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>
43 #include <asm/system.h>
44 #include <asm/uaccess.h>
48 #include "delegation.h"
52 #include "dns_resolve.h"
56 #define NFSDBG_FACILITY NFSDBG_VFS
58 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
60 /* Default is to see 64-bit inode numbers */
61 static bool enable_ino64
= NFS_64_BIT_INODE_NUMBERS_ENABLED
;
63 static void nfs_invalidate_inode(struct inode
*);
64 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
66 static struct kmem_cache
* nfs_inode_cachep
;
68 static inline unsigned long
69 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
71 return nfs_fileid_to_ino_t(fattr
->fileid
);
75 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
76 * @word: long word containing the bit lock
78 int nfs_wait_bit_killable(void *word
)
80 if (fatal_signal_pending(current
))
87 * nfs_compat_user_ino64 - returns the user-visible inode number
88 * @fileid: 64-bit fileid
90 * This function returns a 32-bit inode number if the boot parameter
91 * nfs.enable_ino64 is zero.
93 u64
nfs_compat_user_ino64(u64 fileid
)
104 if (sizeof(ino
) < sizeof(fileid
))
105 ino
^= fileid
>> (sizeof(fileid
)-sizeof(ino
)) * 8;
109 static void nfs_clear_inode(struct inode
*inode
)
112 * The following should never happen...
114 BUG_ON(nfs_have_writebacks(inode
));
115 BUG_ON(!list_empty(&NFS_I(inode
)->open_files
));
116 nfs_zap_acl_cache(inode
);
117 nfs_access_zap_cache(inode
);
118 nfs_fscache_release_inode_cookie(inode
);
121 void nfs_evict_inode(struct inode
*inode
)
123 truncate_inode_pages(&inode
->i_data
, 0);
124 end_writeback(inode
);
125 nfs_clear_inode(inode
);
129 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
131 int nfs_sync_mapping(struct address_space
*mapping
)
135 if (mapping
->nrpages
!= 0) {
136 unmap_mapping_range(mapping
, 0, 0, 0);
137 ret
= nfs_wb_all(mapping
->host
);
143 * Invalidate the local caches
145 static void nfs_zap_caches_locked(struct inode
*inode
)
147 struct nfs_inode
*nfsi
= NFS_I(inode
);
148 int mode
= inode
->i_mode
;
150 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
152 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
153 nfsi
->attrtimeo_timestamp
= jiffies
;
155 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
156 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
157 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
159 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
162 void nfs_zap_caches(struct inode
*inode
)
164 spin_lock(&inode
->i_lock
);
165 nfs_zap_caches_locked(inode
);
166 spin_unlock(&inode
->i_lock
);
169 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
171 if (mapping
->nrpages
!= 0) {
172 spin_lock(&inode
->i_lock
);
173 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
174 spin_unlock(&inode
->i_lock
);
178 void nfs_zap_acl_cache(struct inode
*inode
)
180 void (*clear_acl_cache
)(struct inode
*);
182 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
183 if (clear_acl_cache
!= NULL
)
184 clear_acl_cache(inode
);
185 spin_lock(&inode
->i_lock
);
186 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
187 spin_unlock(&inode
->i_lock
);
190 void nfs_invalidate_atime(struct inode
*inode
)
192 spin_lock(&inode
->i_lock
);
193 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
194 spin_unlock(&inode
->i_lock
);
198 * Invalidate, but do not unhash, the inode.
199 * NB: must be called with inode->i_lock held!
201 static void nfs_invalidate_inode(struct inode
*inode
)
203 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
204 nfs_zap_caches_locked(inode
);
207 struct nfs_find_desc
{
209 struct nfs_fattr
*fattr
;
213 * In NFSv3 we can have 64bit inode numbers. In order to support
214 * this, and re-exported directories (also seen in NFSv2)
215 * we are forced to allow 2 different inodes to have the same
219 nfs_find_actor(struct inode
*inode
, void *opaque
)
221 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
222 struct nfs_fh
*fh
= desc
->fh
;
223 struct nfs_fattr
*fattr
= desc
->fattr
;
225 if (NFS_FILEID(inode
) != fattr
->fileid
)
227 if (nfs_compare_fh(NFS_FH(inode
), fh
))
229 if (is_bad_inode(inode
) || NFS_STALE(inode
))
235 nfs_init_locked(struct inode
*inode
, void *opaque
)
237 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
238 struct nfs_fattr
*fattr
= desc
->fattr
;
240 set_nfs_fileid(inode
, fattr
->fileid
);
241 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
246 * This is our front-end to iget that looks up inodes by file handle
247 * instead of inode number.
250 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
252 struct nfs_find_desc desc
= {
256 struct inode
*inode
= ERR_PTR(-ENOENT
);
259 nfs_attr_check_mountpoint(sb
, fattr
);
261 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
262 !nfs_attr_use_mounted_on_fileid(fattr
))
264 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
267 hash
= nfs_fattr_to_ino_t(fattr
);
269 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
271 inode
= ERR_PTR(-ENOMEM
);
275 if (inode
->i_state
& I_NEW
) {
276 struct nfs_inode
*nfsi
= NFS_I(inode
);
277 unsigned long now
= jiffies
;
279 /* We set i_ino for the few things that still rely on it,
283 /* We can't support update_atime(), since the server will reset it */
284 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
285 inode
->i_mode
= fattr
->mode
;
286 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
287 && nfs_server_capable(inode
, NFS_CAP_MODE
))
288 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
289 | NFS_INO_INVALID_ACCESS
290 | NFS_INO_INVALID_ACL
;
291 /* Why so? Because we want revalidate for devices/FIFOs, and
292 * that's precisely what we have in nfs_file_inode_operations.
294 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
295 if (S_ISREG(inode
->i_mode
)) {
296 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
297 inode
->i_data
.a_ops
= &nfs_file_aops
;
298 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
299 } else if (S_ISDIR(inode
->i_mode
)) {
300 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
301 inode
->i_fop
= &nfs_dir_operations
;
302 inode
->i_data
.a_ops
= &nfs_dir_aops
;
303 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
))
304 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_I(inode
)->flags
);
305 /* Deal with crossing mountpoints */
306 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
307 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
308 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
309 inode
->i_op
= &nfs_referral_inode_operations
;
311 inode
->i_op
= &nfs_mountpoint_inode_operations
;
313 inode
->i_flags
|= S_AUTOMOUNT
;
315 } else if (S_ISLNK(inode
->i_mode
))
316 inode
->i_op
= &nfs_symlink_inode_operations
;
318 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
320 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
321 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
322 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
323 inode
->i_version
= 0;
329 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
331 nfsi
->read_cache_jiffies
= fattr
->time_start
;
332 nfsi
->attr_gencount
= fattr
->gencount
;
333 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
334 inode
->i_atime
= fattr
->atime
;
335 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
336 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
337 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
338 inode
->i_mtime
= fattr
->mtime
;
339 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
340 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
341 | NFS_INO_INVALID_DATA
;
342 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
343 inode
->i_ctime
= fattr
->ctime
;
344 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
345 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
346 | NFS_INO_INVALID_ACCESS
347 | NFS_INO_INVALID_ACL
;
348 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
349 inode
->i_version
= fattr
->change_attr
;
350 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
351 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
352 | NFS_INO_INVALID_DATA
;
353 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
354 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
356 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
357 | NFS_INO_INVALID_DATA
358 | NFS_INO_REVAL_PAGECACHE
;
359 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
360 set_nlink(inode
, fattr
->nlink
);
361 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
362 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
363 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
364 inode
->i_uid
= fattr
->uid
;
365 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
366 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
367 | NFS_INO_INVALID_ACCESS
368 | NFS_INO_INVALID_ACL
;
369 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
370 inode
->i_gid
= fattr
->gid
;
371 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
372 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
373 | NFS_INO_INVALID_ACCESS
374 | NFS_INO_INVALID_ACL
;
375 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
376 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
377 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
379 * report the blocks in 512byte units
381 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
383 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
384 nfsi
->attrtimeo_timestamp
= now
;
385 nfsi
->access_cache
= RB_ROOT
;
387 nfs_fscache_init_inode_cookie(inode
);
389 unlock_new_inode(inode
);
391 nfs_refresh_inode(inode
, fattr
);
392 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
394 (long long)NFS_FILEID(inode
),
395 atomic_read(&inode
->i_count
));
401 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
405 #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)
408 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
410 struct inode
*inode
= dentry
->d_inode
;
411 struct nfs_fattr
*fattr
;
414 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
416 /* skip mode change if it's just for clearing setuid/setgid */
417 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
418 attr
->ia_valid
&= ~ATTR_MODE
;
420 if (attr
->ia_valid
& ATTR_SIZE
) {
421 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
422 attr
->ia_valid
&= ~ATTR_SIZE
;
425 /* Optimization: if the end result is no change, don't RPC */
426 attr
->ia_valid
&= NFS_VALID_ATTRS
;
427 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
430 /* Write all dirty data */
431 if (S_ISREG(inode
->i_mode
))
434 fattr
= nfs_alloc_fattr();
438 * Return any delegations if we're going to change ACLs
440 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
441 nfs_inode_return_delegation(inode
);
442 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
444 nfs_refresh_inode(inode
, fattr
);
445 nfs_free_fattr(fattr
);
451 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
452 * @inode: inode of the file used
453 * @offset: file offset to start truncating
455 * This is a copy of the common vmtruncate, but with the locking
456 * corrected to take into account the fact that NFS requires
457 * inode->i_size to be updated under the inode->i_lock.
459 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
464 err
= inode_newsize_ok(inode
, offset
);
468 spin_lock(&inode
->i_lock
);
469 oldsize
= inode
->i_size
;
470 i_size_write(inode
, offset
);
471 spin_unlock(&inode
->i_lock
);
473 truncate_pagecache(inode
, oldsize
, offset
);
479 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
480 * @inode: pointer to struct inode
481 * @attr: pointer to struct iattr
483 * Note: we do this in the *proc.c in order to ensure that
484 * it works for things like exclusive creates too.
486 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
488 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
489 spin_lock(&inode
->i_lock
);
490 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
491 int mode
= attr
->ia_mode
& S_IALLUGO
;
492 mode
|= inode
->i_mode
& ~S_IALLUGO
;
493 inode
->i_mode
= mode
;
495 if ((attr
->ia_valid
& ATTR_UID
) != 0)
496 inode
->i_uid
= attr
->ia_uid
;
497 if ((attr
->ia_valid
& ATTR_GID
) != 0)
498 inode
->i_gid
= attr
->ia_gid
;
499 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
500 spin_unlock(&inode
->i_lock
);
502 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
503 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
504 nfs_vmtruncate(inode
, attr
->ia_size
);
508 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
510 struct inode
*inode
= dentry
->d_inode
;
511 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
514 /* Flush out writes to the server in order to update c/mtime. */
515 if (S_ISREG(inode
->i_mode
)) {
516 err
= filemap_write_and_wait(inode
->i_mapping
);
522 * We may force a getattr if the user cares about atime.
524 * Note that we only have to check the vfsmount flags here:
525 * - NFS always sets S_NOATIME by so checking it would give a
527 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
528 * no point in checking those.
530 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
531 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
535 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
537 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
539 generic_fillattr(inode
, stat
);
540 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
546 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
548 atomic_set(&l_ctx
->count
, 1);
549 l_ctx
->lockowner
= current
->files
;
550 l_ctx
->pid
= current
->tgid
;
551 INIT_LIST_HEAD(&l_ctx
->list
);
554 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
556 struct nfs_lock_context
*pos
;
558 list_for_each_entry(pos
, &ctx
->lock_context
.list
, list
) {
559 if (pos
->lockowner
!= current
->files
)
561 if (pos
->pid
!= current
->tgid
)
563 atomic_inc(&pos
->count
);
569 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
571 struct nfs_lock_context
*res
, *new = NULL
;
572 struct inode
*inode
= ctx
->dentry
->d_inode
;
574 spin_lock(&inode
->i_lock
);
575 res
= __nfs_find_lock_context(ctx
);
577 spin_unlock(&inode
->i_lock
);
578 new = kmalloc(sizeof(*new), GFP_KERNEL
);
581 nfs_init_lock_context(new);
582 spin_lock(&inode
->i_lock
);
583 res
= __nfs_find_lock_context(ctx
);
585 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
586 new->open_context
= ctx
;
591 spin_unlock(&inode
->i_lock
);
596 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
598 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
599 struct inode
*inode
= ctx
->dentry
->d_inode
;
601 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
603 list_del(&l_ctx
->list
);
604 spin_unlock(&inode
->i_lock
);
609 * nfs_close_context - Common close_context() routine NFSv2/v3
610 * @ctx: pointer to context
611 * @is_sync: is this a synchronous close
613 * always ensure that the attributes are up to date if we're mounted
614 * with close-to-open semantics
616 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
619 struct nfs_server
*server
;
621 if (!(ctx
->mode
& FMODE_WRITE
))
625 inode
= ctx
->dentry
->d_inode
;
626 if (!list_empty(&NFS_I(inode
)->open_files
))
628 server
= NFS_SERVER(inode
);
629 if (server
->flags
& NFS_MOUNT_NOCTO
)
631 nfs_revalidate_inode(server
, inode
);
634 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
636 struct nfs_open_context
*ctx
;
637 struct rpc_cred
*cred
= rpc_lookup_cred();
639 return ERR_CAST(cred
);
641 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
644 return ERR_PTR(-ENOMEM
);
646 nfs_sb_active(dentry
->d_sb
);
647 ctx
->dentry
= dget(dentry
);
653 nfs_init_lock_context(&ctx
->lock_context
);
654 ctx
->lock_context
.open_context
= ctx
;
655 INIT_LIST_HEAD(&ctx
->list
);
659 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
662 atomic_inc(&ctx
->lock_context
.count
);
666 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
668 struct inode
*inode
= ctx
->dentry
->d_inode
;
669 struct super_block
*sb
= ctx
->dentry
->d_sb
;
671 if (!list_empty(&ctx
->list
)) {
672 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
674 list_del(&ctx
->list
);
675 spin_unlock(&inode
->i_lock
);
676 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
679 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
680 if (ctx
->cred
!= NULL
)
681 put_rpccred(ctx
->cred
);
687 void put_nfs_open_context(struct nfs_open_context
*ctx
)
689 __put_nfs_open_context(ctx
, 0);
693 * Ensure that mmap has a recent RPC credential for use when writing out
696 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
698 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
699 struct nfs_inode
*nfsi
= NFS_I(inode
);
701 filp
->private_data
= get_nfs_open_context(ctx
);
702 spin_lock(&inode
->i_lock
);
703 list_add(&ctx
->list
, &nfsi
->open_files
);
704 spin_unlock(&inode
->i_lock
);
708 * Given an inode, search for an open context with the desired characteristics
710 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
712 struct nfs_inode
*nfsi
= NFS_I(inode
);
713 struct nfs_open_context
*pos
, *ctx
= NULL
;
715 spin_lock(&inode
->i_lock
);
716 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
717 if (cred
!= NULL
&& pos
->cred
!= cred
)
719 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
721 ctx
= get_nfs_open_context(pos
);
724 spin_unlock(&inode
->i_lock
);
728 static void nfs_file_clear_open_context(struct file
*filp
)
730 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
731 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
734 filp
->private_data
= NULL
;
735 spin_lock(&inode
->i_lock
);
736 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
737 spin_unlock(&inode
->i_lock
);
738 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
743 * These allocate and release file read/write context information.
745 int nfs_open(struct inode
*inode
, struct file
*filp
)
747 struct nfs_open_context
*ctx
;
749 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
752 nfs_file_set_open_context(filp
, ctx
);
753 put_nfs_open_context(ctx
);
754 nfs_fscache_set_inode_cookie(inode
, filp
);
758 int nfs_release(struct inode
*inode
, struct file
*filp
)
760 nfs_file_clear_open_context(filp
);
765 * This function is called whenever some part of NFS notices that
766 * the cached attributes have to be refreshed.
769 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
771 int status
= -ESTALE
;
772 struct nfs_fattr
*fattr
= NULL
;
773 struct nfs_inode
*nfsi
= NFS_I(inode
);
775 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
776 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
778 if (is_bad_inode(inode
))
780 if (NFS_STALE(inode
))
784 fattr
= nfs_alloc_fattr();
788 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
789 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
);
791 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
793 (long long)NFS_FILEID(inode
), status
);
794 if (status
== -ESTALE
) {
795 nfs_zap_caches(inode
);
796 if (!S_ISDIR(inode
->i_mode
))
797 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
802 status
= nfs_refresh_inode(inode
, fattr
);
804 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
806 (long long)NFS_FILEID(inode
), status
);
810 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
811 nfs_zap_acl_cache(inode
);
813 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
815 (long long)NFS_FILEID(inode
));
818 nfs_free_fattr(fattr
);
822 int nfs_attribute_timeout(struct inode
*inode
)
824 struct nfs_inode
*nfsi
= NFS_I(inode
);
826 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
829 static int nfs_attribute_cache_expired(struct inode
*inode
)
831 if (nfs_have_delegated_attributes(inode
))
833 return nfs_attribute_timeout(inode
);
837 * nfs_revalidate_inode - Revalidate the inode attributes
838 * @server - pointer to nfs_server struct
839 * @inode - pointer to inode struct
841 * Updates inode attribute information by retrieving the data from the server.
843 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
845 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
846 && !nfs_attribute_cache_expired(inode
))
847 return NFS_STALE(inode
) ? -ESTALE
: 0;
848 return __nfs_revalidate_inode(server
, inode
);
851 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
853 struct nfs_inode
*nfsi
= NFS_I(inode
);
855 if (mapping
->nrpages
!= 0) {
856 int ret
= invalidate_inode_pages2(mapping
);
860 spin_lock(&inode
->i_lock
);
861 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
862 if (S_ISDIR(inode
->i_mode
))
863 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
864 spin_unlock(&inode
->i_lock
);
865 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
866 nfs_fscache_reset_inode_cookie(inode
);
867 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
868 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
873 * nfs_revalidate_mapping - Revalidate the pagecache
874 * @inode - pointer to host inode
875 * @mapping - pointer to mapping
877 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
879 struct nfs_inode
*nfsi
= NFS_I(inode
);
882 if ((nfsi
->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
883 || nfs_attribute_cache_expired(inode
)
884 || NFS_STALE(inode
)) {
885 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
889 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
890 ret
= nfs_invalidate_mapping(inode
, mapping
);
895 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
897 struct nfs_inode
*nfsi
= NFS_I(inode
);
898 unsigned long ret
= 0;
900 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
901 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
902 && inode
->i_version
== fattr
->pre_change_attr
) {
903 inode
->i_version
= fattr
->change_attr
;
904 if (S_ISDIR(inode
->i_mode
))
905 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
906 ret
|= NFS_INO_INVALID_ATTR
;
908 /* If we have atomic WCC data, we may update some attributes */
909 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
910 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
911 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
912 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
913 ret
|= NFS_INO_INVALID_ATTR
;
916 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
917 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
918 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
919 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
920 if (S_ISDIR(inode
->i_mode
))
921 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
922 ret
|= NFS_INO_INVALID_ATTR
;
924 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
925 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
926 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
927 && nfsi
->npages
== 0) {
928 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
929 ret
|= NFS_INO_INVALID_ATTR
;
935 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
936 * @inode - pointer to inode
937 * @fattr - updated attributes
939 * Verifies the attribute cache. If we have just changed the attributes,
940 * so that fattr carries weak cache consistency data, then it may
941 * also update the ctime/mtime/change_attribute.
943 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
945 struct nfs_inode
*nfsi
= NFS_I(inode
);
946 loff_t cur_size
, new_isize
;
947 unsigned long invalid
= 0;
950 /* Has the inode gone and changed behind our back? */
951 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
953 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
956 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
957 inode
->i_version
!= fattr
->change_attr
)
958 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
960 /* Verify a few of the more important attributes */
961 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
962 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
964 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
965 cur_size
= i_size_read(inode
);
966 new_isize
= nfs_size_to_loff_t(fattr
->size
);
967 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
968 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
971 /* Have any file permissions changed? */
972 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
973 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
974 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && inode
->i_uid
!= fattr
->uid
)
975 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
976 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && inode
->i_gid
!= fattr
->gid
)
977 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
979 /* Has the link count changed? */
980 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
981 invalid
|= NFS_INO_INVALID_ATTR
;
983 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
984 invalid
|= NFS_INO_INVALID_ATIME
;
987 nfsi
->cache_validity
|= invalid
;
989 nfsi
->read_cache_jiffies
= fattr
->time_start
;
993 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
995 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
997 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1000 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1002 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1004 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1007 static atomic_long_t nfs_attr_generation_counter
;
1009 static unsigned long nfs_read_attr_generation_counter(void)
1011 return atomic_long_read(&nfs_attr_generation_counter
);
1014 unsigned long nfs_inc_attr_generation_counter(void)
1016 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1019 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1022 fattr
->time_start
= jiffies
;
1023 fattr
->gencount
= nfs_inc_attr_generation_counter();
1024 fattr
->owner_name
= NULL
;
1025 fattr
->group_name
= NULL
;
1028 struct nfs_fattr
*nfs_alloc_fattr(void)
1030 struct nfs_fattr
*fattr
;
1032 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1034 nfs_fattr_init(fattr
);
1038 struct nfs_fh
*nfs_alloc_fhandle(void)
1042 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1049 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1050 * @inode - pointer to inode
1051 * @fattr - attributes
1053 * Attempt to divine whether or not an RPC call reply carrying stale
1054 * attributes got scheduled after another call carrying updated ones.
1056 * To do so, the function first assumes that a more recent ctime means
1057 * that the attributes in fattr are newer, however it also attempt to
1058 * catch the case where ctime either didn't change, or went backwards
1059 * (if someone reset the clock on the server) by looking at whether
1060 * or not this RPC call was started after the inode was last updated.
1061 * Note also the check for wraparound of 'attr_gencount'
1063 * The function returns 'true' if it thinks the attributes in 'fattr' are
1064 * more recent than the ones cached in the inode.
1067 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1069 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1071 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1072 nfs_ctime_need_update(inode
, fattr
) ||
1073 nfs_size_need_update(inode
, fattr
) ||
1074 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1077 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1079 if (nfs_inode_attrs_need_update(inode
, fattr
))
1080 return nfs_update_inode(inode
, fattr
);
1081 return nfs_check_inode_attributes(inode
, fattr
);
1085 * nfs_refresh_inode - try to update the inode attribute cache
1086 * @inode - pointer to inode
1087 * @fattr - updated attributes
1089 * Check that an RPC call that returned attributes has not overlapped with
1090 * other recent updates of the inode metadata, then decide whether it is
1091 * safe to do a full update of the inode attributes, or whether just to
1092 * call nfs_check_inode_attributes.
1094 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1098 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1100 spin_lock(&inode
->i_lock
);
1101 status
= nfs_refresh_inode_locked(inode
, fattr
);
1102 spin_unlock(&inode
->i_lock
);
1107 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1109 struct nfs_inode
*nfsi
= NFS_I(inode
);
1111 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1112 if (S_ISDIR(inode
->i_mode
))
1113 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1114 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1116 return nfs_refresh_inode_locked(inode
, fattr
);
1120 * nfs_post_op_update_inode - try to update the inode attribute cache
1121 * @inode - pointer to inode
1122 * @fattr - updated attributes
1124 * After an operation that has changed the inode metadata, mark the
1125 * attribute cache as being invalid, then try to update it.
1127 * NB: if the server didn't return any post op attributes, this
1128 * function will force the retrieval of attributes before the next
1129 * NFS request. Thus it should be used only for operations that
1130 * are expected to change one or more attributes, to avoid
1131 * unnecessary NFS requests and trips through nfs_update_inode().
1133 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1137 spin_lock(&inode
->i_lock
);
1138 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1139 spin_unlock(&inode
->i_lock
);
1144 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1145 * @inode - pointer to inode
1146 * @fattr - updated attributes
1148 * After an operation that has changed the inode metadata, mark the
1149 * attribute cache as being invalid, then try to update it. Fake up
1150 * weak cache consistency data, if none exist.
1152 * This function is mainly designed to be used by the ->write_done() functions.
1154 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1158 spin_lock(&inode
->i_lock
);
1159 /* Don't do a WCC update if these attributes are already stale */
1160 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1161 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1162 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1163 | NFS_ATTR_FATTR_PRESIZE
1164 | NFS_ATTR_FATTR_PREMTIME
1165 | NFS_ATTR_FATTR_PRECTIME
);
1168 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1169 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1170 fattr
->pre_change_attr
= inode
->i_version
;
1171 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1173 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1174 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1175 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1176 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1178 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1179 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1180 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1181 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1183 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1184 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1185 fattr
->pre_size
= i_size_read(inode
);
1186 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1189 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1190 spin_unlock(&inode
->i_lock
);
1195 * Many nfs protocol calls return the new file attributes after
1196 * an operation. Here we update the inode to reflect the state
1197 * of the server's inode.
1199 * This is a bit tricky because we have to make sure all dirty pages
1200 * have been sent off to the server before calling invalidate_inode_pages.
1201 * To make sure no other process adds more write requests while we try
1202 * our best to flush them, we make them sleep during the attribute refresh.
1204 * A very similar scenario holds for the dir cache.
1206 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1208 struct nfs_server
*server
;
1209 struct nfs_inode
*nfsi
= NFS_I(inode
);
1210 loff_t cur_isize
, new_isize
;
1211 unsigned long invalid
= 0;
1212 unsigned long now
= jiffies
;
1213 unsigned long save_cache_validity
;
1215 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1216 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1217 atomic_read(&inode
->i_count
), fattr
->valid
);
1219 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1223 * Make sure the inode's type hasn't changed.
1225 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1228 server
= NFS_SERVER(inode
);
1229 /* Update the fsid? */
1230 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1231 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1232 !IS_AUTOMOUNT(inode
))
1233 server
->fsid
= fattr
->fsid
;
1236 * Update the read time so we don't revalidate too often.
1238 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1240 save_cache_validity
= nfsi
->cache_validity
;
1241 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1242 | NFS_INO_INVALID_ATIME
1243 | NFS_INO_REVAL_FORCED
1244 | NFS_INO_REVAL_PAGECACHE
);
1246 /* Do atomic weak cache consistency updates */
1247 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1249 /* More cache consistency checks */
1250 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1251 if (inode
->i_version
!= fattr
->change_attr
) {
1252 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1253 inode
->i_sb
->s_id
, inode
->i_ino
);
1254 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1255 if (S_ISDIR(inode
->i_mode
))
1256 nfs_force_lookup_revalidate(inode
);
1257 inode
->i_version
= fattr
->change_attr
;
1259 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1260 invalid
|= save_cache_validity
;
1262 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1263 /* NFSv2/v3: Check if the mtime agrees */
1264 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1265 dprintk("NFS: mtime change on server for file %s/%ld\n",
1266 inode
->i_sb
->s_id
, inode
->i_ino
);
1267 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1268 if (S_ISDIR(inode
->i_mode
))
1269 nfs_force_lookup_revalidate(inode
);
1270 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1272 } else if (server
->caps
& NFS_CAP_MTIME
)
1273 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1274 | NFS_INO_INVALID_DATA
1275 | NFS_INO_REVAL_PAGECACHE
1276 | NFS_INO_REVAL_FORCED
);
1278 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1279 /* If ctime has changed we should definitely clear access+acl caches */
1280 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1281 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1282 /* and probably clear data for a directory too as utimes can cause
1283 * havoc with our cache.
1285 if (S_ISDIR(inode
->i_mode
)) {
1286 invalid
|= NFS_INO_INVALID_DATA
;
1287 nfs_force_lookup_revalidate(inode
);
1289 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1291 } else if (server
->caps
& NFS_CAP_CTIME
)
1292 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1293 | NFS_INO_INVALID_ACCESS
1294 | NFS_INO_INVALID_ACL
1295 | NFS_INO_REVAL_FORCED
);
1297 /* Check if our cached file size is stale */
1298 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1299 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1300 cur_isize
= i_size_read(inode
);
1301 if (new_isize
!= cur_isize
) {
1302 /* Do we perhaps have any outstanding writes, or has
1303 * the file grown beyond our last write? */
1304 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1305 new_isize
> cur_isize
) {
1306 i_size_write(inode
, new_isize
);
1307 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1309 dprintk("NFS: isize change on server for file %s/%ld "
1313 (long long)cur_isize
,
1314 (long long)new_isize
);
1317 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1318 | NFS_INO_REVAL_PAGECACHE
1319 | NFS_INO_REVAL_FORCED
);
1322 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1323 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1324 else if (server
->caps
& NFS_CAP_ATIME
)
1325 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1326 | NFS_INO_REVAL_FORCED
);
1328 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1329 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1330 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1331 newmode
|= fattr
->mode
& S_IALLUGO
;
1332 inode
->i_mode
= newmode
;
1333 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1335 } else if (server
->caps
& NFS_CAP_MODE
)
1336 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1337 | NFS_INO_INVALID_ACCESS
1338 | NFS_INO_INVALID_ACL
1339 | NFS_INO_REVAL_FORCED
);
1341 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1342 if (inode
->i_uid
!= fattr
->uid
) {
1343 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1344 inode
->i_uid
= fattr
->uid
;
1346 } else if (server
->caps
& NFS_CAP_OWNER
)
1347 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1348 | NFS_INO_INVALID_ACCESS
1349 | NFS_INO_INVALID_ACL
1350 | NFS_INO_REVAL_FORCED
);
1352 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1353 if (inode
->i_gid
!= fattr
->gid
) {
1354 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1355 inode
->i_gid
= fattr
->gid
;
1357 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1358 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1359 | NFS_INO_INVALID_ACCESS
1360 | NFS_INO_INVALID_ACL
1361 | NFS_INO_REVAL_FORCED
);
1363 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1364 if (inode
->i_nlink
!= fattr
->nlink
) {
1365 invalid
|= NFS_INO_INVALID_ATTR
;
1366 if (S_ISDIR(inode
->i_mode
))
1367 invalid
|= NFS_INO_INVALID_DATA
;
1368 set_nlink(inode
, fattr
->nlink
);
1370 } else if (server
->caps
& NFS_CAP_NLINK
)
1371 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1372 | NFS_INO_REVAL_FORCED
);
1374 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1376 * report the blocks in 512byte units
1378 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1380 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1381 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1383 /* Update attrtimeo value if we're out of the unstable period */
1384 if (invalid
& NFS_INO_INVALID_ATTR
) {
1385 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1386 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1387 nfsi
->attrtimeo_timestamp
= now
;
1388 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1390 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1391 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1392 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1393 nfsi
->attrtimeo_timestamp
= now
;
1396 invalid
&= ~NFS_INO_INVALID_ATTR
;
1397 /* Don't invalidate the data if we were to blame */
1398 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1399 || S_ISLNK(inode
->i_mode
)))
1400 invalid
&= ~NFS_INO_INVALID_DATA
;
1401 if (!nfs_have_delegation(inode
, FMODE_READ
) ||
1402 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1403 nfsi
->cache_validity
|= invalid
;
1408 * Big trouble! The inode has become a different object.
1410 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1411 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1414 * No need to worry about unhashing the dentry, as the
1415 * lookup validation will know that the inode is bad.
1416 * (But we fall through to invalidate the caches.)
1418 nfs_invalidate_inode(inode
);
1422 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1423 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1424 NFS_SERVER(inode
)->nfs_client
->cl_hostname
, inode
->i_sb
->s_id
,
1425 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1430 #ifdef CONFIG_NFS_V4
1433 * Clean out any remaining NFSv4 state that might be left over due
1434 * to open() calls that passed nfs_atomic_lookup, but failed to call
1437 void nfs4_evict_inode(struct inode
*inode
)
1439 truncate_inode_pages(&inode
->i_data
, 0);
1440 end_writeback(inode
);
1441 pnfs_return_layout(inode
);
1442 pnfs_destroy_layout(NFS_I(inode
));
1443 /* If we are holding a delegation, return it! */
1444 nfs_inode_return_delegation_noreclaim(inode
);
1445 /* First call standard NFS clear_inode() code */
1446 nfs_clear_inode(inode
);
1450 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1452 struct nfs_inode
*nfsi
;
1453 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1457 nfsi
->cache_validity
= 0UL;
1458 #ifdef CONFIG_NFS_V3_ACL
1459 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1460 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1462 #ifdef CONFIG_NFS_V4
1463 nfsi
->nfs4_acl
= NULL
;
1464 #endif /* CONFIG_NFS_V4 */
1465 return &nfsi
->vfs_inode
;
1468 static void nfs_i_callback(struct rcu_head
*head
)
1470 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1471 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1474 void nfs_destroy_inode(struct inode
*inode
)
1476 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1479 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1481 #ifdef CONFIG_NFS_V4
1482 INIT_LIST_HEAD(&nfsi
->open_states
);
1483 nfsi
->delegation
= NULL
;
1484 nfsi
->delegation_state
= 0;
1485 init_rwsem(&nfsi
->rwsem
);
1486 nfsi
->layout
= NULL
;
1487 atomic_set(&nfsi
->commits_outstanding
, 0);
1491 static void init_once(void *foo
)
1493 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1495 inode_init_once(&nfsi
->vfs_inode
);
1496 INIT_LIST_HEAD(&nfsi
->open_files
);
1497 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1498 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1499 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
1502 atomic_set(&nfsi
->silly_count
, 1);
1503 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1504 init_waitqueue_head(&nfsi
->waitqueue
);
1505 nfs4_init_once(nfsi
);
1508 static int __init
nfs_init_inodecache(void)
1510 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1511 sizeof(struct nfs_inode
),
1512 0, (SLAB_RECLAIM_ACCOUNT
|
1515 if (nfs_inode_cachep
== NULL
)
1521 static void nfs_destroy_inodecache(void)
1523 kmem_cache_destroy(nfs_inode_cachep
);
1526 struct workqueue_struct
*nfsiod_workqueue
;
1529 * start up the nfsiod workqueue
1531 static int nfsiod_start(void)
1533 struct workqueue_struct
*wq
;
1534 dprintk("RPC: creating workqueue nfsiod\n");
1535 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1538 nfsiod_workqueue
= wq
;
1543 * Destroy the nfsiod workqueue
1545 static void nfsiod_stop(void)
1547 struct workqueue_struct
*wq
;
1549 wq
= nfsiod_workqueue
;
1552 nfsiod_workqueue
= NULL
;
1553 destroy_workqueue(wq
);
1557 EXPORT_SYMBOL_GPL(nfs_net_id
);
1559 static int nfs_net_init(struct net
*net
)
1561 nfs_clients_init(net
);
1562 return nfs_dns_resolver_cache_init(net
);
1565 static void nfs_net_exit(struct net
*net
)
1567 nfs_dns_resolver_cache_destroy(net
);
1568 nfs_cleanup_cb_ident_idr(net
);
1571 static struct pernet_operations nfs_net_ops
= {
1572 .init
= nfs_net_init
,
1573 .exit
= nfs_net_exit
,
1575 .size
= sizeof(struct nfs_net
),
1581 static int __init
init_nfs_fs(void)
1585 err
= nfs_idmap_init();
1589 err
= nfs_dns_resolver_init();
1593 err
= register_pernet_subsys(&nfs_net_ops
);
1597 err
= nfs_fscache_register();
1601 err
= nfsiod_start();
1605 err
= nfs_fs_proc_init();
1609 err
= nfs_init_nfspagecache();
1613 err
= nfs_init_inodecache();
1617 err
= nfs_init_readpagecache();
1621 err
= nfs_init_writepagecache();
1625 err
= nfs_init_directcache();
1629 #ifdef CONFIG_PROC_FS
1630 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1632 if ((err
= register_nfs_fs()) != 0)
1636 #ifdef CONFIG_PROC_FS
1637 rpc_proc_unregister(&init_net
, "nfs");
1639 nfs_destroy_directcache();
1641 nfs_destroy_writepagecache();
1643 nfs_destroy_readpagecache();
1645 nfs_destroy_inodecache();
1647 nfs_destroy_nfspagecache();
1653 nfs_fscache_unregister();
1655 unregister_pernet_subsys(&nfs_net_ops
);
1657 nfs_dns_resolver_destroy();
1664 static void __exit
exit_nfs_fs(void)
1666 nfs_destroy_directcache();
1667 nfs_destroy_writepagecache();
1668 nfs_destroy_readpagecache();
1669 nfs_destroy_inodecache();
1670 nfs_destroy_nfspagecache();
1671 nfs_fscache_unregister();
1672 unregister_pernet_subsys(&nfs_net_ops
);
1673 nfs_dns_resolver_destroy();
1675 #ifdef CONFIG_PROC_FS
1676 rpc_proc_unregister(&init_net
, "nfs");
1678 unregister_nfs_fs();
1683 /* Not quite true; I just maintain it */
1684 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1685 MODULE_LICENSE("GPL");
1686 module_param(enable_ino64
, bool, 0644);
1688 module_init(init_nfs_fs
)
1689 module_exit(exit_nfs_fs
)