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
);
261 * This is our front-end to iget that looks up inodes by file handle
262 * instead of inode number.
265 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
267 struct nfs_find_desc desc
= {
271 struct inode
*inode
= ERR_PTR(-ENOENT
);
274 nfs_attr_check_mountpoint(sb
, fattr
);
276 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
277 !nfs_attr_use_mounted_on_fileid(fattr
))
279 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
282 hash
= nfs_fattr_to_ino_t(fattr
);
284 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
286 inode
= ERR_PTR(-ENOMEM
);
290 if (inode
->i_state
& I_NEW
) {
291 struct nfs_inode
*nfsi
= NFS_I(inode
);
292 unsigned long now
= jiffies
;
294 /* We set i_ino for the few things that still rely on it,
298 /* We can't support update_atime(), since the server will reset it */
299 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
300 inode
->i_mode
= fattr
->mode
;
301 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
302 && nfs_server_capable(inode
, NFS_CAP_MODE
))
303 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
304 /* Why so? Because we want revalidate for devices/FIFOs, and
305 * that's precisely what we have in nfs_file_inode_operations.
307 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
308 if (S_ISREG(inode
->i_mode
)) {
309 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
310 inode
->i_data
.a_ops
= &nfs_file_aops
;
311 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
312 } else if (S_ISDIR(inode
->i_mode
)) {
313 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
314 inode
->i_fop
= &nfs_dir_operations
;
315 inode
->i_data
.a_ops
= &nfs_dir_aops
;
316 /* Deal with crossing mountpoints */
317 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
318 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
319 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
320 inode
->i_op
= &nfs_referral_inode_operations
;
322 inode
->i_op
= &nfs_mountpoint_inode_operations
;
324 inode
->i_flags
|= S_AUTOMOUNT
;
326 } else if (S_ISLNK(inode
->i_mode
))
327 inode
->i_op
= &nfs_symlink_inode_operations
;
329 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
331 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
332 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
333 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
334 inode
->i_version
= 0;
337 inode
->i_uid
= make_kuid(&init_user_ns
, -2);
338 inode
->i_gid
= make_kgid(&init_user_ns
, -2);
340 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
344 nfsi
->read_cache_jiffies
= fattr
->time_start
;
345 nfsi
->attr_gencount
= fattr
->gencount
;
346 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
347 inode
->i_atime
= fattr
->atime
;
348 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
349 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
350 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
351 inode
->i_mtime
= fattr
->mtime
;
352 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
353 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
354 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
355 inode
->i_ctime
= fattr
->ctime
;
356 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
357 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
358 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
359 inode
->i_version
= fattr
->change_attr
;
360 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
361 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
362 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
363 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
365 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
366 | NFS_INO_REVAL_PAGECACHE
;
367 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
368 set_nlink(inode
, fattr
->nlink
);
369 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
370 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
371 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
372 inode
->i_uid
= fattr
->uid
;
373 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
374 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
375 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
376 inode
->i_gid
= fattr
->gid
;
377 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
378 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
379 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
380 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
381 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
383 * report the blocks in 512byte units
385 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
387 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
388 nfsi
->attrtimeo_timestamp
= now
;
389 nfsi
->access_cache
= RB_ROOT
;
391 nfs_fscache_init_inode_cookie(inode
);
393 unlock_new_inode(inode
);
395 nfs_refresh_inode(inode
, fattr
);
396 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
398 (long long)NFS_FILEID(inode
),
399 nfs_display_fhandle_hash(fh
),
400 atomic_read(&inode
->i_count
));
406 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
409 EXPORT_SYMBOL_GPL(nfs_fhget
);
411 #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)
414 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
416 struct inode
*inode
= dentry
->d_inode
;
417 struct nfs_fattr
*fattr
;
420 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
422 /* skip mode change if it's just for clearing setuid/setgid */
423 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
424 attr
->ia_valid
&= ~ATTR_MODE
;
426 if (attr
->ia_valid
& ATTR_SIZE
) {
427 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
428 attr
->ia_valid
&= ~ATTR_SIZE
;
431 /* Optimization: if the end result is no change, don't RPC */
432 attr
->ia_valid
&= NFS_VALID_ATTRS
;
433 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
436 /* Write all dirty data */
437 if (S_ISREG(inode
->i_mode
)) {
438 nfs_inode_dio_wait(inode
);
442 fattr
= nfs_alloc_fattr();
446 * Return any delegations if we're going to change ACLs
448 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
449 NFS_PROTO(inode
)->return_delegation(inode
);
450 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
452 nfs_refresh_inode(inode
, fattr
);
453 nfs_free_fattr(fattr
);
457 EXPORT_SYMBOL_GPL(nfs_setattr
);
460 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
461 * @inode: inode of the file used
462 * @offset: file offset to start truncating
464 * This is a copy of the common vmtruncate, but with the locking
465 * corrected to take into account the fact that NFS requires
466 * inode->i_size to be updated under the inode->i_lock.
468 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
473 err
= inode_newsize_ok(inode
, offset
);
477 spin_lock(&inode
->i_lock
);
478 oldsize
= inode
->i_size
;
479 i_size_write(inode
, offset
);
480 spin_unlock(&inode
->i_lock
);
482 truncate_pagecache(inode
, oldsize
, offset
);
488 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
489 * @inode: pointer to struct inode
490 * @attr: pointer to struct iattr
492 * Note: we do this in the *proc.c in order to ensure that
493 * it works for things like exclusive creates too.
495 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
497 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
498 spin_lock(&inode
->i_lock
);
499 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
500 int mode
= attr
->ia_mode
& S_IALLUGO
;
501 mode
|= inode
->i_mode
& ~S_IALLUGO
;
502 inode
->i_mode
= mode
;
504 if ((attr
->ia_valid
& ATTR_UID
) != 0)
505 inode
->i_uid
= attr
->ia_uid
;
506 if ((attr
->ia_valid
& ATTR_GID
) != 0)
507 inode
->i_gid
= attr
->ia_gid
;
508 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
509 spin_unlock(&inode
->i_lock
);
511 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
512 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
513 nfs_vmtruncate(inode
, attr
->ia_size
);
516 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode
);
518 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
520 struct inode
*inode
= dentry
->d_inode
;
521 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
524 /* Flush out writes to the server in order to update c/mtime. */
525 if (S_ISREG(inode
->i_mode
)) {
526 nfs_inode_dio_wait(inode
);
527 err
= filemap_write_and_wait(inode
->i_mapping
);
533 * We may force a getattr if the user cares about atime.
535 * Note that we only have to check the vfsmount flags here:
536 * - NFS always sets S_NOATIME by so checking it would give a
538 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
539 * no point in checking those.
541 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
542 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
546 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
548 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
550 generic_fillattr(inode
, stat
);
551 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
556 EXPORT_SYMBOL_GPL(nfs_getattr
);
558 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
560 atomic_set(&l_ctx
->count
, 1);
561 l_ctx
->lockowner
.l_owner
= current
->files
;
562 l_ctx
->lockowner
.l_pid
= current
->tgid
;
563 INIT_LIST_HEAD(&l_ctx
->list
);
564 nfs_iocounter_init(&l_ctx
->io_count
);
567 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
569 struct nfs_lock_context
*head
= &ctx
->lock_context
;
570 struct nfs_lock_context
*pos
= head
;
573 if (pos
->lockowner
.l_owner
!= current
->files
)
575 if (pos
->lockowner
.l_pid
!= current
->tgid
)
577 atomic_inc(&pos
->count
);
579 } while ((pos
= list_entry(pos
->list
.next
, typeof(*pos
), list
)) != head
);
583 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
585 struct nfs_lock_context
*res
, *new = NULL
;
586 struct inode
*inode
= ctx
->dentry
->d_inode
;
588 spin_lock(&inode
->i_lock
);
589 res
= __nfs_find_lock_context(ctx
);
591 spin_unlock(&inode
->i_lock
);
592 new = kmalloc(sizeof(*new), GFP_KERNEL
);
594 return ERR_PTR(-ENOMEM
);
595 nfs_init_lock_context(new);
596 spin_lock(&inode
->i_lock
);
597 res
= __nfs_find_lock_context(ctx
);
599 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
600 new->open_context
= ctx
;
605 spin_unlock(&inode
->i_lock
);
610 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
612 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
613 struct inode
*inode
= ctx
->dentry
->d_inode
;
615 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
617 list_del(&l_ctx
->list
);
618 spin_unlock(&inode
->i_lock
);
623 * nfs_close_context - Common close_context() routine NFSv2/v3
624 * @ctx: pointer to context
625 * @is_sync: is this a synchronous close
627 * always ensure that the attributes are up to date if we're mounted
628 * with close-to-open semantics
630 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
633 struct nfs_server
*server
;
635 if (!(ctx
->mode
& FMODE_WRITE
))
639 inode
= ctx
->dentry
->d_inode
;
640 if (!list_empty(&NFS_I(inode
)->open_files
))
642 server
= NFS_SERVER(inode
);
643 if (server
->flags
& NFS_MOUNT_NOCTO
)
645 nfs_revalidate_inode(server
, inode
);
647 EXPORT_SYMBOL_GPL(nfs_close_context
);
649 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
651 struct nfs_open_context
*ctx
;
652 struct rpc_cred
*cred
= rpc_lookup_cred();
654 return ERR_CAST(cred
);
656 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
659 return ERR_PTR(-ENOMEM
);
661 nfs_sb_active(dentry
->d_sb
);
662 ctx
->dentry
= dget(dentry
);
668 nfs_init_lock_context(&ctx
->lock_context
);
669 ctx
->lock_context
.open_context
= ctx
;
670 INIT_LIST_HEAD(&ctx
->list
);
671 ctx
->mdsthreshold
= NULL
;
674 EXPORT_SYMBOL_GPL(alloc_nfs_open_context
);
676 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
679 atomic_inc(&ctx
->lock_context
.count
);
682 EXPORT_SYMBOL_GPL(get_nfs_open_context
);
684 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
686 struct inode
*inode
= ctx
->dentry
->d_inode
;
687 struct super_block
*sb
= ctx
->dentry
->d_sb
;
689 if (!list_empty(&ctx
->list
)) {
690 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
692 list_del(&ctx
->list
);
693 spin_unlock(&inode
->i_lock
);
694 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
697 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
698 if (ctx
->cred
!= NULL
)
699 put_rpccred(ctx
->cred
);
702 kfree(ctx
->mdsthreshold
);
706 void put_nfs_open_context(struct nfs_open_context
*ctx
)
708 __put_nfs_open_context(ctx
, 0);
710 EXPORT_SYMBOL_GPL(put_nfs_open_context
);
713 * Ensure that mmap has a recent RPC credential for use when writing out
716 void nfs_inode_attach_open_context(struct nfs_open_context
*ctx
)
718 struct inode
*inode
= ctx
->dentry
->d_inode
;
719 struct nfs_inode
*nfsi
= NFS_I(inode
);
721 spin_lock(&inode
->i_lock
);
722 list_add(&ctx
->list
, &nfsi
->open_files
);
723 spin_unlock(&inode
->i_lock
);
725 EXPORT_SYMBOL_GPL(nfs_inode_attach_open_context
);
727 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
729 filp
->private_data
= get_nfs_open_context(ctx
);
730 if (list_empty(&ctx
->list
))
731 nfs_inode_attach_open_context(ctx
);
733 EXPORT_SYMBOL_GPL(nfs_file_set_open_context
);
736 * Given an inode, search for an open context with the desired characteristics
738 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
740 struct nfs_inode
*nfsi
= NFS_I(inode
);
741 struct nfs_open_context
*pos
, *ctx
= NULL
;
743 spin_lock(&inode
->i_lock
);
744 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
745 if (cred
!= NULL
&& pos
->cred
!= cred
)
747 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
749 ctx
= get_nfs_open_context(pos
);
752 spin_unlock(&inode
->i_lock
);
756 static void nfs_file_clear_open_context(struct file
*filp
)
758 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
761 struct inode
*inode
= ctx
->dentry
->d_inode
;
763 filp
->private_data
= NULL
;
764 spin_lock(&inode
->i_lock
);
765 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
766 spin_unlock(&inode
->i_lock
);
767 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
772 * These allocate and release file read/write context information.
774 int nfs_open(struct inode
*inode
, struct file
*filp
)
776 struct nfs_open_context
*ctx
;
778 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
781 nfs_file_set_open_context(filp
, ctx
);
782 put_nfs_open_context(ctx
);
783 nfs_fscache_set_inode_cookie(inode
, filp
);
787 int nfs_release(struct inode
*inode
, struct file
*filp
)
789 nfs_file_clear_open_context(filp
);
794 * This function is called whenever some part of NFS notices that
795 * the cached attributes have to be refreshed.
798 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
800 int status
= -ESTALE
;
801 struct nfs_fattr
*fattr
= NULL
;
802 struct nfs_inode
*nfsi
= NFS_I(inode
);
804 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
805 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
807 if (is_bad_inode(inode
))
809 if (NFS_STALE(inode
))
813 fattr
= nfs_alloc_fattr();
817 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
818 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
);
820 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
822 (long long)NFS_FILEID(inode
), status
);
823 if (status
== -ESTALE
) {
824 nfs_zap_caches(inode
);
825 if (!S_ISDIR(inode
->i_mode
))
826 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
831 status
= nfs_refresh_inode(inode
, fattr
);
833 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
835 (long long)NFS_FILEID(inode
), status
);
839 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
840 nfs_zap_acl_cache(inode
);
842 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
844 (long long)NFS_FILEID(inode
));
847 nfs_free_fattr(fattr
);
851 int nfs_attribute_timeout(struct inode
*inode
)
853 struct nfs_inode
*nfsi
= NFS_I(inode
);
855 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
858 static int nfs_attribute_cache_expired(struct inode
*inode
)
860 if (nfs_have_delegated_attributes(inode
))
862 return nfs_attribute_timeout(inode
);
866 * nfs_revalidate_inode - Revalidate the inode attributes
867 * @server - pointer to nfs_server struct
868 * @inode - pointer to inode struct
870 * Updates inode attribute information by retrieving the data from the server.
872 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
874 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
875 && !nfs_attribute_cache_expired(inode
))
876 return NFS_STALE(inode
) ? -ESTALE
: 0;
877 return __nfs_revalidate_inode(server
, inode
);
879 EXPORT_SYMBOL_GPL(nfs_revalidate_inode
);
881 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
883 struct nfs_inode
*nfsi
= NFS_I(inode
);
885 if (mapping
->nrpages
!= 0) {
886 int ret
= invalidate_inode_pages2(mapping
);
890 spin_lock(&inode
->i_lock
);
891 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
892 if (S_ISDIR(inode
->i_mode
))
893 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
894 spin_unlock(&inode
->i_lock
);
895 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
896 nfs_fscache_wait_on_invalidate(inode
);
897 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
898 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
902 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
904 if (nfs_have_delegated_attributes(inode
))
906 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
907 || nfs_attribute_timeout(inode
)
912 * nfs_revalidate_mapping - Revalidate the pagecache
913 * @inode - pointer to host inode
914 * @mapping - pointer to mapping
916 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
918 struct nfs_inode
*nfsi
= NFS_I(inode
);
921 /* swapfiles are not supposed to be shared. */
922 if (IS_SWAPFILE(inode
))
925 if (nfs_mapping_need_revalidate_inode(inode
)) {
926 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
930 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
931 ret
= nfs_invalidate_mapping(inode
, mapping
);
936 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
938 struct nfs_inode
*nfsi
= NFS_I(inode
);
939 unsigned long ret
= 0;
941 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
942 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
943 && inode
->i_version
== fattr
->pre_change_attr
) {
944 inode
->i_version
= fattr
->change_attr
;
945 if (S_ISDIR(inode
->i_mode
))
946 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
947 ret
|= NFS_INO_INVALID_ATTR
;
949 /* If we have atomic WCC data, we may update some attributes */
950 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
951 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
952 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
953 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
954 ret
|= NFS_INO_INVALID_ATTR
;
957 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
958 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
959 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
960 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
961 if (S_ISDIR(inode
->i_mode
))
962 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
963 ret
|= NFS_INO_INVALID_ATTR
;
965 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
966 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
967 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
968 && nfsi
->npages
== 0) {
969 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
970 ret
|= NFS_INO_INVALID_ATTR
;
973 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
974 nfs_fscache_invalidate(inode
);
980 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
981 * @inode - pointer to inode
982 * @fattr - updated attributes
984 * Verifies the attribute cache. If we have just changed the attributes,
985 * so that fattr carries weak cache consistency data, then it may
986 * also update the ctime/mtime/change_attribute.
988 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
990 struct nfs_inode
*nfsi
= NFS_I(inode
);
991 loff_t cur_size
, new_isize
;
992 unsigned long invalid
= 0;
995 if (nfs_have_delegated_attributes(inode
))
997 /* Has the inode gone and changed behind our back? */
998 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1000 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1003 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1004 inode
->i_version
!= fattr
->change_attr
)
1005 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1007 /* Verify a few of the more important attributes */
1008 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
1009 invalid
|= NFS_INO_INVALID_ATTR
;
1011 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1012 cur_size
= i_size_read(inode
);
1013 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1014 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1015 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1018 /* Have any file permissions changed? */
1019 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
1020 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1021 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && !uid_eq(inode
->i_uid
, fattr
->uid
))
1022 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1023 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && !gid_eq(inode
->i_gid
, fattr
->gid
))
1024 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1026 /* Has the link count changed? */
1027 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
1028 invalid
|= NFS_INO_INVALID_ATTR
;
1030 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1031 invalid
|= NFS_INO_INVALID_ATIME
;
1034 nfsi
->cache_validity
|= invalid
;
1036 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1040 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1042 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
1044 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1047 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1049 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1051 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1054 static atomic_long_t nfs_attr_generation_counter
;
1056 static unsigned long nfs_read_attr_generation_counter(void)
1058 return atomic_long_read(&nfs_attr_generation_counter
);
1061 unsigned long nfs_inc_attr_generation_counter(void)
1063 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1066 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1069 fattr
->time_start
= jiffies
;
1070 fattr
->gencount
= nfs_inc_attr_generation_counter();
1071 fattr
->owner_name
= NULL
;
1072 fattr
->group_name
= NULL
;
1074 EXPORT_SYMBOL_GPL(nfs_fattr_init
);
1076 struct nfs_fattr
*nfs_alloc_fattr(void)
1078 struct nfs_fattr
*fattr
;
1080 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1082 nfs_fattr_init(fattr
);
1085 EXPORT_SYMBOL_GPL(nfs_alloc_fattr
);
1087 struct nfs_fh
*nfs_alloc_fhandle(void)
1091 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1096 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle
);
1100 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1101 * in the same way that wireshark does
1105 * For debugging only.
1107 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1109 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1110 * not on the result */
1111 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1115 * _nfs_display_fhandle - display an NFS file handle on the console
1117 * @fh: file handle to display
1118 * @caption: display caption
1120 * For debugging only.
1122 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1126 if (fh
== NULL
|| fh
->size
== 0) {
1127 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1131 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1132 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1133 for (i
= 0; i
< fh
->size
; i
+= 16) {
1134 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1136 switch ((fh
->size
- i
- 1) >> 2) {
1138 printk(KERN_DEFAULT
" %08x\n",
1142 printk(KERN_DEFAULT
" %08x %08x\n",
1143 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1146 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1147 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1148 be32_to_cpup(pos
+ 2));
1151 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1152 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1153 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1160 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1161 * @inode - pointer to inode
1162 * @fattr - attributes
1164 * Attempt to divine whether or not an RPC call reply carrying stale
1165 * attributes got scheduled after another call carrying updated ones.
1167 * To do so, the function first assumes that a more recent ctime means
1168 * that the attributes in fattr are newer, however it also attempt to
1169 * catch the case where ctime either didn't change, or went backwards
1170 * (if someone reset the clock on the server) by looking at whether
1171 * or not this RPC call was started after the inode was last updated.
1172 * Note also the check for wraparound of 'attr_gencount'
1174 * The function returns 'true' if it thinks the attributes in 'fattr' are
1175 * more recent than the ones cached in the inode.
1178 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1180 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1182 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1183 nfs_ctime_need_update(inode
, fattr
) ||
1184 nfs_size_need_update(inode
, fattr
) ||
1185 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1188 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1190 if (nfs_inode_attrs_need_update(inode
, fattr
))
1191 return nfs_update_inode(inode
, fattr
);
1192 return nfs_check_inode_attributes(inode
, fattr
);
1196 * nfs_refresh_inode - try to update the inode attribute cache
1197 * @inode - pointer to inode
1198 * @fattr - updated attributes
1200 * Check that an RPC call that returned attributes has not overlapped with
1201 * other recent updates of the inode metadata, then decide whether it is
1202 * safe to do a full update of the inode attributes, or whether just to
1203 * call nfs_check_inode_attributes.
1205 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1209 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1211 spin_lock(&inode
->i_lock
);
1212 status
= nfs_refresh_inode_locked(inode
, fattr
);
1213 spin_unlock(&inode
->i_lock
);
1217 EXPORT_SYMBOL_GPL(nfs_refresh_inode
);
1219 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1221 struct nfs_inode
*nfsi
= NFS_I(inode
);
1223 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1224 if (S_ISDIR(inode
->i_mode
)) {
1225 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1226 nfs_fscache_invalidate(inode
);
1228 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1230 return nfs_refresh_inode_locked(inode
, fattr
);
1234 * nfs_post_op_update_inode - try to update the inode attribute cache
1235 * @inode - pointer to inode
1236 * @fattr - updated attributes
1238 * After an operation that has changed the inode metadata, mark the
1239 * attribute cache as being invalid, then try to update it.
1241 * NB: if the server didn't return any post op attributes, this
1242 * function will force the retrieval of attributes before the next
1243 * NFS request. Thus it should be used only for operations that
1244 * are expected to change one or more attributes, to avoid
1245 * unnecessary NFS requests and trips through nfs_update_inode().
1247 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1251 spin_lock(&inode
->i_lock
);
1252 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1253 spin_unlock(&inode
->i_lock
);
1256 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode
);
1259 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1260 * @inode - pointer to inode
1261 * @fattr - updated attributes
1263 * After an operation that has changed the inode metadata, mark the
1264 * attribute cache as being invalid, then try to update it. Fake up
1265 * weak cache consistency data, if none exist.
1267 * This function is mainly designed to be used by the ->write_done() functions.
1269 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1273 spin_lock(&inode
->i_lock
);
1274 /* Don't do a WCC update if these attributes are already stale */
1275 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1276 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1277 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1278 | NFS_ATTR_FATTR_PRESIZE
1279 | NFS_ATTR_FATTR_PREMTIME
1280 | NFS_ATTR_FATTR_PRECTIME
);
1283 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1284 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1285 fattr
->pre_change_attr
= inode
->i_version
;
1286 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1288 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1289 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1290 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1291 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1293 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1294 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1295 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1296 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1298 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1299 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1300 fattr
->pre_size
= i_size_read(inode
);
1301 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1304 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1305 spin_unlock(&inode
->i_lock
);
1308 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc
);
1311 * Many nfs protocol calls return the new file attributes after
1312 * an operation. Here we update the inode to reflect the state
1313 * of the server's inode.
1315 * This is a bit tricky because we have to make sure all dirty pages
1316 * have been sent off to the server before calling invalidate_inode_pages.
1317 * To make sure no other process adds more write requests while we try
1318 * our best to flush them, we make them sleep during the attribute refresh.
1320 * A very similar scenario holds for the dir cache.
1322 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1324 struct nfs_server
*server
;
1325 struct nfs_inode
*nfsi
= NFS_I(inode
);
1326 loff_t cur_isize
, new_isize
;
1327 unsigned long invalid
= 0;
1328 unsigned long now
= jiffies
;
1329 unsigned long save_cache_validity
;
1331 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1332 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1333 nfs_display_fhandle_hash(NFS_FH(inode
)),
1334 atomic_read(&inode
->i_count
), fattr
->valid
);
1336 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1337 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1338 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1339 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1340 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1341 (long long)fattr
->fileid
);
1346 * Make sure the inode's type hasn't changed.
1348 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1350 * Big trouble! The inode has become a different object.
1352 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1353 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1357 server
= NFS_SERVER(inode
);
1358 /* Update the fsid? */
1359 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1360 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1361 !IS_AUTOMOUNT(inode
))
1362 server
->fsid
= fattr
->fsid
;
1365 * Update the read time so we don't revalidate too often.
1367 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1369 save_cache_validity
= nfsi
->cache_validity
;
1370 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1371 | NFS_INO_INVALID_ATIME
1372 | NFS_INO_REVAL_FORCED
1373 | NFS_INO_REVAL_PAGECACHE
);
1375 /* Do atomic weak cache consistency updates */
1376 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1378 /* More cache consistency checks */
1379 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1380 if (inode
->i_version
!= fattr
->change_attr
) {
1381 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1382 inode
->i_sb
->s_id
, inode
->i_ino
);
1383 invalid
|= NFS_INO_INVALID_ATTR
1384 | NFS_INO_INVALID_DATA
1385 | NFS_INO_INVALID_ACCESS
1386 | NFS_INO_INVALID_ACL
1387 | NFS_INO_REVAL_PAGECACHE
;
1388 if (S_ISDIR(inode
->i_mode
))
1389 nfs_force_lookup_revalidate(inode
);
1390 inode
->i_version
= fattr
->change_attr
;
1392 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1393 invalid
|= save_cache_validity
;
1395 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1396 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1397 } else if (server
->caps
& NFS_CAP_MTIME
)
1398 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1399 | NFS_INO_REVAL_FORCED
);
1401 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1402 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1403 } else if (server
->caps
& NFS_CAP_CTIME
)
1404 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1405 | NFS_INO_REVAL_FORCED
);
1407 /* Check if our cached file size is stale */
1408 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1409 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1410 cur_isize
= i_size_read(inode
);
1411 if (new_isize
!= cur_isize
) {
1412 /* Do we perhaps have any outstanding writes, or has
1413 * the file grown beyond our last write? */
1414 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1415 new_isize
> cur_isize
) {
1416 i_size_write(inode
, new_isize
);
1417 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1419 dprintk("NFS: isize change on server for file %s/%ld "
1423 (long long)cur_isize
,
1424 (long long)new_isize
);
1427 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1428 | NFS_INO_REVAL_PAGECACHE
1429 | NFS_INO_REVAL_FORCED
);
1432 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1433 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1434 else if (server
->caps
& NFS_CAP_ATIME
)
1435 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1436 | NFS_INO_REVAL_FORCED
);
1438 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1439 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1440 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1441 newmode
|= fattr
->mode
& S_IALLUGO
;
1442 inode
->i_mode
= newmode
;
1443 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1445 } else if (server
->caps
& NFS_CAP_MODE
)
1446 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1447 | NFS_INO_INVALID_ACCESS
1448 | NFS_INO_INVALID_ACL
1449 | NFS_INO_REVAL_FORCED
);
1451 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1452 if (!uid_eq(inode
->i_uid
, fattr
->uid
)) {
1453 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1454 inode
->i_uid
= fattr
->uid
;
1456 } else if (server
->caps
& NFS_CAP_OWNER
)
1457 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1458 | NFS_INO_INVALID_ACCESS
1459 | NFS_INO_INVALID_ACL
1460 | NFS_INO_REVAL_FORCED
);
1462 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1463 if (!gid_eq(inode
->i_gid
, fattr
->gid
)) {
1464 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1465 inode
->i_gid
= fattr
->gid
;
1467 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1468 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1469 | NFS_INO_INVALID_ACCESS
1470 | NFS_INO_INVALID_ACL
1471 | NFS_INO_REVAL_FORCED
);
1473 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1474 if (inode
->i_nlink
!= fattr
->nlink
) {
1475 invalid
|= NFS_INO_INVALID_ATTR
;
1476 if (S_ISDIR(inode
->i_mode
))
1477 invalid
|= NFS_INO_INVALID_DATA
;
1478 set_nlink(inode
, fattr
->nlink
);
1480 } else if (server
->caps
& NFS_CAP_NLINK
)
1481 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1482 | NFS_INO_REVAL_FORCED
);
1484 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1486 * report the blocks in 512byte units
1488 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1490 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1491 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1493 /* Update attrtimeo value if we're out of the unstable period */
1494 if (invalid
& NFS_INO_INVALID_ATTR
) {
1495 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1496 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1497 nfsi
->attrtimeo_timestamp
= now
;
1498 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1500 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1501 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1502 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1503 nfsi
->attrtimeo_timestamp
= now
;
1506 invalid
&= ~NFS_INO_INVALID_ATTR
;
1507 /* Don't invalidate the data if we were to blame */
1508 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1509 || S_ISLNK(inode
->i_mode
)))
1510 invalid
&= ~NFS_INO_INVALID_DATA
;
1511 if (!NFS_PROTO(inode
)->have_delegation(inode
, FMODE_READ
) ||
1512 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1513 nfsi
->cache_validity
|= invalid
;
1515 if (invalid
& NFS_INO_INVALID_DATA
)
1516 nfs_fscache_invalidate(inode
);
1521 * No need to worry about unhashing the dentry, as the
1522 * lookup validation will know that the inode is bad.
1523 * (But we fall through to invalidate the caches.)
1525 nfs_invalidate_inode(inode
);
1529 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1531 struct nfs_inode
*nfsi
;
1532 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1536 nfsi
->cache_validity
= 0UL;
1537 #ifdef CONFIG_NFS_V3_ACL
1538 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1539 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1541 #if IS_ENABLED(CONFIG_NFS_V4)
1542 nfsi
->nfs4_acl
= NULL
;
1543 #endif /* CONFIG_NFS_V4 */
1544 return &nfsi
->vfs_inode
;
1546 EXPORT_SYMBOL_GPL(nfs_alloc_inode
);
1548 static void nfs_i_callback(struct rcu_head
*head
)
1550 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1551 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1554 void nfs_destroy_inode(struct inode
*inode
)
1556 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1558 EXPORT_SYMBOL_GPL(nfs_destroy_inode
);
1560 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1562 #if IS_ENABLED(CONFIG_NFS_V4)
1563 INIT_LIST_HEAD(&nfsi
->open_states
);
1564 nfsi
->delegation
= NULL
;
1565 nfsi
->delegation_state
= 0;
1566 init_rwsem(&nfsi
->rwsem
);
1567 nfsi
->layout
= NULL
;
1571 static void init_once(void *foo
)
1573 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1575 inode_init_once(&nfsi
->vfs_inode
);
1576 INIT_LIST_HEAD(&nfsi
->open_files
);
1577 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1578 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1579 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1581 nfsi
->commit_info
.ncommit
= 0;
1582 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1583 atomic_set(&nfsi
->silly_count
, 1);
1584 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1585 init_waitqueue_head(&nfsi
->waitqueue
);
1586 nfs4_init_once(nfsi
);
1589 static int __init
nfs_init_inodecache(void)
1591 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1592 sizeof(struct nfs_inode
),
1593 0, (SLAB_RECLAIM_ACCOUNT
|
1596 if (nfs_inode_cachep
== NULL
)
1602 static void nfs_destroy_inodecache(void)
1605 * Make sure all delayed rcu free inodes are flushed before we
1609 kmem_cache_destroy(nfs_inode_cachep
);
1612 struct workqueue_struct
*nfsiod_workqueue
;
1613 EXPORT_SYMBOL_GPL(nfsiod_workqueue
);
1616 * start up the nfsiod workqueue
1618 static int nfsiod_start(void)
1620 struct workqueue_struct
*wq
;
1621 dprintk("RPC: creating workqueue nfsiod\n");
1622 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1625 nfsiod_workqueue
= wq
;
1630 * Destroy the nfsiod workqueue
1632 static void nfsiod_stop(void)
1634 struct workqueue_struct
*wq
;
1636 wq
= nfsiod_workqueue
;
1639 nfsiod_workqueue
= NULL
;
1640 destroy_workqueue(wq
);
1644 EXPORT_SYMBOL_GPL(nfs_net_id
);
1646 static int nfs_net_init(struct net
*net
)
1648 nfs_clients_init(net
);
1649 return nfs_dns_resolver_cache_init(net
);
1652 static void nfs_net_exit(struct net
*net
)
1654 nfs_dns_resolver_cache_destroy(net
);
1655 nfs_cleanup_cb_ident_idr(net
);
1658 static struct pernet_operations nfs_net_ops
= {
1659 .init
= nfs_net_init
,
1660 .exit
= nfs_net_exit
,
1662 .size
= sizeof(struct nfs_net
),
1668 static int __init
init_nfs_fs(void)
1672 err
= nfs_dns_resolver_init();
1676 err
= register_pernet_subsys(&nfs_net_ops
);
1680 err
= nfs_fscache_register();
1684 err
= nfsiod_start();
1688 err
= nfs_fs_proc_init();
1692 err
= nfs_init_nfspagecache();
1696 err
= nfs_init_inodecache();
1700 err
= nfs_init_readpagecache();
1704 err
= nfs_init_writepagecache();
1708 err
= nfs_init_directcache();
1712 #ifdef CONFIG_PROC_FS
1713 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1715 if ((err
= register_nfs_fs()) != 0)
1720 #ifdef CONFIG_PROC_FS
1721 rpc_proc_unregister(&init_net
, "nfs");
1723 nfs_destroy_directcache();
1725 nfs_destroy_writepagecache();
1727 nfs_destroy_readpagecache();
1729 nfs_destroy_inodecache();
1731 nfs_destroy_nfspagecache();
1737 nfs_fscache_unregister();
1739 unregister_pernet_subsys(&nfs_net_ops
);
1741 nfs_dns_resolver_destroy();
1746 static void __exit
exit_nfs_fs(void)
1748 nfs_destroy_directcache();
1749 nfs_destroy_writepagecache();
1750 nfs_destroy_readpagecache();
1751 nfs_destroy_inodecache();
1752 nfs_destroy_nfspagecache();
1753 nfs_fscache_unregister();
1754 unregister_pernet_subsys(&nfs_net_ops
);
1755 nfs_dns_resolver_destroy();
1756 #ifdef CONFIG_PROC_FS
1757 rpc_proc_unregister(&init_net
, "nfs");
1759 unregister_nfs_fs();
1764 /* Not quite true; I just maintain it */
1765 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1766 MODULE_LICENSE("GPL");
1767 module_param(enable_ino64
, bool, 0644);
1769 module_init(init_nfs_fs
)
1770 module_exit(exit_nfs_fs
)