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"
55 #define NFSDBG_FACILITY NFSDBG_VFS
57 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
59 /* Default is to see 64-bit inode numbers */
60 static bool enable_ino64
= NFS_64_BIT_INODE_NUMBERS_ENABLED
;
62 static void nfs_invalidate_inode(struct inode
*);
63 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
65 static struct kmem_cache
* nfs_inode_cachep
;
67 static inline unsigned long
68 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
70 return nfs_fileid_to_ino_t(fattr
->fileid
);
74 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
75 * @word: long word containing the bit lock
77 int nfs_wait_bit_killable(void *word
)
79 if (fatal_signal_pending(current
))
81 freezable_schedule_unsafe();
84 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable
);
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 int nfs_drop_inode(struct inode
*inode
)
111 return NFS_STALE(inode
) || generic_drop_inode(inode
);
113 EXPORT_SYMBOL_GPL(nfs_drop_inode
);
115 void nfs_clear_inode(struct inode
*inode
)
118 * The following should never happen...
120 WARN_ON_ONCE(nfs_have_writebacks(inode
));
121 WARN_ON_ONCE(!list_empty(&NFS_I(inode
)->open_files
));
122 nfs_zap_acl_cache(inode
);
123 nfs_access_zap_cache(inode
);
124 nfs_fscache_release_inode_cookie(inode
);
126 EXPORT_SYMBOL_GPL(nfs_clear_inode
);
128 void nfs_evict_inode(struct inode
*inode
)
130 truncate_inode_pages(&inode
->i_data
, 0);
132 nfs_clear_inode(inode
);
136 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
138 int nfs_sync_mapping(struct address_space
*mapping
)
142 if (mapping
->nrpages
!= 0) {
143 unmap_mapping_range(mapping
, 0, 0, 0);
144 ret
= nfs_wb_all(mapping
->host
);
150 * Invalidate the local caches
152 static void nfs_zap_caches_locked(struct inode
*inode
)
154 struct nfs_inode
*nfsi
= NFS_I(inode
);
155 int mode
= inode
->i_mode
;
157 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
159 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
160 nfsi
->attrtimeo_timestamp
= jiffies
;
162 memset(NFS_I(inode
)->cookieverf
, 0, sizeof(NFS_I(inode
)->cookieverf
));
163 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
)) {
164 nfs_fscache_invalidate(inode
);
165 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
166 | NFS_INO_INVALID_LABEL
167 | NFS_INO_INVALID_DATA
168 | NFS_INO_INVALID_ACCESS
169 | NFS_INO_INVALID_ACL
170 | NFS_INO_REVAL_PAGECACHE
;
172 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
173 | NFS_INO_INVALID_LABEL
174 | NFS_INO_INVALID_ACCESS
175 | NFS_INO_INVALID_ACL
176 | NFS_INO_REVAL_PAGECACHE
;
179 void nfs_zap_caches(struct inode
*inode
)
181 spin_lock(&inode
->i_lock
);
182 nfs_zap_caches_locked(inode
);
183 spin_unlock(&inode
->i_lock
);
186 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
188 if (mapping
->nrpages
!= 0) {
189 spin_lock(&inode
->i_lock
);
190 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
191 nfs_fscache_invalidate(inode
);
192 spin_unlock(&inode
->i_lock
);
196 void nfs_zap_acl_cache(struct inode
*inode
)
198 void (*clear_acl_cache
)(struct inode
*);
200 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
201 if (clear_acl_cache
!= NULL
)
202 clear_acl_cache(inode
);
203 spin_lock(&inode
->i_lock
);
204 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
205 spin_unlock(&inode
->i_lock
);
207 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache
);
209 void nfs_invalidate_atime(struct inode
*inode
)
211 spin_lock(&inode
->i_lock
);
212 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
213 spin_unlock(&inode
->i_lock
);
215 EXPORT_SYMBOL_GPL(nfs_invalidate_atime
);
218 * Invalidate, but do not unhash, the inode.
219 * NB: must be called with inode->i_lock held!
221 static void nfs_invalidate_inode(struct inode
*inode
)
223 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
224 nfs_zap_caches_locked(inode
);
227 struct nfs_find_desc
{
229 struct nfs_fattr
*fattr
;
233 * In NFSv3 we can have 64bit inode numbers. In order to support
234 * this, and re-exported directories (also seen in NFSv2)
235 * we are forced to allow 2 different inodes to have the same
239 nfs_find_actor(struct inode
*inode
, void *opaque
)
241 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
242 struct nfs_fh
*fh
= desc
->fh
;
243 struct nfs_fattr
*fattr
= desc
->fattr
;
245 if (NFS_FILEID(inode
) != fattr
->fileid
)
247 if ((S_IFMT
& inode
->i_mode
) != (S_IFMT
& fattr
->mode
))
249 if (nfs_compare_fh(NFS_FH(inode
), fh
))
251 if (is_bad_inode(inode
) || NFS_STALE(inode
))
257 nfs_init_locked(struct inode
*inode
, void *opaque
)
259 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
260 struct nfs_fattr
*fattr
= desc
->fattr
;
262 set_nfs_fileid(inode
, fattr
->fileid
);
263 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
267 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
268 void nfs_setsecurity(struct inode
*inode
, struct nfs_fattr
*fattr
,
269 struct nfs4_label
*label
)
276 if (nfs_server_capable(inode
, NFS_CAP_SECURITY_LABEL
) == 0)
279 if (NFS_SERVER(inode
)->nfs_client
->cl_minorversion
< 2)
282 if ((fattr
->valid
& NFS_ATTR_FATTR_V4_SECURITY_LABEL
) && inode
->i_security
) {
283 error
= security_inode_notifysecctx(inode
, label
->label
,
286 printk(KERN_ERR
"%s() %s %d "
287 "security_inode_notifysecctx() %d\n",
289 (char *)label
->label
,
294 struct nfs4_label
*nfs4_label_alloc(struct nfs_server
*server
, gfp_t flags
)
296 struct nfs4_label
*label
= NULL
;
297 int minor_version
= server
->nfs_client
->cl_minorversion
;
299 if (minor_version
< 2)
302 if (!(server
->caps
& NFS_CAP_SECURITY_LABEL
))
305 label
= kzalloc(sizeof(struct nfs4_label
), flags
);
307 return ERR_PTR(-ENOMEM
);
309 label
->label
= kzalloc(NFS4_MAXLABELLEN
, flags
);
310 if (label
->label
== NULL
) {
312 return ERR_PTR(-ENOMEM
);
314 label
->len
= NFS4_MAXLABELLEN
;
318 EXPORT_SYMBOL_GPL(nfs4_label_alloc
);
320 void inline nfs_setsecurity(struct inode
*inode
, struct nfs_fattr
*fattr
,
321 struct nfs4_label
*label
)
325 EXPORT_SYMBOL_GPL(nfs_setsecurity
);
328 * This is our front-end to iget that looks up inodes by file handle
329 * instead of inode number.
332 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
, struct nfs4_label
*label
)
334 struct nfs_find_desc desc
= {
338 struct inode
*inode
= ERR_PTR(-ENOENT
);
341 nfs_attr_check_mountpoint(sb
, fattr
);
343 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
344 !nfs_attr_use_mounted_on_fileid(fattr
))
346 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
349 hash
= nfs_fattr_to_ino_t(fattr
);
351 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
353 inode
= ERR_PTR(-ENOMEM
);
357 if (inode
->i_state
& I_NEW
) {
358 struct nfs_inode
*nfsi
= NFS_I(inode
);
359 unsigned long now
= jiffies
;
361 /* We set i_ino for the few things that still rely on it,
365 /* We can't support update_atime(), since the server will reset it */
366 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
367 inode
->i_mode
= fattr
->mode
;
368 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
369 && nfs_server_capable(inode
, NFS_CAP_MODE
))
370 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
371 /* Why so? Because we want revalidate for devices/FIFOs, and
372 * that's precisely what we have in nfs_file_inode_operations.
374 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
375 if (S_ISREG(inode
->i_mode
)) {
376 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
377 inode
->i_data
.a_ops
= &nfs_file_aops
;
378 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
379 } else if (S_ISDIR(inode
->i_mode
)) {
380 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
381 inode
->i_fop
= &nfs_dir_operations
;
382 inode
->i_data
.a_ops
= &nfs_dir_aops
;
383 /* Deal with crossing mountpoints */
384 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
385 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
386 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
387 inode
->i_op
= &nfs_referral_inode_operations
;
389 inode
->i_op
= &nfs_mountpoint_inode_operations
;
391 inode
->i_flags
|= S_AUTOMOUNT
;
393 } else if (S_ISLNK(inode
->i_mode
))
394 inode
->i_op
= &nfs_symlink_inode_operations
;
396 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
398 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
399 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
400 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
401 inode
->i_version
= 0;
404 inode
->i_uid
= make_kuid(&init_user_ns
, -2);
405 inode
->i_gid
= make_kgid(&init_user_ns
, -2);
407 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
411 nfsi
->read_cache_jiffies
= fattr
->time_start
;
412 nfsi
->attr_gencount
= fattr
->gencount
;
413 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
414 inode
->i_atime
= fattr
->atime
;
415 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
416 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
417 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
418 inode
->i_mtime
= fattr
->mtime
;
419 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
420 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
421 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
422 inode
->i_ctime
= fattr
->ctime
;
423 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
424 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
425 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
426 inode
->i_version
= fattr
->change_attr
;
427 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
428 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
429 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
430 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
432 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
433 | NFS_INO_REVAL_PAGECACHE
;
434 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
435 set_nlink(inode
, fattr
->nlink
);
436 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
437 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
438 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
439 inode
->i_uid
= fattr
->uid
;
440 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
441 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
442 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
443 inode
->i_gid
= fattr
->gid
;
444 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
445 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
446 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
447 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
448 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
450 * report the blocks in 512byte units
452 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
455 nfs_setsecurity(inode
, fattr
, label
);
457 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
458 nfsi
->attrtimeo_timestamp
= now
;
459 nfsi
->access_cache
= RB_ROOT
;
461 nfs_fscache_init_inode_cookie(inode
);
463 unlock_new_inode(inode
);
465 nfs_refresh_inode(inode
, fattr
);
466 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
468 (long long)NFS_FILEID(inode
),
469 nfs_display_fhandle_hash(fh
),
470 atomic_read(&inode
->i_count
));
476 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
479 EXPORT_SYMBOL_GPL(nfs_fhget
);
481 #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)
484 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
486 struct inode
*inode
= dentry
->d_inode
;
487 struct nfs_fattr
*fattr
;
490 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
492 /* skip mode change if it's just for clearing setuid/setgid */
493 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
494 attr
->ia_valid
&= ~ATTR_MODE
;
496 if (attr
->ia_valid
& ATTR_SIZE
) {
497 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
498 attr
->ia_valid
&= ~ATTR_SIZE
;
501 /* Optimization: if the end result is no change, don't RPC */
502 attr
->ia_valid
&= NFS_VALID_ATTRS
;
503 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
506 /* Write all dirty data */
507 if (S_ISREG(inode
->i_mode
)) {
508 nfs_inode_dio_wait(inode
);
512 fattr
= nfs_alloc_fattr();
516 * Return any delegations if we're going to change ACLs
518 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
519 NFS_PROTO(inode
)->return_delegation(inode
);
520 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
522 error
= nfs_refresh_inode(inode
, fattr
);
523 nfs_free_fattr(fattr
);
527 EXPORT_SYMBOL_GPL(nfs_setattr
);
530 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
531 * @inode: inode of the file used
532 * @offset: file offset to start truncating
534 * This is a copy of the common vmtruncate, but with the locking
535 * corrected to take into account the fact that NFS requires
536 * inode->i_size to be updated under the inode->i_lock.
538 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
543 err
= inode_newsize_ok(inode
, offset
);
547 spin_lock(&inode
->i_lock
);
548 oldsize
= inode
->i_size
;
549 i_size_write(inode
, offset
);
550 spin_unlock(&inode
->i_lock
);
552 truncate_pagecache(inode
, oldsize
, offset
);
558 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
559 * @inode: pointer to struct inode
560 * @attr: pointer to struct iattr
562 * Note: we do this in the *proc.c in order to ensure that
563 * it works for things like exclusive creates too.
565 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
567 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
568 spin_lock(&inode
->i_lock
);
569 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
570 int mode
= attr
->ia_mode
& S_IALLUGO
;
571 mode
|= inode
->i_mode
& ~S_IALLUGO
;
572 inode
->i_mode
= mode
;
574 if ((attr
->ia_valid
& ATTR_UID
) != 0)
575 inode
->i_uid
= attr
->ia_uid
;
576 if ((attr
->ia_valid
& ATTR_GID
) != 0)
577 inode
->i_gid
= attr
->ia_gid
;
578 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
579 spin_unlock(&inode
->i_lock
);
581 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
582 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
583 nfs_vmtruncate(inode
, attr
->ia_size
);
586 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode
);
588 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
590 struct inode
*inode
= dentry
->d_inode
;
591 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
594 /* Flush out writes to the server in order to update c/mtime. */
595 if (S_ISREG(inode
->i_mode
)) {
596 nfs_inode_dio_wait(inode
);
597 err
= filemap_write_and_wait(inode
->i_mapping
);
603 * We may force a getattr if the user cares about atime.
605 * Note that we only have to check the vfsmount flags here:
606 * - NFS always sets S_NOATIME by so checking it would give a
608 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
609 * no point in checking those.
611 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
612 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
616 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
618 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
620 generic_fillattr(inode
, stat
);
621 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
626 EXPORT_SYMBOL_GPL(nfs_getattr
);
628 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
630 atomic_set(&l_ctx
->count
, 1);
631 l_ctx
->lockowner
.l_owner
= current
->files
;
632 l_ctx
->lockowner
.l_pid
= current
->tgid
;
633 INIT_LIST_HEAD(&l_ctx
->list
);
634 nfs_iocounter_init(&l_ctx
->io_count
);
637 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
639 struct nfs_lock_context
*head
= &ctx
->lock_context
;
640 struct nfs_lock_context
*pos
= head
;
643 if (pos
->lockowner
.l_owner
!= current
->files
)
645 if (pos
->lockowner
.l_pid
!= current
->tgid
)
647 atomic_inc(&pos
->count
);
649 } while ((pos
= list_entry(pos
->list
.next
, typeof(*pos
), list
)) != head
);
653 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
655 struct nfs_lock_context
*res
, *new = NULL
;
656 struct inode
*inode
= ctx
->dentry
->d_inode
;
658 spin_lock(&inode
->i_lock
);
659 res
= __nfs_find_lock_context(ctx
);
661 spin_unlock(&inode
->i_lock
);
662 new = kmalloc(sizeof(*new), GFP_KERNEL
);
664 return ERR_PTR(-ENOMEM
);
665 nfs_init_lock_context(new);
666 spin_lock(&inode
->i_lock
);
667 res
= __nfs_find_lock_context(ctx
);
669 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
670 new->open_context
= ctx
;
675 spin_unlock(&inode
->i_lock
);
680 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
682 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
683 struct inode
*inode
= ctx
->dentry
->d_inode
;
685 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
687 list_del(&l_ctx
->list
);
688 spin_unlock(&inode
->i_lock
);
693 * nfs_close_context - Common close_context() routine NFSv2/v3
694 * @ctx: pointer to context
695 * @is_sync: is this a synchronous close
697 * always ensure that the attributes are up to date if we're mounted
698 * with close-to-open semantics
700 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
703 struct nfs_server
*server
;
705 if (!(ctx
->mode
& FMODE_WRITE
))
709 inode
= ctx
->dentry
->d_inode
;
710 if (!list_empty(&NFS_I(inode
)->open_files
))
712 server
= NFS_SERVER(inode
);
713 if (server
->flags
& NFS_MOUNT_NOCTO
)
715 nfs_revalidate_inode(server
, inode
);
717 EXPORT_SYMBOL_GPL(nfs_close_context
);
719 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
721 struct nfs_open_context
*ctx
;
722 struct rpc_cred
*cred
= rpc_lookup_cred();
724 return ERR_CAST(cred
);
726 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
729 return ERR_PTR(-ENOMEM
);
731 nfs_sb_active(dentry
->d_sb
);
732 ctx
->dentry
= dget(dentry
);
738 nfs_init_lock_context(&ctx
->lock_context
);
739 ctx
->lock_context
.open_context
= ctx
;
740 INIT_LIST_HEAD(&ctx
->list
);
741 ctx
->mdsthreshold
= NULL
;
744 EXPORT_SYMBOL_GPL(alloc_nfs_open_context
);
746 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
749 atomic_inc(&ctx
->lock_context
.count
);
752 EXPORT_SYMBOL_GPL(get_nfs_open_context
);
754 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
756 struct inode
*inode
= ctx
->dentry
->d_inode
;
757 struct super_block
*sb
= ctx
->dentry
->d_sb
;
759 if (!list_empty(&ctx
->list
)) {
760 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
762 list_del(&ctx
->list
);
763 spin_unlock(&inode
->i_lock
);
764 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
767 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
768 if (ctx
->cred
!= NULL
)
769 put_rpccred(ctx
->cred
);
772 kfree(ctx
->mdsthreshold
);
776 void put_nfs_open_context(struct nfs_open_context
*ctx
)
778 __put_nfs_open_context(ctx
, 0);
780 EXPORT_SYMBOL_GPL(put_nfs_open_context
);
783 * Ensure that mmap has a recent RPC credential for use when writing out
786 void nfs_inode_attach_open_context(struct nfs_open_context
*ctx
)
788 struct inode
*inode
= ctx
->dentry
->d_inode
;
789 struct nfs_inode
*nfsi
= NFS_I(inode
);
791 spin_lock(&inode
->i_lock
);
792 list_add(&ctx
->list
, &nfsi
->open_files
);
793 spin_unlock(&inode
->i_lock
);
795 EXPORT_SYMBOL_GPL(nfs_inode_attach_open_context
);
797 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
799 filp
->private_data
= get_nfs_open_context(ctx
);
800 if (list_empty(&ctx
->list
))
801 nfs_inode_attach_open_context(ctx
);
803 EXPORT_SYMBOL_GPL(nfs_file_set_open_context
);
806 * Given an inode, search for an open context with the desired characteristics
808 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
810 struct nfs_inode
*nfsi
= NFS_I(inode
);
811 struct nfs_open_context
*pos
, *ctx
= NULL
;
813 spin_lock(&inode
->i_lock
);
814 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
815 if (cred
!= NULL
&& pos
->cred
!= cred
)
817 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
819 ctx
= get_nfs_open_context(pos
);
822 spin_unlock(&inode
->i_lock
);
826 static void nfs_file_clear_open_context(struct file
*filp
)
828 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
831 struct inode
*inode
= ctx
->dentry
->d_inode
;
833 filp
->private_data
= NULL
;
834 spin_lock(&inode
->i_lock
);
835 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
836 spin_unlock(&inode
->i_lock
);
837 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
842 * These allocate and release file read/write context information.
844 int nfs_open(struct inode
*inode
, struct file
*filp
)
846 struct nfs_open_context
*ctx
;
848 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
851 nfs_file_set_open_context(filp
, ctx
);
852 put_nfs_open_context(ctx
);
853 nfs_fscache_set_inode_cookie(inode
, filp
);
857 int nfs_release(struct inode
*inode
, struct file
*filp
)
859 nfs_file_clear_open_context(filp
);
864 * This function is called whenever some part of NFS notices that
865 * the cached attributes have to be refreshed.
868 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
870 int status
= -ESTALE
;
871 struct nfs4_label
*label
= NULL
;
872 struct nfs_fattr
*fattr
= NULL
;
873 struct nfs_inode
*nfsi
= NFS_I(inode
);
875 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
876 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
878 if (is_bad_inode(inode
))
880 if (NFS_STALE(inode
))
884 fattr
= nfs_alloc_fattr();
888 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
890 label
= nfs4_label_alloc(NFS_SERVER(inode
), GFP_KERNEL
);
892 status
= PTR_ERR(label
);
896 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
, label
);
898 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
900 (long long)NFS_FILEID(inode
), status
);
901 if (status
== -ESTALE
) {
902 nfs_zap_caches(inode
);
903 if (!S_ISDIR(inode
->i_mode
))
904 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
909 status
= nfs_refresh_inode(inode
, fattr
);
911 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
913 (long long)NFS_FILEID(inode
), status
);
917 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
918 nfs_zap_acl_cache(inode
);
920 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
922 (long long)NFS_FILEID(inode
));
925 nfs4_label_free(label
);
927 nfs_free_fattr(fattr
);
931 int nfs_attribute_timeout(struct inode
*inode
)
933 struct nfs_inode
*nfsi
= NFS_I(inode
);
935 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
938 int nfs_attribute_cache_expired(struct inode
*inode
)
940 if (nfs_have_delegated_attributes(inode
))
942 return nfs_attribute_timeout(inode
);
946 * nfs_revalidate_inode - Revalidate the inode attributes
947 * @server - pointer to nfs_server struct
948 * @inode - pointer to inode struct
950 * Updates inode attribute information by retrieving the data from the server.
952 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
954 if (!(NFS_I(inode
)->cache_validity
&
955 (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_LABEL
))
956 && !nfs_attribute_cache_expired(inode
))
957 return NFS_STALE(inode
) ? -ESTALE
: 0;
958 return __nfs_revalidate_inode(server
, inode
);
960 EXPORT_SYMBOL_GPL(nfs_revalidate_inode
);
962 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
964 struct nfs_inode
*nfsi
= NFS_I(inode
);
967 if (mapping
->nrpages
!= 0) {
968 if (S_ISREG(inode
->i_mode
)) {
969 ret
= nfs_sync_mapping(mapping
);
973 ret
= invalidate_inode_pages2(mapping
);
977 spin_lock(&inode
->i_lock
);
978 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
979 if (S_ISDIR(inode
->i_mode
))
980 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
981 spin_unlock(&inode
->i_lock
);
982 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
983 nfs_fscache_wait_on_invalidate(inode
);
984 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
985 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
989 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
991 if (nfs_have_delegated_attributes(inode
))
993 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
994 || nfs_attribute_timeout(inode
)
999 * nfs_revalidate_mapping - Revalidate the pagecache
1000 * @inode - pointer to host inode
1001 * @mapping - pointer to mapping
1003 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
1005 struct nfs_inode
*nfsi
= NFS_I(inode
);
1008 /* swapfiles are not supposed to be shared. */
1009 if (IS_SWAPFILE(inode
))
1012 if (nfs_mapping_need_revalidate_inode(inode
)) {
1013 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1017 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
1018 ret
= nfs_invalidate_mapping(inode
, mapping
);
1023 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1025 struct nfs_inode
*nfsi
= NFS_I(inode
);
1026 unsigned long ret
= 0;
1028 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
1029 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
1030 && inode
->i_version
== fattr
->pre_change_attr
) {
1031 inode
->i_version
= fattr
->change_attr
;
1032 if (S_ISDIR(inode
->i_mode
))
1033 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1034 ret
|= NFS_INO_INVALID_ATTR
;
1036 /* If we have atomic WCC data, we may update some attributes */
1037 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
1038 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
1039 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1040 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1041 ret
|= NFS_INO_INVALID_ATTR
;
1044 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
1045 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
1046 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1047 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1048 if (S_ISDIR(inode
->i_mode
))
1049 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1050 ret
|= NFS_INO_INVALID_ATTR
;
1052 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
1053 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
1054 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
1055 && nfsi
->npages
== 0) {
1056 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
1057 ret
|= NFS_INO_INVALID_ATTR
;
1060 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
1061 nfs_fscache_invalidate(inode
);
1067 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1068 * @inode - pointer to inode
1069 * @fattr - updated attributes
1071 * Verifies the attribute cache. If we have just changed the attributes,
1072 * so that fattr carries weak cache consistency data, then it may
1073 * also update the ctime/mtime/change_attribute.
1075 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1077 struct nfs_inode
*nfsi
= NFS_I(inode
);
1078 loff_t cur_size
, new_isize
;
1079 unsigned long invalid
= 0;
1082 if (nfs_have_delegated_attributes(inode
))
1084 /* Has the inode gone and changed behind our back? */
1085 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1087 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1090 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1091 inode
->i_version
!= fattr
->change_attr
)
1092 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1094 /* Verify a few of the more important attributes */
1095 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
1096 invalid
|= NFS_INO_INVALID_ATTR
;
1098 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1099 cur_size
= i_size_read(inode
);
1100 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1101 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1102 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1105 /* Have any file permissions changed? */
1106 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
1107 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1108 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && !uid_eq(inode
->i_uid
, fattr
->uid
))
1109 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1110 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && !gid_eq(inode
->i_gid
, fattr
->gid
))
1111 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1113 /* Has the link count changed? */
1114 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
1115 invalid
|= NFS_INO_INVALID_ATTR
;
1117 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1118 invalid
|= NFS_INO_INVALID_ATIME
;
1121 nfsi
->cache_validity
|= invalid
;
1123 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1127 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1129 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
1131 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1134 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1136 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1138 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1141 static atomic_long_t nfs_attr_generation_counter
;
1143 static unsigned long nfs_read_attr_generation_counter(void)
1145 return atomic_long_read(&nfs_attr_generation_counter
);
1148 unsigned long nfs_inc_attr_generation_counter(void)
1150 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1153 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1156 fattr
->time_start
= jiffies
;
1157 fattr
->gencount
= nfs_inc_attr_generation_counter();
1158 fattr
->owner_name
= NULL
;
1159 fattr
->group_name
= NULL
;
1161 EXPORT_SYMBOL_GPL(nfs_fattr_init
);
1163 struct nfs_fattr
*nfs_alloc_fattr(void)
1165 struct nfs_fattr
*fattr
;
1167 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1169 nfs_fattr_init(fattr
);
1172 EXPORT_SYMBOL_GPL(nfs_alloc_fattr
);
1174 struct nfs_fh
*nfs_alloc_fhandle(void)
1178 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1183 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle
);
1187 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1188 * in the same way that wireshark does
1192 * For debugging only.
1194 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1196 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1197 * not on the result */
1198 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1202 * _nfs_display_fhandle - display an NFS file handle on the console
1204 * @fh: file handle to display
1205 * @caption: display caption
1207 * For debugging only.
1209 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1213 if (fh
== NULL
|| fh
->size
== 0) {
1214 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1218 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1219 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1220 for (i
= 0; i
< fh
->size
; i
+= 16) {
1221 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1223 switch ((fh
->size
- i
- 1) >> 2) {
1225 printk(KERN_DEFAULT
" %08x\n",
1229 printk(KERN_DEFAULT
" %08x %08x\n",
1230 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1233 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1234 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1235 be32_to_cpup(pos
+ 2));
1238 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1239 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1240 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1247 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1248 * @inode - pointer to inode
1249 * @fattr - attributes
1251 * Attempt to divine whether or not an RPC call reply carrying stale
1252 * attributes got scheduled after another call carrying updated ones.
1254 * To do so, the function first assumes that a more recent ctime means
1255 * that the attributes in fattr are newer, however it also attempt to
1256 * catch the case where ctime either didn't change, or went backwards
1257 * (if someone reset the clock on the server) by looking at whether
1258 * or not this RPC call was started after the inode was last updated.
1259 * Note also the check for wraparound of 'attr_gencount'
1261 * The function returns 'true' if it thinks the attributes in 'fattr' are
1262 * more recent than the ones cached in the inode.
1265 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1267 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1269 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1270 nfs_ctime_need_update(inode
, fattr
) ||
1271 nfs_size_need_update(inode
, fattr
) ||
1272 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1275 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1277 if (nfs_inode_attrs_need_update(inode
, fattr
))
1278 return nfs_update_inode(inode
, fattr
);
1279 return nfs_check_inode_attributes(inode
, fattr
);
1283 * nfs_refresh_inode - try to update the inode attribute cache
1284 * @inode - pointer to inode
1285 * @fattr - updated attributes
1287 * Check that an RPC call that returned attributes has not overlapped with
1288 * other recent updates of the inode metadata, then decide whether it is
1289 * safe to do a full update of the inode attributes, or whether just to
1290 * call nfs_check_inode_attributes.
1292 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1296 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1298 spin_lock(&inode
->i_lock
);
1299 status
= nfs_refresh_inode_locked(inode
, fattr
);
1300 spin_unlock(&inode
->i_lock
);
1304 EXPORT_SYMBOL_GPL(nfs_refresh_inode
);
1306 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1308 struct nfs_inode
*nfsi
= NFS_I(inode
);
1310 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1311 if (S_ISDIR(inode
->i_mode
)) {
1312 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1313 nfs_fscache_invalidate(inode
);
1315 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1317 return nfs_refresh_inode_locked(inode
, fattr
);
1321 * nfs_post_op_update_inode - try to update the inode attribute cache
1322 * @inode - pointer to inode
1323 * @fattr - updated attributes
1325 * After an operation that has changed the inode metadata, mark the
1326 * attribute cache as being invalid, then try to update it.
1328 * NB: if the server didn't return any post op attributes, this
1329 * function will force the retrieval of attributes before the next
1330 * NFS request. Thus it should be used only for operations that
1331 * are expected to change one or more attributes, to avoid
1332 * unnecessary NFS requests and trips through nfs_update_inode().
1334 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1338 spin_lock(&inode
->i_lock
);
1339 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1340 spin_unlock(&inode
->i_lock
);
1344 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode
);
1347 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1348 * @inode - pointer to inode
1349 * @fattr - updated attributes
1351 * After an operation that has changed the inode metadata, mark the
1352 * attribute cache as being invalid, then try to update it. Fake up
1353 * weak cache consistency data, if none exist.
1355 * This function is mainly designed to be used by the ->write_done() functions.
1357 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1361 spin_lock(&inode
->i_lock
);
1362 /* Don't do a WCC update if these attributes are already stale */
1363 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1364 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1365 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1366 | NFS_ATTR_FATTR_PRESIZE
1367 | NFS_ATTR_FATTR_PREMTIME
1368 | NFS_ATTR_FATTR_PRECTIME
);
1371 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1372 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1373 fattr
->pre_change_attr
= inode
->i_version
;
1374 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1376 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1377 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1378 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1379 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1381 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1382 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1383 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1384 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1386 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1387 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1388 fattr
->pre_size
= i_size_read(inode
);
1389 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1392 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1393 spin_unlock(&inode
->i_lock
);
1396 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc
);
1399 * Many nfs protocol calls return the new file attributes after
1400 * an operation. Here we update the inode to reflect the state
1401 * of the server's inode.
1403 * This is a bit tricky because we have to make sure all dirty pages
1404 * have been sent off to the server before calling invalidate_inode_pages.
1405 * To make sure no other process adds more write requests while we try
1406 * our best to flush them, we make them sleep during the attribute refresh.
1408 * A very similar scenario holds for the dir cache.
1410 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1412 struct nfs_server
*server
;
1413 struct nfs_inode
*nfsi
= NFS_I(inode
);
1414 loff_t cur_isize
, new_isize
;
1415 unsigned long invalid
= 0;
1416 unsigned long now
= jiffies
;
1417 unsigned long save_cache_validity
;
1419 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1420 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1421 nfs_display_fhandle_hash(NFS_FH(inode
)),
1422 atomic_read(&inode
->i_count
), fattr
->valid
);
1424 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1425 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1426 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1427 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1428 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1429 (long long)fattr
->fileid
);
1434 * Make sure the inode's type hasn't changed.
1436 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1438 * Big trouble! The inode has become a different object.
1440 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1441 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1445 server
= NFS_SERVER(inode
);
1446 /* Update the fsid? */
1447 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1448 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1449 !IS_AUTOMOUNT(inode
))
1450 server
->fsid
= fattr
->fsid
;
1453 * Update the read time so we don't revalidate too often.
1455 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1457 save_cache_validity
= nfsi
->cache_validity
;
1458 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1459 | NFS_INO_INVALID_ATIME
1460 | NFS_INO_REVAL_FORCED
1461 | NFS_INO_REVAL_PAGECACHE
);
1463 /* Do atomic weak cache consistency updates */
1464 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1466 /* More cache consistency checks */
1467 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1468 if (inode
->i_version
!= fattr
->change_attr
) {
1469 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1470 inode
->i_sb
->s_id
, inode
->i_ino
);
1471 invalid
|= NFS_INO_INVALID_ATTR
1472 | NFS_INO_INVALID_DATA
1473 | NFS_INO_INVALID_ACCESS
1474 | NFS_INO_INVALID_ACL
1475 | NFS_INO_REVAL_PAGECACHE
;
1476 if (S_ISDIR(inode
->i_mode
))
1477 nfs_force_lookup_revalidate(inode
);
1478 inode
->i_version
= fattr
->change_attr
;
1480 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1481 invalid
|= save_cache_validity
;
1483 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1484 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1485 } else if (server
->caps
& NFS_CAP_MTIME
)
1486 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1487 | NFS_INO_REVAL_FORCED
);
1489 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1490 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1491 } else if (server
->caps
& NFS_CAP_CTIME
)
1492 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1493 | NFS_INO_REVAL_FORCED
);
1495 /* Check if our cached file size is stale */
1496 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1497 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1498 cur_isize
= i_size_read(inode
);
1499 if (new_isize
!= cur_isize
) {
1500 /* Do we perhaps have any outstanding writes, or has
1501 * the file grown beyond our last write? */
1502 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1503 new_isize
> cur_isize
) {
1504 i_size_write(inode
, new_isize
);
1505 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1507 dprintk("NFS: isize change on server for file %s/%ld "
1511 (long long)cur_isize
,
1512 (long long)new_isize
);
1515 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1516 | NFS_INO_REVAL_PAGECACHE
1517 | NFS_INO_REVAL_FORCED
);
1520 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1521 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1522 else if (server
->caps
& NFS_CAP_ATIME
)
1523 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1524 | NFS_INO_REVAL_FORCED
);
1526 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1527 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1528 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1529 newmode
|= fattr
->mode
& S_IALLUGO
;
1530 inode
->i_mode
= newmode
;
1531 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1533 } else if (server
->caps
& NFS_CAP_MODE
)
1534 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1535 | NFS_INO_INVALID_ACCESS
1536 | NFS_INO_INVALID_ACL
1537 | NFS_INO_REVAL_FORCED
);
1539 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1540 if (!uid_eq(inode
->i_uid
, fattr
->uid
)) {
1541 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1542 inode
->i_uid
= fattr
->uid
;
1544 } else if (server
->caps
& NFS_CAP_OWNER
)
1545 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1546 | NFS_INO_INVALID_ACCESS
1547 | NFS_INO_INVALID_ACL
1548 | NFS_INO_REVAL_FORCED
);
1550 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1551 if (!gid_eq(inode
->i_gid
, fattr
->gid
)) {
1552 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1553 inode
->i_gid
= fattr
->gid
;
1555 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1556 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1557 | NFS_INO_INVALID_ACCESS
1558 | NFS_INO_INVALID_ACL
1559 | NFS_INO_REVAL_FORCED
);
1561 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1562 if (inode
->i_nlink
!= fattr
->nlink
) {
1563 invalid
|= NFS_INO_INVALID_ATTR
;
1564 if (S_ISDIR(inode
->i_mode
))
1565 invalid
|= NFS_INO_INVALID_DATA
;
1566 set_nlink(inode
, fattr
->nlink
);
1568 } else if (server
->caps
& NFS_CAP_NLINK
)
1569 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1570 | NFS_INO_REVAL_FORCED
);
1572 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1574 * report the blocks in 512byte units
1576 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1578 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1579 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1581 /* Update attrtimeo value if we're out of the unstable period */
1582 if (invalid
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_LABEL
)) {
1583 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1584 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1585 nfsi
->attrtimeo_timestamp
= now
;
1586 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1588 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1589 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1590 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1591 nfsi
->attrtimeo_timestamp
= now
;
1594 invalid
&= ~NFS_INO_INVALID_ATTR
;
1595 invalid
&= ~NFS_INO_INVALID_LABEL
;
1596 /* Don't invalidate the data if we were to blame */
1597 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1598 || S_ISLNK(inode
->i_mode
)))
1599 invalid
&= ~NFS_INO_INVALID_DATA
;
1600 if (!NFS_PROTO(inode
)->have_delegation(inode
, FMODE_READ
) ||
1601 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1602 nfsi
->cache_validity
|= invalid
;
1604 if (invalid
& NFS_INO_INVALID_DATA
)
1605 nfs_fscache_invalidate(inode
);
1610 * No need to worry about unhashing the dentry, as the
1611 * lookup validation will know that the inode is bad.
1612 * (But we fall through to invalidate the caches.)
1614 nfs_invalidate_inode(inode
);
1618 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1620 struct nfs_inode
*nfsi
;
1621 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1625 nfsi
->cache_validity
= 0UL;
1626 #ifdef CONFIG_NFS_V3_ACL
1627 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1628 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1630 #if IS_ENABLED(CONFIG_NFS_V4)
1631 nfsi
->nfs4_acl
= NULL
;
1632 #endif /* CONFIG_NFS_V4 */
1633 return &nfsi
->vfs_inode
;
1635 EXPORT_SYMBOL_GPL(nfs_alloc_inode
);
1637 static void nfs_i_callback(struct rcu_head
*head
)
1639 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1640 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1643 void nfs_destroy_inode(struct inode
*inode
)
1645 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1647 EXPORT_SYMBOL_GPL(nfs_destroy_inode
);
1649 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1651 #if IS_ENABLED(CONFIG_NFS_V4)
1652 INIT_LIST_HEAD(&nfsi
->open_states
);
1653 nfsi
->delegation
= NULL
;
1654 nfsi
->delegation_state
= 0;
1655 init_rwsem(&nfsi
->rwsem
);
1656 nfsi
->layout
= NULL
;
1660 static void init_once(void *foo
)
1662 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1664 inode_init_once(&nfsi
->vfs_inode
);
1665 INIT_LIST_HEAD(&nfsi
->open_files
);
1666 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1667 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1668 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1670 nfsi
->commit_info
.ncommit
= 0;
1671 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1672 atomic_set(&nfsi
->silly_count
, 1);
1673 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1674 init_waitqueue_head(&nfsi
->waitqueue
);
1675 nfs4_init_once(nfsi
);
1678 static int __init
nfs_init_inodecache(void)
1680 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1681 sizeof(struct nfs_inode
),
1682 0, (SLAB_RECLAIM_ACCOUNT
|
1685 if (nfs_inode_cachep
== NULL
)
1691 static void nfs_destroy_inodecache(void)
1694 * Make sure all delayed rcu free inodes are flushed before we
1698 kmem_cache_destroy(nfs_inode_cachep
);
1701 struct workqueue_struct
*nfsiod_workqueue
;
1702 EXPORT_SYMBOL_GPL(nfsiod_workqueue
);
1705 * start up the nfsiod workqueue
1707 static int nfsiod_start(void)
1709 struct workqueue_struct
*wq
;
1710 dprintk("RPC: creating workqueue nfsiod\n");
1711 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1714 nfsiod_workqueue
= wq
;
1719 * Destroy the nfsiod workqueue
1721 static void nfsiod_stop(void)
1723 struct workqueue_struct
*wq
;
1725 wq
= nfsiod_workqueue
;
1728 nfsiod_workqueue
= NULL
;
1729 destroy_workqueue(wq
);
1733 EXPORT_SYMBOL_GPL(nfs_net_id
);
1735 static int nfs_net_init(struct net
*net
)
1737 nfs_clients_init(net
);
1741 static void nfs_net_exit(struct net
*net
)
1743 nfs_cleanup_cb_ident_idr(net
);
1746 static struct pernet_operations nfs_net_ops
= {
1747 .init
= nfs_net_init
,
1748 .exit
= nfs_net_exit
,
1750 .size
= sizeof(struct nfs_net
),
1756 static int __init
init_nfs_fs(void)
1760 err
= register_pernet_subsys(&nfs_net_ops
);
1764 err
= nfs_fscache_register();
1768 err
= nfsiod_start();
1772 err
= nfs_fs_proc_init();
1776 err
= nfs_init_nfspagecache();
1780 err
= nfs_init_inodecache();
1784 err
= nfs_init_readpagecache();
1788 err
= nfs_init_writepagecache();
1792 err
= nfs_init_directcache();
1796 #ifdef CONFIG_PROC_FS
1797 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1799 if ((err
= register_nfs_fs()) != 0)
1804 #ifdef CONFIG_PROC_FS
1805 rpc_proc_unregister(&init_net
, "nfs");
1807 nfs_destroy_directcache();
1809 nfs_destroy_writepagecache();
1811 nfs_destroy_readpagecache();
1813 nfs_destroy_inodecache();
1815 nfs_destroy_nfspagecache();
1821 nfs_fscache_unregister();
1823 unregister_pernet_subsys(&nfs_net_ops
);
1828 static void __exit
exit_nfs_fs(void)
1830 nfs_destroy_directcache();
1831 nfs_destroy_writepagecache();
1832 nfs_destroy_readpagecache();
1833 nfs_destroy_inodecache();
1834 nfs_destroy_nfspagecache();
1835 nfs_fscache_unregister();
1836 unregister_pernet_subsys(&nfs_net_ops
);
1837 #ifdef CONFIG_PROC_FS
1838 rpc_proc_unregister(&init_net
, "nfs");
1840 unregister_nfs_fs();
1845 /* Not quite true; I just maintain it */
1846 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1847 MODULE_LICENSE("GPL");
1848 module_param(enable_ino64
, bool, 0644);
1850 module_init(init_nfs_fs
)
1851 module_exit(exit_nfs_fs
)