4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/time.h>
20 #include <linux/kernel.h>
22 #include <linux/string.h>
23 #include <linux/stat.h>
24 #include <linux/errno.h>
25 #include <linux/unistd.h>
26 #include <linux/sunrpc/clnt.h>
27 #include <linux/sunrpc/stats.h>
28 #include <linux/sunrpc/metrics.h>
29 #include <linux/nfs_fs.h>
30 #include <linux/nfs_mount.h>
31 #include <linux/nfs4_mount.h>
32 #include <linux/lockd/bind.h>
33 #include <linux/seq_file.h>
34 #include <linux/mount.h>
35 #include <linux/vfs.h>
36 #include <linux/inet.h>
37 #include <linux/nfs_xdr.h>
38 #include <linux/slab.h>
39 #include <linux/compat.h>
40 #include <linux/freezer.h>
41 #include <linux/crc32.h>
43 #include <asm/uaccess.h>
47 #include "delegation.h"
51 #include "dns_resolve.h"
56 #define NFSDBG_FACILITY NFSDBG_VFS
58 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
60 /* Default is to see 64-bit inode numbers */
61 static bool enable_ino64
= NFS_64_BIT_INODE_NUMBERS_ENABLED
;
63 static void nfs_invalidate_inode(struct inode
*);
64 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
66 static struct kmem_cache
* nfs_inode_cachep
;
68 static inline unsigned long
69 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
71 return nfs_fileid_to_ino_t(fattr
->fileid
);
75 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
76 * @word: long word containing the bit lock
78 int nfs_wait_bit_killable(void *word
)
80 if (fatal_signal_pending(current
))
85 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable
);
88 * nfs_compat_user_ino64 - returns the user-visible inode number
89 * @fileid: 64-bit fileid
91 * This function returns a 32-bit inode number if the boot parameter
92 * nfs.enable_ino64 is zero.
94 u64
nfs_compat_user_ino64(u64 fileid
)
105 if (sizeof(ino
) < sizeof(fileid
))
106 ino
^= fileid
>> (sizeof(fileid
)-sizeof(ino
)) * 8;
110 int nfs_drop_inode(struct inode
*inode
)
112 return NFS_STALE(inode
) || generic_drop_inode(inode
);
114 EXPORT_SYMBOL_GPL(nfs_drop_inode
);
116 void nfs_clear_inode(struct inode
*inode
)
119 * The following should never happen...
121 WARN_ON_ONCE(nfs_have_writebacks(inode
));
122 WARN_ON_ONCE(!list_empty(&NFS_I(inode
)->open_files
));
123 nfs_zap_acl_cache(inode
);
124 nfs_access_zap_cache(inode
);
125 nfs_fscache_release_inode_cookie(inode
);
127 EXPORT_SYMBOL_GPL(nfs_clear_inode
);
129 void nfs_evict_inode(struct inode
*inode
)
131 truncate_inode_pages(&inode
->i_data
, 0);
133 nfs_clear_inode(inode
);
137 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
139 int nfs_sync_mapping(struct address_space
*mapping
)
143 if (mapping
->nrpages
!= 0) {
144 unmap_mapping_range(mapping
, 0, 0, 0);
145 ret
= nfs_wb_all(mapping
->host
);
151 * Invalidate the local caches
153 static void nfs_zap_caches_locked(struct inode
*inode
)
155 struct nfs_inode
*nfsi
= NFS_I(inode
);
156 int mode
= inode
->i_mode
;
158 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
160 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
161 nfsi
->attrtimeo_timestamp
= jiffies
;
163 memset(NFS_I(inode
)->cookieverf
, 0, sizeof(NFS_I(inode
)->cookieverf
));
164 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
)) {
165 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
166 nfs_fscache_invalidate(inode
);
168 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
172 void nfs_zap_caches(struct inode
*inode
)
174 spin_lock(&inode
->i_lock
);
175 nfs_zap_caches_locked(inode
);
176 spin_unlock(&inode
->i_lock
);
179 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
181 if (mapping
->nrpages
!= 0) {
182 spin_lock(&inode
->i_lock
);
183 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
184 nfs_fscache_invalidate(inode
);
185 spin_unlock(&inode
->i_lock
);
189 void nfs_zap_acl_cache(struct inode
*inode
)
191 void (*clear_acl_cache
)(struct inode
*);
193 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
194 if (clear_acl_cache
!= NULL
)
195 clear_acl_cache(inode
);
196 spin_lock(&inode
->i_lock
);
197 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
198 spin_unlock(&inode
->i_lock
);
200 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache
);
202 void nfs_invalidate_atime(struct inode
*inode
)
204 spin_lock(&inode
->i_lock
);
205 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
206 spin_unlock(&inode
->i_lock
);
208 EXPORT_SYMBOL_GPL(nfs_invalidate_atime
);
211 * Invalidate, but do not unhash, the inode.
212 * NB: must be called with inode->i_lock held!
214 static void nfs_invalidate_inode(struct inode
*inode
)
216 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
217 nfs_zap_caches_locked(inode
);
220 struct nfs_find_desc
{
222 struct nfs_fattr
*fattr
;
226 * In NFSv3 we can have 64bit inode numbers. In order to support
227 * this, and re-exported directories (also seen in NFSv2)
228 * we are forced to allow 2 different inodes to have the same
232 nfs_find_actor(struct inode
*inode
, void *opaque
)
234 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
235 struct nfs_fh
*fh
= desc
->fh
;
236 struct nfs_fattr
*fattr
= desc
->fattr
;
238 if (NFS_FILEID(inode
) != fattr
->fileid
)
240 if ((S_IFMT
& inode
->i_mode
) != (S_IFMT
& fattr
->mode
))
242 if (nfs_compare_fh(NFS_FH(inode
), fh
))
244 if (is_bad_inode(inode
) || NFS_STALE(inode
))
250 nfs_init_locked(struct inode
*inode
, void *opaque
)
252 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
253 struct nfs_fattr
*fattr
= desc
->fattr
;
255 set_nfs_fileid(inode
, fattr
->fileid
);
256 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
260 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
261 struct nfs4_label
*nfs4_label_alloc(struct nfs_server
*server
, gfp_t flags
)
263 struct nfs4_label
*label
= NULL
;
264 int minor_version
= server
->nfs_client
->cl_minorversion
;
266 if (minor_version
< 2)
269 if (!(server
->caps
& NFS_CAP_SECURITY_LABEL
))
272 label
= kzalloc(sizeof(struct nfs4_label
), flags
);
274 return ERR_PTR(-ENOMEM
);
276 label
->label
= kzalloc(NFS4_MAXLABELLEN
, flags
);
277 if (label
->label
== NULL
) {
279 return ERR_PTR(-ENOMEM
);
281 label
->len
= NFS4_MAXLABELLEN
;
285 EXPORT_SYMBOL_GPL(nfs4_label_alloc
);
289 * This is our front-end to iget that looks up inodes by file handle
290 * instead of inode number.
293 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
, struct nfs4_label
*label
)
295 struct nfs_find_desc desc
= {
299 struct inode
*inode
= ERR_PTR(-ENOENT
);
302 nfs_attr_check_mountpoint(sb
, fattr
);
304 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
305 !nfs_attr_use_mounted_on_fileid(fattr
))
307 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
310 hash
= nfs_fattr_to_ino_t(fattr
);
312 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
314 inode
= ERR_PTR(-ENOMEM
);
318 if (inode
->i_state
& I_NEW
) {
319 struct nfs_inode
*nfsi
= NFS_I(inode
);
320 unsigned long now
= jiffies
;
322 /* We set i_ino for the few things that still rely on it,
326 /* We can't support update_atime(), since the server will reset it */
327 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
328 inode
->i_mode
= fattr
->mode
;
329 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
330 && nfs_server_capable(inode
, NFS_CAP_MODE
))
331 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
332 /* Why so? Because we want revalidate for devices/FIFOs, and
333 * that's precisely what we have in nfs_file_inode_operations.
335 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
336 if (S_ISREG(inode
->i_mode
)) {
337 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
338 inode
->i_data
.a_ops
= &nfs_file_aops
;
339 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
340 } else if (S_ISDIR(inode
->i_mode
)) {
341 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
342 inode
->i_fop
= &nfs_dir_operations
;
343 inode
->i_data
.a_ops
= &nfs_dir_aops
;
344 /* Deal with crossing mountpoints */
345 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
346 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
347 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
348 inode
->i_op
= &nfs_referral_inode_operations
;
350 inode
->i_op
= &nfs_mountpoint_inode_operations
;
352 inode
->i_flags
|= S_AUTOMOUNT
;
354 } else if (S_ISLNK(inode
->i_mode
))
355 inode
->i_op
= &nfs_symlink_inode_operations
;
357 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
359 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
360 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
361 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
362 inode
->i_version
= 0;
365 inode
->i_uid
= make_kuid(&init_user_ns
, -2);
366 inode
->i_gid
= make_kgid(&init_user_ns
, -2);
368 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
372 nfsi
->read_cache_jiffies
= fattr
->time_start
;
373 nfsi
->attr_gencount
= fattr
->gencount
;
374 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
375 inode
->i_atime
= fattr
->atime
;
376 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
377 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
378 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
379 inode
->i_mtime
= fattr
->mtime
;
380 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
381 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
382 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
383 inode
->i_ctime
= fattr
->ctime
;
384 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
385 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
386 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
387 inode
->i_version
= fattr
->change_attr
;
388 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
389 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
390 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
391 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
393 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
394 | NFS_INO_REVAL_PAGECACHE
;
395 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
396 set_nlink(inode
, fattr
->nlink
);
397 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
398 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
399 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
400 inode
->i_uid
= fattr
->uid
;
401 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
402 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
403 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
404 inode
->i_gid
= fattr
->gid
;
405 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
406 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
407 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
408 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
409 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
411 * report the blocks in 512byte units
413 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
415 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
416 nfsi
->attrtimeo_timestamp
= now
;
417 nfsi
->access_cache
= RB_ROOT
;
419 nfs_fscache_init_inode_cookie(inode
);
421 unlock_new_inode(inode
);
423 nfs_refresh_inode(inode
, fattr
);
424 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
426 (long long)NFS_FILEID(inode
),
427 nfs_display_fhandle_hash(fh
),
428 atomic_read(&inode
->i_count
));
434 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
437 EXPORT_SYMBOL_GPL(nfs_fhget
);
439 #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)
442 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
444 struct inode
*inode
= dentry
->d_inode
;
445 struct nfs_fattr
*fattr
;
448 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
450 /* skip mode change if it's just for clearing setuid/setgid */
451 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
452 attr
->ia_valid
&= ~ATTR_MODE
;
454 if (attr
->ia_valid
& ATTR_SIZE
) {
455 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
456 attr
->ia_valid
&= ~ATTR_SIZE
;
459 /* Optimization: if the end result is no change, don't RPC */
460 attr
->ia_valid
&= NFS_VALID_ATTRS
;
461 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
464 /* Write all dirty data */
465 if (S_ISREG(inode
->i_mode
)) {
466 nfs_inode_dio_wait(inode
);
470 fattr
= nfs_alloc_fattr();
474 * Return any delegations if we're going to change ACLs
476 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
477 NFS_PROTO(inode
)->return_delegation(inode
);
478 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
480 nfs_refresh_inode(inode
, fattr
);
481 nfs_free_fattr(fattr
);
485 EXPORT_SYMBOL_GPL(nfs_setattr
);
488 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
489 * @inode: inode of the file used
490 * @offset: file offset to start truncating
492 * This is a copy of the common vmtruncate, but with the locking
493 * corrected to take into account the fact that NFS requires
494 * inode->i_size to be updated under the inode->i_lock.
496 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
501 err
= inode_newsize_ok(inode
, offset
);
505 spin_lock(&inode
->i_lock
);
506 oldsize
= inode
->i_size
;
507 i_size_write(inode
, offset
);
508 spin_unlock(&inode
->i_lock
);
510 truncate_pagecache(inode
, oldsize
, offset
);
516 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
517 * @inode: pointer to struct inode
518 * @attr: pointer to struct iattr
520 * Note: we do this in the *proc.c in order to ensure that
521 * it works for things like exclusive creates too.
523 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
525 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
526 spin_lock(&inode
->i_lock
);
527 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
528 int mode
= attr
->ia_mode
& S_IALLUGO
;
529 mode
|= inode
->i_mode
& ~S_IALLUGO
;
530 inode
->i_mode
= mode
;
532 if ((attr
->ia_valid
& ATTR_UID
) != 0)
533 inode
->i_uid
= attr
->ia_uid
;
534 if ((attr
->ia_valid
& ATTR_GID
) != 0)
535 inode
->i_gid
= attr
->ia_gid
;
536 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
537 spin_unlock(&inode
->i_lock
);
539 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
540 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
541 nfs_vmtruncate(inode
, attr
->ia_size
);
544 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode
);
546 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
548 struct inode
*inode
= dentry
->d_inode
;
549 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
552 /* Flush out writes to the server in order to update c/mtime. */
553 if (S_ISREG(inode
->i_mode
)) {
554 nfs_inode_dio_wait(inode
);
555 err
= filemap_write_and_wait(inode
->i_mapping
);
561 * We may force a getattr if the user cares about atime.
563 * Note that we only have to check the vfsmount flags here:
564 * - NFS always sets S_NOATIME by so checking it would give a
566 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
567 * no point in checking those.
569 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
570 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
574 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
576 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
578 generic_fillattr(inode
, stat
);
579 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
584 EXPORT_SYMBOL_GPL(nfs_getattr
);
586 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
588 atomic_set(&l_ctx
->count
, 1);
589 l_ctx
->lockowner
.l_owner
= current
->files
;
590 l_ctx
->lockowner
.l_pid
= current
->tgid
;
591 INIT_LIST_HEAD(&l_ctx
->list
);
592 nfs_iocounter_init(&l_ctx
->io_count
);
595 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
597 struct nfs_lock_context
*head
= &ctx
->lock_context
;
598 struct nfs_lock_context
*pos
= head
;
601 if (pos
->lockowner
.l_owner
!= current
->files
)
603 if (pos
->lockowner
.l_pid
!= current
->tgid
)
605 atomic_inc(&pos
->count
);
607 } while ((pos
= list_entry(pos
->list
.next
, typeof(*pos
), list
)) != head
);
611 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
613 struct nfs_lock_context
*res
, *new = NULL
;
614 struct inode
*inode
= ctx
->dentry
->d_inode
;
616 spin_lock(&inode
->i_lock
);
617 res
= __nfs_find_lock_context(ctx
);
619 spin_unlock(&inode
->i_lock
);
620 new = kmalloc(sizeof(*new), GFP_KERNEL
);
622 return ERR_PTR(-ENOMEM
);
623 nfs_init_lock_context(new);
624 spin_lock(&inode
->i_lock
);
625 res
= __nfs_find_lock_context(ctx
);
627 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
628 new->open_context
= ctx
;
633 spin_unlock(&inode
->i_lock
);
638 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
640 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
641 struct inode
*inode
= ctx
->dentry
->d_inode
;
643 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
645 list_del(&l_ctx
->list
);
646 spin_unlock(&inode
->i_lock
);
651 * nfs_close_context - Common close_context() routine NFSv2/v3
652 * @ctx: pointer to context
653 * @is_sync: is this a synchronous close
655 * always ensure that the attributes are up to date if we're mounted
656 * with close-to-open semantics
658 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
661 struct nfs_server
*server
;
663 if (!(ctx
->mode
& FMODE_WRITE
))
667 inode
= ctx
->dentry
->d_inode
;
668 if (!list_empty(&NFS_I(inode
)->open_files
))
670 server
= NFS_SERVER(inode
);
671 if (server
->flags
& NFS_MOUNT_NOCTO
)
673 nfs_revalidate_inode(server
, inode
);
675 EXPORT_SYMBOL_GPL(nfs_close_context
);
677 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
679 struct nfs_open_context
*ctx
;
680 struct rpc_cred
*cred
= rpc_lookup_cred();
682 return ERR_CAST(cred
);
684 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
687 return ERR_PTR(-ENOMEM
);
689 nfs_sb_active(dentry
->d_sb
);
690 ctx
->dentry
= dget(dentry
);
696 nfs_init_lock_context(&ctx
->lock_context
);
697 ctx
->lock_context
.open_context
= ctx
;
698 INIT_LIST_HEAD(&ctx
->list
);
699 ctx
->mdsthreshold
= NULL
;
702 EXPORT_SYMBOL_GPL(alloc_nfs_open_context
);
704 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
707 atomic_inc(&ctx
->lock_context
.count
);
710 EXPORT_SYMBOL_GPL(get_nfs_open_context
);
712 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
714 struct inode
*inode
= ctx
->dentry
->d_inode
;
715 struct super_block
*sb
= ctx
->dentry
->d_sb
;
717 if (!list_empty(&ctx
->list
)) {
718 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
720 list_del(&ctx
->list
);
721 spin_unlock(&inode
->i_lock
);
722 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
725 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
726 if (ctx
->cred
!= NULL
)
727 put_rpccred(ctx
->cred
);
730 kfree(ctx
->mdsthreshold
);
734 void put_nfs_open_context(struct nfs_open_context
*ctx
)
736 __put_nfs_open_context(ctx
, 0);
738 EXPORT_SYMBOL_GPL(put_nfs_open_context
);
741 * Ensure that mmap has a recent RPC credential for use when writing out
744 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
746 struct inode
*inode
= file_inode(filp
);
747 struct nfs_inode
*nfsi
= NFS_I(inode
);
749 filp
->private_data
= get_nfs_open_context(ctx
);
750 spin_lock(&inode
->i_lock
);
751 list_add(&ctx
->list
, &nfsi
->open_files
);
752 spin_unlock(&inode
->i_lock
);
754 EXPORT_SYMBOL_GPL(nfs_file_set_open_context
);
757 * Given an inode, search for an open context with the desired characteristics
759 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
761 struct nfs_inode
*nfsi
= NFS_I(inode
);
762 struct nfs_open_context
*pos
, *ctx
= NULL
;
764 spin_lock(&inode
->i_lock
);
765 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
766 if (cred
!= NULL
&& pos
->cred
!= cred
)
768 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
770 ctx
= get_nfs_open_context(pos
);
773 spin_unlock(&inode
->i_lock
);
777 static void nfs_file_clear_open_context(struct file
*filp
)
779 struct inode
*inode
= file_inode(filp
);
780 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
783 filp
->private_data
= NULL
;
784 spin_lock(&inode
->i_lock
);
785 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
786 spin_unlock(&inode
->i_lock
);
787 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
792 * These allocate and release file read/write context information.
794 int nfs_open(struct inode
*inode
, struct file
*filp
)
796 struct nfs_open_context
*ctx
;
798 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
801 nfs_file_set_open_context(filp
, ctx
);
802 put_nfs_open_context(ctx
);
803 nfs_fscache_set_inode_cookie(inode
, filp
);
807 int nfs_release(struct inode
*inode
, struct file
*filp
)
809 nfs_file_clear_open_context(filp
);
814 * This function is called whenever some part of NFS notices that
815 * the cached attributes have to be refreshed.
818 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
820 int status
= -ESTALE
;
821 struct nfs4_label
*label
= NULL
;
822 struct nfs_fattr
*fattr
= NULL
;
823 struct nfs_inode
*nfsi
= NFS_I(inode
);
825 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
826 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
828 if (is_bad_inode(inode
))
830 if (NFS_STALE(inode
))
834 fattr
= nfs_alloc_fattr();
838 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
840 label
= nfs4_label_alloc(NFS_SERVER(inode
), GFP_KERNEL
);
842 status
= PTR_ERR(label
);
846 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
, label
);
848 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
850 (long long)NFS_FILEID(inode
), status
);
851 if (status
== -ESTALE
) {
852 nfs_zap_caches(inode
);
853 if (!S_ISDIR(inode
->i_mode
))
854 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
859 status
= nfs_refresh_inode(inode
, fattr
);
861 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
863 (long long)NFS_FILEID(inode
), status
);
867 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
868 nfs_zap_acl_cache(inode
);
870 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
872 (long long)NFS_FILEID(inode
));
875 nfs4_label_free(label
);
877 nfs_free_fattr(fattr
);
881 int nfs_attribute_timeout(struct inode
*inode
)
883 struct nfs_inode
*nfsi
= NFS_I(inode
);
885 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
888 static int nfs_attribute_cache_expired(struct inode
*inode
)
890 if (nfs_have_delegated_attributes(inode
))
892 return nfs_attribute_timeout(inode
);
896 * nfs_revalidate_inode - Revalidate the inode attributes
897 * @server - pointer to nfs_server struct
898 * @inode - pointer to inode struct
900 * Updates inode attribute information by retrieving the data from the server.
902 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
904 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
905 && !nfs_attribute_cache_expired(inode
))
906 return NFS_STALE(inode
) ? -ESTALE
: 0;
907 return __nfs_revalidate_inode(server
, inode
);
909 EXPORT_SYMBOL_GPL(nfs_revalidate_inode
);
911 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
913 struct nfs_inode
*nfsi
= NFS_I(inode
);
915 if (mapping
->nrpages
!= 0) {
916 int ret
= invalidate_inode_pages2(mapping
);
920 spin_lock(&inode
->i_lock
);
921 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
922 if (S_ISDIR(inode
->i_mode
))
923 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
924 spin_unlock(&inode
->i_lock
);
925 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
926 nfs_fscache_wait_on_invalidate(inode
);
927 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
928 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
932 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
934 if (nfs_have_delegated_attributes(inode
))
936 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
937 || nfs_attribute_timeout(inode
)
942 * nfs_revalidate_mapping - Revalidate the pagecache
943 * @inode - pointer to host inode
944 * @mapping - pointer to mapping
946 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
948 struct nfs_inode
*nfsi
= NFS_I(inode
);
951 /* swapfiles are not supposed to be shared. */
952 if (IS_SWAPFILE(inode
))
955 if (nfs_mapping_need_revalidate_inode(inode
)) {
956 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
960 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
961 ret
= nfs_invalidate_mapping(inode
, mapping
);
966 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
968 struct nfs_inode
*nfsi
= NFS_I(inode
);
969 unsigned long ret
= 0;
971 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
972 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
973 && inode
->i_version
== fattr
->pre_change_attr
) {
974 inode
->i_version
= fattr
->change_attr
;
975 if (S_ISDIR(inode
->i_mode
))
976 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
977 ret
|= NFS_INO_INVALID_ATTR
;
979 /* If we have atomic WCC data, we may update some attributes */
980 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
981 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
982 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
983 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
984 ret
|= NFS_INO_INVALID_ATTR
;
987 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
988 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
989 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
990 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
991 if (S_ISDIR(inode
->i_mode
))
992 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
993 ret
|= NFS_INO_INVALID_ATTR
;
995 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
996 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
997 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
998 && nfsi
->npages
== 0) {
999 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
1000 ret
|= NFS_INO_INVALID_ATTR
;
1003 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
1004 nfs_fscache_invalidate(inode
);
1010 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1011 * @inode - pointer to inode
1012 * @fattr - updated attributes
1014 * Verifies the attribute cache. If we have just changed the attributes,
1015 * so that fattr carries weak cache consistency data, then it may
1016 * also update the ctime/mtime/change_attribute.
1018 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1020 struct nfs_inode
*nfsi
= NFS_I(inode
);
1021 loff_t cur_size
, new_isize
;
1022 unsigned long invalid
= 0;
1025 if (nfs_have_delegated_attributes(inode
))
1027 /* Has the inode gone and changed behind our back? */
1028 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1030 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1033 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1034 inode
->i_version
!= fattr
->change_attr
)
1035 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1037 /* Verify a few of the more important attributes */
1038 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
1039 invalid
|= NFS_INO_INVALID_ATTR
;
1041 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1042 cur_size
= i_size_read(inode
);
1043 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1044 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1045 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1048 /* Have any file permissions changed? */
1049 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
1050 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1051 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && !uid_eq(inode
->i_uid
, fattr
->uid
))
1052 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1053 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && !gid_eq(inode
->i_gid
, fattr
->gid
))
1054 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1056 /* Has the link count changed? */
1057 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
1058 invalid
|= NFS_INO_INVALID_ATTR
;
1060 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1061 invalid
|= NFS_INO_INVALID_ATIME
;
1064 nfsi
->cache_validity
|= invalid
;
1066 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1070 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1072 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
1074 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1077 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1079 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1081 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1084 static atomic_long_t nfs_attr_generation_counter
;
1086 static unsigned long nfs_read_attr_generation_counter(void)
1088 return atomic_long_read(&nfs_attr_generation_counter
);
1091 unsigned long nfs_inc_attr_generation_counter(void)
1093 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1096 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1099 fattr
->time_start
= jiffies
;
1100 fattr
->gencount
= nfs_inc_attr_generation_counter();
1101 fattr
->owner_name
= NULL
;
1102 fattr
->group_name
= NULL
;
1104 EXPORT_SYMBOL_GPL(nfs_fattr_init
);
1106 struct nfs_fattr
*nfs_alloc_fattr(void)
1108 struct nfs_fattr
*fattr
;
1110 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1112 nfs_fattr_init(fattr
);
1115 EXPORT_SYMBOL_GPL(nfs_alloc_fattr
);
1117 struct nfs_fh
*nfs_alloc_fhandle(void)
1121 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1126 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle
);
1130 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1131 * in the same way that wireshark does
1135 * For debugging only.
1137 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1139 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1140 * not on the result */
1141 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1145 * _nfs_display_fhandle - display an NFS file handle on the console
1147 * @fh: file handle to display
1148 * @caption: display caption
1150 * For debugging only.
1152 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1156 if (fh
== NULL
|| fh
->size
== 0) {
1157 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1161 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1162 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1163 for (i
= 0; i
< fh
->size
; i
+= 16) {
1164 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1166 switch ((fh
->size
- i
- 1) >> 2) {
1168 printk(KERN_DEFAULT
" %08x\n",
1172 printk(KERN_DEFAULT
" %08x %08x\n",
1173 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1176 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1177 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1178 be32_to_cpup(pos
+ 2));
1181 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1182 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1183 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1190 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1191 * @inode - pointer to inode
1192 * @fattr - attributes
1194 * Attempt to divine whether or not an RPC call reply carrying stale
1195 * attributes got scheduled after another call carrying updated ones.
1197 * To do so, the function first assumes that a more recent ctime means
1198 * that the attributes in fattr are newer, however it also attempt to
1199 * catch the case where ctime either didn't change, or went backwards
1200 * (if someone reset the clock on the server) by looking at whether
1201 * or not this RPC call was started after the inode was last updated.
1202 * Note also the check for wraparound of 'attr_gencount'
1204 * The function returns 'true' if it thinks the attributes in 'fattr' are
1205 * more recent than the ones cached in the inode.
1208 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1210 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1212 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1213 nfs_ctime_need_update(inode
, fattr
) ||
1214 nfs_size_need_update(inode
, fattr
) ||
1215 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1218 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1220 if (nfs_inode_attrs_need_update(inode
, fattr
))
1221 return nfs_update_inode(inode
, fattr
);
1222 return nfs_check_inode_attributes(inode
, fattr
);
1226 * nfs_refresh_inode - try to update the inode attribute cache
1227 * @inode - pointer to inode
1228 * @fattr - updated attributes
1230 * Check that an RPC call that returned attributes has not overlapped with
1231 * other recent updates of the inode metadata, then decide whether it is
1232 * safe to do a full update of the inode attributes, or whether just to
1233 * call nfs_check_inode_attributes.
1235 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1239 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1241 spin_lock(&inode
->i_lock
);
1242 status
= nfs_refresh_inode_locked(inode
, fattr
);
1243 spin_unlock(&inode
->i_lock
);
1247 EXPORT_SYMBOL_GPL(nfs_refresh_inode
);
1249 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1251 struct nfs_inode
*nfsi
= NFS_I(inode
);
1253 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1254 if (S_ISDIR(inode
->i_mode
)) {
1255 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1256 nfs_fscache_invalidate(inode
);
1258 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1260 return nfs_refresh_inode_locked(inode
, fattr
);
1264 * nfs_post_op_update_inode - try to update the inode attribute cache
1265 * @inode - pointer to inode
1266 * @fattr - updated attributes
1268 * After an operation that has changed the inode metadata, mark the
1269 * attribute cache as being invalid, then try to update it.
1271 * NB: if the server didn't return any post op attributes, this
1272 * function will force the retrieval of attributes before the next
1273 * NFS request. Thus it should be used only for operations that
1274 * are expected to change one or more attributes, to avoid
1275 * unnecessary NFS requests and trips through nfs_update_inode().
1277 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1281 spin_lock(&inode
->i_lock
);
1282 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1283 spin_unlock(&inode
->i_lock
);
1286 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode
);
1289 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1290 * @inode - pointer to inode
1291 * @fattr - updated attributes
1293 * After an operation that has changed the inode metadata, mark the
1294 * attribute cache as being invalid, then try to update it. Fake up
1295 * weak cache consistency data, if none exist.
1297 * This function is mainly designed to be used by the ->write_done() functions.
1299 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1303 spin_lock(&inode
->i_lock
);
1304 /* Don't do a WCC update if these attributes are already stale */
1305 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1306 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1307 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1308 | NFS_ATTR_FATTR_PRESIZE
1309 | NFS_ATTR_FATTR_PREMTIME
1310 | NFS_ATTR_FATTR_PRECTIME
);
1313 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1314 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1315 fattr
->pre_change_attr
= inode
->i_version
;
1316 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1318 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1319 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1320 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1321 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1323 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1324 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1325 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1326 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1328 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1329 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1330 fattr
->pre_size
= i_size_read(inode
);
1331 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1334 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1335 spin_unlock(&inode
->i_lock
);
1338 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc
);
1341 * Many nfs protocol calls return the new file attributes after
1342 * an operation. Here we update the inode to reflect the state
1343 * of the server's inode.
1345 * This is a bit tricky because we have to make sure all dirty pages
1346 * have been sent off to the server before calling invalidate_inode_pages.
1347 * To make sure no other process adds more write requests while we try
1348 * our best to flush them, we make them sleep during the attribute refresh.
1350 * A very similar scenario holds for the dir cache.
1352 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1354 struct nfs_server
*server
;
1355 struct nfs_inode
*nfsi
= NFS_I(inode
);
1356 loff_t cur_isize
, new_isize
;
1357 unsigned long invalid
= 0;
1358 unsigned long now
= jiffies
;
1359 unsigned long save_cache_validity
;
1361 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1362 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1363 nfs_display_fhandle_hash(NFS_FH(inode
)),
1364 atomic_read(&inode
->i_count
), fattr
->valid
);
1366 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1367 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1368 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1369 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1370 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1371 (long long)fattr
->fileid
);
1376 * Make sure the inode's type hasn't changed.
1378 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1380 * Big trouble! The inode has become a different object.
1382 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1383 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1387 server
= NFS_SERVER(inode
);
1388 /* Update the fsid? */
1389 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1390 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1391 !IS_AUTOMOUNT(inode
))
1392 server
->fsid
= fattr
->fsid
;
1395 * Update the read time so we don't revalidate too often.
1397 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1399 save_cache_validity
= nfsi
->cache_validity
;
1400 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1401 | NFS_INO_INVALID_ATIME
1402 | NFS_INO_REVAL_FORCED
1403 | NFS_INO_REVAL_PAGECACHE
);
1405 /* Do atomic weak cache consistency updates */
1406 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1408 /* More cache consistency checks */
1409 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1410 if (inode
->i_version
!= fattr
->change_attr
) {
1411 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1412 inode
->i_sb
->s_id
, inode
->i_ino
);
1413 invalid
|= NFS_INO_INVALID_ATTR
1414 | NFS_INO_INVALID_DATA
1415 | NFS_INO_INVALID_ACCESS
1416 | NFS_INO_INVALID_ACL
1417 | NFS_INO_REVAL_PAGECACHE
;
1418 if (S_ISDIR(inode
->i_mode
))
1419 nfs_force_lookup_revalidate(inode
);
1420 inode
->i_version
= fattr
->change_attr
;
1422 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1423 invalid
|= save_cache_validity
;
1425 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1426 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1427 } else if (server
->caps
& NFS_CAP_MTIME
)
1428 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1429 | NFS_INO_REVAL_FORCED
);
1431 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1432 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1433 } else if (server
->caps
& NFS_CAP_CTIME
)
1434 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1435 | NFS_INO_REVAL_FORCED
);
1437 /* Check if our cached file size is stale */
1438 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1439 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1440 cur_isize
= i_size_read(inode
);
1441 if (new_isize
!= cur_isize
) {
1442 /* Do we perhaps have any outstanding writes, or has
1443 * the file grown beyond our last write? */
1444 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1445 new_isize
> cur_isize
) {
1446 i_size_write(inode
, new_isize
);
1447 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1449 dprintk("NFS: isize change on server for file %s/%ld "
1453 (long long)cur_isize
,
1454 (long long)new_isize
);
1457 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1458 | NFS_INO_REVAL_PAGECACHE
1459 | NFS_INO_REVAL_FORCED
);
1462 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1463 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1464 else if (server
->caps
& NFS_CAP_ATIME
)
1465 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1466 | NFS_INO_REVAL_FORCED
);
1468 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1469 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1470 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1471 newmode
|= fattr
->mode
& S_IALLUGO
;
1472 inode
->i_mode
= newmode
;
1473 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1475 } else if (server
->caps
& NFS_CAP_MODE
)
1476 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1477 | NFS_INO_INVALID_ACCESS
1478 | NFS_INO_INVALID_ACL
1479 | NFS_INO_REVAL_FORCED
);
1481 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1482 if (!uid_eq(inode
->i_uid
, fattr
->uid
)) {
1483 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1484 inode
->i_uid
= fattr
->uid
;
1486 } else if (server
->caps
& NFS_CAP_OWNER
)
1487 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1488 | NFS_INO_INVALID_ACCESS
1489 | NFS_INO_INVALID_ACL
1490 | NFS_INO_REVAL_FORCED
);
1492 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1493 if (!gid_eq(inode
->i_gid
, fattr
->gid
)) {
1494 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1495 inode
->i_gid
= fattr
->gid
;
1497 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1498 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1499 | NFS_INO_INVALID_ACCESS
1500 | NFS_INO_INVALID_ACL
1501 | NFS_INO_REVAL_FORCED
);
1503 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1504 if (inode
->i_nlink
!= fattr
->nlink
) {
1505 invalid
|= NFS_INO_INVALID_ATTR
;
1506 if (S_ISDIR(inode
->i_mode
))
1507 invalid
|= NFS_INO_INVALID_DATA
;
1508 set_nlink(inode
, fattr
->nlink
);
1510 } else if (server
->caps
& NFS_CAP_NLINK
)
1511 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1512 | NFS_INO_REVAL_FORCED
);
1514 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1516 * report the blocks in 512byte units
1518 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1520 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1521 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1523 /* Update attrtimeo value if we're out of the unstable period */
1524 if (invalid
& NFS_INO_INVALID_ATTR
) {
1525 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1526 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1527 nfsi
->attrtimeo_timestamp
= now
;
1528 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1530 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1531 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1532 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1533 nfsi
->attrtimeo_timestamp
= now
;
1536 invalid
&= ~NFS_INO_INVALID_ATTR
;
1537 /* Don't invalidate the data if we were to blame */
1538 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1539 || S_ISLNK(inode
->i_mode
)))
1540 invalid
&= ~NFS_INO_INVALID_DATA
;
1541 if (!NFS_PROTO(inode
)->have_delegation(inode
, FMODE_READ
) ||
1542 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1543 nfsi
->cache_validity
|= invalid
;
1545 if (invalid
& NFS_INO_INVALID_DATA
)
1546 nfs_fscache_invalidate(inode
);
1551 * No need to worry about unhashing the dentry, as the
1552 * lookup validation will know that the inode is bad.
1553 * (But we fall through to invalidate the caches.)
1555 nfs_invalidate_inode(inode
);
1559 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1561 struct nfs_inode
*nfsi
;
1562 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1566 nfsi
->cache_validity
= 0UL;
1567 #ifdef CONFIG_NFS_V3_ACL
1568 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1569 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1571 #if IS_ENABLED(CONFIG_NFS_V4)
1572 nfsi
->nfs4_acl
= NULL
;
1573 #endif /* CONFIG_NFS_V4 */
1574 return &nfsi
->vfs_inode
;
1576 EXPORT_SYMBOL_GPL(nfs_alloc_inode
);
1578 static void nfs_i_callback(struct rcu_head
*head
)
1580 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1581 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1584 void nfs_destroy_inode(struct inode
*inode
)
1586 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1588 EXPORT_SYMBOL_GPL(nfs_destroy_inode
);
1590 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1592 #if IS_ENABLED(CONFIG_NFS_V4)
1593 INIT_LIST_HEAD(&nfsi
->open_states
);
1594 nfsi
->delegation
= NULL
;
1595 nfsi
->delegation_state
= 0;
1596 init_rwsem(&nfsi
->rwsem
);
1597 nfsi
->layout
= NULL
;
1601 static void init_once(void *foo
)
1603 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1605 inode_init_once(&nfsi
->vfs_inode
);
1606 INIT_LIST_HEAD(&nfsi
->open_files
);
1607 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1608 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1609 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1611 nfsi
->commit_info
.ncommit
= 0;
1612 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1613 atomic_set(&nfsi
->silly_count
, 1);
1614 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1615 init_waitqueue_head(&nfsi
->waitqueue
);
1616 nfs4_init_once(nfsi
);
1619 static int __init
nfs_init_inodecache(void)
1621 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1622 sizeof(struct nfs_inode
),
1623 0, (SLAB_RECLAIM_ACCOUNT
|
1626 if (nfs_inode_cachep
== NULL
)
1632 static void nfs_destroy_inodecache(void)
1635 * Make sure all delayed rcu free inodes are flushed before we
1639 kmem_cache_destroy(nfs_inode_cachep
);
1642 struct workqueue_struct
*nfsiod_workqueue
;
1643 EXPORT_SYMBOL_GPL(nfsiod_workqueue
);
1646 * start up the nfsiod workqueue
1648 static int nfsiod_start(void)
1650 struct workqueue_struct
*wq
;
1651 dprintk("RPC: creating workqueue nfsiod\n");
1652 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1655 nfsiod_workqueue
= wq
;
1660 * Destroy the nfsiod workqueue
1662 static void nfsiod_stop(void)
1664 struct workqueue_struct
*wq
;
1666 wq
= nfsiod_workqueue
;
1669 nfsiod_workqueue
= NULL
;
1670 destroy_workqueue(wq
);
1674 EXPORT_SYMBOL_GPL(nfs_net_id
);
1676 static int nfs_net_init(struct net
*net
)
1678 nfs_clients_init(net
);
1679 return nfs_dns_resolver_cache_init(net
);
1682 static void nfs_net_exit(struct net
*net
)
1684 nfs_dns_resolver_cache_destroy(net
);
1685 nfs_cleanup_cb_ident_idr(net
);
1688 static struct pernet_operations nfs_net_ops
= {
1689 .init
= nfs_net_init
,
1690 .exit
= nfs_net_exit
,
1692 .size
= sizeof(struct nfs_net
),
1698 static int __init
init_nfs_fs(void)
1702 err
= nfs_dns_resolver_init();
1706 err
= register_pernet_subsys(&nfs_net_ops
);
1710 err
= nfs_fscache_register();
1714 err
= nfsiod_start();
1718 err
= nfs_fs_proc_init();
1722 err
= nfs_init_nfspagecache();
1726 err
= nfs_init_inodecache();
1730 err
= nfs_init_readpagecache();
1734 err
= nfs_init_writepagecache();
1738 err
= nfs_init_directcache();
1742 #ifdef CONFIG_PROC_FS
1743 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1745 if ((err
= register_nfs_fs()) != 0)
1750 #ifdef CONFIG_PROC_FS
1751 rpc_proc_unregister(&init_net
, "nfs");
1753 nfs_destroy_directcache();
1755 nfs_destroy_writepagecache();
1757 nfs_destroy_readpagecache();
1759 nfs_destroy_inodecache();
1761 nfs_destroy_nfspagecache();
1767 nfs_fscache_unregister();
1769 unregister_pernet_subsys(&nfs_net_ops
);
1771 nfs_dns_resolver_destroy();
1776 static void __exit
exit_nfs_fs(void)
1778 nfs_destroy_directcache();
1779 nfs_destroy_writepagecache();
1780 nfs_destroy_readpagecache();
1781 nfs_destroy_inodecache();
1782 nfs_destroy_nfspagecache();
1783 nfs_fscache_unregister();
1784 unregister_pernet_subsys(&nfs_net_ops
);
1785 nfs_dns_resolver_destroy();
1786 #ifdef CONFIG_PROC_FS
1787 rpc_proc_unregister(&init_net
, "nfs");
1789 unregister_nfs_fs();
1794 /* Not quite true; I just maintain it */
1795 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1796 MODULE_LICENSE("GPL");
1797 module_param(enable_ino64
, bool, 0644);
1799 module_init(init_nfs_fs
)
1800 module_exit(exit_nfs_fs
)