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/vfs.h>
36 #include <linux/inet.h>
37 #include <linux/nfs_xdr.h>
38 #include <linux/slab.h>
39 #include <linux/compat.h>
40 #include <linux/freezer.h>
41 #include <linux/crc32.h>
43 #include <asm/uaccess.h>
47 #include "delegation.h"
51 #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
))
85 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable
);
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 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
);
121 EXPORT_SYMBOL_GPL(nfs_clear_inode
);
123 void nfs_evict_inode(struct inode
*inode
)
125 truncate_inode_pages(&inode
->i_data
, 0);
127 nfs_clear_inode(inode
);
131 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
133 int nfs_sync_mapping(struct address_space
*mapping
)
137 if (mapping
->nrpages
!= 0) {
138 unmap_mapping_range(mapping
, 0, 0, 0);
139 ret
= nfs_wb_all(mapping
->host
);
145 * Invalidate the local caches
147 static void nfs_zap_caches_locked(struct inode
*inode
)
149 struct nfs_inode
*nfsi
= NFS_I(inode
);
150 int mode
= inode
->i_mode
;
152 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
154 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
155 nfsi
->attrtimeo_timestamp
= jiffies
;
157 memset(NFS_I(inode
)->cookieverf
, 0, sizeof(NFS_I(inode
)->cookieverf
));
158 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
159 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
161 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
164 void nfs_zap_caches(struct inode
*inode
)
166 spin_lock(&inode
->i_lock
);
167 nfs_zap_caches_locked(inode
);
168 spin_unlock(&inode
->i_lock
);
171 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
173 if (mapping
->nrpages
!= 0) {
174 spin_lock(&inode
->i_lock
);
175 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
176 spin_unlock(&inode
->i_lock
);
180 void nfs_zap_acl_cache(struct inode
*inode
)
182 void (*clear_acl_cache
)(struct inode
*);
184 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
185 if (clear_acl_cache
!= NULL
)
186 clear_acl_cache(inode
);
187 spin_lock(&inode
->i_lock
);
188 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
189 spin_unlock(&inode
->i_lock
);
191 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache
);
193 void nfs_invalidate_atime(struct inode
*inode
)
195 spin_lock(&inode
->i_lock
);
196 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
197 spin_unlock(&inode
->i_lock
);
199 EXPORT_SYMBOL_GPL(nfs_invalidate_atime
);
202 * Invalidate, but do not unhash, the inode.
203 * NB: must be called with inode->i_lock held!
205 static void nfs_invalidate_inode(struct inode
*inode
)
207 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
208 nfs_zap_caches_locked(inode
);
211 struct nfs_find_desc
{
213 struct nfs_fattr
*fattr
;
217 * In NFSv3 we can have 64bit inode numbers. In order to support
218 * this, and re-exported directories (also seen in NFSv2)
219 * we are forced to allow 2 different inodes to have the same
223 nfs_find_actor(struct inode
*inode
, void *opaque
)
225 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
226 struct nfs_fh
*fh
= desc
->fh
;
227 struct nfs_fattr
*fattr
= desc
->fattr
;
229 if (NFS_FILEID(inode
) != fattr
->fileid
)
231 if (nfs_compare_fh(NFS_FH(inode
), fh
))
233 if (is_bad_inode(inode
) || NFS_STALE(inode
))
239 nfs_init_locked(struct inode
*inode
, void *opaque
)
241 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
242 struct nfs_fattr
*fattr
= desc
->fattr
;
244 set_nfs_fileid(inode
, fattr
->fileid
);
245 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
250 * This is our front-end to iget that looks up inodes by file handle
251 * instead of inode number.
254 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
256 struct nfs_find_desc desc
= {
260 struct inode
*inode
= ERR_PTR(-ENOENT
);
263 nfs_attr_check_mountpoint(sb
, fattr
);
265 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
266 !nfs_attr_use_mounted_on_fileid(fattr
))
268 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
271 hash
= nfs_fattr_to_ino_t(fattr
);
273 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
275 inode
= ERR_PTR(-ENOMEM
);
279 if (inode
->i_state
& I_NEW
) {
280 struct nfs_inode
*nfsi
= NFS_I(inode
);
281 unsigned long now
= jiffies
;
283 /* We set i_ino for the few things that still rely on it,
287 /* We can't support update_atime(), since the server will reset it */
288 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
289 inode
->i_mode
= fattr
->mode
;
290 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
291 && nfs_server_capable(inode
, NFS_CAP_MODE
))
292 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
293 /* Why so? Because we want revalidate for devices/FIFOs, and
294 * that's precisely what we have in nfs_file_inode_operations.
296 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
297 if (S_ISREG(inode
->i_mode
)) {
298 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
299 inode
->i_data
.a_ops
= &nfs_file_aops
;
300 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
301 } else if (S_ISDIR(inode
->i_mode
)) {
302 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
303 inode
->i_fop
= &nfs_dir_operations
;
304 inode
->i_data
.a_ops
= &nfs_dir_aops
;
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
));
333 nfsi
->read_cache_jiffies
= fattr
->time_start
;
334 nfsi
->attr_gencount
= fattr
->gencount
;
335 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
336 inode
->i_atime
= fattr
->atime
;
337 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
338 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
339 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
340 inode
->i_mtime
= fattr
->mtime
;
341 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
342 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
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 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
348 inode
->i_version
= fattr
->change_attr
;
349 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
350 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
351 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
352 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
354 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
355 | NFS_INO_REVAL_PAGECACHE
;
356 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
357 set_nlink(inode
, fattr
->nlink
);
358 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
359 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
360 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
361 inode
->i_uid
= fattr
->uid
;
362 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
363 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
364 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
365 inode
->i_gid
= fattr
->gid
;
366 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
367 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
368 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
369 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
370 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
372 * report the blocks in 512byte units
374 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
376 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
377 nfsi
->attrtimeo_timestamp
= now
;
378 nfsi
->access_cache
= RB_ROOT
;
380 nfs_fscache_init_inode_cookie(inode
);
382 unlock_new_inode(inode
);
384 nfs_refresh_inode(inode
, fattr
);
385 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
387 (long long)NFS_FILEID(inode
),
388 nfs_display_fhandle_hash(fh
),
389 atomic_read(&inode
->i_count
));
395 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
398 EXPORT_SYMBOL_GPL(nfs_fhget
);
400 #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)
403 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
405 struct inode
*inode
= dentry
->d_inode
;
406 struct nfs_fattr
*fattr
;
409 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
411 /* skip mode change if it's just for clearing setuid/setgid */
412 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
413 attr
->ia_valid
&= ~ATTR_MODE
;
415 if (attr
->ia_valid
& ATTR_SIZE
) {
416 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
417 attr
->ia_valid
&= ~ATTR_SIZE
;
420 /* Optimization: if the end result is no change, don't RPC */
421 attr
->ia_valid
&= NFS_VALID_ATTRS
;
422 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
425 /* Write all dirty data */
426 if (S_ISREG(inode
->i_mode
)) {
427 nfs_inode_dio_wait(inode
);
431 fattr
= nfs_alloc_fattr();
435 * Return any delegations if we're going to change ACLs
437 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
438 NFS_PROTO(inode
)->return_delegation(inode
);
439 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
441 nfs_refresh_inode(inode
, fattr
);
442 nfs_free_fattr(fattr
);
446 EXPORT_SYMBOL_GPL(nfs_setattr
);
449 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
450 * @inode: inode of the file used
451 * @offset: file offset to start truncating
453 * This is a copy of the common vmtruncate, but with the locking
454 * corrected to take into account the fact that NFS requires
455 * inode->i_size to be updated under the inode->i_lock.
457 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
462 err
= inode_newsize_ok(inode
, offset
);
466 spin_lock(&inode
->i_lock
);
467 oldsize
= inode
->i_size
;
468 i_size_write(inode
, offset
);
469 spin_unlock(&inode
->i_lock
);
471 truncate_pagecache(inode
, oldsize
, offset
);
477 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
478 * @inode: pointer to struct inode
479 * @attr: pointer to struct iattr
481 * Note: we do this in the *proc.c in order to ensure that
482 * it works for things like exclusive creates too.
484 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
486 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
487 spin_lock(&inode
->i_lock
);
488 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
489 int mode
= attr
->ia_mode
& S_IALLUGO
;
490 mode
|= inode
->i_mode
& ~S_IALLUGO
;
491 inode
->i_mode
= mode
;
493 if ((attr
->ia_valid
& ATTR_UID
) != 0)
494 inode
->i_uid
= attr
->ia_uid
;
495 if ((attr
->ia_valid
& ATTR_GID
) != 0)
496 inode
->i_gid
= attr
->ia_gid
;
497 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
498 spin_unlock(&inode
->i_lock
);
500 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
501 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
502 nfs_vmtruncate(inode
, attr
->ia_size
);
505 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode
);
507 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
509 struct inode
*inode
= dentry
->d_inode
;
510 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
513 /* Flush out writes to the server in order to update c/mtime. */
514 if (S_ISREG(inode
->i_mode
)) {
515 nfs_inode_dio_wait(inode
);
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
));
545 EXPORT_SYMBOL_GPL(nfs_getattr
);
547 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
549 atomic_set(&l_ctx
->count
, 1);
550 l_ctx
->lockowner
.l_owner
= current
->files
;
551 l_ctx
->lockowner
.l_pid
= current
->tgid
;
552 INIT_LIST_HEAD(&l_ctx
->list
);
555 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
557 struct nfs_lock_context
*pos
;
559 list_for_each_entry(pos
, &ctx
->lock_context
.list
, list
) {
560 if (pos
->lockowner
.l_owner
!= current
->files
)
562 if (pos
->lockowner
.l_pid
!= current
->tgid
)
564 atomic_inc(&pos
->count
);
570 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
572 struct nfs_lock_context
*res
, *new = NULL
;
573 struct inode
*inode
= ctx
->dentry
->d_inode
;
575 spin_lock(&inode
->i_lock
);
576 res
= __nfs_find_lock_context(ctx
);
578 spin_unlock(&inode
->i_lock
);
579 new = kmalloc(sizeof(*new), GFP_KERNEL
);
581 return ERR_PTR(-ENOMEM
);
582 nfs_init_lock_context(new);
583 spin_lock(&inode
->i_lock
);
584 res
= __nfs_find_lock_context(ctx
);
586 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
587 new->open_context
= ctx
;
592 spin_unlock(&inode
->i_lock
);
597 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
599 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
600 struct inode
*inode
= ctx
->dentry
->d_inode
;
602 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
604 list_del(&l_ctx
->list
);
605 spin_unlock(&inode
->i_lock
);
610 * nfs_close_context - Common close_context() routine NFSv2/v3
611 * @ctx: pointer to context
612 * @is_sync: is this a synchronous close
614 * always ensure that the attributes are up to date if we're mounted
615 * with close-to-open semantics
617 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
620 struct nfs_server
*server
;
622 if (!(ctx
->mode
& FMODE_WRITE
))
626 inode
= ctx
->dentry
->d_inode
;
627 if (!list_empty(&NFS_I(inode
)->open_files
))
629 server
= NFS_SERVER(inode
);
630 if (server
->flags
& NFS_MOUNT_NOCTO
)
632 nfs_revalidate_inode(server
, inode
);
634 EXPORT_SYMBOL_GPL(nfs_close_context
);
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
);
658 ctx
->mdsthreshold
= NULL
;
661 EXPORT_SYMBOL_GPL(alloc_nfs_open_context
);
663 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
666 atomic_inc(&ctx
->lock_context
.count
);
669 EXPORT_SYMBOL_GPL(get_nfs_open_context
);
671 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
673 struct inode
*inode
= ctx
->dentry
->d_inode
;
674 struct super_block
*sb
= ctx
->dentry
->d_sb
;
676 if (!list_empty(&ctx
->list
)) {
677 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
679 list_del(&ctx
->list
);
680 spin_unlock(&inode
->i_lock
);
681 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
684 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
685 if (ctx
->cred
!= NULL
)
686 put_rpccred(ctx
->cred
);
689 kfree(ctx
->mdsthreshold
);
693 void put_nfs_open_context(struct nfs_open_context
*ctx
)
695 __put_nfs_open_context(ctx
, 0);
697 EXPORT_SYMBOL_GPL(put_nfs_open_context
);
700 * Ensure that mmap has a recent RPC credential for use when writing out
703 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
705 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
706 struct nfs_inode
*nfsi
= NFS_I(inode
);
708 filp
->private_data
= get_nfs_open_context(ctx
);
709 spin_lock(&inode
->i_lock
);
710 list_add(&ctx
->list
, &nfsi
->open_files
);
711 spin_unlock(&inode
->i_lock
);
713 EXPORT_SYMBOL_GPL(nfs_file_set_open_context
);
716 * Given an inode, search for an open context with the desired characteristics
718 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
720 struct nfs_inode
*nfsi
= NFS_I(inode
);
721 struct nfs_open_context
*pos
, *ctx
= NULL
;
723 spin_lock(&inode
->i_lock
);
724 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
725 if (cred
!= NULL
&& pos
->cred
!= cred
)
727 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
729 ctx
= get_nfs_open_context(pos
);
732 spin_unlock(&inode
->i_lock
);
736 static void nfs_file_clear_open_context(struct file
*filp
)
738 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
739 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
742 filp
->private_data
= NULL
;
743 spin_lock(&inode
->i_lock
);
744 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
745 spin_unlock(&inode
->i_lock
);
746 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
751 * These allocate and release file read/write context information.
753 int nfs_open(struct inode
*inode
, struct file
*filp
)
755 struct nfs_open_context
*ctx
;
757 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
760 nfs_file_set_open_context(filp
, ctx
);
761 put_nfs_open_context(ctx
);
762 nfs_fscache_set_inode_cookie(inode
, filp
);
766 int nfs_release(struct inode
*inode
, struct file
*filp
)
768 nfs_file_clear_open_context(filp
);
773 * This function is called whenever some part of NFS notices that
774 * the cached attributes have to be refreshed.
777 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
779 int status
= -ESTALE
;
780 struct nfs_fattr
*fattr
= NULL
;
781 struct nfs_inode
*nfsi
= NFS_I(inode
);
783 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
784 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
786 if (is_bad_inode(inode
))
788 if (NFS_STALE(inode
))
792 fattr
= nfs_alloc_fattr();
796 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
797 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
);
799 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
801 (long long)NFS_FILEID(inode
), status
);
802 if (status
== -ESTALE
) {
803 nfs_zap_caches(inode
);
804 if (!S_ISDIR(inode
->i_mode
))
805 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
810 status
= nfs_refresh_inode(inode
, fattr
);
812 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
814 (long long)NFS_FILEID(inode
), status
);
818 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
819 nfs_zap_acl_cache(inode
);
821 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
823 (long long)NFS_FILEID(inode
));
826 nfs_free_fattr(fattr
);
830 int nfs_attribute_timeout(struct inode
*inode
)
832 struct nfs_inode
*nfsi
= NFS_I(inode
);
834 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
837 static int nfs_attribute_cache_expired(struct inode
*inode
)
839 if (nfs_have_delegated_attributes(inode
))
841 return nfs_attribute_timeout(inode
);
845 * nfs_revalidate_inode - Revalidate the inode attributes
846 * @server - pointer to nfs_server struct
847 * @inode - pointer to inode struct
849 * Updates inode attribute information by retrieving the data from the server.
851 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
853 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
854 && !nfs_attribute_cache_expired(inode
))
855 return NFS_STALE(inode
) ? -ESTALE
: 0;
856 return __nfs_revalidate_inode(server
, inode
);
858 EXPORT_SYMBOL_GPL(nfs_revalidate_inode
);
860 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
862 struct nfs_inode
*nfsi
= NFS_I(inode
);
864 if (mapping
->nrpages
!= 0) {
865 int ret
= invalidate_inode_pages2(mapping
);
869 spin_lock(&inode
->i_lock
);
870 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
871 if (S_ISDIR(inode
->i_mode
))
872 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
873 spin_unlock(&inode
->i_lock
);
874 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
875 nfs_fscache_reset_inode_cookie(inode
);
876 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
877 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
881 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
883 if (nfs_have_delegated_attributes(inode
))
885 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
886 || nfs_attribute_timeout(inode
)
891 * nfs_revalidate_mapping - Revalidate the pagecache
892 * @inode - pointer to host inode
893 * @mapping - pointer to mapping
895 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
897 struct nfs_inode
*nfsi
= NFS_I(inode
);
900 /* swapfiles are not supposed to be shared. */
901 if (IS_SWAPFILE(inode
))
904 if (nfs_mapping_need_revalidate_inode(inode
)) {
905 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
909 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
910 ret
= nfs_invalidate_mapping(inode
, mapping
);
915 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
917 struct nfs_inode
*nfsi
= NFS_I(inode
);
918 unsigned long ret
= 0;
920 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
921 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
922 && inode
->i_version
== fattr
->pre_change_attr
) {
923 inode
->i_version
= fattr
->change_attr
;
924 if (S_ISDIR(inode
->i_mode
))
925 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
926 ret
|= NFS_INO_INVALID_ATTR
;
928 /* If we have atomic WCC data, we may update some attributes */
929 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
930 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
931 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
932 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
933 ret
|= NFS_INO_INVALID_ATTR
;
936 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
937 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
938 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
939 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
940 if (S_ISDIR(inode
->i_mode
))
941 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
942 ret
|= NFS_INO_INVALID_ATTR
;
944 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
945 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
946 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
947 && nfsi
->npages
== 0) {
948 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
949 ret
|= NFS_INO_INVALID_ATTR
;
955 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
956 * @inode - pointer to inode
957 * @fattr - updated attributes
959 * Verifies the attribute cache. If we have just changed the attributes,
960 * so that fattr carries weak cache consistency data, then it may
961 * also update the ctime/mtime/change_attribute.
963 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
965 struct nfs_inode
*nfsi
= NFS_I(inode
);
966 loff_t cur_size
, new_isize
;
967 unsigned long invalid
= 0;
970 if (nfs_have_delegated_attributes(inode
))
972 /* Has the inode gone and changed behind our back? */
973 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
975 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
978 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
979 inode
->i_version
!= fattr
->change_attr
)
980 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
982 /* Verify a few of the more important attributes */
983 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
984 invalid
|= NFS_INO_INVALID_ATTR
;
986 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
987 cur_size
= i_size_read(inode
);
988 new_isize
= nfs_size_to_loff_t(fattr
->size
);
989 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
990 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
993 /* Have any file permissions changed? */
994 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
995 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
996 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && inode
->i_uid
!= fattr
->uid
)
997 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
998 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && inode
->i_gid
!= fattr
->gid
)
999 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1001 /* Has the link count changed? */
1002 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
1003 invalid
|= NFS_INO_INVALID_ATTR
;
1005 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1006 invalid
|= NFS_INO_INVALID_ATIME
;
1009 nfsi
->cache_validity
|= invalid
;
1011 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1015 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1017 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
1019 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1022 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1024 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1026 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1029 static atomic_long_t nfs_attr_generation_counter
;
1031 static unsigned long nfs_read_attr_generation_counter(void)
1033 return atomic_long_read(&nfs_attr_generation_counter
);
1036 unsigned long nfs_inc_attr_generation_counter(void)
1038 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1041 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1044 fattr
->time_start
= jiffies
;
1045 fattr
->gencount
= nfs_inc_attr_generation_counter();
1046 fattr
->owner_name
= NULL
;
1047 fattr
->group_name
= NULL
;
1049 EXPORT_SYMBOL_GPL(nfs_fattr_init
);
1051 struct nfs_fattr
*nfs_alloc_fattr(void)
1053 struct nfs_fattr
*fattr
;
1055 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1057 nfs_fattr_init(fattr
);
1060 EXPORT_SYMBOL_GPL(nfs_alloc_fattr
);
1062 struct nfs_fh
*nfs_alloc_fhandle(void)
1066 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1071 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle
);
1075 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1076 * in the same way that wireshark does
1080 * For debugging only.
1082 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1084 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1085 * not on the result */
1086 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1090 * _nfs_display_fhandle - display an NFS file handle on the console
1092 * @fh: file handle to display
1093 * @caption: display caption
1095 * For debugging only.
1097 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1101 if (fh
== NULL
|| fh
->size
== 0) {
1102 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1106 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1107 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1108 for (i
= 0; i
< fh
->size
; i
+= 16) {
1109 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1111 switch ((fh
->size
- i
- 1) >> 2) {
1113 printk(KERN_DEFAULT
" %08x\n",
1117 printk(KERN_DEFAULT
" %08x %08x\n",
1118 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1121 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1122 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1123 be32_to_cpup(pos
+ 2));
1126 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1127 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1128 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1135 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1136 * @inode - pointer to inode
1137 * @fattr - attributes
1139 * Attempt to divine whether or not an RPC call reply carrying stale
1140 * attributes got scheduled after another call carrying updated ones.
1142 * To do so, the function first assumes that a more recent ctime means
1143 * that the attributes in fattr are newer, however it also attempt to
1144 * catch the case where ctime either didn't change, or went backwards
1145 * (if someone reset the clock on the server) by looking at whether
1146 * or not this RPC call was started after the inode was last updated.
1147 * Note also the check for wraparound of 'attr_gencount'
1149 * The function returns 'true' if it thinks the attributes in 'fattr' are
1150 * more recent than the ones cached in the inode.
1153 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1155 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1157 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1158 nfs_ctime_need_update(inode
, fattr
) ||
1159 nfs_size_need_update(inode
, fattr
) ||
1160 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1163 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1165 if (nfs_inode_attrs_need_update(inode
, fattr
))
1166 return nfs_update_inode(inode
, fattr
);
1167 return nfs_check_inode_attributes(inode
, fattr
);
1171 * nfs_refresh_inode - try to update the inode attribute cache
1172 * @inode - pointer to inode
1173 * @fattr - updated attributes
1175 * Check that an RPC call that returned attributes has not overlapped with
1176 * other recent updates of the inode metadata, then decide whether it is
1177 * safe to do a full update of the inode attributes, or whether just to
1178 * call nfs_check_inode_attributes.
1180 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1184 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1186 spin_lock(&inode
->i_lock
);
1187 status
= nfs_refresh_inode_locked(inode
, fattr
);
1188 spin_unlock(&inode
->i_lock
);
1192 EXPORT_SYMBOL_GPL(nfs_refresh_inode
);
1194 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1196 struct nfs_inode
*nfsi
= NFS_I(inode
);
1198 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1199 if (S_ISDIR(inode
->i_mode
))
1200 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1201 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1203 return nfs_refresh_inode_locked(inode
, fattr
);
1207 * nfs_post_op_update_inode - 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.
1214 * NB: if the server didn't return any post op attributes, this
1215 * function will force the retrieval of attributes before the next
1216 * NFS request. Thus it should be used only for operations that
1217 * are expected to change one or more attributes, to avoid
1218 * unnecessary NFS requests and trips through nfs_update_inode().
1220 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1224 spin_lock(&inode
->i_lock
);
1225 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1226 spin_unlock(&inode
->i_lock
);
1229 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode
);
1232 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1233 * @inode - pointer to inode
1234 * @fattr - updated attributes
1236 * After an operation that has changed the inode metadata, mark the
1237 * attribute cache as being invalid, then try to update it. Fake up
1238 * weak cache consistency data, if none exist.
1240 * This function is mainly designed to be used by the ->write_done() functions.
1242 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1246 spin_lock(&inode
->i_lock
);
1247 /* Don't do a WCC update if these attributes are already stale */
1248 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1249 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1250 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1251 | NFS_ATTR_FATTR_PRESIZE
1252 | NFS_ATTR_FATTR_PREMTIME
1253 | NFS_ATTR_FATTR_PRECTIME
);
1256 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1257 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1258 fattr
->pre_change_attr
= inode
->i_version
;
1259 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1261 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1262 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1263 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1264 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1266 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1267 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1268 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1269 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1271 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1272 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1273 fattr
->pre_size
= i_size_read(inode
);
1274 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1277 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1278 spin_unlock(&inode
->i_lock
);
1281 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc
);
1284 * Many nfs protocol calls return the new file attributes after
1285 * an operation. Here we update the inode to reflect the state
1286 * of the server's inode.
1288 * This is a bit tricky because we have to make sure all dirty pages
1289 * have been sent off to the server before calling invalidate_inode_pages.
1290 * To make sure no other process adds more write requests while we try
1291 * our best to flush them, we make them sleep during the attribute refresh.
1293 * A very similar scenario holds for the dir cache.
1295 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1297 struct nfs_server
*server
;
1298 struct nfs_inode
*nfsi
= NFS_I(inode
);
1299 loff_t cur_isize
, new_isize
;
1300 unsigned long invalid
= 0;
1301 unsigned long now
= jiffies
;
1302 unsigned long save_cache_validity
;
1304 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1305 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1306 nfs_display_fhandle_hash(NFS_FH(inode
)),
1307 atomic_read(&inode
->i_count
), fattr
->valid
);
1309 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1310 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1311 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1312 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1313 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1314 (long long)fattr
->fileid
);
1319 * Make sure the inode's type hasn't changed.
1321 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1323 * Big trouble! The inode has become a different object.
1325 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1326 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1330 server
= NFS_SERVER(inode
);
1331 /* Update the fsid? */
1332 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1333 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1334 !IS_AUTOMOUNT(inode
))
1335 server
->fsid
= fattr
->fsid
;
1338 * Update the read time so we don't revalidate too often.
1340 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1342 save_cache_validity
= nfsi
->cache_validity
;
1343 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1344 | NFS_INO_INVALID_ATIME
1345 | NFS_INO_REVAL_FORCED
1346 | NFS_INO_REVAL_PAGECACHE
);
1348 /* Do atomic weak cache consistency updates */
1349 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1351 /* More cache consistency checks */
1352 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1353 if (inode
->i_version
!= fattr
->change_attr
) {
1354 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1355 inode
->i_sb
->s_id
, inode
->i_ino
);
1356 invalid
|= NFS_INO_INVALID_ATTR
1357 | NFS_INO_INVALID_DATA
1358 | NFS_INO_INVALID_ACCESS
1359 | NFS_INO_INVALID_ACL
1360 | NFS_INO_REVAL_PAGECACHE
;
1361 if (S_ISDIR(inode
->i_mode
))
1362 nfs_force_lookup_revalidate(inode
);
1363 inode
->i_version
= fattr
->change_attr
;
1365 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1366 invalid
|= save_cache_validity
;
1368 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1369 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1370 } else if (server
->caps
& NFS_CAP_MTIME
)
1371 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1372 | NFS_INO_REVAL_FORCED
);
1374 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1375 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1376 } else if (server
->caps
& NFS_CAP_CTIME
)
1377 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1378 | NFS_INO_REVAL_FORCED
);
1380 /* Check if our cached file size is stale */
1381 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1382 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1383 cur_isize
= i_size_read(inode
);
1384 if (new_isize
!= cur_isize
) {
1385 /* Do we perhaps have any outstanding writes, or has
1386 * the file grown beyond our last write? */
1387 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1388 new_isize
> cur_isize
) {
1389 i_size_write(inode
, new_isize
);
1390 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1392 dprintk("NFS: isize change on server for file %s/%ld "
1396 (long long)cur_isize
,
1397 (long long)new_isize
);
1400 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1401 | NFS_INO_REVAL_PAGECACHE
1402 | NFS_INO_REVAL_FORCED
);
1405 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1406 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1407 else if (server
->caps
& NFS_CAP_ATIME
)
1408 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1409 | NFS_INO_REVAL_FORCED
);
1411 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1412 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1413 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1414 newmode
|= fattr
->mode
& S_IALLUGO
;
1415 inode
->i_mode
= newmode
;
1416 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1418 } else if (server
->caps
& NFS_CAP_MODE
)
1419 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1420 | NFS_INO_INVALID_ACCESS
1421 | NFS_INO_INVALID_ACL
1422 | NFS_INO_REVAL_FORCED
);
1424 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1425 if (inode
->i_uid
!= fattr
->uid
) {
1426 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1427 inode
->i_uid
= fattr
->uid
;
1429 } else if (server
->caps
& NFS_CAP_OWNER
)
1430 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1431 | NFS_INO_INVALID_ACCESS
1432 | NFS_INO_INVALID_ACL
1433 | NFS_INO_REVAL_FORCED
);
1435 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1436 if (inode
->i_gid
!= fattr
->gid
) {
1437 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1438 inode
->i_gid
= fattr
->gid
;
1440 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1441 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1442 | NFS_INO_INVALID_ACCESS
1443 | NFS_INO_INVALID_ACL
1444 | NFS_INO_REVAL_FORCED
);
1446 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1447 if (inode
->i_nlink
!= fattr
->nlink
) {
1448 invalid
|= NFS_INO_INVALID_ATTR
;
1449 if (S_ISDIR(inode
->i_mode
))
1450 invalid
|= NFS_INO_INVALID_DATA
;
1451 set_nlink(inode
, fattr
->nlink
);
1453 } else if (server
->caps
& NFS_CAP_NLINK
)
1454 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1455 | NFS_INO_REVAL_FORCED
);
1457 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1459 * report the blocks in 512byte units
1461 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1463 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1464 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1466 /* Update attrtimeo value if we're out of the unstable period */
1467 if (invalid
& NFS_INO_INVALID_ATTR
) {
1468 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1469 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1470 nfsi
->attrtimeo_timestamp
= now
;
1471 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1473 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1474 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1475 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1476 nfsi
->attrtimeo_timestamp
= now
;
1479 invalid
&= ~NFS_INO_INVALID_ATTR
;
1480 /* Don't invalidate the data if we were to blame */
1481 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1482 || S_ISLNK(inode
->i_mode
)))
1483 invalid
&= ~NFS_INO_INVALID_DATA
;
1484 if (!NFS_PROTO(inode
)->have_delegation(inode
, FMODE_READ
) ||
1485 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1486 nfsi
->cache_validity
|= invalid
;
1491 * No need to worry about unhashing the dentry, as the
1492 * lookup validation will know that the inode is bad.
1493 * (But we fall through to invalidate the caches.)
1495 nfs_invalidate_inode(inode
);
1499 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1501 struct nfs_inode
*nfsi
;
1502 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1506 nfsi
->cache_validity
= 0UL;
1507 #ifdef CONFIG_NFS_V3_ACL
1508 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1509 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1511 #if IS_ENABLED(CONFIG_NFS_V4)
1512 nfsi
->nfs4_acl
= NULL
;
1513 #endif /* CONFIG_NFS_V4 */
1514 return &nfsi
->vfs_inode
;
1516 EXPORT_SYMBOL_GPL(nfs_alloc_inode
);
1518 static void nfs_i_callback(struct rcu_head
*head
)
1520 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1521 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1524 void nfs_destroy_inode(struct inode
*inode
)
1526 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1528 EXPORT_SYMBOL_GPL(nfs_destroy_inode
);
1530 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1532 #if IS_ENABLED(CONFIG_NFS_V4)
1533 INIT_LIST_HEAD(&nfsi
->open_states
);
1534 nfsi
->delegation
= NULL
;
1535 nfsi
->delegation_state
= 0;
1536 init_rwsem(&nfsi
->rwsem
);
1537 nfsi
->layout
= NULL
;
1541 static void init_once(void *foo
)
1543 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1545 inode_init_once(&nfsi
->vfs_inode
);
1546 INIT_LIST_HEAD(&nfsi
->open_files
);
1547 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1548 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1549 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1551 nfsi
->commit_info
.ncommit
= 0;
1552 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1553 atomic_set(&nfsi
->silly_count
, 1);
1554 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1555 init_waitqueue_head(&nfsi
->waitqueue
);
1556 nfs4_init_once(nfsi
);
1559 static int __init
nfs_init_inodecache(void)
1561 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1562 sizeof(struct nfs_inode
),
1563 0, (SLAB_RECLAIM_ACCOUNT
|
1566 if (nfs_inode_cachep
== NULL
)
1572 static void nfs_destroy_inodecache(void)
1575 * Make sure all delayed rcu free inodes are flushed before we
1579 kmem_cache_destroy(nfs_inode_cachep
);
1582 struct workqueue_struct
*nfsiod_workqueue
;
1583 EXPORT_SYMBOL_GPL(nfsiod_workqueue
);
1586 * start up the nfsiod workqueue
1588 static int nfsiod_start(void)
1590 struct workqueue_struct
*wq
;
1591 dprintk("RPC: creating workqueue nfsiod\n");
1592 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1595 nfsiod_workqueue
= wq
;
1600 * Destroy the nfsiod workqueue
1602 static void nfsiod_stop(void)
1604 struct workqueue_struct
*wq
;
1606 wq
= nfsiod_workqueue
;
1609 nfsiod_workqueue
= NULL
;
1610 destroy_workqueue(wq
);
1614 EXPORT_SYMBOL_GPL(nfs_net_id
);
1616 static int nfs_net_init(struct net
*net
)
1618 nfs_clients_init(net
);
1619 return nfs_dns_resolver_cache_init(net
);
1622 static void nfs_net_exit(struct net
*net
)
1624 nfs_dns_resolver_cache_destroy(net
);
1625 nfs_cleanup_cb_ident_idr(net
);
1628 static struct pernet_operations nfs_net_ops
= {
1629 .init
= nfs_net_init
,
1630 .exit
= nfs_net_exit
,
1632 .size
= sizeof(struct nfs_net
),
1638 static int __init
init_nfs_fs(void)
1642 err
= nfs_dns_resolver_init();
1646 err
= register_pernet_subsys(&nfs_net_ops
);
1650 err
= nfs_fscache_register();
1654 err
= nfsiod_start();
1658 err
= nfs_fs_proc_init();
1662 err
= nfs_init_nfspagecache();
1666 err
= nfs_init_inodecache();
1670 err
= nfs_init_readpagecache();
1674 err
= nfs_init_writepagecache();
1678 err
= nfs_init_directcache();
1682 #ifdef CONFIG_PROC_FS
1683 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1685 if ((err
= register_nfs_fs()) != 0)
1690 #ifdef CONFIG_PROC_FS
1691 rpc_proc_unregister(&init_net
, "nfs");
1693 nfs_destroy_directcache();
1695 nfs_destroy_writepagecache();
1697 nfs_destroy_readpagecache();
1699 nfs_destroy_inodecache();
1701 nfs_destroy_nfspagecache();
1707 nfs_fscache_unregister();
1709 unregister_pernet_subsys(&nfs_net_ops
);
1711 nfs_dns_resolver_destroy();
1716 static void __exit
exit_nfs_fs(void)
1718 nfs_destroy_directcache();
1719 nfs_destroy_writepagecache();
1720 nfs_destroy_readpagecache();
1721 nfs_destroy_inodecache();
1722 nfs_destroy_nfspagecache();
1723 nfs_fscache_unregister();
1724 unregister_pernet_subsys(&nfs_net_ops
);
1725 nfs_dns_resolver_destroy();
1726 #ifdef CONFIG_PROC_FS
1727 rpc_proc_unregister(&init_net
, "nfs");
1729 unregister_nfs_fs();
1734 /* Not quite true; I just maintain it */
1735 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1736 MODULE_LICENSE("GPL");
1737 module_param(enable_ino64
, bool, 0644);
1739 module_init(init_nfs_fs
)
1740 module_exit(exit_nfs_fs
)