlinks: invalidate page cache */
1240 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1241 spin_lock(&inode->i_lock);
55296809 1242 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
decf491f
TM		 1243 spin_unlock(&inode->i_lock);
1244 }
1da177e4 1245 }
913a70fc 1246 nfsi->cache_change_attribute = jiffies;
1da177e4
LT		 1247 atomic_dec(&nfsi->data_updates);
1248}
1249
1da177e4 1250/**
33801147 1251 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1da177e4
LT		 1252 * @inode - pointer to inode
1253 * @fattr - updated attributes
1254 *
1255 * Verifies the attribute cache. If we have just changed the attributes,
1256 * so that fattr carries weak cache consistency data, then it may
1257 * also update the ctime/mtime/change_attribute.
1258 */
33801147 1259static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1da177e4
LT		 1260{
1261 struct nfs_inode *nfsi = NFS_I(inode);
1262 loff_t cur_size, new_isize;
1263 int data_unstable;
1264
dc59250c 1265
1da177e4
LT		 1266 /* Are we in the process of updating data on the server? */
1267 data_unstable = nfs_caches_unstable(inode);
1268
1269 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1270 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1271 && nfsi->change_attr == fattr->pre_change_attr)
1272 nfsi->change_attr = fattr->change_attr;
fe51beec 1273 if (nfsi->change_attr != fattr->change_attr) {
55296809 1274 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1275 if (!data_unstable)
55296809 1276 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1277 }
1da177e4
LT		 1278 }
1279
dc59250c 1280 if ((fattr->valid & NFS_ATTR_FATTR) == 0) {
1da177e4 1281 return 0;
dc59250c 1282 }
1da177e4
LT		 1283
1284 /* Has the inode gone and changed behind our back? */
1285 if (nfsi->fileid != fattr->fileid
dc59250c 1286 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1da177e4 1287 return -EIO;
dc59250c 1288 }
1da177e4
LT		 1289
1290 cur_size = i_size_read(inode);
1291 new_isize = nfs_size_to_loff_t(fattr->size);
1292
1293 /* If we have atomic WCC data, we may update some attributes */
1294 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1295 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1296 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1297 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1298 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1299 }
1300
1301 /* Verify a few of the more important attributes */
fe51beec 1302 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1303 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1304 if (!data_unstable)
55296809 1305 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec
TM		 1306 }
1307 if (cur_size != new_isize) {
55296809 1308 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1309 if (nfsi->npages == 0)
55296809 1310 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1311 }
1da177e4
LT		 1312
1313 /* Have any file permissions changed? */
1314 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1315 || inode->i_uid != fattr->uid
1316 || inode->i_gid != fattr->gid)
55296809 1317 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1318
1319 /* Has the link count changed? */
1320 if (inode->i_nlink != fattr->nlink)
55296809 1321 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1322
1323 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1324 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4 1325
33801147 1326 nfsi->read_cache_jiffies = fattr->time_start;
1da177e4
LT		 1327 return 0;
1328}
1329
33801147
TM		 1330/**
1331 * nfs_refresh_inode - try to update the inode attribute cache
1332 * @inode - pointer to inode
1333 * @fattr - updated attributes
1334 *
1335 * Check that an RPC call that returned attributes has not overlapped with
1336 * other recent updates of the inode metadata, then decide whether it is
1337 * safe to do a full update of the inode attributes, or whether just to
1338 * call nfs_check_inode_attributes.
1339 */
1340int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1341{
1342 struct nfs_inode *nfsi = NFS_I(inode);
1343 int status;
1344
1345 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1346 return 0;
1347 spin_lock(&inode->i_lock);
1348 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1349 if (nfs_verify_change_attribute(inode, fattr->time_start))
1350 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1351 if (time_after(fattr->time_start, nfsi->last_updated))
1352 status = nfs_update_inode(inode, fattr, fattr->time_start);
1353 else
1354 status = nfs_check_inode_attributes(inode, fattr);
1355
1356 spin_unlock(&inode->i_lock);
1357 return status;
1358}
1359
decf491f
TM		 1360/**
1361 * nfs_post_op_update_inode - try to update the inode attribute cache
1362 * @inode - pointer to inode
1363 * @fattr - updated attributes
1364 *
1365 * After an operation that has changed the inode metadata, mark the
1366 * attribute cache as being invalid, then try to update it.
1367 */
1368int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1369{
1370 struct nfs_inode *nfsi = NFS_I(inode);
1371 int status = 0;
1372
1373 spin_lock(&inode->i_lock);
1374 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1375 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1376 goto out;
1377 }
1378 status = nfs_update_inode(inode, fattr, fattr->time_start);
1379 if (time_after_eq(fattr->time_start, nfsi->cache_change_attribute))
1380 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1381 nfsi->cache_change_attribute = jiffies;
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 */
1399static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1400{
1401 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1402 loff_t cur_isize, new_isize;
1da177e4 1403 unsigned int invalid = 0;
1da177e4
LT		 1404 int data_unstable;
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? */
b3c52da3
TM		 1435 data_unstable = ! (nfs_verify_change_attribute(inode, verifier) ||
1436 (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE));
1da177e4 1437
951a143b 1438 /* Check if our cached file size is stale */
1da177e4
LT		 1439 new_isize = nfs_size_to_loff_t(fattr->size);
1440 cur_isize = i_size_read(inode);
951a143b
TM		 1441 if (new_isize != cur_isize) {
1442 /* Do we perhaps have any outstanding writes? */
1443 if (nfsi->npages == 0) {
1444 /* No, but did we race with nfs_end_data_update()? */
913a70fc 1445 if (time_after_eq(verifier, nfsi->cache_change_attribute)) {
1da177e4 1446 inode->i_size = new_isize;
951a143b 1447 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1448 }
951a143b
TM		 1449 invalid |= NFS_INO_INVALID_ATTR;
1450 } else if (new_isize > cur_isize) {
1da177e4
LT		 1451 inode->i_size = new_isize;
1452 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1453 }
951a143b
TM		 1454 dprintk("NFS: isize change on server for file %s/%ld\n",
1455 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1456 }
1457
951a143b 1458 /* Check if the mtime agrees */
1da177e4
LT		 1459 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1460 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1461 dprintk("NFS: mtime change on server for file %s/%ld\n",
1462 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1463 if (!data_unstable)
1464 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1465 }
1466
1467 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1468 && nfsi->change_attr != fattr->change_attr) {
951a143b 1469 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1da177e4 1470 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1471 nfsi->change_attr = fattr->change_attr;
1472 if (!data_unstable)
ada70d94 1473 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1474 }
1475
ada70d94
TM		 1476 /* If ctime has changed we should definitely clear access+acl caches */
1477 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1478 if (!data_unstable)
1479 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1480 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1481 }
1da177e4
LT		 1482 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1483
1484 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1485 inode->i_uid != fattr->uid ||
1486 inode->i_gid != fattr->gid)
ada70d94 1487 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1488
1489 inode->i_mode = fattr->mode;
1490 inode->i_nlink = fattr->nlink;
1491 inode->i_uid = fattr->uid;
1492 inode->i_gid = fattr->gid;
1493
1494 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1495 /*
1496 * report the blocks in 512byte units
1497 */
1498 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1499 inode->i_blksize = inode->i_sb->s_blocksize;
1500 } else {
1501 inode->i_blocks = fattr->du.nfs2.blocks;
1502 inode->i_blksize = fattr->du.nfs2.blocksize;
1503 }
1504
1505 /* Update attrtimeo value if we're out of the unstable period */
1506 if (invalid & NFS_INO_INVALID_ATTR) {
1507 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1508 nfsi->attrtimeo_timestamp = jiffies;
1509 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1510 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1511 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1512 nfsi->attrtimeo_timestamp = jiffies;
1513 }
1514 /* Don't invalidate the data if we were to blame */
1515 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1516 || S_ISLNK(inode->i_mode)))
1517 invalid &= ~NFS_INO_INVALID_DATA;
1518 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1519 nfsi->cache_validity |= invalid;
1da177e4
LT		 1520
1521 return 0;
1522 out_changed:
1523 /*
1524 * Big trouble! The inode has become a different object.
1525 */
1526#ifdef NFS_PARANOIA
1527 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1528 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1529#endif
b37b03b7 1530 out_err:
1da177e4
LT		 1531 /*
1532 * No need to worry about unhashing the dentry, as the
1533 * lookup validation will know that the inode is bad.
1534 * (But we fall through to invalidate the caches.)
1535 */
1536 nfs_invalidate_inode(inode);
1da177e4
LT		 1537 return -ESTALE;
1538}
1539
1540/*
1541 * File system information
1542 */
1543
1544static int nfs_set_super(struct super_block *s, void *data)
1545{
1546 s->s_fs_info = data;
1547 return set_anon_super(s, data);
1548}
1549
1550static int nfs_compare_super(struct super_block *sb, void *data)
1551{
1552 struct nfs_server *server = data;
1553 struct nfs_server *old = NFS_SB(sb);
1554
1555 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1556 return 0;
1557 if (old->addr.sin_port != server->addr.sin_port)
1558 return 0;
1559 return !nfs_compare_fh(&old->fh, &server->fh);
1560}
1561
1562static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1563 int flags, const char *dev_name, void *raw_data)
1564{
1565 int error;
9085bbcb 1566 struct nfs_server *server = NULL;
1da177e4
LT		 1567 struct super_block *s;
1568 struct nfs_fh *root;
1569 struct nfs_mount_data *data = raw_data;
1570
9085bbcb
TM		 1571 s = ERR_PTR(-EINVAL);
1572 if (data == NULL) {
1573 dprintk("%s: missing data argument\n", __FUNCTION__);
1574 goto out_err;
1575 }
1576 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1577 dprintk("%s: bad mount version\n", __FUNCTION__);
1578 goto out_err;
1da177e4 1579 }
9085bbcb
TM		 1580 switch (data->version) {
1581 case 1:
1582 data->namlen = 0;
1583 case 2:
1584 data->bsize = 0;
1585 case 3:
1586 if (data->flags & NFS_MOUNT_VER3) {
1587 dprintk("%s: mount structure version %d does not support NFSv3\n",
1588 __FUNCTION__,
1589 data->version);
1590 goto out_err;
1591 }
1592 data->root.size = NFS2_FHSIZE;
1593 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1594 case 4:
1595 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1596 dprintk("%s: mount structure version %d does not support strong security\n",
1597 __FUNCTION__,
1598 data->version);
1599 goto out_err;
1600 }
1601 case 5:
1602 memset(data->context, 0, sizeof(data->context));
1603 }
1604#ifndef CONFIG_NFS_V3
1605 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1606 s = ERR_PTR(-EPROTONOSUPPORT);
1607 if (data->flags & NFS_MOUNT_VER3) {
1608 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1609 goto out_err;
1610 }
1611#endif /* CONFIG_NFS_V3 */
1da177e4 1612
9085bbcb 1613 s = ERR_PTR(-ENOMEM);
1da177e4
LT		 1614 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1615 if (!server)
9085bbcb 1616 goto out_err;
1da177e4
LT		 1617 memset(server, 0, sizeof(struct nfs_server));
1618 /* Zero out the NFS state stuff */
1619 init_nfsv4_state(server);
b7fa0554 1620 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1621
1da177e4
LT		 1622 root = &server->fh;
1623 if (data->flags & NFS_MOUNT_VER3)
1624 root->size = data->root.size;
1625 else
1626 root->size = NFS2_FHSIZE;
9085bbcb 1627 s = ERR_PTR(-EINVAL);
1da177e4 1628 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1629 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1630 goto out_err;
1da177e4
LT		 1631 }
1632 memcpy(root->data, data->root.data, root->size);
1633
1634 /* We now require that the mount process passes the remote address */
1635 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1636 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1637 dprintk("%s: mount program didn't pass remote address!\n",
1638 __FUNCTION__);
1639 goto out_err;
1da177e4
LT		 1640 }
1641
9085bbcb
TM		 1642 /* Fire up rpciod if not yet running */
1643 s = ERR_PTR(rpciod_up());
1644 if (IS_ERR(s)) {
1645 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1646 __FUNCTION__, PTR_ERR(s));
1647 goto out_err;
1da177e4
LT		 1648 }
1649
9085bbcb
TM		 1650 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1651 if (IS_ERR(s) || s->s_root)
1652 goto out_rpciod_down;
1da177e4 1653
9085bbcb 1654 s->s_flags = flags;
1da177e4
LT		 1655
1656 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1657 if (error) {
1658 up_write(&s->s_umount);
1659 deactivate_super(s);
1660 return ERR_PTR(error);
1661 }
1662 s->s_flags |= MS_ACTIVE;
1663 return s;
9085bbcb
TM		 1664out_rpciod_down:
1665 rpciod_down();
1666out_err:
1667 kfree(server);
1668 return s;
1da177e4
LT		 1669}
1670
1671static void nfs_kill_super(struct super_block *s)
1672{
1673 struct nfs_server *server = NFS_SB(s);
1674
1675 kill_anon_super(s);
1676
6a19275a 1677 if (!IS_ERR(server->client))
1da177e4 1678 rpc_shutdown_client(server->client);
6a19275a 1679 if (!IS_ERR(server->client_sys))
1da177e4 1680 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1681 if (!IS_ERR(server->client_acl))
1682 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1683
1684 if (!(server->flags & NFS_MOUNT_NONLM))
1685 lockd_down(); /* release rpc.lockd */
1686
1687 rpciod_down(); /* release rpciod */
1688
f99d49ad 1689 kfree(server->hostname);
1da177e4
LT		 1690 kfree(server);
1691}
1692
1693static struct file_system_type nfs_fs_type = {
1694 .owner = THIS_MODULE,
1695 .name = "nfs",
1696 .get_sb = nfs_get_sb,
1697 .kill_sb = nfs_kill_super,
1698 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1699};
1700
1701#ifdef CONFIG_NFS_V4
1702
1703static void nfs4_clear_inode(struct inode *);
1704
1705
1706static struct super_operations nfs4_sops = {
1707 .alloc_inode = nfs_alloc_inode,
1708 .destroy_inode = nfs_destroy_inode,
1709 .write_inode = nfs_write_inode,
1710 .delete_inode = nfs_delete_inode,
1711 .statfs = nfs_statfs,
1712 .clear_inode = nfs4_clear_inode,
1713 .umount_begin = nfs_umount_begin,
1714 .show_options = nfs_show_options,
1715};
1716
1717/*
1718 * Clean out any remaining NFSv4 state that might be left over due
1719 * to open() calls that passed nfs_atomic_lookup, but failed to call
1720 * nfs_open().
1721 */
1722static void nfs4_clear_inode(struct inode *inode)
1723{
1724 struct nfs_inode *nfsi = NFS_I(inode);
1725
1726 /* If we are holding a delegation, return it! */
cae7a073 1727 nfs_inode_return_delegation(inode);
1da177e4
LT		 1728 /* First call standard NFS clear_inode() code */
1729 nfs_clear_inode(inode);
1730 /* Now clear out any remaining state */
1731 while (!list_empty(&nfsi->open_states)) {
1732 struct nfs4_state *state;
1733
1734 state = list_entry(nfsi->open_states.next,
1735 struct nfs4_state,
1736 inode_states);
1737 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1738 __FUNCTION__,
1739 inode->i_sb->s_id,
1740 (long long)NFS_FILEID(inode),
1741 state);
1742 BUG_ON(atomic_read(&state->count) != 1);
1743 nfs4_close_state(state, state->state);
1744 }
1745}
1746
1747
1748static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1749{
1750 struct nfs_server *server;
1751 struct nfs4_client *clp = NULL;
1752 struct rpc_xprt *xprt = NULL;
1753 struct rpc_clnt *clnt = NULL;
1754 struct rpc_timeout timeparms;
1755 rpc_authflavor_t authflavour;
eab5c084 1756 int err = -EIO;
1da177e4
LT		 1757
1758 sb->s_blocksize_bits = 0;
1759 sb->s_blocksize = 0;
1760 server = NFS_SB(sb);
1761 if (data->rsize != 0)
1762 server->rsize = nfs_block_size(data->rsize, NULL);
1763 if (data->wsize != 0)
1764 server->wsize = nfs_block_size(data->wsize, NULL);
1765 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1766 server->caps = NFS_CAP_ATOMIC_OPEN;
1767
1768 server->acregmin = data->acregmin*HZ;
1769 server->acregmax = data->acregmax*HZ;
1770 server->acdirmin = data->acdirmin*HZ;
1771 server->acdirmax = data->acdirmax*HZ;
1772
1773 server->rpc_ops = &nfs_v4_clientops;
1da177e4 1774
eab5c084 1775 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1da177e4
LT		 1776
1777 clp = nfs4_get_client(&server->addr.sin_addr);
1778 if (!clp) {
9085bbcb 1779 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1780 return -EIO;
1781 }
1782
1783 /* Now create transport and client */
1784 authflavour = RPC_AUTH_UNIX;
1785 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1786 if (data->auth_flavourlen != 1) {
1787 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1788 __FUNCTION__, data->auth_flavourlen);
1789 err = -EINVAL;
1790 goto out_fail;
1791 }
1da177e4
LT		 1792 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1793 err = -EFAULT;
1794 goto out_fail;
1795 }
1796 }
1797
1798 down_write(&clp->cl_sem);
6a19275a 1799 if (IS_ERR(clp->cl_rpcclient)) {
eab5c084 1800 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1da177e4
LT		 1801 if (IS_ERR(xprt)) {
1802 up_write(&clp->cl_sem);
1da177e4 1803 err = PTR_ERR(xprt);
9085bbcb
TM		 1804 dprintk("%s: cannot create RPC transport. Error = %d\n",
1805 __FUNCTION__, err);
1da177e4
LT		 1806 goto out_fail;
1807 }
1808 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1809 server->rpc_ops->version, authflavour);
1810 if (IS_ERR(clnt)) {
1811 up_write(&clp->cl_sem);
1da177e4 1812 err = PTR_ERR(clnt);
9085bbcb
TM		 1813 dprintk("%s: cannot create RPC client. Error = %d\n",
1814 __FUNCTION__, err);
1da177e4
LT		 1815 goto out_fail;
1816 }
1817 clnt->cl_intr = 1;
1818 clnt->cl_softrtry = 1;
1819 clnt->cl_chatty = 1;
1820 clp->cl_rpcclient = clnt;
1821 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1822 if (IS_ERR(clp->cl_cred)) {
1823 up_write(&clp->cl_sem);
1824 err = PTR_ERR(clp->cl_cred);
1825 clp->cl_cred = NULL;
1826 goto out_fail;
1827 }
1828 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1829 nfs_idmap_new(clp);
1830 }
1831 if (list_empty(&clp->cl_superblocks)) {
1832 err = nfs4_init_client(clp);
1833 if (err != 0) {
1834 up_write(&clp->cl_sem);
1835 goto out_fail;
1836 }
1837 }
1838 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1839 clnt = rpc_clone_client(clp->cl_rpcclient);
1840 if (!IS_ERR(clnt))
1841 server->nfs4_state = clp;
1842 up_write(&clp->cl_sem);
1843 clp = NULL;
1844
1845 if (IS_ERR(clnt)) {
9085bbcb
TM		 1846 err = PTR_ERR(clnt);
1847 dprintk("%s: cannot create RPC client. Error = %d\n",
1848 __FUNCTION__, err);
1849 return err;
1da177e4
LT		 1850 }
1851
1852 server->client = clnt;
1853
1854 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1855 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1856 return -ENOMEM;
1857 }
1858
1859 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1860 struct rpc_auth *auth;
1861
1862 auth = rpcauth_create(authflavour, clnt);
1863 if (IS_ERR(auth)) {
9085bbcb 1864 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1865 return PTR_ERR(auth);
1da177e4
LT		 1866 }
1867 }
1868
1869 sb->s_time_gran = 1;
1870
1871 sb->s_op = &nfs4_sops;
1872 err = nfs_sb_init(sb, authflavour);
1873 if (err == 0)
1874 return 0;
1875out_fail:
1876 if (clp)
1877 nfs4_put_client(clp);
1878 return err;
1879}
1880
1881static int nfs4_compare_super(struct super_block *sb, void *data)
1882{
1883 struct nfs_server *server = data;
1884 struct nfs_server *old = NFS_SB(sb);
1885
1886 if (strcmp(server->hostname, old->hostname) != 0)
1887 return 0;
1888 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1889 return 0;
1890 return 1;
1891}
1892
1893static void *
1894nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1895{
1896 void *p = NULL;
1897
1898 if (!src->len)
1899 return ERR_PTR(-EINVAL);
1900 if (src->len < maxlen)
1901 maxlen = src->len;
1902 if (dst == NULL) {
1903 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1904 if (p == NULL)
1905 return ERR_PTR(-ENOMEM);
1906 }
1907 if (copy_from_user(dst, src->data, maxlen)) {
f99d49ad 1908 kfree(p);
1da177e4
LT		 1909 return ERR_PTR(-EFAULT);
1910 }
1911 dst[maxlen] = '\0';
1912 return dst;
1913}
1914
1915static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1916 int flags, const char *dev_name, void *raw_data)
1917{
1918 int error;
1919 struct nfs_server *server;
1920 struct super_block *s;
1921 struct nfs4_mount_data *data = raw_data;
1922 void *p;
1923
9085bbcb
TM		 1924 if (data == NULL) {
1925 dprintk("%s: missing data argument\n", __FUNCTION__);
1926 return ERR_PTR(-EINVAL);
1927 }
1928 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1929 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1930 return ERR_PTR(-EINVAL);
1931 }
1932
1933 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1934 if (!server)
1935 return ERR_PTR(-ENOMEM);
1936 memset(server, 0, sizeof(struct nfs_server));
1937 /* Zero out the NFS state stuff */
1938 init_nfsv4_state(server);
b7fa0554 1939 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1940
1da177e4
LT		 1941 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1942 if (IS_ERR(p))
1943 goto out_err;
1944 server->hostname = p;
1945
1946 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1947 if (IS_ERR(p))
1948 goto out_err;
1949 server->mnt_path = p;
1950
1951 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1952 sizeof(server->ip_addr) - 1);
1953 if (IS_ERR(p))
1954 goto out_err;
1955
1956 /* We now require that the mount process passes the remote address */
1957 if (data->host_addrlen != sizeof(server->addr)) {
1958 s = ERR_PTR(-EINVAL);
1959 goto out_free;
1960 }
1961 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1962 s = ERR_PTR(-EFAULT);
1963 goto out_free;
1964 }
1965 if (server->addr.sin_family != AF_INET ||
1966 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1967 dprintk("%s: mount program didn't pass remote IP address!\n",
1968 __FUNCTION__);
1da177e4
LT		 1969 s = ERR_PTR(-EINVAL);
1970 goto out_free;
1971 }
1972
9085bbcb
TM		 1973 /* Fire up rpciod if not yet running */
1974 s = ERR_PTR(rpciod_up());
1975 if (IS_ERR(s)) {
1976 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1977 __FUNCTION__, PTR_ERR(s));
1978 goto out_free;
1979 }
1980
1da177e4
LT		 1981 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1982
1983 if (IS_ERR(s) || s->s_root)
1984 goto out_free;
1985
1986 s->s_flags = flags;
1987
1da177e4
LT		 1988 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1989 if (error) {
1990 up_write(&s->s_umount);
1991 deactivate_super(s);
1992 return ERR_PTR(error);
1993 }
1994 s->s_flags |= MS_ACTIVE;
1995 return s;
1996out_err:
1997 s = (struct super_block *)p;
1998out_free:
f99d49ad
JJ		 1999 kfree(server->mnt_path);
2000 kfree(server->hostname);
1da177e4
LT		 2001 kfree(server);
2002 return s;
2003}
2004
2005static void nfs4_kill_super(struct super_block *sb)
2006{
2007 struct nfs_server *server = NFS_SB(sb);
2008
2009 nfs_return_all_delegations(sb);
2010 kill_anon_super(sb);
2011
2012 nfs4_renewd_prepare_shutdown(server);
2013
2014 if (server->client != NULL && !IS_ERR(server->client))
2015 rpc_shutdown_client(server->client);
2016 rpciod_down(); /* release rpciod */
2017
2018 destroy_nfsv4_state(server);
2019
f99d49ad 2020 kfree(server->hostname);
1da177e4
LT		 2021 kfree(server);
2022}
2023
2024static struct file_system_type nfs4_fs_type = {
2025 .owner = THIS_MODULE,
2026 .name = "nfs4",
2027 .get_sb = nfs4_get_sb,
2028 .kill_sb = nfs4_kill_super,
2029 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2030};
2031
2032#define nfs4_init_once(nfsi) \
2033 do { \
2034 INIT_LIST_HEAD(&(nfsi)->open_states); \
2035 nfsi->delegation = NULL; \
2036 nfsi->delegation_state = 0; \
2037 init_rwsem(&nfsi->rwsem); \
2038 } while(0)
2039#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
2040#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
2041#else
2042#define nfs4_init_once(nfsi) \
2043 do { } while (0)
2044#define register_nfs4fs() (0)
2045#define unregister_nfs4fs()
2046#endif
2047
2048extern int nfs_init_nfspagecache(void);
2049extern void nfs_destroy_nfspagecache(void);
2050extern int nfs_init_readpagecache(void);
2051extern void nfs_destroy_readpagecache(void);
2052extern int nfs_init_writepagecache(void);
2053extern void nfs_destroy_writepagecache(void);
2054#ifdef CONFIG_NFS_DIRECTIO
2055extern int nfs_init_directcache(void);
2056extern void nfs_destroy_directcache(void);
2057#endif
2058
2059static kmem_cache_t * nfs_inode_cachep;
2060
2061static struct inode *nfs_alloc_inode(struct super_block *sb)
2062{
2063 struct nfs_inode *nfsi;
2064 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2065 if (!nfsi)
2066 return NULL;
55296809
CL		 2067 nfsi->flags = 0UL;
2068 nfsi->cache_validity = 0UL;
223db122 2069 nfsi->cache_change_attribute = jiffies;
458818ed
TM		 2070#ifdef CONFIG_NFS_V3_ACL
2071 nfsi->acl_access = ERR_PTR(-EAGAIN);
2072 nfsi->acl_default = ERR_PTR(-EAGAIN);
2073#endif
e50a1c2e
BF		 2074#ifdef CONFIG_NFS_V4
2075 nfsi->nfs4_acl = NULL;
2076#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2077 return &nfsi->vfs_inode;
2078}
2079
2080static void nfs_destroy_inode(struct inode *inode)
2081{
2082 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2083}
2084
2085static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2086{
2087 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2088
2089 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2090 SLAB_CTOR_CONSTRUCTOR) {
2091 inode_init_once(&nfsi->vfs_inode);
2092 spin_lock_init(&nfsi->req_lock);
2093 INIT_LIST_HEAD(&nfsi->dirty);
2094 INIT_LIST_HEAD(&nfsi->commit);
2095 INIT_LIST_HEAD(&nfsi->open_files);
2096 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2097 atomic_set(&nfsi->data_updates, 0);
2098 nfsi->ndirty = 0;
2099 nfsi->ncommit = 0;
2100 nfsi->npages = 0;
1da177e4
LT		 2101 nfs4_init_once(nfsi);
2102 }
2103}
2104
75c96f85 2105static int nfs_init_inodecache(void)
1da177e4
LT		 2106{
2107 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2108 sizeof(struct nfs_inode),
2109 0, SLAB_RECLAIM_ACCOUNT,
2110 init_once, NULL);
2111 if (nfs_inode_cachep == NULL)
2112 return -ENOMEM;
2113
2114 return 0;
2115}
2116
75c96f85 2117static void nfs_destroy_inodecache(void)
1da177e4
LT		 2118{
2119 if (kmem_cache_destroy(nfs_inode_cachep))
2120 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2121}
2122
2123/*
2124 * Initialize NFS
2125 */
2126static int __init init_nfs_fs(void)
2127{
2128 int err;
2129
2130 err = nfs_init_nfspagecache();
2131 if (err)
2132 goto out4;
2133
2134 err = nfs_init_inodecache();
2135 if (err)
2136 goto out3;
2137
2138 err = nfs_init_readpagecache();
2139 if (err)
2140 goto out2;
2141
2142 err = nfs_init_writepagecache();
2143 if (err)
2144 goto out1;
2145
2146#ifdef CONFIG_NFS_DIRECTIO
2147 err = nfs_init_directcache();
2148 if (err)
2149 goto out0;
2150#endif
2151
2152#ifdef CONFIG_PROC_FS
2153 rpc_proc_register(&nfs_rpcstat);
2154#endif
2155 err = register_filesystem(&nfs_fs_type);
2156 if (err)
2157 goto out;
2158 if ((err = register_nfs4fs()) != 0)
2159 goto out;
2160 return 0;
2161out:
2162#ifdef CONFIG_PROC_FS
2163 rpc_proc_unregister("nfs");
2164#endif
2165 nfs_destroy_writepagecache();
2166#ifdef CONFIG_NFS_DIRECTIO
2167out0:
2168 nfs_destroy_directcache();
2169#endif
2170out1:
2171 nfs_destroy_readpagecache();
2172out2:
2173 nfs_destroy_inodecache();
2174out3:
2175 nfs_destroy_nfspagecache();
2176out4:
2177 return err;
2178}
2179
2180static void __exit exit_nfs_fs(void)
2181{
2182#ifdef CONFIG_NFS_DIRECTIO
2183 nfs_destroy_directcache();
2184#endif
2185 nfs_destroy_writepagecache();
2186 nfs_destroy_readpagecache();
2187 nfs_destroy_inodecache();
2188 nfs_destroy_nfspagecache();
2189#ifdef CONFIG_PROC_FS
2190 rpc_proc_unregister("nfs");
2191#endif
2192 unregister_filesystem(&nfs_fs_type);
2193 unregister_nfs4fs();
2194}
2195
2196/* Not quite true; I just maintain it */
2197MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2198MODULE_LICENSE("GPL");
2199
2200module_init(init_nfs_fs)
2201module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)