4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/time.h>
20 #include <linux/kernel.h>
22 #include <linux/string.h>
23 #include <linux/stat.h>
24 #include <linux/errno.h>
25 #include <linux/unistd.h>
26 #include <linux/sunrpc/clnt.h>
27 #include <linux/sunrpc/stats.h>
28 #include <linux/sunrpc/metrics.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/seq_file.h>
34 #include <linux/mount.h>
35 #include <linux/nfs_idmap.h>
36 #include <linux/vfs.h>
37 #include <linux/inet.h>
38 #include <linux/nfs_xdr.h>
39 #include <linux/slab.h>
40 #include <linux/compat.h>
41 #include <linux/freezer.h>
42 #include <linux/crc32.h>
44 #include <asm/system.h>
45 #include <asm/uaccess.h>
49 #include "delegation.h"
53 #include "dns_resolve.h"
57 #define NFSDBG_FACILITY NFSDBG_VFS
59 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
61 /* Default is to see 64-bit inode numbers */
62 static bool enable_ino64
= NFS_64_BIT_INODE_NUMBERS_ENABLED
;
64 static void nfs_invalidate_inode(struct inode
*);
65 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
67 static struct kmem_cache
* nfs_inode_cachep
;
69 static inline unsigned long
70 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
72 return nfs_fileid_to_ino_t(fattr
->fileid
);
76 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
77 * @word: long word containing the bit lock
79 int nfs_wait_bit_killable(void *word
)
81 if (fatal_signal_pending(current
))
88 * nfs_compat_user_ino64 - returns the user-visible inode number
89 * @fileid: 64-bit fileid
91 * This function returns a 32-bit inode number if the boot parameter
92 * nfs.enable_ino64 is zero.
94 u64
nfs_compat_user_ino64(u64 fileid
)
105 if (sizeof(ino
) < sizeof(fileid
))
106 ino
^= fileid
>> (sizeof(fileid
)-sizeof(ino
)) * 8;
110 static void nfs_clear_inode(struct inode
*inode
)
113 * The following should never happen...
115 BUG_ON(nfs_have_writebacks(inode
));
116 BUG_ON(!list_empty(&NFS_I(inode
)->open_files
));
117 nfs_zap_acl_cache(inode
);
118 nfs_access_zap_cache(inode
);
119 nfs_fscache_release_inode_cookie(inode
);
122 void nfs_evict_inode(struct inode
*inode
)
124 truncate_inode_pages(&inode
->i_data
, 0);
125 end_writeback(inode
);
126 nfs_clear_inode(inode
);
130 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
132 int nfs_sync_mapping(struct address_space
*mapping
)
136 if (mapping
->nrpages
!= 0) {
137 unmap_mapping_range(mapping
, 0, 0, 0);
138 ret
= nfs_wb_all(mapping
->host
);
144 * Invalidate the local caches
146 static void nfs_zap_caches_locked(struct inode
*inode
)
148 struct nfs_inode
*nfsi
= NFS_I(inode
);
149 int mode
= inode
->i_mode
;
151 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
153 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
154 nfsi
->attrtimeo_timestamp
= jiffies
;
156 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
157 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
158 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
160 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
163 void nfs_zap_caches(struct inode
*inode
)
165 spin_lock(&inode
->i_lock
);
166 nfs_zap_caches_locked(inode
);
167 spin_unlock(&inode
->i_lock
);
170 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
172 if (mapping
->nrpages
!= 0) {
173 spin_lock(&inode
->i_lock
);
174 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
175 spin_unlock(&inode
->i_lock
);
179 void nfs_zap_acl_cache(struct inode
*inode
)
181 void (*clear_acl_cache
)(struct inode
*);
183 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
184 if (clear_acl_cache
!= NULL
)
185 clear_acl_cache(inode
);
186 spin_lock(&inode
->i_lock
);
187 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
188 spin_unlock(&inode
->i_lock
);
191 void nfs_invalidate_atime(struct inode
*inode
)
193 spin_lock(&inode
->i_lock
);
194 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
195 spin_unlock(&inode
->i_lock
);
199 * Invalidate, but do not unhash, the inode.
200 * NB: must be called with inode->i_lock held!
202 static void nfs_invalidate_inode(struct inode
*inode
)
204 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
205 nfs_zap_caches_locked(inode
);
208 struct nfs_find_desc
{
210 struct nfs_fattr
*fattr
;
214 * In NFSv3 we can have 64bit inode numbers. In order to support
215 * this, and re-exported directories (also seen in NFSv2)
216 * we are forced to allow 2 different inodes to have the same
220 nfs_find_actor(struct inode
*inode
, void *opaque
)
222 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
223 struct nfs_fh
*fh
= desc
->fh
;
224 struct nfs_fattr
*fattr
= desc
->fattr
;
226 if (NFS_FILEID(inode
) != fattr
->fileid
)
228 if (nfs_compare_fh(NFS_FH(inode
), fh
))
230 if (is_bad_inode(inode
) || NFS_STALE(inode
))
236 nfs_init_locked(struct inode
*inode
, void *opaque
)
238 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
239 struct nfs_fattr
*fattr
= desc
->fattr
;
241 set_nfs_fileid(inode
, fattr
->fileid
);
242 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
247 * This is our front-end to iget that looks up inodes by file handle
248 * instead of inode number.
251 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
253 struct nfs_find_desc desc
= {
257 struct inode
*inode
= ERR_PTR(-ENOENT
);
260 nfs_attr_check_mountpoint(sb
, fattr
);
262 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
263 !nfs_attr_use_mounted_on_fileid(fattr
))
265 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
268 hash
= nfs_fattr_to_ino_t(fattr
);
270 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
272 inode
= ERR_PTR(-ENOMEM
);
276 if (inode
->i_state
& I_NEW
) {
277 struct nfs_inode
*nfsi
= NFS_I(inode
);
278 unsigned long now
= jiffies
;
280 /* We set i_ino for the few things that still rely on it,
284 /* We can't support update_atime(), since the server will reset it */
285 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
286 inode
->i_mode
= fattr
->mode
;
287 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
288 && nfs_server_capable(inode
, NFS_CAP_MODE
))
289 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
290 | NFS_INO_INVALID_ACCESS
291 | NFS_INO_INVALID_ACL
;
292 /* Why so? Because we want revalidate for devices/FIFOs, and
293 * that's precisely what we have in nfs_file_inode_operations.
295 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
296 if (S_ISREG(inode
->i_mode
)) {
297 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
298 inode
->i_data
.a_ops
= &nfs_file_aops
;
299 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
300 } else if (S_ISDIR(inode
->i_mode
)) {
301 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
302 inode
->i_fop
= &nfs_dir_operations
;
303 inode
->i_data
.a_ops
= &nfs_dir_aops
;
304 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
))
305 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_I(inode
)->flags
);
306 /* Deal with crossing mountpoints */
307 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
308 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
309 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
310 inode
->i_op
= &nfs_referral_inode_operations
;
312 inode
->i_op
= &nfs_mountpoint_inode_operations
;
314 inode
->i_flags
|= S_AUTOMOUNT
;
316 } else if (S_ISLNK(inode
->i_mode
))
317 inode
->i_op
= &nfs_symlink_inode_operations
;
319 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
321 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
322 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
323 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
324 inode
->i_version
= 0;
330 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
332 nfsi
->read_cache_jiffies
= fattr
->time_start
;
333 nfsi
->attr_gencount
= fattr
->gencount
;
334 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
335 inode
->i_atime
= fattr
->atime
;
336 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
337 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
338 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
339 inode
->i_mtime
= fattr
->mtime
;
340 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
341 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
342 | NFS_INO_INVALID_DATA
;
343 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
344 inode
->i_ctime
= fattr
->ctime
;
345 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
346 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
347 | NFS_INO_INVALID_ACCESS
348 | NFS_INO_INVALID_ACL
;
349 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
350 inode
->i_version
= fattr
->change_attr
;
351 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
352 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
353 | NFS_INO_INVALID_DATA
;
354 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
355 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
357 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
358 | NFS_INO_INVALID_DATA
359 | NFS_INO_REVAL_PAGECACHE
;
360 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
361 set_nlink(inode
, fattr
->nlink
);
362 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
363 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
364 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
365 inode
->i_uid
= fattr
->uid
;
366 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
367 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
368 | NFS_INO_INVALID_ACCESS
369 | NFS_INO_INVALID_ACL
;
370 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
371 inode
->i_gid
= fattr
->gid
;
372 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
373 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
374 | NFS_INO_INVALID_ACCESS
375 | NFS_INO_INVALID_ACL
;
376 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
377 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
378 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
380 * report the blocks in 512byte units
382 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
384 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
385 nfsi
->attrtimeo_timestamp
= now
;
386 nfsi
->access_cache
= RB_ROOT
;
388 nfs_fscache_init_inode_cookie(inode
);
390 unlock_new_inode(inode
);
392 nfs_refresh_inode(inode
, fattr
);
393 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
395 (long long)NFS_FILEID(inode
),
396 nfs_display_fhandle_hash(fh
),
397 atomic_read(&inode
->i_count
));
403 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
407 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET|ATTR_FILE|ATTR_OPEN)
410 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
412 struct inode
*inode
= dentry
->d_inode
;
413 struct nfs_fattr
*fattr
;
416 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
418 /* skip mode change if it's just for clearing setuid/setgid */
419 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
420 attr
->ia_valid
&= ~ATTR_MODE
;
422 if (attr
->ia_valid
& ATTR_SIZE
) {
423 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
424 attr
->ia_valid
&= ~ATTR_SIZE
;
427 /* Optimization: if the end result is no change, don't RPC */
428 attr
->ia_valid
&= NFS_VALID_ATTRS
;
429 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
432 /* Write all dirty data */
433 if (S_ISREG(inode
->i_mode
))
436 fattr
= nfs_alloc_fattr();
440 * Return any delegations if we're going to change ACLs
442 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
443 nfs_inode_return_delegation(inode
);
444 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
446 nfs_refresh_inode(inode
, fattr
);
447 nfs_free_fattr(fattr
);
453 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
454 * @inode: inode of the file used
455 * @offset: file offset to start truncating
457 * This is a copy of the common vmtruncate, but with the locking
458 * corrected to take into account the fact that NFS requires
459 * inode->i_size to be updated under the inode->i_lock.
461 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
466 err
= inode_newsize_ok(inode
, offset
);
470 spin_lock(&inode
->i_lock
);
471 oldsize
= inode
->i_size
;
472 i_size_write(inode
, offset
);
473 spin_unlock(&inode
->i_lock
);
475 truncate_pagecache(inode
, oldsize
, offset
);
481 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
482 * @inode: pointer to struct inode
483 * @attr: pointer to struct iattr
485 * Note: we do this in the *proc.c in order to ensure that
486 * it works for things like exclusive creates too.
488 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
490 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
491 spin_lock(&inode
->i_lock
);
492 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
493 int mode
= attr
->ia_mode
& S_IALLUGO
;
494 mode
|= inode
->i_mode
& ~S_IALLUGO
;
495 inode
->i_mode
= mode
;
497 if ((attr
->ia_valid
& ATTR_UID
) != 0)
498 inode
->i_uid
= attr
->ia_uid
;
499 if ((attr
->ia_valid
& ATTR_GID
) != 0)
500 inode
->i_gid
= attr
->ia_gid
;
501 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
502 spin_unlock(&inode
->i_lock
);
504 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
505 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
506 nfs_vmtruncate(inode
, attr
->ia_size
);
510 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
512 struct inode
*inode
= dentry
->d_inode
;
513 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
516 /* Flush out writes to the server in order to update c/mtime. */
517 if (S_ISREG(inode
->i_mode
)) {
518 err
= filemap_write_and_wait(inode
->i_mapping
);
524 * We may force a getattr if the user cares about atime.
526 * Note that we only have to check the vfsmount flags here:
527 * - NFS always sets S_NOATIME by so checking it would give a
529 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
530 * no point in checking those.
532 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
533 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
537 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
539 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
541 generic_fillattr(inode
, stat
);
542 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
548 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
550 atomic_set(&l_ctx
->count
, 1);
551 l_ctx
->lockowner
= current
->files
;
552 l_ctx
->pid
= current
->tgid
;
553 INIT_LIST_HEAD(&l_ctx
->list
);
556 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
558 struct nfs_lock_context
*pos
;
560 list_for_each_entry(pos
, &ctx
->lock_context
.list
, list
) {
561 if (pos
->lockowner
!= current
->files
)
563 if (pos
->pid
!= current
->tgid
)
565 atomic_inc(&pos
->count
);
571 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
573 struct nfs_lock_context
*res
, *new = NULL
;
574 struct inode
*inode
= ctx
->dentry
->d_inode
;
576 spin_lock(&inode
->i_lock
);
577 res
= __nfs_find_lock_context(ctx
);
579 spin_unlock(&inode
->i_lock
);
580 new = kmalloc(sizeof(*new), GFP_KERNEL
);
583 nfs_init_lock_context(new);
584 spin_lock(&inode
->i_lock
);
585 res
= __nfs_find_lock_context(ctx
);
587 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
588 new->open_context
= ctx
;
593 spin_unlock(&inode
->i_lock
);
598 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
600 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
601 struct inode
*inode
= ctx
->dentry
->d_inode
;
603 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
605 list_del(&l_ctx
->list
);
606 spin_unlock(&inode
->i_lock
);
611 * nfs_close_context - Common close_context() routine NFSv2/v3
612 * @ctx: pointer to context
613 * @is_sync: is this a synchronous close
615 * always ensure that the attributes are up to date if we're mounted
616 * with close-to-open semantics
618 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
621 struct nfs_server
*server
;
623 if (!(ctx
->mode
& FMODE_WRITE
))
627 inode
= ctx
->dentry
->d_inode
;
628 if (!list_empty(&NFS_I(inode
)->open_files
))
630 server
= NFS_SERVER(inode
);
631 if (server
->flags
& NFS_MOUNT_NOCTO
)
633 nfs_revalidate_inode(server
, inode
);
636 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
638 struct nfs_open_context
*ctx
;
639 struct rpc_cred
*cred
= rpc_lookup_cred();
641 return ERR_CAST(cred
);
643 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
646 return ERR_PTR(-ENOMEM
);
648 nfs_sb_active(dentry
->d_sb
);
649 ctx
->dentry
= dget(dentry
);
655 nfs_init_lock_context(&ctx
->lock_context
);
656 ctx
->lock_context
.open_context
= ctx
;
657 INIT_LIST_HEAD(&ctx
->list
);
661 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
664 atomic_inc(&ctx
->lock_context
.count
);
668 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
670 struct inode
*inode
= ctx
->dentry
->d_inode
;
671 struct super_block
*sb
= ctx
->dentry
->d_sb
;
673 if (!list_empty(&ctx
->list
)) {
674 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
676 list_del(&ctx
->list
);
677 spin_unlock(&inode
->i_lock
);
678 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
681 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
682 if (ctx
->cred
!= NULL
)
683 put_rpccred(ctx
->cred
);
689 void put_nfs_open_context(struct nfs_open_context
*ctx
)
691 __put_nfs_open_context(ctx
, 0);
695 * Ensure that mmap has a recent RPC credential for use when writing out
698 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
700 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
701 struct nfs_inode
*nfsi
= NFS_I(inode
);
703 filp
->private_data
= get_nfs_open_context(ctx
);
704 spin_lock(&inode
->i_lock
);
705 list_add(&ctx
->list
, &nfsi
->open_files
);
706 spin_unlock(&inode
->i_lock
);
710 * Given an inode, search for an open context with the desired characteristics
712 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
714 struct nfs_inode
*nfsi
= NFS_I(inode
);
715 struct nfs_open_context
*pos
, *ctx
= NULL
;
717 spin_lock(&inode
->i_lock
);
718 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
719 if (cred
!= NULL
&& pos
->cred
!= cred
)
721 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
723 ctx
= get_nfs_open_context(pos
);
726 spin_unlock(&inode
->i_lock
);
730 static void nfs_file_clear_open_context(struct file
*filp
)
732 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
733 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
736 filp
->private_data
= NULL
;
737 spin_lock(&inode
->i_lock
);
738 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
739 spin_unlock(&inode
->i_lock
);
740 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
745 * These allocate and release file read/write context information.
747 int nfs_open(struct inode
*inode
, struct file
*filp
)
749 struct nfs_open_context
*ctx
;
751 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
754 nfs_file_set_open_context(filp
, ctx
);
755 put_nfs_open_context(ctx
);
756 nfs_fscache_set_inode_cookie(inode
, filp
);
760 int nfs_release(struct inode
*inode
, struct file
*filp
)
762 nfs_file_clear_open_context(filp
);
767 * This function is called whenever some part of NFS notices that
768 * the cached attributes have to be refreshed.
771 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
773 int status
= -ESTALE
;
774 struct nfs_fattr
*fattr
= NULL
;
775 struct nfs_inode
*nfsi
= NFS_I(inode
);
777 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
778 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
780 if (is_bad_inode(inode
))
782 if (NFS_STALE(inode
))
786 fattr
= nfs_alloc_fattr();
790 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
791 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
);
793 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
795 (long long)NFS_FILEID(inode
), status
);
796 if (status
== -ESTALE
) {
797 nfs_zap_caches(inode
);
798 if (!S_ISDIR(inode
->i_mode
))
799 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
804 status
= nfs_refresh_inode(inode
, fattr
);
806 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
808 (long long)NFS_FILEID(inode
), status
);
812 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
813 nfs_zap_acl_cache(inode
);
815 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
817 (long long)NFS_FILEID(inode
));
820 nfs_free_fattr(fattr
);
824 int nfs_attribute_timeout(struct inode
*inode
)
826 struct nfs_inode
*nfsi
= NFS_I(inode
);
828 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
831 static int nfs_attribute_cache_expired(struct inode
*inode
)
833 if (nfs_have_delegated_attributes(inode
))
835 return nfs_attribute_timeout(inode
);
839 * nfs_revalidate_inode - Revalidate the inode attributes
840 * @server - pointer to nfs_server struct
841 * @inode - pointer to inode struct
843 * Updates inode attribute information by retrieving the data from the server.
845 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
847 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
848 && !nfs_attribute_cache_expired(inode
))
849 return NFS_STALE(inode
) ? -ESTALE
: 0;
850 return __nfs_revalidate_inode(server
, inode
);
853 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
855 struct nfs_inode
*nfsi
= NFS_I(inode
);
857 if (mapping
->nrpages
!= 0) {
858 int ret
= invalidate_inode_pages2(mapping
);
862 spin_lock(&inode
->i_lock
);
863 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
864 if (S_ISDIR(inode
->i_mode
))
865 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
866 spin_unlock(&inode
->i_lock
);
867 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
868 nfs_fscache_reset_inode_cookie(inode
);
869 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
870 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
875 * nfs_revalidate_mapping - Revalidate the pagecache
876 * @inode - pointer to host inode
877 * @mapping - pointer to mapping
879 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
881 struct nfs_inode
*nfsi
= NFS_I(inode
);
884 if ((nfsi
->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
885 || nfs_attribute_cache_expired(inode
)
886 || NFS_STALE(inode
)) {
887 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
891 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
892 ret
= nfs_invalidate_mapping(inode
, mapping
);
897 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
899 struct nfs_inode
*nfsi
= NFS_I(inode
);
900 unsigned long ret
= 0;
902 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
903 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
904 && inode
->i_version
== fattr
->pre_change_attr
) {
905 inode
->i_version
= fattr
->change_attr
;
906 if (S_ISDIR(inode
->i_mode
))
907 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
908 ret
|= NFS_INO_INVALID_ATTR
;
910 /* If we have atomic WCC data, we may update some attributes */
911 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
912 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
913 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
914 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
915 ret
|= NFS_INO_INVALID_ATTR
;
918 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
919 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
920 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
921 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
922 if (S_ISDIR(inode
->i_mode
))
923 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
924 ret
|= NFS_INO_INVALID_ATTR
;
926 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
927 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
928 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
929 && nfsi
->npages
== 0) {
930 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
931 ret
|= NFS_INO_INVALID_ATTR
;
937 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
938 * @inode - pointer to inode
939 * @fattr - updated attributes
941 * Verifies the attribute cache. If we have just changed the attributes,
942 * so that fattr carries weak cache consistency data, then it may
943 * also update the ctime/mtime/change_attribute.
945 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
947 struct nfs_inode
*nfsi
= NFS_I(inode
);
948 loff_t cur_size
, new_isize
;
949 unsigned long invalid
= 0;
952 /* Has the inode gone and changed behind our back? */
953 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
955 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
958 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
959 inode
->i_version
!= fattr
->change_attr
)
960 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
962 /* Verify a few of the more important attributes */
963 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
964 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
966 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
967 cur_size
= i_size_read(inode
);
968 new_isize
= nfs_size_to_loff_t(fattr
->size
);
969 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
970 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
973 /* Have any file permissions changed? */
974 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
975 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
976 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && inode
->i_uid
!= fattr
->uid
)
977 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
978 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && inode
->i_gid
!= fattr
->gid
)
979 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
981 /* Has the link count changed? */
982 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
983 invalid
|= NFS_INO_INVALID_ATTR
;
985 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
986 invalid
|= NFS_INO_INVALID_ATIME
;
989 nfsi
->cache_validity
|= invalid
;
991 nfsi
->read_cache_jiffies
= fattr
->time_start
;
995 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
997 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
999 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1002 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1004 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1006 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1009 static atomic_long_t nfs_attr_generation_counter
;
1011 static unsigned long nfs_read_attr_generation_counter(void)
1013 return atomic_long_read(&nfs_attr_generation_counter
);
1016 unsigned long nfs_inc_attr_generation_counter(void)
1018 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1021 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1024 fattr
->time_start
= jiffies
;
1025 fattr
->gencount
= nfs_inc_attr_generation_counter();
1026 fattr
->owner_name
= NULL
;
1027 fattr
->group_name
= NULL
;
1030 struct nfs_fattr
*nfs_alloc_fattr(void)
1032 struct nfs_fattr
*fattr
;
1034 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1036 nfs_fattr_init(fattr
);
1040 struct nfs_fh
*nfs_alloc_fhandle(void)
1044 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1052 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1053 * in the same way that wireshark does
1057 * For debugging only.
1059 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1061 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1062 * not on the result */
1063 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1067 * _nfs_display_fhandle - display an NFS file handle on the console
1069 * @fh: file handle to display
1070 * @caption: display caption
1072 * For debugging only.
1074 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1078 if (fh
== NULL
|| fh
->size
== 0) {
1079 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1083 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1084 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1085 for (i
= 0; i
< fh
->size
; i
+= 16) {
1086 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1088 switch ((fh
->size
- i
- 1) >> 2) {
1090 printk(KERN_DEFAULT
" %08x\n",
1094 printk(KERN_DEFAULT
" %08x %08x\n",
1095 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1098 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1099 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1100 be32_to_cpup(pos
+ 2));
1103 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1104 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1105 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1112 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1113 * @inode - pointer to inode
1114 * @fattr - attributes
1116 * Attempt to divine whether or not an RPC call reply carrying stale
1117 * attributes got scheduled after another call carrying updated ones.
1119 * To do so, the function first assumes that a more recent ctime means
1120 * that the attributes in fattr are newer, however it also attempt to
1121 * catch the case where ctime either didn't change, or went backwards
1122 * (if someone reset the clock on the server) by looking at whether
1123 * or not this RPC call was started after the inode was last updated.
1124 * Note also the check for wraparound of 'attr_gencount'
1126 * The function returns 'true' if it thinks the attributes in 'fattr' are
1127 * more recent than the ones cached in the inode.
1130 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1132 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1134 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1135 nfs_ctime_need_update(inode
, fattr
) ||
1136 nfs_size_need_update(inode
, fattr
) ||
1137 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1140 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1142 if (nfs_inode_attrs_need_update(inode
, fattr
))
1143 return nfs_update_inode(inode
, fattr
);
1144 return nfs_check_inode_attributes(inode
, fattr
);
1148 * nfs_refresh_inode - try to update the inode attribute cache
1149 * @inode - pointer to inode
1150 * @fattr - updated attributes
1152 * Check that an RPC call that returned attributes has not overlapped with
1153 * other recent updates of the inode metadata, then decide whether it is
1154 * safe to do a full update of the inode attributes, or whether just to
1155 * call nfs_check_inode_attributes.
1157 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1161 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1163 spin_lock(&inode
->i_lock
);
1164 status
= nfs_refresh_inode_locked(inode
, fattr
);
1165 spin_unlock(&inode
->i_lock
);
1170 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1172 struct nfs_inode
*nfsi
= NFS_I(inode
);
1174 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1175 if (S_ISDIR(inode
->i_mode
))
1176 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1177 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1179 return nfs_refresh_inode_locked(inode
, fattr
);
1183 * nfs_post_op_update_inode - try to update the inode attribute cache
1184 * @inode - pointer to inode
1185 * @fattr - updated attributes
1187 * After an operation that has changed the inode metadata, mark the
1188 * attribute cache as being invalid, then try to update it.
1190 * NB: if the server didn't return any post op attributes, this
1191 * function will force the retrieval of attributes before the next
1192 * NFS request. Thus it should be used only for operations that
1193 * are expected to change one or more attributes, to avoid
1194 * unnecessary NFS requests and trips through nfs_update_inode().
1196 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1200 spin_lock(&inode
->i_lock
);
1201 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1202 spin_unlock(&inode
->i_lock
);
1207 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1208 * @inode - pointer to inode
1209 * @fattr - updated attributes
1211 * After an operation that has changed the inode metadata, mark the
1212 * attribute cache as being invalid, then try to update it. Fake up
1213 * weak cache consistency data, if none exist.
1215 * This function is mainly designed to be used by the ->write_done() functions.
1217 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1221 spin_lock(&inode
->i_lock
);
1222 /* Don't do a WCC update if these attributes are already stale */
1223 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1224 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1225 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1226 | NFS_ATTR_FATTR_PRESIZE
1227 | NFS_ATTR_FATTR_PREMTIME
1228 | NFS_ATTR_FATTR_PRECTIME
);
1231 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1232 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1233 fattr
->pre_change_attr
= inode
->i_version
;
1234 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1236 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1237 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1238 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1239 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1241 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1242 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1243 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1244 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1246 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1247 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1248 fattr
->pre_size
= i_size_read(inode
);
1249 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1252 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1253 spin_unlock(&inode
->i_lock
);
1258 * Many nfs protocol calls return the new file attributes after
1259 * an operation. Here we update the inode to reflect the state
1260 * of the server's inode.
1262 * This is a bit tricky because we have to make sure all dirty pages
1263 * have been sent off to the server before calling invalidate_inode_pages.
1264 * To make sure no other process adds more write requests while we try
1265 * our best to flush them, we make them sleep during the attribute refresh.
1267 * A very similar scenario holds for the dir cache.
1269 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1271 struct nfs_server
*server
;
1272 struct nfs_inode
*nfsi
= NFS_I(inode
);
1273 loff_t cur_isize
, new_isize
;
1274 unsigned long invalid
= 0;
1275 unsigned long now
= jiffies
;
1276 unsigned long save_cache_validity
;
1278 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1279 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1280 nfs_display_fhandle_hash(NFS_FH(inode
)),
1281 atomic_read(&inode
->i_count
), fattr
->valid
);
1283 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1287 * Make sure the inode's type hasn't changed.
1289 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1292 server
= NFS_SERVER(inode
);
1293 /* Update the fsid? */
1294 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1295 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1296 !IS_AUTOMOUNT(inode
))
1297 server
->fsid
= fattr
->fsid
;
1300 * Update the read time so we don't revalidate too often.
1302 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1304 save_cache_validity
= nfsi
->cache_validity
;
1305 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1306 | NFS_INO_INVALID_ATIME
1307 | NFS_INO_REVAL_FORCED
1308 | NFS_INO_REVAL_PAGECACHE
);
1310 /* Do atomic weak cache consistency updates */
1311 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1313 /* More cache consistency checks */
1314 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1315 if (inode
->i_version
!= fattr
->change_attr
) {
1316 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1317 inode
->i_sb
->s_id
, inode
->i_ino
);
1318 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1319 if (S_ISDIR(inode
->i_mode
))
1320 nfs_force_lookup_revalidate(inode
);
1321 inode
->i_version
= fattr
->change_attr
;
1323 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1324 invalid
|= save_cache_validity
;
1326 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1327 /* NFSv2/v3: Check if the mtime agrees */
1328 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1329 dprintk("NFS: mtime change on server for file %s/%ld\n",
1330 inode
->i_sb
->s_id
, inode
->i_ino
);
1331 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1332 if (S_ISDIR(inode
->i_mode
))
1333 nfs_force_lookup_revalidate(inode
);
1334 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1336 } else if (server
->caps
& NFS_CAP_MTIME
)
1337 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1338 | NFS_INO_INVALID_DATA
1339 | NFS_INO_REVAL_PAGECACHE
1340 | NFS_INO_REVAL_FORCED
);
1342 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1343 /* If ctime has changed we should definitely clear access+acl caches */
1344 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1345 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1346 /* and probably clear data for a directory too as utimes can cause
1347 * havoc with our cache.
1349 if (S_ISDIR(inode
->i_mode
)) {
1350 invalid
|= NFS_INO_INVALID_DATA
;
1351 nfs_force_lookup_revalidate(inode
);
1353 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1355 } else if (server
->caps
& NFS_CAP_CTIME
)
1356 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1357 | NFS_INO_INVALID_ACCESS
1358 | NFS_INO_INVALID_ACL
1359 | NFS_INO_REVAL_FORCED
);
1361 /* Check if our cached file size is stale */
1362 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1363 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1364 cur_isize
= i_size_read(inode
);
1365 if (new_isize
!= cur_isize
) {
1366 /* Do we perhaps have any outstanding writes, or has
1367 * the file grown beyond our last write? */
1368 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1369 new_isize
> cur_isize
) {
1370 i_size_write(inode
, new_isize
);
1371 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1373 dprintk("NFS: isize change on server for file %s/%ld "
1377 (long long)cur_isize
,
1378 (long long)new_isize
);
1381 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1382 | NFS_INO_REVAL_PAGECACHE
1383 | NFS_INO_REVAL_FORCED
);
1386 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1387 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1388 else if (server
->caps
& NFS_CAP_ATIME
)
1389 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1390 | NFS_INO_REVAL_FORCED
);
1392 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1393 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1394 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1395 newmode
|= fattr
->mode
& S_IALLUGO
;
1396 inode
->i_mode
= newmode
;
1397 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1399 } else if (server
->caps
& NFS_CAP_MODE
)
1400 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1401 | NFS_INO_INVALID_ACCESS
1402 | NFS_INO_INVALID_ACL
1403 | NFS_INO_REVAL_FORCED
);
1405 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1406 if (inode
->i_uid
!= fattr
->uid
) {
1407 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1408 inode
->i_uid
= fattr
->uid
;
1410 } else if (server
->caps
& NFS_CAP_OWNER
)
1411 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1412 | NFS_INO_INVALID_ACCESS
1413 | NFS_INO_INVALID_ACL
1414 | NFS_INO_REVAL_FORCED
);
1416 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1417 if (inode
->i_gid
!= fattr
->gid
) {
1418 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1419 inode
->i_gid
= fattr
->gid
;
1421 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1422 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1423 | NFS_INO_INVALID_ACCESS
1424 | NFS_INO_INVALID_ACL
1425 | NFS_INO_REVAL_FORCED
);
1427 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1428 if (inode
->i_nlink
!= fattr
->nlink
) {
1429 invalid
|= NFS_INO_INVALID_ATTR
;
1430 if (S_ISDIR(inode
->i_mode
))
1431 invalid
|= NFS_INO_INVALID_DATA
;
1432 set_nlink(inode
, fattr
->nlink
);
1434 } else if (server
->caps
& NFS_CAP_NLINK
)
1435 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1436 | NFS_INO_REVAL_FORCED
);
1438 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1440 * report the blocks in 512byte units
1442 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1444 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1445 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1447 /* Update attrtimeo value if we're out of the unstable period */
1448 if (invalid
& NFS_INO_INVALID_ATTR
) {
1449 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1450 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1451 nfsi
->attrtimeo_timestamp
= now
;
1452 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1454 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1455 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1456 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1457 nfsi
->attrtimeo_timestamp
= now
;
1460 invalid
&= ~NFS_INO_INVALID_ATTR
;
1461 /* Don't invalidate the data if we were to blame */
1462 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1463 || S_ISLNK(inode
->i_mode
)))
1464 invalid
&= ~NFS_INO_INVALID_DATA
;
1465 if (!nfs_have_delegation(inode
, FMODE_READ
) ||
1466 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1467 nfsi
->cache_validity
|= invalid
;
1472 * Big trouble! The inode has become a different object.
1474 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1475 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1478 * No need to worry about unhashing the dentry, as the
1479 * lookup validation will know that the inode is bad.
1480 * (But we fall through to invalidate the caches.)
1482 nfs_invalidate_inode(inode
);
1486 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1487 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1488 NFS_SERVER(inode
)->nfs_client
->cl_hostname
, inode
->i_sb
->s_id
,
1489 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1494 #ifdef CONFIG_NFS_V4
1497 * Clean out any remaining NFSv4 state that might be left over due
1498 * to open() calls that passed nfs_atomic_lookup, but failed to call
1501 void nfs4_evict_inode(struct inode
*inode
)
1503 truncate_inode_pages(&inode
->i_data
, 0);
1504 end_writeback(inode
);
1505 pnfs_return_layout(inode
);
1506 pnfs_destroy_layout(NFS_I(inode
));
1507 /* If we are holding a delegation, return it! */
1508 nfs_inode_return_delegation_noreclaim(inode
);
1509 /* First call standard NFS clear_inode() code */
1510 nfs_clear_inode(inode
);
1514 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1516 struct nfs_inode
*nfsi
;
1517 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1521 nfsi
->cache_validity
= 0UL;
1522 #ifdef CONFIG_NFS_V3_ACL
1523 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1524 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1526 #ifdef CONFIG_NFS_V4
1527 nfsi
->nfs4_acl
= NULL
;
1528 #endif /* CONFIG_NFS_V4 */
1529 return &nfsi
->vfs_inode
;
1532 static void nfs_i_callback(struct rcu_head
*head
)
1534 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1535 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1538 void nfs_destroy_inode(struct inode
*inode
)
1540 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1543 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1545 #ifdef CONFIG_NFS_V4
1546 INIT_LIST_HEAD(&nfsi
->open_states
);
1547 nfsi
->delegation
= NULL
;
1548 nfsi
->delegation_state
= 0;
1549 init_rwsem(&nfsi
->rwsem
);
1550 nfsi
->layout
= NULL
;
1551 atomic_set(&nfsi
->commits_outstanding
, 0);
1555 static void init_once(void *foo
)
1557 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1559 inode_init_once(&nfsi
->vfs_inode
);
1560 INIT_LIST_HEAD(&nfsi
->open_files
);
1561 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1562 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1563 INIT_LIST_HEAD(&nfsi
->commit_list
);
1566 atomic_set(&nfsi
->silly_count
, 1);
1567 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1568 init_waitqueue_head(&nfsi
->waitqueue
);
1569 nfs4_init_once(nfsi
);
1572 static int __init
nfs_init_inodecache(void)
1574 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1575 sizeof(struct nfs_inode
),
1576 0, (SLAB_RECLAIM_ACCOUNT
|
1579 if (nfs_inode_cachep
== NULL
)
1585 static void nfs_destroy_inodecache(void)
1587 kmem_cache_destroy(nfs_inode_cachep
);
1590 struct workqueue_struct
*nfsiod_workqueue
;
1593 * start up the nfsiod workqueue
1595 static int nfsiod_start(void)
1597 struct workqueue_struct
*wq
;
1598 dprintk("RPC: creating workqueue nfsiod\n");
1599 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1602 nfsiod_workqueue
= wq
;
1607 * Destroy the nfsiod workqueue
1609 static void nfsiod_stop(void)
1611 struct workqueue_struct
*wq
;
1613 wq
= nfsiod_workqueue
;
1616 nfsiod_workqueue
= NULL
;
1617 destroy_workqueue(wq
);
1621 EXPORT_SYMBOL_GPL(nfs_net_id
);
1623 static int nfs_net_init(struct net
*net
)
1625 nfs_clients_init(net
);
1626 return nfs_dns_resolver_cache_init(net
);
1629 static void nfs_net_exit(struct net
*net
)
1631 nfs_dns_resolver_cache_destroy(net
);
1632 nfs_cleanup_cb_ident_idr(net
);
1635 static struct pernet_operations nfs_net_ops
= {
1636 .init
= nfs_net_init
,
1637 .exit
= nfs_net_exit
,
1639 .size
= sizeof(struct nfs_net
),
1645 static int __init
init_nfs_fs(void)
1649 err
= nfs_idmap_init();
1653 err
= nfs_dns_resolver_init();
1657 err
= register_pernet_subsys(&nfs_net_ops
);
1661 err
= nfs_fscache_register();
1665 err
= nfsiod_start();
1669 err
= nfs_fs_proc_init();
1673 err
= nfs_init_nfspagecache();
1677 err
= nfs_init_inodecache();
1681 err
= nfs_init_readpagecache();
1685 err
= nfs_init_writepagecache();
1689 err
= nfs_init_directcache();
1693 #ifdef CONFIG_PROC_FS
1694 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1696 if ((err
= register_nfs_fs()) != 0)
1700 #ifdef CONFIG_PROC_FS
1701 rpc_proc_unregister(&init_net
, "nfs");
1703 nfs_destroy_directcache();
1705 nfs_destroy_writepagecache();
1707 nfs_destroy_readpagecache();
1709 nfs_destroy_inodecache();
1711 nfs_destroy_nfspagecache();
1717 nfs_fscache_unregister();
1719 unregister_pernet_subsys(&nfs_net_ops
);
1721 nfs_dns_resolver_destroy();
1728 static void __exit
exit_nfs_fs(void)
1730 nfs_destroy_directcache();
1731 nfs_destroy_writepagecache();
1732 nfs_destroy_readpagecache();
1733 nfs_destroy_inodecache();
1734 nfs_destroy_nfspagecache();
1735 nfs_fscache_unregister();
1736 unregister_pernet_subsys(&nfs_net_ops
);
1737 nfs_dns_resolver_destroy();
1739 #ifdef CONFIG_PROC_FS
1740 rpc_proc_unregister(&init_net
, "nfs");
1742 unregister_nfs_fs();
1747 /* Not quite true; I just maintain it */
1748 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1749 MODULE_LICENSE("GPL");
1750 module_param(enable_ino64
, bool, 0644);
1752 module_init(init_nfs_fs
)
1753 module_exit(exit_nfs_fs
)