4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/time.h>
20 #include <linux/kernel.h>
22 #include <linux/string.h>
23 #include <linux/stat.h>
24 #include <linux/errno.h>
25 #include <linux/unistd.h>
26 #include <linux/sunrpc/clnt.h>
27 #include <linux/sunrpc/stats.h>
28 #include <linux/sunrpc/metrics.h>
29 #include <linux/nfs_fs.h>
30 #include <linux/nfs_mount.h>
31 #include <linux/nfs4_mount.h>
32 #include <linux/lockd/bind.h>
33 #include <linux/seq_file.h>
34 #include <linux/mount.h>
35 #include <linux/nfs_idmap.h>
36 #include <linux/vfs.h>
37 #include <linux/inet.h>
38 #include <linux/nfs_xdr.h>
39 #include <linux/slab.h>
40 #include <linux/compat.h>
41 #include <linux/freezer.h>
42 #include <linux/crc32.h>
44 #include <asm/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);
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 /* Why so? Because we want revalidate for devices/FIFOs, and
290 * that's precisely what we have in nfs_file_inode_operations.
292 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
293 if (S_ISREG(inode
->i_mode
)) {
294 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
295 inode
->i_data
.a_ops
= &nfs_file_aops
;
296 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
297 } else if (S_ISDIR(inode
->i_mode
)) {
298 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
299 inode
->i_fop
= &nfs_dir_operations
;
300 inode
->i_data
.a_ops
= &nfs_dir_aops
;
301 /* Deal with crossing mountpoints */
302 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
303 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
304 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
305 inode
->i_op
= &nfs_referral_inode_operations
;
307 inode
->i_op
= &nfs_mountpoint_inode_operations
;
309 inode
->i_flags
|= S_AUTOMOUNT
;
311 } else if (S_ISLNK(inode
->i_mode
))
312 inode
->i_op
= &nfs_symlink_inode_operations
;
314 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
316 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
317 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
318 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
319 inode
->i_version
= 0;
325 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
329 nfsi
->read_cache_jiffies
= fattr
->time_start
;
330 nfsi
->attr_gencount
= fattr
->gencount
;
331 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
332 inode
->i_atime
= fattr
->atime
;
333 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
334 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
335 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
336 inode
->i_mtime
= fattr
->mtime
;
337 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
338 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
339 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
340 inode
->i_ctime
= fattr
->ctime
;
341 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
342 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
343 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
344 inode
->i_version
= fattr
->change_attr
;
345 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
346 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
347 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
348 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
350 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
351 | NFS_INO_REVAL_PAGECACHE
;
352 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
353 set_nlink(inode
, fattr
->nlink
);
354 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
355 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
356 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
357 inode
->i_uid
= fattr
->uid
;
358 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
359 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
360 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
361 inode
->i_gid
= fattr
->gid
;
362 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
363 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
364 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
365 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
366 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
368 * report the blocks in 512byte units
370 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
372 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
373 nfsi
->attrtimeo_timestamp
= now
;
374 nfsi
->access_cache
= RB_ROOT
;
376 nfs_fscache_init_inode_cookie(inode
);
378 unlock_new_inode(inode
);
380 nfs_refresh_inode(inode
, fattr
);
381 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
383 (long long)NFS_FILEID(inode
),
384 nfs_display_fhandle_hash(fh
),
385 atomic_read(&inode
->i_count
));
391 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
395 #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)
398 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
400 struct inode
*inode
= dentry
->d_inode
;
401 struct nfs_fattr
*fattr
;
404 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
406 /* skip mode change if it's just for clearing setuid/setgid */
407 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
408 attr
->ia_valid
&= ~ATTR_MODE
;
410 if (attr
->ia_valid
& ATTR_SIZE
) {
411 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
412 attr
->ia_valid
&= ~ATTR_SIZE
;
415 /* Optimization: if the end result is no change, don't RPC */
416 attr
->ia_valid
&= NFS_VALID_ATTRS
;
417 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
420 /* Write all dirty data */
421 if (S_ISREG(inode
->i_mode
))
424 fattr
= nfs_alloc_fattr();
428 * Return any delegations if we're going to change ACLs
430 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
431 nfs_inode_return_delegation(inode
);
432 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
434 nfs_refresh_inode(inode
, fattr
);
435 nfs_free_fattr(fattr
);
441 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
442 * @inode: inode of the file used
443 * @offset: file offset to start truncating
445 * This is a copy of the common vmtruncate, but with the locking
446 * corrected to take into account the fact that NFS requires
447 * inode->i_size to be updated under the inode->i_lock.
449 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
454 err
= inode_newsize_ok(inode
, offset
);
458 spin_lock(&inode
->i_lock
);
459 oldsize
= inode
->i_size
;
460 i_size_write(inode
, offset
);
461 spin_unlock(&inode
->i_lock
);
463 truncate_pagecache(inode
, oldsize
, offset
);
469 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
470 * @inode: pointer to struct inode
471 * @attr: pointer to struct iattr
473 * Note: we do this in the *proc.c in order to ensure that
474 * it works for things like exclusive creates too.
476 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
478 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
479 spin_lock(&inode
->i_lock
);
480 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
481 int mode
= attr
->ia_mode
& S_IALLUGO
;
482 mode
|= inode
->i_mode
& ~S_IALLUGO
;
483 inode
->i_mode
= mode
;
485 if ((attr
->ia_valid
& ATTR_UID
) != 0)
486 inode
->i_uid
= attr
->ia_uid
;
487 if ((attr
->ia_valid
& ATTR_GID
) != 0)
488 inode
->i_gid
= attr
->ia_gid
;
489 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
490 spin_unlock(&inode
->i_lock
);
492 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
493 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
494 nfs_vmtruncate(inode
, attr
->ia_size
);
498 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
500 struct inode
*inode
= dentry
->d_inode
;
501 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
504 /* Flush out writes to the server in order to update c/mtime. */
505 if (S_ISREG(inode
->i_mode
)) {
506 err
= filemap_write_and_wait(inode
->i_mapping
);
512 * We may force a getattr if the user cares about atime.
514 * Note that we only have to check the vfsmount flags here:
515 * - NFS always sets S_NOATIME by so checking it would give a
517 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
518 * no point in checking those.
520 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
521 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
525 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
527 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
529 generic_fillattr(inode
, stat
);
530 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
536 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
538 atomic_set(&l_ctx
->count
, 1);
539 l_ctx
->lockowner
= current
->files
;
540 l_ctx
->pid
= current
->tgid
;
541 INIT_LIST_HEAD(&l_ctx
->list
);
544 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
546 struct nfs_lock_context
*pos
;
548 list_for_each_entry(pos
, &ctx
->lock_context
.list
, list
) {
549 if (pos
->lockowner
!= current
->files
)
551 if (pos
->pid
!= current
->tgid
)
553 atomic_inc(&pos
->count
);
559 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
561 struct nfs_lock_context
*res
, *new = NULL
;
562 struct inode
*inode
= ctx
->dentry
->d_inode
;
564 spin_lock(&inode
->i_lock
);
565 res
= __nfs_find_lock_context(ctx
);
567 spin_unlock(&inode
->i_lock
);
568 new = kmalloc(sizeof(*new), GFP_KERNEL
);
571 nfs_init_lock_context(new);
572 spin_lock(&inode
->i_lock
);
573 res
= __nfs_find_lock_context(ctx
);
575 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
576 new->open_context
= ctx
;
581 spin_unlock(&inode
->i_lock
);
586 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
588 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
589 struct inode
*inode
= ctx
->dentry
->d_inode
;
591 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
593 list_del(&l_ctx
->list
);
594 spin_unlock(&inode
->i_lock
);
599 * nfs_close_context - Common close_context() routine NFSv2/v3
600 * @ctx: pointer to context
601 * @is_sync: is this a synchronous close
603 * always ensure that the attributes are up to date if we're mounted
604 * with close-to-open semantics
606 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
609 struct nfs_server
*server
;
611 if (!(ctx
->mode
& FMODE_WRITE
))
615 inode
= ctx
->dentry
->d_inode
;
616 if (!list_empty(&NFS_I(inode
)->open_files
))
618 server
= NFS_SERVER(inode
);
619 if (server
->flags
& NFS_MOUNT_NOCTO
)
621 nfs_revalidate_inode(server
, inode
);
624 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
626 struct nfs_open_context
*ctx
;
627 struct rpc_cred
*cred
= rpc_lookup_cred();
629 return ERR_CAST(cred
);
631 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
634 return ERR_PTR(-ENOMEM
);
636 nfs_sb_active(dentry
->d_sb
);
637 ctx
->dentry
= dget(dentry
);
643 nfs_init_lock_context(&ctx
->lock_context
);
644 ctx
->lock_context
.open_context
= ctx
;
645 INIT_LIST_HEAD(&ctx
->list
);
646 ctx
->mdsthreshold
= NULL
;
650 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
653 atomic_inc(&ctx
->lock_context
.count
);
657 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
659 struct inode
*inode
= ctx
->dentry
->d_inode
;
660 struct super_block
*sb
= ctx
->dentry
->d_sb
;
662 if (!list_empty(&ctx
->list
)) {
663 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
665 list_del(&ctx
->list
);
666 spin_unlock(&inode
->i_lock
);
667 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
670 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
671 if (ctx
->cred
!= NULL
)
672 put_rpccred(ctx
->cred
);
675 kfree(ctx
->mdsthreshold
);
679 void put_nfs_open_context(struct nfs_open_context
*ctx
)
681 __put_nfs_open_context(ctx
, 0);
685 * Ensure that mmap has a recent RPC credential for use when writing out
688 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
690 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
691 struct nfs_inode
*nfsi
= NFS_I(inode
);
693 filp
->private_data
= get_nfs_open_context(ctx
);
694 spin_lock(&inode
->i_lock
);
695 list_add(&ctx
->list
, &nfsi
->open_files
);
696 spin_unlock(&inode
->i_lock
);
700 * Given an inode, search for an open context with the desired characteristics
702 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
704 struct nfs_inode
*nfsi
= NFS_I(inode
);
705 struct nfs_open_context
*pos
, *ctx
= NULL
;
707 spin_lock(&inode
->i_lock
);
708 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
709 if (cred
!= NULL
&& pos
->cred
!= cred
)
711 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
713 ctx
= get_nfs_open_context(pos
);
716 spin_unlock(&inode
->i_lock
);
720 static void nfs_file_clear_open_context(struct file
*filp
)
722 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
723 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
726 filp
->private_data
= NULL
;
727 spin_lock(&inode
->i_lock
);
728 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
729 spin_unlock(&inode
->i_lock
);
730 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
735 * These allocate and release file read/write context information.
737 int nfs_open(struct inode
*inode
, struct file
*filp
)
739 struct nfs_open_context
*ctx
;
741 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
744 nfs_file_set_open_context(filp
, ctx
);
745 put_nfs_open_context(ctx
);
746 nfs_fscache_set_inode_cookie(inode
, filp
);
750 int nfs_release(struct inode
*inode
, struct file
*filp
)
752 nfs_file_clear_open_context(filp
);
757 * This function is called whenever some part of NFS notices that
758 * the cached attributes have to be refreshed.
761 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
763 int status
= -ESTALE
;
764 struct nfs_fattr
*fattr
= NULL
;
765 struct nfs_inode
*nfsi
= NFS_I(inode
);
767 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
768 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
770 if (is_bad_inode(inode
))
772 if (NFS_STALE(inode
))
776 fattr
= nfs_alloc_fattr();
780 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
781 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
);
783 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
785 (long long)NFS_FILEID(inode
), status
);
786 if (status
== -ESTALE
) {
787 nfs_zap_caches(inode
);
788 if (!S_ISDIR(inode
->i_mode
))
789 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
794 status
= nfs_refresh_inode(inode
, fattr
);
796 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
798 (long long)NFS_FILEID(inode
), status
);
802 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
803 nfs_zap_acl_cache(inode
);
805 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
807 (long long)NFS_FILEID(inode
));
810 nfs_free_fattr(fattr
);
814 int nfs_attribute_timeout(struct inode
*inode
)
816 struct nfs_inode
*nfsi
= NFS_I(inode
);
818 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
821 static int nfs_attribute_cache_expired(struct inode
*inode
)
823 if (nfs_have_delegated_attributes(inode
))
825 return nfs_attribute_timeout(inode
);
829 * nfs_revalidate_inode - Revalidate the inode attributes
830 * @server - pointer to nfs_server struct
831 * @inode - pointer to inode struct
833 * Updates inode attribute information by retrieving the data from the server.
835 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
837 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
838 && !nfs_attribute_cache_expired(inode
))
839 return NFS_STALE(inode
) ? -ESTALE
: 0;
840 return __nfs_revalidate_inode(server
, inode
);
843 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
845 struct nfs_inode
*nfsi
= NFS_I(inode
);
847 if (mapping
->nrpages
!= 0) {
848 int ret
= invalidate_inode_pages2(mapping
);
852 spin_lock(&inode
->i_lock
);
853 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
854 if (S_ISDIR(inode
->i_mode
))
855 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
856 spin_unlock(&inode
->i_lock
);
857 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
858 nfs_fscache_reset_inode_cookie(inode
);
859 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
860 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
864 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
866 if (nfs_have_delegated_attributes(inode
))
868 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
869 || nfs_attribute_timeout(inode
)
874 * nfs_revalidate_mapping - Revalidate the pagecache
875 * @inode - pointer to host inode
876 * @mapping - pointer to mapping
878 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
880 struct nfs_inode
*nfsi
= NFS_I(inode
);
883 if (nfs_mapping_need_revalidate_inode(inode
)) {
884 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
888 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
889 ret
= nfs_invalidate_mapping(inode
, mapping
);
894 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
896 struct nfs_inode
*nfsi
= NFS_I(inode
);
897 unsigned long ret
= 0;
899 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
900 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
901 && inode
->i_version
== fattr
->pre_change_attr
) {
902 inode
->i_version
= fattr
->change_attr
;
903 if (S_ISDIR(inode
->i_mode
))
904 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
905 ret
|= NFS_INO_INVALID_ATTR
;
907 /* If we have atomic WCC data, we may update some attributes */
908 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
909 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
910 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
911 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
912 ret
|= NFS_INO_INVALID_ATTR
;
915 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
916 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
917 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
918 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
919 if (S_ISDIR(inode
->i_mode
))
920 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
921 ret
|= NFS_INO_INVALID_ATTR
;
923 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
924 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
925 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
926 && nfsi
->npages
== 0) {
927 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
928 ret
|= NFS_INO_INVALID_ATTR
;
934 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
935 * @inode - pointer to inode
936 * @fattr - updated attributes
938 * Verifies the attribute cache. If we have just changed the attributes,
939 * so that fattr carries weak cache consistency data, then it may
940 * also update the ctime/mtime/change_attribute.
942 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
944 struct nfs_inode
*nfsi
= NFS_I(inode
);
945 loff_t cur_size
, new_isize
;
946 unsigned long invalid
= 0;
949 if (nfs_have_delegated_attributes(inode
))
951 /* Has the inode gone and changed behind our back? */
952 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
954 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
957 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
958 inode
->i_version
!= fattr
->change_attr
)
959 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
961 /* Verify a few of the more important attributes */
962 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
963 invalid
|= NFS_INO_INVALID_ATTR
;
965 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
966 cur_size
= i_size_read(inode
);
967 new_isize
= nfs_size_to_loff_t(fattr
->size
);
968 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
969 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
972 /* Have any file permissions changed? */
973 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
974 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
975 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && inode
->i_uid
!= fattr
->uid
)
976 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
977 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && inode
->i_gid
!= fattr
->gid
)
978 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
980 /* Has the link count changed? */
981 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
982 invalid
|= NFS_INO_INVALID_ATTR
;
984 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
985 invalid
|= NFS_INO_INVALID_ATIME
;
988 nfsi
->cache_validity
|= invalid
;
990 nfsi
->read_cache_jiffies
= fattr
->time_start
;
994 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
996 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
998 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1001 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1003 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1005 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1008 static atomic_long_t nfs_attr_generation_counter
;
1010 static unsigned long nfs_read_attr_generation_counter(void)
1012 return atomic_long_read(&nfs_attr_generation_counter
);
1015 unsigned long nfs_inc_attr_generation_counter(void)
1017 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1020 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1023 fattr
->time_start
= jiffies
;
1024 fattr
->gencount
= nfs_inc_attr_generation_counter();
1025 fattr
->owner_name
= NULL
;
1026 fattr
->group_name
= NULL
;
1029 struct nfs_fattr
*nfs_alloc_fattr(void)
1031 struct nfs_fattr
*fattr
;
1033 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1035 nfs_fattr_init(fattr
);
1039 struct nfs_fh
*nfs_alloc_fhandle(void)
1043 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1051 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1052 * in the same way that wireshark does
1056 * For debugging only.
1058 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1060 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1061 * not on the result */
1062 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1066 * _nfs_display_fhandle - display an NFS file handle on the console
1068 * @fh: file handle to display
1069 * @caption: display caption
1071 * For debugging only.
1073 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1077 if (fh
== NULL
|| fh
->size
== 0) {
1078 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1082 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1083 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1084 for (i
= 0; i
< fh
->size
; i
+= 16) {
1085 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1087 switch ((fh
->size
- i
- 1) >> 2) {
1089 printk(KERN_DEFAULT
" %08x\n",
1093 printk(KERN_DEFAULT
" %08x %08x\n",
1094 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1097 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1098 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1099 be32_to_cpup(pos
+ 2));
1102 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1103 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1104 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1111 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1112 * @inode - pointer to inode
1113 * @fattr - attributes
1115 * Attempt to divine whether or not an RPC call reply carrying stale
1116 * attributes got scheduled after another call carrying updated ones.
1118 * To do so, the function first assumes that a more recent ctime means
1119 * that the attributes in fattr are newer, however it also attempt to
1120 * catch the case where ctime either didn't change, or went backwards
1121 * (if someone reset the clock on the server) by looking at whether
1122 * or not this RPC call was started after the inode was last updated.
1123 * Note also the check for wraparound of 'attr_gencount'
1125 * The function returns 'true' if it thinks the attributes in 'fattr' are
1126 * more recent than the ones cached in the inode.
1129 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1131 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1133 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1134 nfs_ctime_need_update(inode
, fattr
) ||
1135 nfs_size_need_update(inode
, fattr
) ||
1136 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1139 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1141 if (nfs_inode_attrs_need_update(inode
, fattr
))
1142 return nfs_update_inode(inode
, fattr
);
1143 return nfs_check_inode_attributes(inode
, fattr
);
1147 * nfs_refresh_inode - try to update the inode attribute cache
1148 * @inode - pointer to inode
1149 * @fattr - updated attributes
1151 * Check that an RPC call that returned attributes has not overlapped with
1152 * other recent updates of the inode metadata, then decide whether it is
1153 * safe to do a full update of the inode attributes, or whether just to
1154 * call nfs_check_inode_attributes.
1156 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1160 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1162 spin_lock(&inode
->i_lock
);
1163 status
= nfs_refresh_inode_locked(inode
, fattr
);
1164 spin_unlock(&inode
->i_lock
);
1169 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1171 struct nfs_inode
*nfsi
= NFS_I(inode
);
1173 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1174 if (S_ISDIR(inode
->i_mode
))
1175 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1176 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1178 return nfs_refresh_inode_locked(inode
, fattr
);
1182 * nfs_post_op_update_inode - try to update the inode attribute cache
1183 * @inode - pointer to inode
1184 * @fattr - updated attributes
1186 * After an operation that has changed the inode metadata, mark the
1187 * attribute cache as being invalid, then try to update it.
1189 * NB: if the server didn't return any post op attributes, this
1190 * function will force the retrieval of attributes before the next
1191 * NFS request. Thus it should be used only for operations that
1192 * are expected to change one or more attributes, to avoid
1193 * unnecessary NFS requests and trips through nfs_update_inode().
1195 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1199 spin_lock(&inode
->i_lock
);
1200 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1201 spin_unlock(&inode
->i_lock
);
1206 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1207 * @inode - pointer to inode
1208 * @fattr - updated attributes
1210 * After an operation that has changed the inode metadata, mark the
1211 * attribute cache as being invalid, then try to update it. Fake up
1212 * weak cache consistency data, if none exist.
1214 * This function is mainly designed to be used by the ->write_done() functions.
1216 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1220 spin_lock(&inode
->i_lock
);
1221 /* Don't do a WCC update if these attributes are already stale */
1222 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1223 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1224 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1225 | NFS_ATTR_FATTR_PRESIZE
1226 | NFS_ATTR_FATTR_PREMTIME
1227 | NFS_ATTR_FATTR_PRECTIME
);
1230 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1231 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1232 fattr
->pre_change_attr
= inode
->i_version
;
1233 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1235 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1236 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1237 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1238 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1240 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1241 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1242 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1243 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1245 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1246 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1247 fattr
->pre_size
= i_size_read(inode
);
1248 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1251 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1252 spin_unlock(&inode
->i_lock
);
1257 * Many nfs protocol calls return the new file attributes after
1258 * an operation. Here we update the inode to reflect the state
1259 * of the server's inode.
1261 * This is a bit tricky because we have to make sure all dirty pages
1262 * have been sent off to the server before calling invalidate_inode_pages.
1263 * To make sure no other process adds more write requests while we try
1264 * our best to flush them, we make them sleep during the attribute refresh.
1266 * A very similar scenario holds for the dir cache.
1268 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1270 struct nfs_server
*server
;
1271 struct nfs_inode
*nfsi
= NFS_I(inode
);
1272 loff_t cur_isize
, new_isize
;
1273 unsigned long invalid
= 0;
1274 unsigned long now
= jiffies
;
1275 unsigned long save_cache_validity
;
1277 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1278 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1279 nfs_display_fhandle_hash(NFS_FH(inode
)),
1280 atomic_read(&inode
->i_count
), fattr
->valid
);
1282 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1283 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1284 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1285 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1286 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1287 (long long)fattr
->fileid
);
1292 * Make sure the inode's type hasn't changed.
1294 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1296 * Big trouble! The inode has become a different object.
1298 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1299 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1303 server
= NFS_SERVER(inode
);
1304 /* Update the fsid? */
1305 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1306 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1307 !IS_AUTOMOUNT(inode
))
1308 server
->fsid
= fattr
->fsid
;
1311 * Update the read time so we don't revalidate too often.
1313 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1315 save_cache_validity
= nfsi
->cache_validity
;
1316 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1317 | NFS_INO_INVALID_ATIME
1318 | NFS_INO_REVAL_FORCED
1319 | NFS_INO_REVAL_PAGECACHE
);
1321 /* Do atomic weak cache consistency updates */
1322 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1324 /* More cache consistency checks */
1325 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1326 if (inode
->i_version
!= fattr
->change_attr
) {
1327 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1328 inode
->i_sb
->s_id
, inode
->i_ino
);
1329 invalid
|= NFS_INO_INVALID_ATTR
1330 | NFS_INO_INVALID_DATA
1331 | NFS_INO_INVALID_ACCESS
1332 | NFS_INO_INVALID_ACL
1333 | NFS_INO_REVAL_PAGECACHE
;
1334 if (S_ISDIR(inode
->i_mode
))
1335 nfs_force_lookup_revalidate(inode
);
1336 inode
->i_version
= fattr
->change_attr
;
1338 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1339 invalid
|= save_cache_validity
;
1341 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1342 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1343 } else if (server
->caps
& NFS_CAP_MTIME
)
1344 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1345 | NFS_INO_REVAL_FORCED
);
1347 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1348 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1349 } else if (server
->caps
& NFS_CAP_CTIME
)
1350 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1351 | NFS_INO_REVAL_FORCED
);
1353 /* Check if our cached file size is stale */
1354 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1355 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1356 cur_isize
= i_size_read(inode
);
1357 if (new_isize
!= cur_isize
) {
1358 /* Do we perhaps have any outstanding writes, or has
1359 * the file grown beyond our last write? */
1360 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1361 new_isize
> cur_isize
) {
1362 i_size_write(inode
, new_isize
);
1363 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1365 dprintk("NFS: isize change on server for file %s/%ld "
1369 (long long)cur_isize
,
1370 (long long)new_isize
);
1373 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1374 | NFS_INO_REVAL_PAGECACHE
1375 | NFS_INO_REVAL_FORCED
);
1378 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1379 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1380 else if (server
->caps
& NFS_CAP_ATIME
)
1381 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1382 | NFS_INO_REVAL_FORCED
);
1384 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1385 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1386 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1387 newmode
|= fattr
->mode
& S_IALLUGO
;
1388 inode
->i_mode
= newmode
;
1389 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1391 } else if (server
->caps
& NFS_CAP_MODE
)
1392 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1393 | NFS_INO_INVALID_ACCESS
1394 | NFS_INO_INVALID_ACL
1395 | NFS_INO_REVAL_FORCED
);
1397 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1398 if (inode
->i_uid
!= fattr
->uid
) {
1399 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1400 inode
->i_uid
= fattr
->uid
;
1402 } else if (server
->caps
& NFS_CAP_OWNER
)
1403 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1404 | NFS_INO_INVALID_ACCESS
1405 | NFS_INO_INVALID_ACL
1406 | NFS_INO_REVAL_FORCED
);
1408 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1409 if (inode
->i_gid
!= fattr
->gid
) {
1410 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1411 inode
->i_gid
= fattr
->gid
;
1413 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1414 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1415 | NFS_INO_INVALID_ACCESS
1416 | NFS_INO_INVALID_ACL
1417 | NFS_INO_REVAL_FORCED
);
1419 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1420 if (inode
->i_nlink
!= fattr
->nlink
) {
1421 invalid
|= NFS_INO_INVALID_ATTR
;
1422 if (S_ISDIR(inode
->i_mode
))
1423 invalid
|= NFS_INO_INVALID_DATA
;
1424 set_nlink(inode
, fattr
->nlink
);
1426 } else if (server
->caps
& NFS_CAP_NLINK
)
1427 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1428 | NFS_INO_REVAL_FORCED
);
1430 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1432 * report the blocks in 512byte units
1434 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1436 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1437 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1439 /* Update attrtimeo value if we're out of the unstable period */
1440 if (invalid
& NFS_INO_INVALID_ATTR
) {
1441 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1442 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1443 nfsi
->attrtimeo_timestamp
= now
;
1444 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1446 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1447 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1448 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1449 nfsi
->attrtimeo_timestamp
= now
;
1452 invalid
&= ~NFS_INO_INVALID_ATTR
;
1453 /* Don't invalidate the data if we were to blame */
1454 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1455 || S_ISLNK(inode
->i_mode
)))
1456 invalid
&= ~NFS_INO_INVALID_DATA
;
1457 if (!nfs_have_delegation(inode
, FMODE_READ
) ||
1458 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1459 nfsi
->cache_validity
|= invalid
;
1464 * No need to worry about unhashing the dentry, as the
1465 * lookup validation will know that the inode is bad.
1466 * (But we fall through to invalidate the caches.)
1468 nfs_invalidate_inode(inode
);
1473 #ifdef CONFIG_NFS_V4
1476 * Clean out any remaining NFSv4 state that might be left over due
1477 * to open() calls that passed nfs_atomic_lookup, but failed to call
1480 void nfs4_evict_inode(struct inode
*inode
)
1482 truncate_inode_pages(&inode
->i_data
, 0);
1484 pnfs_return_layout(inode
);
1485 pnfs_destroy_layout(NFS_I(inode
));
1486 /* If we are holding a delegation, return it! */
1487 nfs_inode_return_delegation_noreclaim(inode
);
1488 /* First call standard NFS clear_inode() code */
1489 nfs_clear_inode(inode
);
1493 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1495 struct nfs_inode
*nfsi
;
1496 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1500 nfsi
->cache_validity
= 0UL;
1501 #ifdef CONFIG_NFS_V3_ACL
1502 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1503 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1505 #ifdef CONFIG_NFS_V4
1506 nfsi
->nfs4_acl
= NULL
;
1507 #endif /* CONFIG_NFS_V4 */
1508 return &nfsi
->vfs_inode
;
1511 static void nfs_i_callback(struct rcu_head
*head
)
1513 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1514 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1517 void nfs_destroy_inode(struct inode
*inode
)
1519 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1522 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1524 #ifdef CONFIG_NFS_V4
1525 INIT_LIST_HEAD(&nfsi
->open_states
);
1526 nfsi
->delegation
= NULL
;
1527 nfsi
->delegation_state
= 0;
1528 init_rwsem(&nfsi
->rwsem
);
1529 nfsi
->layout
= NULL
;
1530 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1534 static void init_once(void *foo
)
1536 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1538 inode_init_once(&nfsi
->vfs_inode
);
1539 INIT_LIST_HEAD(&nfsi
->open_files
);
1540 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1541 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1542 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1544 nfsi
->commit_info
.ncommit
= 0;
1545 atomic_set(&nfsi
->silly_count
, 1);
1546 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1547 init_waitqueue_head(&nfsi
->waitqueue
);
1548 nfs4_init_once(nfsi
);
1551 static int __init
nfs_init_inodecache(void)
1553 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1554 sizeof(struct nfs_inode
),
1555 0, (SLAB_RECLAIM_ACCOUNT
|
1558 if (nfs_inode_cachep
== NULL
)
1564 static void nfs_destroy_inodecache(void)
1566 kmem_cache_destroy(nfs_inode_cachep
);
1569 struct workqueue_struct
*nfsiod_workqueue
;
1572 * start up the nfsiod workqueue
1574 static int nfsiod_start(void)
1576 struct workqueue_struct
*wq
;
1577 dprintk("RPC: creating workqueue nfsiod\n");
1578 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1581 nfsiod_workqueue
= wq
;
1586 * Destroy the nfsiod workqueue
1588 static void nfsiod_stop(void)
1590 struct workqueue_struct
*wq
;
1592 wq
= nfsiod_workqueue
;
1595 nfsiod_workqueue
= NULL
;
1596 destroy_workqueue(wq
);
1600 EXPORT_SYMBOL_GPL(nfs_net_id
);
1602 static int nfs_net_init(struct net
*net
)
1604 nfs_clients_init(net
);
1605 return nfs_dns_resolver_cache_init(net
);
1608 static void nfs_net_exit(struct net
*net
)
1610 nfs_dns_resolver_cache_destroy(net
);
1611 nfs_cleanup_cb_ident_idr(net
);
1614 static struct pernet_operations nfs_net_ops
= {
1615 .init
= nfs_net_init
,
1616 .exit
= nfs_net_exit
,
1618 .size
= sizeof(struct nfs_net
),
1624 static int __init
init_nfs_fs(void)
1628 err
= nfs_idmap_init();
1632 err
= nfs_dns_resolver_init();
1636 err
= register_pernet_subsys(&nfs_net_ops
);
1640 err
= nfs_fscache_register();
1644 err
= nfsiod_start();
1648 err
= nfs_fs_proc_init();
1652 err
= nfs_init_nfspagecache();
1656 err
= nfs_init_inodecache();
1660 err
= nfs_init_readpagecache();
1664 err
= nfs_init_writepagecache();
1668 err
= nfs_init_directcache();
1672 #ifdef CONFIG_PROC_FS
1673 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1675 if ((err
= register_nfs_fs()) != 0)
1679 #ifdef CONFIG_PROC_FS
1680 rpc_proc_unregister(&init_net
, "nfs");
1682 nfs_destroy_directcache();
1684 nfs_destroy_writepagecache();
1686 nfs_destroy_readpagecache();
1688 nfs_destroy_inodecache();
1690 nfs_destroy_nfspagecache();
1696 nfs_fscache_unregister();
1698 unregister_pernet_subsys(&nfs_net_ops
);
1700 nfs_dns_resolver_destroy();
1707 static void __exit
exit_nfs_fs(void)
1709 nfs_destroy_directcache();
1710 nfs_destroy_writepagecache();
1711 nfs_destroy_readpagecache();
1712 nfs_destroy_inodecache();
1713 nfs_destroy_nfspagecache();
1714 nfs_fscache_unregister();
1715 unregister_pernet_subsys(&nfs_net_ops
);
1716 nfs_dns_resolver_destroy();
1718 #ifdef CONFIG_PROC_FS
1719 rpc_proc_unregister(&init_net
, "nfs");
1721 unregister_nfs_fs();
1726 /* Not quite true; I just maintain it */
1727 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1728 MODULE_LICENSE("GPL");
1729 module_param(enable_ino64
, bool, 0644);
1731 module_init(init_nfs_fs
)
1732 module_exit(exit_nfs_fs
)