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 nfs_fscache_invalidate(inode
);
166 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
167 | NFS_INO_INVALID_LABEL
168 | NFS_INO_INVALID_DATA
169 | NFS_INO_INVALID_ACCESS
170 | NFS_INO_INVALID_ACL
171 | NFS_INO_REVAL_PAGECACHE
;
173 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
174 | NFS_INO_INVALID_LABEL
175 | NFS_INO_INVALID_ACCESS
176 | NFS_INO_INVALID_ACL
177 | NFS_INO_REVAL_PAGECACHE
;
180 void nfs_zap_caches(struct inode
*inode
)
182 spin_lock(&inode
->i_lock
);
183 nfs_zap_caches_locked(inode
);
184 spin_unlock(&inode
->i_lock
);
187 void nfs_zap_mapping(struct inode
*inode
, struct address_space
*mapping
)
189 if (mapping
->nrpages
!= 0) {
190 spin_lock(&inode
->i_lock
);
191 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_DATA
;
192 nfs_fscache_invalidate(inode
);
193 spin_unlock(&inode
->i_lock
);
197 void nfs_zap_acl_cache(struct inode
*inode
)
199 void (*clear_acl_cache
)(struct inode
*);
201 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
202 if (clear_acl_cache
!= NULL
)
203 clear_acl_cache(inode
);
204 spin_lock(&inode
->i_lock
);
205 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
206 spin_unlock(&inode
->i_lock
);
208 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache
);
210 void nfs_invalidate_atime(struct inode
*inode
)
212 spin_lock(&inode
->i_lock
);
213 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ATIME
;
214 spin_unlock(&inode
->i_lock
);
216 EXPORT_SYMBOL_GPL(nfs_invalidate_atime
);
219 * Invalidate, but do not unhash, the inode.
220 * NB: must be called with inode->i_lock held!
222 static void nfs_invalidate_inode(struct inode
*inode
)
224 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
225 nfs_zap_caches_locked(inode
);
228 struct nfs_find_desc
{
230 struct nfs_fattr
*fattr
;
234 * In NFSv3 we can have 64bit inode numbers. In order to support
235 * this, and re-exported directories (also seen in NFSv2)
236 * we are forced to allow 2 different inodes to have the same
240 nfs_find_actor(struct inode
*inode
, void *opaque
)
242 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
243 struct nfs_fh
*fh
= desc
->fh
;
244 struct nfs_fattr
*fattr
= desc
->fattr
;
246 if (NFS_FILEID(inode
) != fattr
->fileid
)
248 if ((S_IFMT
& inode
->i_mode
) != (S_IFMT
& fattr
->mode
))
250 if (nfs_compare_fh(NFS_FH(inode
), fh
))
252 if (is_bad_inode(inode
) || NFS_STALE(inode
))
258 nfs_init_locked(struct inode
*inode
, void *opaque
)
260 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
261 struct nfs_fattr
*fattr
= desc
->fattr
;
263 set_nfs_fileid(inode
, fattr
->fileid
);
264 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
268 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
269 void nfs_setsecurity(struct inode
*inode
, struct nfs_fattr
*fattr
,
270 struct nfs4_label
*label
)
277 if (nfs_server_capable(inode
, NFS_CAP_SECURITY_LABEL
) == 0)
280 if (NFS_SERVER(inode
)->nfs_client
->cl_minorversion
< 2)
283 if ((fattr
->valid
& NFS_ATTR_FATTR_V4_SECURITY_LABEL
) && inode
->i_security
) {
284 error
= security_inode_notifysecctx(inode
, label
->label
,
287 printk(KERN_ERR
"%s() %s %d "
288 "security_inode_notifysecctx() %d\n",
290 (char *)label
->label
,
295 struct nfs4_label
*nfs4_label_alloc(struct nfs_server
*server
, gfp_t flags
)
297 struct nfs4_label
*label
= NULL
;
298 int minor_version
= server
->nfs_client
->cl_minorversion
;
300 if (minor_version
< 2)
303 if (!(server
->caps
& NFS_CAP_SECURITY_LABEL
))
306 label
= kzalloc(sizeof(struct nfs4_label
), flags
);
308 return ERR_PTR(-ENOMEM
);
310 label
->label
= kzalloc(NFS4_MAXLABELLEN
, flags
);
311 if (label
->label
== NULL
) {
313 return ERR_PTR(-ENOMEM
);
315 label
->len
= NFS4_MAXLABELLEN
;
319 EXPORT_SYMBOL_GPL(nfs4_label_alloc
);
321 void inline nfs_setsecurity(struct inode
*inode
, struct nfs_fattr
*fattr
,
322 struct nfs4_label
*label
)
326 EXPORT_SYMBOL_GPL(nfs_setsecurity
);
329 * This is our front-end to iget that looks up inodes by file handle
330 * instead of inode number.
333 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
, struct nfs4_label
*label
)
335 struct nfs_find_desc desc
= {
339 struct inode
*inode
= ERR_PTR(-ENOENT
);
342 nfs_attr_check_mountpoint(sb
, fattr
);
344 if (((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) == 0) &&
345 !nfs_attr_use_mounted_on_fileid(fattr
))
347 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) == 0)
350 hash
= nfs_fattr_to_ino_t(fattr
);
352 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
354 inode
= ERR_PTR(-ENOMEM
);
358 if (inode
->i_state
& I_NEW
) {
359 struct nfs_inode
*nfsi
= NFS_I(inode
);
360 unsigned long now
= jiffies
;
362 /* We set i_ino for the few things that still rely on it,
366 /* We can't support update_atime(), since the server will reset it */
367 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
368 inode
->i_mode
= fattr
->mode
;
369 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) == 0
370 && nfs_server_capable(inode
, NFS_CAP_MODE
))
371 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
372 /* Why so? Because we want revalidate for devices/FIFOs, and
373 * that's precisely what we have in nfs_file_inode_operations.
375 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_inode_ops
;
376 if (S_ISREG(inode
->i_mode
)) {
377 inode
->i_fop
= NFS_SB(sb
)->nfs_client
->rpc_ops
->file_ops
;
378 inode
->i_data
.a_ops
= &nfs_file_aops
;
379 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
380 } else if (S_ISDIR(inode
->i_mode
)) {
381 inode
->i_op
= NFS_SB(sb
)->nfs_client
->rpc_ops
->dir_inode_ops
;
382 inode
->i_fop
= &nfs_dir_operations
;
383 inode
->i_data
.a_ops
= &nfs_dir_aops
;
384 /* Deal with crossing mountpoints */
385 if (fattr
->valid
& NFS_ATTR_FATTR_MOUNTPOINT
||
386 fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
) {
387 if (fattr
->valid
& NFS_ATTR_FATTR_V4_REFERRAL
)
388 inode
->i_op
= &nfs_referral_inode_operations
;
390 inode
->i_op
= &nfs_mountpoint_inode_operations
;
392 inode
->i_flags
|= S_AUTOMOUNT
;
394 } else if (S_ISLNK(inode
->i_mode
))
395 inode
->i_op
= &nfs_symlink_inode_operations
;
397 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
399 memset(&inode
->i_atime
, 0, sizeof(inode
->i_atime
));
400 memset(&inode
->i_mtime
, 0, sizeof(inode
->i_mtime
));
401 memset(&inode
->i_ctime
, 0, sizeof(inode
->i_ctime
));
402 inode
->i_version
= 0;
405 inode
->i_uid
= make_kuid(&init_user_ns
, -2);
406 inode
->i_gid
= make_kgid(&init_user_ns
, -2);
408 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
412 nfsi
->read_cache_jiffies
= fattr
->time_start
;
413 nfsi
->attr_gencount
= fattr
->gencount
;
414 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
415 inode
->i_atime
= fattr
->atime
;
416 else if (nfs_server_capable(inode
, NFS_CAP_ATIME
))
417 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
418 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
419 inode
->i_mtime
= fattr
->mtime
;
420 else if (nfs_server_capable(inode
, NFS_CAP_MTIME
))
421 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
422 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
423 inode
->i_ctime
= fattr
->ctime
;
424 else if (nfs_server_capable(inode
, NFS_CAP_CTIME
))
425 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
426 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
427 inode
->i_version
= fattr
->change_attr
;
428 else if (nfs_server_capable(inode
, NFS_CAP_CHANGE_ATTR
))
429 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
430 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
431 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
433 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
434 | NFS_INO_REVAL_PAGECACHE
;
435 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
)
436 set_nlink(inode
, fattr
->nlink
);
437 else if (nfs_server_capable(inode
, NFS_CAP_NLINK
))
438 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
439 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
)
440 inode
->i_uid
= fattr
->uid
;
441 else if (nfs_server_capable(inode
, NFS_CAP_OWNER
))
442 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
443 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
)
444 inode
->i_gid
= fattr
->gid
;
445 else if (nfs_server_capable(inode
, NFS_CAP_OWNER_GROUP
))
446 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
447 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
448 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
449 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
451 * report the blocks in 512byte units
453 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
456 nfs_setsecurity(inode
, fattr
, label
);
458 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
459 nfsi
->attrtimeo_timestamp
= now
;
460 nfsi
->access_cache
= RB_ROOT
;
462 nfs_fscache_init_inode_cookie(inode
);
464 unlock_new_inode(inode
);
466 nfs_refresh_inode(inode
, fattr
);
467 nfs_setsecurity(inode
, fattr
, label
);
468 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
470 (long long)NFS_FILEID(inode
),
471 nfs_display_fhandle_hash(fh
),
472 atomic_read(&inode
->i_count
));
478 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
481 EXPORT_SYMBOL_GPL(nfs_fhget
);
483 #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)
486 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
488 struct inode
*inode
= dentry
->d_inode
;
489 struct nfs_fattr
*fattr
;
492 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
494 /* skip mode change if it's just for clearing setuid/setgid */
495 if (attr
->ia_valid
& (ATTR_KILL_SUID
| ATTR_KILL_SGID
))
496 attr
->ia_valid
&= ~ATTR_MODE
;
498 if (attr
->ia_valid
& ATTR_SIZE
) {
499 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
500 attr
->ia_valid
&= ~ATTR_SIZE
;
503 /* Optimization: if the end result is no change, don't RPC */
504 attr
->ia_valid
&= NFS_VALID_ATTRS
;
505 if ((attr
->ia_valid
& ~(ATTR_FILE
|ATTR_OPEN
)) == 0)
508 /* Write all dirty data */
509 if (S_ISREG(inode
->i_mode
)) {
510 nfs_inode_dio_wait(inode
);
514 fattr
= nfs_alloc_fattr();
518 * Return any delegations if we're going to change ACLs
520 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
521 NFS_PROTO(inode
)->return_delegation(inode
);
522 error
= NFS_PROTO(inode
)->setattr(dentry
, fattr
, attr
);
524 error
= nfs_refresh_inode(inode
, fattr
);
525 nfs_free_fattr(fattr
);
529 EXPORT_SYMBOL_GPL(nfs_setattr
);
532 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
533 * @inode: inode of the file used
534 * @offset: file offset to start truncating
536 * This is a copy of the common vmtruncate, but with the locking
537 * corrected to take into account the fact that NFS requires
538 * inode->i_size to be updated under the inode->i_lock.
540 static int nfs_vmtruncate(struct inode
* inode
, loff_t offset
)
545 err
= inode_newsize_ok(inode
, offset
);
549 spin_lock(&inode
->i_lock
);
550 oldsize
= inode
->i_size
;
551 i_size_write(inode
, offset
);
552 spin_unlock(&inode
->i_lock
);
554 truncate_pagecache(inode
, oldsize
, offset
);
560 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
561 * @inode: pointer to struct inode
562 * @attr: pointer to struct iattr
564 * Note: we do this in the *proc.c in order to ensure that
565 * it works for things like exclusive creates too.
567 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
569 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
570 spin_lock(&inode
->i_lock
);
571 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
572 int mode
= attr
->ia_mode
& S_IALLUGO
;
573 mode
|= inode
->i_mode
& ~S_IALLUGO
;
574 inode
->i_mode
= mode
;
576 if ((attr
->ia_valid
& ATTR_UID
) != 0)
577 inode
->i_uid
= attr
->ia_uid
;
578 if ((attr
->ia_valid
& ATTR_GID
) != 0)
579 inode
->i_gid
= attr
->ia_gid
;
580 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
581 spin_unlock(&inode
->i_lock
);
583 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
584 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
585 nfs_vmtruncate(inode
, attr
->ia_size
);
588 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode
);
590 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
592 struct inode
*inode
= dentry
->d_inode
;
593 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
596 /* Flush out writes to the server in order to update c/mtime. */
597 if (S_ISREG(inode
->i_mode
)) {
598 nfs_inode_dio_wait(inode
);
599 err
= filemap_write_and_wait(inode
->i_mapping
);
605 * We may force a getattr if the user cares about atime.
607 * Note that we only have to check the vfsmount flags here:
608 * - NFS always sets S_NOATIME by so checking it would give a
610 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
611 * no point in checking those.
613 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
614 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
618 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
620 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
622 generic_fillattr(inode
, stat
);
623 stat
->ino
= nfs_compat_user_ino64(NFS_FILEID(inode
));
628 EXPORT_SYMBOL_GPL(nfs_getattr
);
630 static void nfs_init_lock_context(struct nfs_lock_context
*l_ctx
)
632 atomic_set(&l_ctx
->count
, 1);
633 l_ctx
->lockowner
.l_owner
= current
->files
;
634 l_ctx
->lockowner
.l_pid
= current
->tgid
;
635 INIT_LIST_HEAD(&l_ctx
->list
);
636 nfs_iocounter_init(&l_ctx
->io_count
);
639 static struct nfs_lock_context
*__nfs_find_lock_context(struct nfs_open_context
*ctx
)
641 struct nfs_lock_context
*head
= &ctx
->lock_context
;
642 struct nfs_lock_context
*pos
= head
;
645 if (pos
->lockowner
.l_owner
!= current
->files
)
647 if (pos
->lockowner
.l_pid
!= current
->tgid
)
649 atomic_inc(&pos
->count
);
651 } while ((pos
= list_entry(pos
->list
.next
, typeof(*pos
), list
)) != head
);
655 struct nfs_lock_context
*nfs_get_lock_context(struct nfs_open_context
*ctx
)
657 struct nfs_lock_context
*res
, *new = NULL
;
658 struct inode
*inode
= ctx
->dentry
->d_inode
;
660 spin_lock(&inode
->i_lock
);
661 res
= __nfs_find_lock_context(ctx
);
663 spin_unlock(&inode
->i_lock
);
664 new = kmalloc(sizeof(*new), GFP_KERNEL
);
666 return ERR_PTR(-ENOMEM
);
667 nfs_init_lock_context(new);
668 spin_lock(&inode
->i_lock
);
669 res
= __nfs_find_lock_context(ctx
);
671 list_add_tail(&new->list
, &ctx
->lock_context
.list
);
672 new->open_context
= ctx
;
677 spin_unlock(&inode
->i_lock
);
682 void nfs_put_lock_context(struct nfs_lock_context
*l_ctx
)
684 struct nfs_open_context
*ctx
= l_ctx
->open_context
;
685 struct inode
*inode
= ctx
->dentry
->d_inode
;
687 if (!atomic_dec_and_lock(&l_ctx
->count
, &inode
->i_lock
))
689 list_del(&l_ctx
->list
);
690 spin_unlock(&inode
->i_lock
);
695 * nfs_close_context - Common close_context() routine NFSv2/v3
696 * @ctx: pointer to context
697 * @is_sync: is this a synchronous close
699 * always ensure that the attributes are up to date if we're mounted
700 * with close-to-open semantics
702 void nfs_close_context(struct nfs_open_context
*ctx
, int is_sync
)
705 struct nfs_server
*server
;
707 if (!(ctx
->mode
& FMODE_WRITE
))
711 inode
= ctx
->dentry
->d_inode
;
712 if (!list_empty(&NFS_I(inode
)->open_files
))
714 server
= NFS_SERVER(inode
);
715 if (server
->flags
& NFS_MOUNT_NOCTO
)
717 nfs_revalidate_inode(server
, inode
);
719 EXPORT_SYMBOL_GPL(nfs_close_context
);
721 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, fmode_t f_mode
)
723 struct nfs_open_context
*ctx
;
724 struct rpc_cred
*cred
= rpc_lookup_cred();
726 return ERR_CAST(cred
);
728 ctx
= kmalloc(sizeof(*ctx
), GFP_KERNEL
);
731 return ERR_PTR(-ENOMEM
);
733 nfs_sb_active(dentry
->d_sb
);
734 ctx
->dentry
= dget(dentry
);
740 nfs_init_lock_context(&ctx
->lock_context
);
741 ctx
->lock_context
.open_context
= ctx
;
742 INIT_LIST_HEAD(&ctx
->list
);
743 ctx
->mdsthreshold
= NULL
;
746 EXPORT_SYMBOL_GPL(alloc_nfs_open_context
);
748 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
751 atomic_inc(&ctx
->lock_context
.count
);
754 EXPORT_SYMBOL_GPL(get_nfs_open_context
);
756 static void __put_nfs_open_context(struct nfs_open_context
*ctx
, int is_sync
)
758 struct inode
*inode
= ctx
->dentry
->d_inode
;
759 struct super_block
*sb
= ctx
->dentry
->d_sb
;
761 if (!list_empty(&ctx
->list
)) {
762 if (!atomic_dec_and_lock(&ctx
->lock_context
.count
, &inode
->i_lock
))
764 list_del(&ctx
->list
);
765 spin_unlock(&inode
->i_lock
);
766 } else if (!atomic_dec_and_test(&ctx
->lock_context
.count
))
769 NFS_PROTO(inode
)->close_context(ctx
, is_sync
);
770 if (ctx
->cred
!= NULL
)
771 put_rpccred(ctx
->cred
);
774 kfree(ctx
->mdsthreshold
);
778 void put_nfs_open_context(struct nfs_open_context
*ctx
)
780 __put_nfs_open_context(ctx
, 0);
782 EXPORT_SYMBOL_GPL(put_nfs_open_context
);
785 * Ensure that mmap has a recent RPC credential for use when writing out
788 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
790 struct inode
*inode
= file_inode(filp
);
791 struct nfs_inode
*nfsi
= NFS_I(inode
);
793 filp
->private_data
= get_nfs_open_context(ctx
);
794 spin_lock(&inode
->i_lock
);
795 list_add(&ctx
->list
, &nfsi
->open_files
);
796 spin_unlock(&inode
->i_lock
);
798 EXPORT_SYMBOL_GPL(nfs_file_set_open_context
);
801 * Given an inode, search for an open context with the desired characteristics
803 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, fmode_t mode
)
805 struct nfs_inode
*nfsi
= NFS_I(inode
);
806 struct nfs_open_context
*pos
, *ctx
= NULL
;
808 spin_lock(&inode
->i_lock
);
809 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
810 if (cred
!= NULL
&& pos
->cred
!= cred
)
812 if ((pos
->mode
& (FMODE_READ
|FMODE_WRITE
)) != mode
)
814 ctx
= get_nfs_open_context(pos
);
817 spin_unlock(&inode
->i_lock
);
821 static void nfs_file_clear_open_context(struct file
*filp
)
823 struct inode
*inode
= file_inode(filp
);
824 struct nfs_open_context
*ctx
= nfs_file_open_context(filp
);
827 filp
->private_data
= NULL
;
828 spin_lock(&inode
->i_lock
);
829 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
830 spin_unlock(&inode
->i_lock
);
831 __put_nfs_open_context(ctx
, filp
->f_flags
& O_DIRECT
? 0 : 1);
836 * These allocate and release file read/write context information.
838 int nfs_open(struct inode
*inode
, struct file
*filp
)
840 struct nfs_open_context
*ctx
;
842 ctx
= alloc_nfs_open_context(filp
->f_path
.dentry
, filp
->f_mode
);
845 nfs_file_set_open_context(filp
, ctx
);
846 put_nfs_open_context(ctx
);
847 nfs_fscache_set_inode_cookie(inode
, filp
);
851 int nfs_release(struct inode
*inode
, struct file
*filp
)
853 nfs_file_clear_open_context(filp
);
858 * This function is called whenever some part of NFS notices that
859 * the cached attributes have to be refreshed.
862 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
864 int status
= -ESTALE
;
865 struct nfs4_label
*label
= NULL
;
866 struct nfs_fattr
*fattr
= NULL
;
867 struct nfs_inode
*nfsi
= NFS_I(inode
);
869 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
870 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
872 if (is_bad_inode(inode
))
874 if (NFS_STALE(inode
))
878 fattr
= nfs_alloc_fattr();
882 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
884 label
= nfs4_label_alloc(NFS_SERVER(inode
), GFP_KERNEL
);
886 status
= PTR_ERR(label
);
890 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), fattr
, label
);
892 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
894 (long long)NFS_FILEID(inode
), status
);
895 if (status
== -ESTALE
) {
896 nfs_zap_caches(inode
);
897 if (!S_ISDIR(inode
->i_mode
))
898 set_bit(NFS_INO_STALE
, &NFS_I(inode
)->flags
);
903 status
= nfs_refresh_inode(inode
, fattr
);
905 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
907 (long long)NFS_FILEID(inode
), status
);
911 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
912 nfs_zap_acl_cache(inode
);
914 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
916 (long long)NFS_FILEID(inode
));
919 nfs4_label_free(label
);
921 nfs_free_fattr(fattr
);
925 int nfs_attribute_timeout(struct inode
*inode
)
927 struct nfs_inode
*nfsi
= NFS_I(inode
);
929 return !time_in_range_open(jiffies
, nfsi
->read_cache_jiffies
, nfsi
->read_cache_jiffies
+ nfsi
->attrtimeo
);
932 static int nfs_attribute_cache_expired(struct inode
*inode
)
934 if (nfs_have_delegated_attributes(inode
))
936 return nfs_attribute_timeout(inode
);
940 * nfs_revalidate_inode - Revalidate the inode attributes
941 * @server - pointer to nfs_server struct
942 * @inode - pointer to inode struct
944 * Updates inode attribute information by retrieving the data from the server.
946 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
948 if (!(NFS_I(inode
)->cache_validity
&
949 (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_LABEL
))
950 && !nfs_attribute_cache_expired(inode
))
951 return NFS_STALE(inode
) ? -ESTALE
: 0;
952 return __nfs_revalidate_inode(server
, inode
);
954 EXPORT_SYMBOL_GPL(nfs_revalidate_inode
);
956 static int nfs_invalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
958 struct nfs_inode
*nfsi
= NFS_I(inode
);
960 if (mapping
->nrpages
!= 0) {
961 int ret
= invalidate_inode_pages2(mapping
);
965 spin_lock(&inode
->i_lock
);
966 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
967 if (S_ISDIR(inode
->i_mode
))
968 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
969 spin_unlock(&inode
->i_lock
);
970 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
971 nfs_fscache_wait_on_invalidate(inode
);
972 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
973 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
977 static bool nfs_mapping_need_revalidate_inode(struct inode
*inode
)
979 if (nfs_have_delegated_attributes(inode
))
981 return (NFS_I(inode
)->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
982 || nfs_attribute_timeout(inode
)
987 * nfs_revalidate_mapping - Revalidate the pagecache
988 * @inode - pointer to host inode
989 * @mapping - pointer to mapping
991 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
993 struct nfs_inode
*nfsi
= NFS_I(inode
);
996 /* swapfiles are not supposed to be shared. */
997 if (IS_SWAPFILE(inode
))
1000 if (nfs_mapping_need_revalidate_inode(inode
)) {
1001 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1005 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
1006 ret
= nfs_invalidate_mapping(inode
, mapping
);
1011 static unsigned long nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1013 struct nfs_inode
*nfsi
= NFS_I(inode
);
1014 unsigned long ret
= 0;
1016 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
)
1017 && (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
)
1018 && inode
->i_version
== fattr
->pre_change_attr
) {
1019 inode
->i_version
= fattr
->change_attr
;
1020 if (S_ISDIR(inode
->i_mode
))
1021 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1022 ret
|= NFS_INO_INVALID_ATTR
;
1024 /* If we have atomic WCC data, we may update some attributes */
1025 if ((fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
)
1026 && (fattr
->valid
& NFS_ATTR_FATTR_CTIME
)
1027 && timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1028 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1029 ret
|= NFS_INO_INVALID_ATTR
;
1032 if ((fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
)
1033 && (fattr
->valid
& NFS_ATTR_FATTR_MTIME
)
1034 && timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1035 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1036 if (S_ISDIR(inode
->i_mode
))
1037 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1038 ret
|= NFS_INO_INVALID_ATTR
;
1040 if ((fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
)
1041 && (fattr
->valid
& NFS_ATTR_FATTR_SIZE
)
1042 && i_size_read(inode
) == nfs_size_to_loff_t(fattr
->pre_size
)
1043 && nfsi
->npages
== 0) {
1044 i_size_write(inode
, nfs_size_to_loff_t(fattr
->size
));
1045 ret
|= NFS_INO_INVALID_ATTR
;
1048 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
)
1049 nfs_fscache_invalidate(inode
);
1055 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1056 * @inode - pointer to inode
1057 * @fattr - updated attributes
1059 * Verifies the attribute cache. If we have just changed the attributes,
1060 * so that fattr carries weak cache consistency data, then it may
1061 * also update the ctime/mtime/change_attribute.
1063 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1065 struct nfs_inode
*nfsi
= NFS_I(inode
);
1066 loff_t cur_size
, new_isize
;
1067 unsigned long invalid
= 0;
1070 if (nfs_have_delegated_attributes(inode
))
1072 /* Has the inode gone and changed behind our back? */
1073 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
)
1075 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1078 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1079 inode
->i_version
!= fattr
->change_attr
)
1080 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1082 /* Verify a few of the more important attributes */
1083 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) && !timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
1084 invalid
|= NFS_INO_INVALID_ATTR
;
1086 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1087 cur_size
= i_size_read(inode
);
1088 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1089 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1090 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1093 /* Have any file permissions changed? */
1094 if ((fattr
->valid
& NFS_ATTR_FATTR_MODE
) && (inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
))
1095 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1096 if ((fattr
->valid
& NFS_ATTR_FATTR_OWNER
) && !uid_eq(inode
->i_uid
, fattr
->uid
))
1097 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1098 if ((fattr
->valid
& NFS_ATTR_FATTR_GROUP
) && !gid_eq(inode
->i_gid
, fattr
->gid
))
1099 invalid
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1101 /* Has the link count changed? */
1102 if ((fattr
->valid
& NFS_ATTR_FATTR_NLINK
) && inode
->i_nlink
!= fattr
->nlink
)
1103 invalid
|= NFS_INO_INVALID_ATTR
;
1105 if ((fattr
->valid
& NFS_ATTR_FATTR_ATIME
) && !timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1106 invalid
|= NFS_INO_INVALID_ATIME
;
1109 nfsi
->cache_validity
|= invalid
;
1111 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1115 static int nfs_ctime_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1117 if (!(fattr
->valid
& NFS_ATTR_FATTR_CTIME
))
1119 return timespec_compare(&fattr
->ctime
, &inode
->i_ctime
) > 0;
1122 static int nfs_size_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1124 if (!(fattr
->valid
& NFS_ATTR_FATTR_SIZE
))
1126 return nfs_size_to_loff_t(fattr
->size
) > i_size_read(inode
);
1129 static atomic_long_t nfs_attr_generation_counter
;
1131 static unsigned long nfs_read_attr_generation_counter(void)
1133 return atomic_long_read(&nfs_attr_generation_counter
);
1136 unsigned long nfs_inc_attr_generation_counter(void)
1138 return atomic_long_inc_return(&nfs_attr_generation_counter
);
1141 void nfs_fattr_init(struct nfs_fattr
*fattr
)
1144 fattr
->time_start
= jiffies
;
1145 fattr
->gencount
= nfs_inc_attr_generation_counter();
1146 fattr
->owner_name
= NULL
;
1147 fattr
->group_name
= NULL
;
1149 EXPORT_SYMBOL_GPL(nfs_fattr_init
);
1151 struct nfs_fattr
*nfs_alloc_fattr(void)
1153 struct nfs_fattr
*fattr
;
1155 fattr
= kmalloc(sizeof(*fattr
), GFP_NOFS
);
1157 nfs_fattr_init(fattr
);
1160 EXPORT_SYMBOL_GPL(nfs_alloc_fattr
);
1162 struct nfs_fh
*nfs_alloc_fhandle(void)
1166 fh
= kmalloc(sizeof(struct nfs_fh
), GFP_NOFS
);
1171 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle
);
1175 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1176 * in the same way that wireshark does
1180 * For debugging only.
1182 u32
_nfs_display_fhandle_hash(const struct nfs_fh
*fh
)
1184 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1185 * not on the result */
1186 return ~crc32(0xFFFFFFFF, &fh
->data
[0], fh
->size
);
1190 * _nfs_display_fhandle - display an NFS file handle on the console
1192 * @fh: file handle to display
1193 * @caption: display caption
1195 * For debugging only.
1197 void _nfs_display_fhandle(const struct nfs_fh
*fh
, const char *caption
)
1201 if (fh
== NULL
|| fh
->size
== 0) {
1202 printk(KERN_DEFAULT
"%s at %p is empty\n", caption
, fh
);
1206 printk(KERN_DEFAULT
"%s at %p is %u bytes, crc: 0x%08x:\n",
1207 caption
, fh
, fh
->size
, _nfs_display_fhandle_hash(fh
));
1208 for (i
= 0; i
< fh
->size
; i
+= 16) {
1209 __be32
*pos
= (__be32
*)&fh
->data
[i
];
1211 switch ((fh
->size
- i
- 1) >> 2) {
1213 printk(KERN_DEFAULT
" %08x\n",
1217 printk(KERN_DEFAULT
" %08x %08x\n",
1218 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1));
1221 printk(KERN_DEFAULT
" %08x %08x %08x\n",
1222 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1223 be32_to_cpup(pos
+ 2));
1226 printk(KERN_DEFAULT
" %08x %08x %08x %08x\n",
1227 be32_to_cpup(pos
), be32_to_cpup(pos
+ 1),
1228 be32_to_cpup(pos
+ 2), be32_to_cpup(pos
+ 3));
1235 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1236 * @inode - pointer to inode
1237 * @fattr - attributes
1239 * Attempt to divine whether or not an RPC call reply carrying stale
1240 * attributes got scheduled after another call carrying updated ones.
1242 * To do so, the function first assumes that a more recent ctime means
1243 * that the attributes in fattr are newer, however it also attempt to
1244 * catch the case where ctime either didn't change, or went backwards
1245 * (if someone reset the clock on the server) by looking at whether
1246 * or not this RPC call was started after the inode was last updated.
1247 * Note also the check for wraparound of 'attr_gencount'
1249 * The function returns 'true' if it thinks the attributes in 'fattr' are
1250 * more recent than the ones cached in the inode.
1253 static int nfs_inode_attrs_need_update(const struct inode
*inode
, const struct nfs_fattr
*fattr
)
1255 const struct nfs_inode
*nfsi
= NFS_I(inode
);
1257 return ((long)fattr
->gencount
- (long)nfsi
->attr_gencount
) > 0 ||
1258 nfs_ctime_need_update(inode
, fattr
) ||
1259 nfs_size_need_update(inode
, fattr
) ||
1260 ((long)nfsi
->attr_gencount
- (long)nfs_read_attr_generation_counter() > 0);
1263 static int nfs_refresh_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1265 if (nfs_inode_attrs_need_update(inode
, fattr
))
1266 return nfs_update_inode(inode
, fattr
);
1267 return nfs_check_inode_attributes(inode
, fattr
);
1271 * nfs_refresh_inode - try to update the inode attribute cache
1272 * @inode - pointer to inode
1273 * @fattr - updated attributes
1275 * Check that an RPC call that returned attributes has not overlapped with
1276 * other recent updates of the inode metadata, then decide whether it is
1277 * safe to do a full update of the inode attributes, or whether just to
1278 * call nfs_check_inode_attributes.
1280 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1284 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1286 spin_lock(&inode
->i_lock
);
1287 status
= nfs_refresh_inode_locked(inode
, fattr
);
1288 spin_unlock(&inode
->i_lock
);
1292 EXPORT_SYMBOL_GPL(nfs_refresh_inode
);
1294 static int nfs_post_op_update_inode_locked(struct inode
*inode
, struct nfs_fattr
*fattr
)
1296 struct nfs_inode
*nfsi
= NFS_I(inode
);
1298 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1299 if (S_ISDIR(inode
->i_mode
)) {
1300 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1301 nfs_fscache_invalidate(inode
);
1303 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1305 return nfs_refresh_inode_locked(inode
, fattr
);
1309 * nfs_post_op_update_inode - try to update the inode attribute cache
1310 * @inode - pointer to inode
1311 * @fattr - updated attributes
1313 * After an operation that has changed the inode metadata, mark the
1314 * attribute cache as being invalid, then try to update it.
1316 * NB: if the server didn't return any post op attributes, this
1317 * function will force the retrieval of attributes before the next
1318 * NFS request. Thus it should be used only for operations that
1319 * are expected to change one or more attributes, to avoid
1320 * unnecessary NFS requests and trips through nfs_update_inode().
1322 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1326 spin_lock(&inode
->i_lock
);
1327 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1328 spin_unlock(&inode
->i_lock
);
1332 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode
);
1335 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1336 * @inode - pointer to inode
1337 * @fattr - updated attributes
1339 * After an operation that has changed the inode metadata, mark the
1340 * attribute cache as being invalid, then try to update it. Fake up
1341 * weak cache consistency data, if none exist.
1343 * This function is mainly designed to be used by the ->write_done() functions.
1345 int nfs_post_op_update_inode_force_wcc(struct inode
*inode
, struct nfs_fattr
*fattr
)
1349 spin_lock(&inode
->i_lock
);
1350 /* Don't do a WCC update if these attributes are already stale */
1351 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0 ||
1352 !nfs_inode_attrs_need_update(inode
, fattr
)) {
1353 fattr
->valid
&= ~(NFS_ATTR_FATTR_PRECHANGE
1354 | NFS_ATTR_FATTR_PRESIZE
1355 | NFS_ATTR_FATTR_PREMTIME
1356 | NFS_ATTR_FATTR_PRECTIME
);
1359 if ((fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) != 0 &&
1360 (fattr
->valid
& NFS_ATTR_FATTR_PRECHANGE
) == 0) {
1361 fattr
->pre_change_attr
= inode
->i_version
;
1362 fattr
->valid
|= NFS_ATTR_FATTR_PRECHANGE
;
1364 if ((fattr
->valid
& NFS_ATTR_FATTR_CTIME
) != 0 &&
1365 (fattr
->valid
& NFS_ATTR_FATTR_PRECTIME
) == 0) {
1366 memcpy(&fattr
->pre_ctime
, &inode
->i_ctime
, sizeof(fattr
->pre_ctime
));
1367 fattr
->valid
|= NFS_ATTR_FATTR_PRECTIME
;
1369 if ((fattr
->valid
& NFS_ATTR_FATTR_MTIME
) != 0 &&
1370 (fattr
->valid
& NFS_ATTR_FATTR_PREMTIME
) == 0) {
1371 memcpy(&fattr
->pre_mtime
, &inode
->i_mtime
, sizeof(fattr
->pre_mtime
));
1372 fattr
->valid
|= NFS_ATTR_FATTR_PREMTIME
;
1374 if ((fattr
->valid
& NFS_ATTR_FATTR_SIZE
) != 0 &&
1375 (fattr
->valid
& NFS_ATTR_FATTR_PRESIZE
) == 0) {
1376 fattr
->pre_size
= i_size_read(inode
);
1377 fattr
->valid
|= NFS_ATTR_FATTR_PRESIZE
;
1380 status
= nfs_post_op_update_inode_locked(inode
, fattr
);
1381 spin_unlock(&inode
->i_lock
);
1384 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc
);
1387 * Many nfs protocol calls return the new file attributes after
1388 * an operation. Here we update the inode to reflect the state
1389 * of the server's inode.
1391 * This is a bit tricky because we have to make sure all dirty pages
1392 * have been sent off to the server before calling invalidate_inode_pages.
1393 * To make sure no other process adds more write requests while we try
1394 * our best to flush them, we make them sleep during the attribute refresh.
1396 * A very similar scenario holds for the dir cache.
1398 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1400 struct nfs_server
*server
;
1401 struct nfs_inode
*nfsi
= NFS_I(inode
);
1402 loff_t cur_isize
, new_isize
;
1403 unsigned long invalid
= 0;
1404 unsigned long now
= jiffies
;
1405 unsigned long save_cache_validity
;
1407 dfprintk(VFS
, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1408 __func__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1409 nfs_display_fhandle_hash(NFS_FH(inode
)),
1410 atomic_read(&inode
->i_count
), fattr
->valid
);
1412 if ((fattr
->valid
& NFS_ATTR_FATTR_FILEID
) && nfsi
->fileid
!= fattr
->fileid
) {
1413 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1414 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1415 NFS_SERVER(inode
)->nfs_client
->cl_hostname
,
1416 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1417 (long long)fattr
->fileid
);
1422 * Make sure the inode's type hasn't changed.
1424 if ((fattr
->valid
& NFS_ATTR_FATTR_TYPE
) && (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1426 * Big trouble! The inode has become a different object.
1428 printk(KERN_DEBUG
"NFS: %s: inode %ld mode changed, %07o to %07o\n",
1429 __func__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1433 server
= NFS_SERVER(inode
);
1434 /* Update the fsid? */
1435 if (S_ISDIR(inode
->i_mode
) && (fattr
->valid
& NFS_ATTR_FATTR_FSID
) &&
1436 !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
) &&
1437 !IS_AUTOMOUNT(inode
))
1438 server
->fsid
= fattr
->fsid
;
1441 * Update the read time so we don't revalidate too often.
1443 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1445 save_cache_validity
= nfsi
->cache_validity
;
1446 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
1447 | NFS_INO_INVALID_ATIME
1448 | NFS_INO_REVAL_FORCED
1449 | NFS_INO_REVAL_PAGECACHE
);
1451 /* Do atomic weak cache consistency updates */
1452 invalid
|= nfs_wcc_update_inode(inode
, fattr
);
1454 /* More cache consistency checks */
1455 if (fattr
->valid
& NFS_ATTR_FATTR_CHANGE
) {
1456 if (inode
->i_version
!= fattr
->change_attr
) {
1457 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1458 inode
->i_sb
->s_id
, inode
->i_ino
);
1459 invalid
|= NFS_INO_INVALID_ATTR
1460 | NFS_INO_INVALID_DATA
1461 | NFS_INO_INVALID_ACCESS
1462 | NFS_INO_INVALID_ACL
1463 | NFS_INO_REVAL_PAGECACHE
;
1464 if (S_ISDIR(inode
->i_mode
))
1465 nfs_force_lookup_revalidate(inode
);
1466 inode
->i_version
= fattr
->change_attr
;
1468 } else if (server
->caps
& NFS_CAP_CHANGE_ATTR
)
1469 invalid
|= save_cache_validity
;
1471 if (fattr
->valid
& NFS_ATTR_FATTR_MTIME
) {
1472 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1473 } else if (server
->caps
& NFS_CAP_MTIME
)
1474 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1475 | NFS_INO_REVAL_FORCED
);
1477 if (fattr
->valid
& NFS_ATTR_FATTR_CTIME
) {
1478 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1479 } else if (server
->caps
& NFS_CAP_CTIME
)
1480 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1481 | NFS_INO_REVAL_FORCED
);
1483 /* Check if our cached file size is stale */
1484 if (fattr
->valid
& NFS_ATTR_FATTR_SIZE
) {
1485 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1486 cur_isize
= i_size_read(inode
);
1487 if (new_isize
!= cur_isize
) {
1488 /* Do we perhaps have any outstanding writes, or has
1489 * the file grown beyond our last write? */
1490 if ((nfsi
->npages
== 0 && !test_bit(NFS_INO_LAYOUTCOMMIT
, &nfsi
->flags
)) ||
1491 new_isize
> cur_isize
) {
1492 i_size_write(inode
, new_isize
);
1493 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1495 dprintk("NFS: isize change on server for file %s/%ld "
1499 (long long)cur_isize
,
1500 (long long)new_isize
);
1503 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1504 | NFS_INO_REVAL_PAGECACHE
1505 | NFS_INO_REVAL_FORCED
);
1508 if (fattr
->valid
& NFS_ATTR_FATTR_ATIME
)
1509 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1510 else if (server
->caps
& NFS_CAP_ATIME
)
1511 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATIME
1512 | NFS_INO_REVAL_FORCED
);
1514 if (fattr
->valid
& NFS_ATTR_FATTR_MODE
) {
1515 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)) {
1516 umode_t newmode
= inode
->i_mode
& S_IFMT
;
1517 newmode
|= fattr
->mode
& S_IALLUGO
;
1518 inode
->i_mode
= newmode
;
1519 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1521 } else if (server
->caps
& NFS_CAP_MODE
)
1522 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1523 | NFS_INO_INVALID_ACCESS
1524 | NFS_INO_INVALID_ACL
1525 | NFS_INO_REVAL_FORCED
);
1527 if (fattr
->valid
& NFS_ATTR_FATTR_OWNER
) {
1528 if (!uid_eq(inode
->i_uid
, fattr
->uid
)) {
1529 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1530 inode
->i_uid
= fattr
->uid
;
1532 } else if (server
->caps
& NFS_CAP_OWNER
)
1533 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1534 | NFS_INO_INVALID_ACCESS
1535 | NFS_INO_INVALID_ACL
1536 | NFS_INO_REVAL_FORCED
);
1538 if (fattr
->valid
& NFS_ATTR_FATTR_GROUP
) {
1539 if (!gid_eq(inode
->i_gid
, fattr
->gid
)) {
1540 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1541 inode
->i_gid
= fattr
->gid
;
1543 } else if (server
->caps
& NFS_CAP_OWNER_GROUP
)
1544 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1545 | NFS_INO_INVALID_ACCESS
1546 | NFS_INO_INVALID_ACL
1547 | NFS_INO_REVAL_FORCED
);
1549 if (fattr
->valid
& NFS_ATTR_FATTR_NLINK
) {
1550 if (inode
->i_nlink
!= fattr
->nlink
) {
1551 invalid
|= NFS_INO_INVALID_ATTR
;
1552 if (S_ISDIR(inode
->i_mode
))
1553 invalid
|= NFS_INO_INVALID_DATA
;
1554 set_nlink(inode
, fattr
->nlink
);
1556 } else if (server
->caps
& NFS_CAP_NLINK
)
1557 invalid
|= save_cache_validity
& (NFS_INO_INVALID_ATTR
1558 | NFS_INO_REVAL_FORCED
);
1560 if (fattr
->valid
& NFS_ATTR_FATTR_SPACE_USED
) {
1562 * report the blocks in 512byte units
1564 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1566 if (fattr
->valid
& NFS_ATTR_FATTR_BLOCKS_USED
)
1567 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1569 /* Update attrtimeo value if we're out of the unstable period */
1570 if (invalid
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_LABEL
)) {
1571 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1572 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1573 nfsi
->attrtimeo_timestamp
= now
;
1574 nfsi
->attr_gencount
= nfs_inc_attr_generation_counter();
1576 if (!time_in_range_open(now
, nfsi
->attrtimeo_timestamp
, nfsi
->attrtimeo_timestamp
+ nfsi
->attrtimeo
)) {
1577 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1578 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1579 nfsi
->attrtimeo_timestamp
= now
;
1582 invalid
&= ~NFS_INO_INVALID_ATTR
;
1583 invalid
&= ~NFS_INO_INVALID_LABEL
;
1584 /* Don't invalidate the data if we were to blame */
1585 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1586 || S_ISLNK(inode
->i_mode
)))
1587 invalid
&= ~NFS_INO_INVALID_DATA
;
1588 if (!NFS_PROTO(inode
)->have_delegation(inode
, FMODE_READ
) ||
1589 (save_cache_validity
& NFS_INO_REVAL_FORCED
))
1590 nfsi
->cache_validity
|= invalid
;
1592 if (invalid
& NFS_INO_INVALID_DATA
)
1593 nfs_fscache_invalidate(inode
);
1598 * No need to worry about unhashing the dentry, as the
1599 * lookup validation will know that the inode is bad.
1600 * (But we fall through to invalidate the caches.)
1602 nfs_invalidate_inode(inode
);
1606 struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1608 struct nfs_inode
*nfsi
;
1609 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, GFP_KERNEL
);
1613 nfsi
->cache_validity
= 0UL;
1614 #ifdef CONFIG_NFS_V3_ACL
1615 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
1616 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
1618 #if IS_ENABLED(CONFIG_NFS_V4)
1619 nfsi
->nfs4_acl
= NULL
;
1620 #endif /* CONFIG_NFS_V4 */
1621 return &nfsi
->vfs_inode
;
1623 EXPORT_SYMBOL_GPL(nfs_alloc_inode
);
1625 static void nfs_i_callback(struct rcu_head
*head
)
1627 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
1628 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1631 void nfs_destroy_inode(struct inode
*inode
)
1633 call_rcu(&inode
->i_rcu
, nfs_i_callback
);
1635 EXPORT_SYMBOL_GPL(nfs_destroy_inode
);
1637 static inline void nfs4_init_once(struct nfs_inode
*nfsi
)
1639 #if IS_ENABLED(CONFIG_NFS_V4)
1640 INIT_LIST_HEAD(&nfsi
->open_states
);
1641 nfsi
->delegation
= NULL
;
1642 nfsi
->delegation_state
= 0;
1643 init_rwsem(&nfsi
->rwsem
);
1644 nfsi
->layout
= NULL
;
1648 static void init_once(void *foo
)
1650 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1652 inode_init_once(&nfsi
->vfs_inode
);
1653 INIT_LIST_HEAD(&nfsi
->open_files
);
1654 INIT_LIST_HEAD(&nfsi
->access_cache_entry_lru
);
1655 INIT_LIST_HEAD(&nfsi
->access_cache_inode_lru
);
1656 INIT_LIST_HEAD(&nfsi
->commit_info
.list
);
1658 nfsi
->commit_info
.ncommit
= 0;
1659 atomic_set(&nfsi
->commit_info
.rpcs_out
, 0);
1660 atomic_set(&nfsi
->silly_count
, 1);
1661 INIT_HLIST_HEAD(&nfsi
->silly_list
);
1662 init_waitqueue_head(&nfsi
->waitqueue
);
1663 nfs4_init_once(nfsi
);
1666 static int __init
nfs_init_inodecache(void)
1668 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1669 sizeof(struct nfs_inode
),
1670 0, (SLAB_RECLAIM_ACCOUNT
|
1673 if (nfs_inode_cachep
== NULL
)
1679 static void nfs_destroy_inodecache(void)
1682 * Make sure all delayed rcu free inodes are flushed before we
1686 kmem_cache_destroy(nfs_inode_cachep
);
1689 struct workqueue_struct
*nfsiod_workqueue
;
1690 EXPORT_SYMBOL_GPL(nfsiod_workqueue
);
1693 * start up the nfsiod workqueue
1695 static int nfsiod_start(void)
1697 struct workqueue_struct
*wq
;
1698 dprintk("RPC: creating workqueue nfsiod\n");
1699 wq
= alloc_workqueue("nfsiod", WQ_MEM_RECLAIM
, 0);
1702 nfsiod_workqueue
= wq
;
1707 * Destroy the nfsiod workqueue
1709 static void nfsiod_stop(void)
1711 struct workqueue_struct
*wq
;
1713 wq
= nfsiod_workqueue
;
1716 nfsiod_workqueue
= NULL
;
1717 destroy_workqueue(wq
);
1721 EXPORT_SYMBOL_GPL(nfs_net_id
);
1723 static int nfs_net_init(struct net
*net
)
1725 nfs_clients_init(net
);
1726 return nfs_dns_resolver_cache_init(net
);
1729 static void nfs_net_exit(struct net
*net
)
1731 nfs_dns_resolver_cache_destroy(net
);
1732 nfs_cleanup_cb_ident_idr(net
);
1735 static struct pernet_operations nfs_net_ops
= {
1736 .init
= nfs_net_init
,
1737 .exit
= nfs_net_exit
,
1739 .size
= sizeof(struct nfs_net
),
1745 static int __init
init_nfs_fs(void)
1749 err
= nfs_dns_resolver_init();
1753 err
= register_pernet_subsys(&nfs_net_ops
);
1757 err
= nfs_fscache_register();
1761 err
= nfsiod_start();
1765 err
= nfs_fs_proc_init();
1769 err
= nfs_init_nfspagecache();
1773 err
= nfs_init_inodecache();
1777 err
= nfs_init_readpagecache();
1781 err
= nfs_init_writepagecache();
1785 err
= nfs_init_directcache();
1789 #ifdef CONFIG_PROC_FS
1790 rpc_proc_register(&init_net
, &nfs_rpcstat
);
1792 if ((err
= register_nfs_fs()) != 0)
1797 #ifdef CONFIG_PROC_FS
1798 rpc_proc_unregister(&init_net
, "nfs");
1800 nfs_destroy_directcache();
1802 nfs_destroy_writepagecache();
1804 nfs_destroy_readpagecache();
1806 nfs_destroy_inodecache();
1808 nfs_destroy_nfspagecache();
1814 nfs_fscache_unregister();
1816 unregister_pernet_subsys(&nfs_net_ops
);
1818 nfs_dns_resolver_destroy();
1823 static void __exit
exit_nfs_fs(void)
1825 nfs_destroy_directcache();
1826 nfs_destroy_writepagecache();
1827 nfs_destroy_readpagecache();
1828 nfs_destroy_inodecache();
1829 nfs_destroy_nfspagecache();
1830 nfs_fscache_unregister();
1831 unregister_pernet_subsys(&nfs_net_ops
);
1832 nfs_dns_resolver_destroy();
1833 #ifdef CONFIG_PROC_FS
1834 rpc_proc_unregister(&init_net
, "nfs");
1836 unregister_nfs_fs();
1841 /* Not quite true; I just maintain it */
1842 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1843 MODULE_LICENSE("GPL");
1844 module_param(enable_ino64
, bool, 0644);
1846 module_init(init_nfs_fs
)
1847 module_exit(exit_nfs_fs
)