4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/time.h>
20 #include <linux/kernel.h>
22 #include <linux/string.h>
23 #include <linux/stat.h>
24 #include <linux/errno.h>
25 #include <linux/unistd.h>
26 #include <linux/sunrpc/clnt.h>
27 #include <linux/sunrpc/stats.h>
28 #include <linux/sunrpc/metrics.h>
29 #include <linux/nfs_fs.h>
30 #include <linux/nfs_mount.h>
31 #include <linux/nfs4_mount.h>
32 #include <linux/lockd/bind.h>
33 #include <linux/seq_file.h>
34 #include <linux/mount.h>
35 #include <linux/vfs.h>
36 #include <linux/inet.h>
37 #include <linux/nfs_xdr.h>
38 #include <linux/slab.h>
39 #include <linux/compat.h>
40 #include <linux/freezer.h>
41 #include <linux/crc32.h>
43 #include <asm/uaccess.h>
47 #include "delegation.h"
51 #include "dns_resolve.h"
56 #define NFSDBG_FACILITY NFSDBG_VFS
58 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
60 /* Default is to see 64-bit inode numbers */
61 static bool enable_ino64 = NFS_64_BIT_INODE_NUMBERS_ENABLED;
63 static void nfs_invalidate_inode(struct inode *);
64 static int nfs_update_inode(struct inode *, struct nfs_fattr *);
66 static struct kmem_cache * nfs_inode_cachep;
68 static inline unsigned long
69 nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
71 return nfs_fileid_to_ino_t(fattr->fileid);
75 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
76 * @word: long word containing the bit lock
78 int nfs_wait_bit_killable(void *word)
80 if (fatal_signal_pending(current))
85 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable);
88 * nfs_compat_user_ino64 - returns the user-visible inode number
89 * @fileid: 64-bit fileid
91 * This function returns a 32-bit inode number if the boot parameter
92 * nfs.enable_ino64 is zero.
94 u64 nfs_compat_user_ino64(u64 fileid)
105 if (sizeof(ino) < sizeof(fileid))
106 ino ^= fileid >> (sizeof(fileid)-sizeof(ino)) * 8;
110 int nfs_drop_inode(struct inode *inode)
112 return NFS_STALE(inode) || generic_drop_inode(inode);
114 EXPORT_SYMBOL_GPL(nfs_drop_inode);
116 void nfs_clear_inode(struct inode *inode)
119 * The following should never happen...
121 WARN_ON_ONCE(nfs_have_writebacks(inode));
122 WARN_ON_ONCE(!list_empty(&NFS_I(inode)->open_files));
123 nfs_zap_acl_cache(inode);
124 nfs_access_zap_cache(inode);
125 nfs_fscache_release_inode_cookie(inode);
127 EXPORT_SYMBOL_GPL(nfs_clear_inode);
129 void nfs_evict_inode(struct inode *inode)
131 truncate_inode_pages(&inode->i_data, 0);
133 nfs_clear_inode(inode);
137 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
139 int nfs_sync_mapping(struct address_space *mapping)
143 if (mapping->nrpages != 0) {
144 unmap_mapping_range(mapping, 0, 0, 0);
145 ret = nfs_wb_all(mapping->host);
151 * Invalidate the local caches
153 static void nfs_zap_caches_locked(struct inode *inode)
155 struct nfs_inode *nfsi = NFS_I(inode);
156 int mode = inode->i_mode;
158 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
160 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
161 nfsi->attrtimeo_timestamp = jiffies;
163 memset(NFS_I(inode)->cookieverf, 0, sizeof(NFS_I(inode)->cookieverf));
164 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)) {
165 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
166 nfs_fscache_invalidate(inode);
168 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
172 void nfs_zap_caches(struct inode *inode)
174 spin_lock(&inode->i_lock);
175 nfs_zap_caches_locked(inode);
176 spin_unlock(&inode->i_lock);
179 void nfs_zap_mapping(struct inode *inode, struct address_space *mapping)
181 if (mapping->nrpages != 0) {
182 spin_lock(&inode->i_lock);
183 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_DATA;
184 nfs_fscache_invalidate(inode);
185 spin_unlock(&inode->i_lock);
189 void nfs_zap_acl_cache(struct inode *inode)
191 void (*clear_acl_cache)(struct inode *);
193 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
194 if (clear_acl_cache != NULL)
195 clear_acl_cache(inode);
196 spin_lock(&inode->i_lock);
197 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
198 spin_unlock(&inode->i_lock);
200 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache);
202 void nfs_invalidate_atime(struct inode *inode)
204 spin_lock(&inode->i_lock);
205 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME;
206 spin_unlock(&inode->i_lock);
208 EXPORT_SYMBOL_GPL(nfs_invalidate_atime);
211 * Invalidate, but do not unhash, the inode.
212 * NB: must be called with inode->i_lock held!
214 static void nfs_invalidate_inode(struct inode *inode)
216 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
217 nfs_zap_caches_locked(inode);
220 struct nfs_find_desc {
222 struct nfs_fattr *fattr;
226 * In NFSv3 we can have 64bit inode numbers. In order to support
227 * this, and re-exported directories (also seen in NFSv2)
228 * we are forced to allow 2 different inodes to have the same
232 nfs_find_actor(struct inode *inode, void *opaque)
234 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
235 struct nfs_fh *fh = desc->fh;
236 struct nfs_fattr *fattr = desc->fattr;
238 if (NFS_FILEID(inode) != fattr->fileid)
240 if ((S_IFMT & inode->i_mode) != (S_IFMT & fattr->mode))
242 if (nfs_compare_fh(NFS_FH(inode), fh))
244 if (is_bad_inode(inode) || NFS_STALE(inode))
250 nfs_init_locked(struct inode *inode, void *opaque)
252 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
253 struct nfs_fattr *fattr = desc->fattr;
255 set_nfs_fileid(inode, fattr->fileid);
256 nfs_copy_fh(NFS_FH(inode), desc->fh);
260 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
261 struct nfs4_label *nfs4_label_alloc(struct nfs_server *server, gfp_t flags)
263 struct nfs4_label *label = NULL;
264 int minor_version = server->nfs_client->cl_minorversion;
266 if (minor_version < 2)
269 if (!(server->caps & NFS_CAP_SECURITY_LABEL))
272 label = kzalloc(sizeof(struct nfs4_label), flags);
274 return ERR_PTR(-ENOMEM);
276 label->label = kzalloc(NFS4_MAXLABELLEN, flags);
277 if (label->label == NULL) {
279 return ERR_PTR(-ENOMEM);
281 label->len = NFS4_MAXLABELLEN;
285 EXPORT_SYMBOL_GPL(nfs4_label_alloc);
289 * This is our front-end to iget that looks up inodes by file handle
290 * instead of inode number.
293 nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
295 struct nfs_find_desc desc = {
299 struct inode *inode = ERR_PTR(-ENOENT);
302 nfs_attr_check_mountpoint(sb, fattr);
304 if (((fattr->valid & NFS_ATTR_FATTR_FILEID) == 0) &&
305 !nfs_attr_use_mounted_on_fileid(fattr))
307 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) == 0)
310 hash = nfs_fattr_to_ino_t(fattr);
312 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
314 inode = ERR_PTR(-ENOMEM);
318 if (inode->i_state & I_NEW) {
319 struct nfs_inode *nfsi = NFS_I(inode);
320 unsigned long now = jiffies;
322 /* We set i_ino for the few things that still rely on it,
326 /* We can't support update_atime(), since the server will reset it */
327 inode->i_flags |= S_NOATIME|S_NOCMTIME;
328 inode->i_mode = fattr->mode;
329 if ((fattr->valid & NFS_ATTR_FATTR_MODE) == 0
330 && nfs_server_capable(inode, NFS_CAP_MODE))
331 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
332 /* Why so? Because we want revalidate for devices/FIFOs, and
333 * that's precisely what we have in nfs_file_inode_operations.
335 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->file_inode_ops;
336 if (S_ISREG(inode->i_mode)) {
337 inode->i_fop = NFS_SB(sb)->nfs_client->rpc_ops->file_ops;
338 inode->i_data.a_ops = &nfs_file_aops;
339 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
340 } else if (S_ISDIR(inode->i_mode)) {
341 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->dir_inode_ops;
342 inode->i_fop = &nfs_dir_operations;
343 inode->i_data.a_ops = &nfs_dir_aops;
344 /* Deal with crossing mountpoints */
345 if (fattr->valid & NFS_ATTR_FATTR_MOUNTPOINT ||
346 fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL) {
347 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)
348 inode->i_op = &nfs_referral_inode_operations;
350 inode->i_op = &nfs_mountpoint_inode_operations;
352 inode->i_flags |= S_AUTOMOUNT;
354 } else if (S_ISLNK(inode->i_mode))
355 inode->i_op = &nfs_symlink_inode_operations;
357 init_special_inode(inode, inode->i_mode, fattr->rdev);
359 memset(&inode->i_atime, 0, sizeof(inode->i_atime));
360 memset(&inode->i_mtime, 0, sizeof(inode->i_mtime));
361 memset(&inode->i_ctime, 0, sizeof(inode->i_ctime));
362 inode->i_version = 0;
365 inode->i_uid = make_kuid(&init_user_ns, -2);
366 inode->i_gid = make_kgid(&init_user_ns, -2);
368 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
372 nfsi->read_cache_jiffies = fattr->time_start;
373 nfsi->attr_gencount = fattr->gencount;
374 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
375 inode->i_atime = fattr->atime;
376 else if (nfs_server_capable(inode, NFS_CAP_ATIME))
377 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
378 if (fattr->valid & NFS_ATTR_FATTR_MTIME)
379 inode->i_mtime = fattr->mtime;
380 else if (nfs_server_capable(inode, NFS_CAP_MTIME))
381 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
382 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
383 inode->i_ctime = fattr->ctime;
384 else if (nfs_server_capable(inode, NFS_CAP_CTIME))
385 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
386 if (fattr->valid & NFS_ATTR_FATTR_CHANGE)
387 inode->i_version = fattr->change_attr;
388 else if (nfs_server_capable(inode, NFS_CAP_CHANGE_ATTR))
389 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
390 if (fattr->valid & NFS_ATTR_FATTR_SIZE)
391 inode->i_size = nfs_size_to_loff_t(fattr->size);
393 nfsi->cache_validity |= NFS_INO_INVALID_ATTR
394 | NFS_INO_REVAL_PAGECACHE;
395 if (fattr->valid & NFS_ATTR_FATTR_NLINK)
396 set_nlink(inode, fattr->nlink);
397 else if (nfs_server_capable(inode, NFS_CAP_NLINK))
398 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
399 if (fattr->valid & NFS_ATTR_FATTR_OWNER)
400 inode->i_uid = fattr->uid;
401 else if (nfs_server_capable(inode, NFS_CAP_OWNER))
402 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
403 if (fattr->valid & NFS_ATTR_FATTR_GROUP)
404 inode->i_gid = fattr->gid;
405 else if (nfs_server_capable(inode, NFS_CAP_OWNER_GROUP))
406 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
407 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
408 inode->i_blocks = fattr->du.nfs2.blocks;
409 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
411 * report the blocks in 512byte units
413 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
415 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
416 nfsi->attrtimeo_timestamp = now;
417 nfsi->access_cache = RB_ROOT;
419 nfs_fscache_init_inode_cookie(inode);
421 unlock_new_inode(inode);
423 nfs_refresh_inode(inode, fattr);
424 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
426 (long long)NFS_FILEID(inode),
427 nfs_display_fhandle_hash(fh),
428 atomic_read(&inode->i_count));
434 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
437 EXPORT_SYMBOL_GPL(nfs_fhget);
439 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET|ATTR_FILE|ATTR_OPEN)
442 nfs_setattr(struct dentry *dentry, struct iattr *attr)
444 struct inode *inode = dentry->d_inode;
445 struct nfs_fattr *fattr;
448 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
450 /* skip mode change if it's just for clearing setuid/setgid */
451 if (attr->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
452 attr->ia_valid &= ~ATTR_MODE;
454 if (attr->ia_valid & ATTR_SIZE) {
455 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
456 attr->ia_valid &= ~ATTR_SIZE;
459 /* Optimization: if the end result is no change, don't RPC */
460 attr->ia_valid &= NFS_VALID_ATTRS;
461 if ((attr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
464 /* Write all dirty data */
465 if (S_ISREG(inode->i_mode)) {
466 nfs_inode_dio_wait(inode);
470 fattr = nfs_alloc_fattr();
474 * Return any delegations if we're going to change ACLs
476 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
477 NFS_PROTO(inode)->return_delegation(inode);
478 error = NFS_PROTO(inode)->setattr(dentry, fattr, attr);
480 nfs_refresh_inode(inode, fattr);
481 nfs_free_fattr(fattr);
485 EXPORT_SYMBOL_GPL(nfs_setattr);
488 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
489 * @inode: inode of the file used
490 * @offset: file offset to start truncating
492 * This is a copy of the common vmtruncate, but with the locking
493 * corrected to take into account the fact that NFS requires
494 * inode->i_size to be updated under the inode->i_lock.
496 static int nfs_vmtruncate(struct inode * inode, loff_t offset)
501 err = inode_newsize_ok(inode, offset);
505 spin_lock(&inode->i_lock);
506 oldsize = inode->i_size;
507 i_size_write(inode, offset);
508 spin_unlock(&inode->i_lock);
510 truncate_pagecache(inode, oldsize, offset);
516 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
517 * @inode: pointer to struct inode
518 * @attr: pointer to struct iattr
520 * Note: we do this in the *proc.c in order to ensure that
521 * it works for things like exclusive creates too.
523 void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
525 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
526 spin_lock(&inode->i_lock);
527 if ((attr->ia_valid & ATTR_MODE) != 0) {
528 int mode = attr->ia_mode & S_IALLUGO;
529 mode |= inode->i_mode & ~S_IALLUGO;
530 inode->i_mode = mode;
532 if ((attr->ia_valid & ATTR_UID) != 0)
533 inode->i_uid = attr->ia_uid;
534 if ((attr->ia_valid & ATTR_GID) != 0)
535 inode->i_gid = attr->ia_gid;
536 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
537 spin_unlock(&inode->i_lock);
539 if ((attr->ia_valid & ATTR_SIZE) != 0) {
540 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
541 nfs_vmtruncate(inode, attr->ia_size);
544 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode);
546 int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
548 struct inode *inode = dentry->d_inode;
549 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
552 /* Flush out writes to the server in order to update c/mtime. */
553 if (S_ISREG(inode->i_mode)) {
554 nfs_inode_dio_wait(inode);
555 err = filemap_write_and_wait(inode->i_mapping);
561 * We may force a getattr if the user cares about atime.
563 * Note that we only have to check the vfsmount flags here:
564 * - NFS always sets S_NOATIME by so checking it would give a
566 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
567 * no point in checking those.
569 if ((mnt->mnt_flags & MNT_NOATIME) ||
570 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
574 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
576 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
578 generic_fillattr(inode, stat);
579 stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode));
584 EXPORT_SYMBOL_GPL(nfs_getattr);
586 static void nfs_init_lock_context(struct nfs_lock_context *l_ctx)
588 atomic_set(&l_ctx->count, 1);
589 l_ctx->lockowner.l_owner = current->files;
590 l_ctx->lockowner.l_pid = current->tgid;
591 INIT_LIST_HEAD(&l_ctx->list);
592 nfs_iocounter_init(&l_ctx->io_count);
595 static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context *ctx)
597 struct nfs_lock_context *head = &ctx->lock_context;
598 struct nfs_lock_context *pos = head;
601 if (pos->lockowner.l_owner != current->files)
603 if (pos->lockowner.l_pid != current->tgid)
605 atomic_inc(&pos->count);
607 } while ((pos = list_entry(pos->list.next, typeof(*pos), list)) != head);
611 struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx)
613 struct nfs_lock_context *res, *new = NULL;
614 struct inode *inode = ctx->dentry->d_inode;
616 spin_lock(&inode->i_lock);
617 res = __nfs_find_lock_context(ctx);
619 spin_unlock(&inode->i_lock);
620 new = kmalloc(sizeof(*new), GFP_KERNEL);
622 return ERR_PTR(-ENOMEM);
623 nfs_init_lock_context(new);
624 spin_lock(&inode->i_lock);
625 res = __nfs_find_lock_context(ctx);
627 list_add_tail(&new->list, &ctx->lock_context.list);
628 new->open_context = ctx;
633 spin_unlock(&inode->i_lock);
638 void nfs_put_lock_context(struct nfs_lock_context *l_ctx)
640 struct nfs_open_context *ctx = l_ctx->open_context;
641 struct inode *inode = ctx->dentry->d_inode;
643 if (!atomic_dec_and_lock(&l_ctx->count, &inode->i_lock))
645 list_del(&l_ctx->list);
646 spin_unlock(&inode->i_lock);
651 * nfs_close_context - Common close_context() routine NFSv2/v3
652 * @ctx: pointer to context
653 * @is_sync: is this a synchronous close
655 * always ensure that the attributes are up to date if we're mounted
656 * with close-to-open semantics
658 void nfs_close_context(struct nfs_open_context *ctx, int is_sync)
661 struct nfs_server *server;
663 if (!(ctx->mode & FMODE_WRITE))
667 inode = ctx->dentry->d_inode;
668 if (!list_empty(&NFS_I(inode)->open_files))
670 server = NFS_SERVER(inode);
671 if (server->flags & NFS_MOUNT_NOCTO)
673 nfs_revalidate_inode(server, inode);
675 EXPORT_SYMBOL_GPL(nfs_close_context);
677 struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, fmode_t f_mode)
679 struct nfs_open_context *ctx;
680 struct rpc_cred *cred = rpc_lookup_cred();
682 return ERR_CAST(cred);
684 ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
687 return ERR_PTR(-ENOMEM);
689 nfs_sb_active(dentry->d_sb);
690 ctx->dentry = dget(dentry);
696 nfs_init_lock_context(&ctx->lock_context);
697 ctx->lock_context.open_context = ctx;
698 INIT_LIST_HEAD(&ctx->list);
699 ctx->mdsthreshold = NULL;
702 EXPORT_SYMBOL_GPL(alloc_nfs_open_context);
704 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
707 atomic_inc(&ctx->lock_context.count);
710 EXPORT_SYMBOL_GPL(get_nfs_open_context);
712 static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
714 struct inode *inode = ctx->dentry->d_inode;
715 struct super_block *sb = ctx->dentry->d_sb;
717 if (!list_empty(&ctx->list)) {
718 if (!atomic_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
720 list_del(&ctx->list);
721 spin_unlock(&inode->i_lock);
722 } else if (!atomic_dec_and_test(&ctx->lock_context.count))
725 NFS_PROTO(inode)->close_context(ctx, is_sync);
726 if (ctx->cred != NULL)
727 put_rpccred(ctx->cred);
730 kfree(ctx->mdsthreshold);
734 void put_nfs_open_context(struct nfs_open_context *ctx)
736 __put_nfs_open_context(ctx, 0);
738 EXPORT_SYMBOL_GPL(put_nfs_open_context);
741 * Ensure that mmap has a recent RPC credential for use when writing out
744 void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
746 struct inode *inode = file_inode(filp);
747 struct nfs_inode *nfsi = NFS_I(inode);
749 filp->private_data = get_nfs_open_context(ctx);
750 spin_lock(&inode->i_lock);
751 list_add(&ctx->list, &nfsi->open_files);
752 spin_unlock(&inode->i_lock);
754 EXPORT_SYMBOL_GPL(nfs_file_set_open_context);
757 * Given an inode, search for an open context with the desired characteristics
759 struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, fmode_t mode)
761 struct nfs_inode *nfsi = NFS_I(inode);
762 struct nfs_open_context *pos, *ctx = NULL;
764 spin_lock(&inode->i_lock);
765 list_for_each_entry(pos, &nfsi->open_files, list) {
766 if (cred != NULL && pos->cred != cred)
768 if ((pos->mode & (FMODE_READ|FMODE_WRITE)) != mode)
770 ctx = get_nfs_open_context(pos);
773 spin_unlock(&inode->i_lock);
777 static void nfs_file_clear_open_context(struct file *filp)
779 struct inode *inode = file_inode(filp);
780 struct nfs_open_context *ctx = nfs_file_open_context(filp);
783 filp->private_data = NULL;
784 spin_lock(&inode->i_lock);
785 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
786 spin_unlock(&inode->i_lock);
787 __put_nfs_open_context(ctx, filp->f_flags & O_DIRECT ? 0 : 1);
792 * These allocate and release file read/write context information.
794 int nfs_open(struct inode *inode, struct file *filp)
796 struct nfs_open_context *ctx;
798 ctx = alloc_nfs_open_context(filp->f_path.dentry, filp->f_mode);
801 nfs_file_set_open_context(filp, ctx);
802 put_nfs_open_context(ctx);
803 nfs_fscache_set_inode_cookie(inode, filp);
807 int nfs_release(struct inode *inode, struct file *filp)
809 nfs_file_clear_open_context(filp);
814 * This function is called whenever some part of NFS notices that
815 * the cached attributes have to be refreshed.
818 __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
820 int status = -ESTALE;
821 struct nfs_fattr *fattr = NULL;
822 struct nfs_inode *nfsi = NFS_I(inode);
824 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
825 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
827 if (is_bad_inode(inode))
829 if (NFS_STALE(inode))
833 fattr = nfs_alloc_fattr();
837 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
838 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), fattr);
840 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
842 (long long)NFS_FILEID(inode), status);
843 if (status == -ESTALE) {
844 nfs_zap_caches(inode);
845 if (!S_ISDIR(inode->i_mode))
846 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
851 status = nfs_refresh_inode(inode, fattr);
853 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
855 (long long)NFS_FILEID(inode), status);
859 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
860 nfs_zap_acl_cache(inode);
862 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
864 (long long)NFS_FILEID(inode));
867 nfs_free_fattr(fattr);
871 int nfs_attribute_timeout(struct inode *inode)
873 struct nfs_inode *nfsi = NFS_I(inode);
875 return !time_in_range_open(jiffies, nfsi->read_cache_jiffies, nfsi->read_cache_jiffies + nfsi->attrtimeo);
878 static int nfs_attribute_cache_expired(struct inode *inode)
880 if (nfs_have_delegated_attributes(inode))
882 return nfs_attribute_timeout(inode);
886 * nfs_revalidate_inode - Revalidate the inode attributes
887 * @server - pointer to nfs_server struct
888 * @inode - pointer to inode struct
890 * Updates inode attribute information by retrieving the data from the server.
892 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
894 if (!(NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATTR)
895 && !nfs_attribute_cache_expired(inode))
896 return NFS_STALE(inode) ? -ESTALE : 0;
897 return __nfs_revalidate_inode(server, inode);
899 EXPORT_SYMBOL_GPL(nfs_revalidate_inode);
901 static int nfs_invalidate_mapping(struct inode *inode, struct address_space *mapping)
903 struct nfs_inode *nfsi = NFS_I(inode);
905 if (mapping->nrpages != 0) {
906 int ret = invalidate_inode_pages2(mapping);
910 spin_lock(&inode->i_lock);
911 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
912 if (S_ISDIR(inode->i_mode))
913 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
914 spin_unlock(&inode->i_lock);
915 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
916 nfs_fscache_wait_on_invalidate(inode);
917 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
918 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
922 static bool nfs_mapping_need_revalidate_inode(struct inode *inode)
924 if (nfs_have_delegated_attributes(inode))
926 return (NFS_I(inode)->cache_validity & NFS_INO_REVAL_PAGECACHE)
927 || nfs_attribute_timeout(inode)
932 * nfs_revalidate_mapping - Revalidate the pagecache
933 * @inode - pointer to host inode
934 * @mapping - pointer to mapping
936 int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
938 struct nfs_inode *nfsi = NFS_I(inode);
941 /* swapfiles are not supposed to be shared. */
942 if (IS_SWAPFILE(inode))
945 if (nfs_mapping_need_revalidate_inode(inode)) {
946 ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
950 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
951 ret = nfs_invalidate_mapping(inode, mapping);
956 static unsigned long nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
958 struct nfs_inode *nfsi = NFS_I(inode);
959 unsigned long ret = 0;
961 if ((fattr->valid & NFS_ATTR_FATTR_PRECHANGE)
962 && (fattr->valid & NFS_ATTR_FATTR_CHANGE)
963 && inode->i_version == fattr->pre_change_attr) {
964 inode->i_version = fattr->change_attr;
965 if (S_ISDIR(inode->i_mode))
966 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
967 ret |= NFS_INO_INVALID_ATTR;
969 /* If we have atomic WCC data, we may update some attributes */
970 if ((fattr->valid & NFS_ATTR_FATTR_PRECTIME)
971 && (fattr->valid & NFS_ATTR_FATTR_CTIME)
972 && timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
973 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
974 ret |= NFS_INO_INVALID_ATTR;
977 if ((fattr->valid & NFS_ATTR_FATTR_PREMTIME)
978 && (fattr->valid & NFS_ATTR_FATTR_MTIME)
979 && timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
980 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
981 if (S_ISDIR(inode->i_mode))
982 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
983 ret |= NFS_INO_INVALID_ATTR;
985 if ((fattr->valid & NFS_ATTR_FATTR_PRESIZE)
986 && (fattr->valid & NFS_ATTR_FATTR_SIZE)
987 && i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size)
988 && nfsi->npages == 0) {
989 i_size_write(inode, nfs_size_to_loff_t(fattr->size));
990 ret |= NFS_INO_INVALID_ATTR;
993 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
994 nfs_fscache_invalidate(inode);
1000 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1001 * @inode - pointer to inode
1002 * @fattr - updated attributes
1004 * Verifies the attribute cache. If we have just changed the attributes,
1005 * so that fattr carries weak cache consistency data, then it may
1006 * also update the ctime/mtime/change_attribute.
1008 static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1010 struct nfs_inode *nfsi = NFS_I(inode);
1011 loff_t cur_size, new_isize;
1012 unsigned long invalid = 0;
1015 if (nfs_have_delegated_attributes(inode))
1017 /* Has the inode gone and changed behind our back? */
1018 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid)
1020 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1023 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1024 inode->i_version != fattr->change_attr)
1025 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1027 /* Verify a few of the more important attributes */
1028 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) && !timespec_equal(&inode->i_mtime, &fattr->mtime))
1029 invalid |= NFS_INO_INVALID_ATTR;
1031 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1032 cur_size = i_size_read(inode);
1033 new_isize = nfs_size_to_loff_t(fattr->size);
1034 if (cur_size != new_isize && nfsi->npages == 0)
1035 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1038 /* Have any file permissions changed? */
1039 if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO))
1040 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1041 if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && !uid_eq(inode->i_uid, fattr->uid))
1042 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1043 if ((fattr->valid & NFS_ATTR_FATTR_GROUP) && !gid_eq(inode->i_gid, fattr->gid))
1044 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1046 /* Has the link count changed? */
1047 if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink)
1048 invalid |= NFS_INO_INVALID_ATTR;
1050 if ((fattr->valid & NFS_ATTR_FATTR_ATIME) && !timespec_equal(&inode->i_atime, &fattr->atime))
1051 invalid |= NFS_INO_INVALID_ATIME;
1054 nfsi->cache_validity |= invalid;
1056 nfsi->read_cache_jiffies = fattr->time_start;
1060 static int nfs_ctime_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1062 if (!(fattr->valid & NFS_ATTR_FATTR_CTIME))
1064 return timespec_compare(&fattr->ctime, &inode->i_ctime) > 0;
1067 static int nfs_size_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1069 if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
1071 return nfs_size_to_loff_t(fattr->size) > i_size_read(inode);
1074 static atomic_long_t nfs_attr_generation_counter;
1076 static unsigned long nfs_read_attr_generation_counter(void)
1078 return atomic_long_read(&nfs_attr_generation_counter);
1081 unsigned long nfs_inc_attr_generation_counter(void)
1083 return atomic_long_inc_return(&nfs_attr_generation_counter);
1086 void nfs_fattr_init(struct nfs_fattr *fattr)
1089 fattr->time_start = jiffies;
1090 fattr->gencount = nfs_inc_attr_generation_counter();
1091 fattr->owner_name = NULL;
1092 fattr->group_name = NULL;
1094 EXPORT_SYMBOL_GPL(nfs_fattr_init);
1096 struct nfs_fattr *nfs_alloc_fattr(void)
1098 struct nfs_fattr *fattr;
1100 fattr = kmalloc(sizeof(*fattr), GFP_NOFS);
1102 nfs_fattr_init(fattr);
1105 EXPORT_SYMBOL_GPL(nfs_alloc_fattr);
1107 struct nfs_fh *nfs_alloc_fhandle(void)
1111 fh = kmalloc(sizeof(struct nfs_fh), GFP_NOFS);
1116 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle);
1120 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1121 * in the same way that wireshark does
1125 * For debugging only.
1127 u32 _nfs_display_fhandle_hash(const struct nfs_fh *fh)
1129 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1130 * not on the result */
1131 return ~crc32(0xFFFFFFFF, &fh->data[0], fh->size);
1135 * _nfs_display_fhandle - display an NFS file handle on the console
1137 * @fh: file handle to display
1138 * @caption: display caption
1140 * For debugging only.
1142 void _nfs_display_fhandle(const struct nfs_fh *fh, const char *caption)
1146 if (fh == NULL || fh->size == 0) {
1147 printk(KERN_DEFAULT "%s at %p is empty\n", caption, fh);
1151 printk(KERN_DEFAULT "%s at %p is %u bytes, crc: 0x%08x:\n",
1152 caption, fh, fh->size, _nfs_display_fhandle_hash(fh));
1153 for (i = 0; i < fh->size; i += 16) {
1154 __be32 *pos = (__be32 *)&fh->data[i];
1156 switch ((fh->size - i - 1) >> 2) {
1158 printk(KERN_DEFAULT " %08x\n",
1162 printk(KERN_DEFAULT " %08x %08x\n",
1163 be32_to_cpup(pos), be32_to_cpup(pos + 1));
1166 printk(KERN_DEFAULT " %08x %08x %08x\n",
1167 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1168 be32_to_cpup(pos + 2));
1171 printk(KERN_DEFAULT " %08x %08x %08x %08x\n",
1172 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1173 be32_to_cpup(pos + 2), be32_to_cpup(pos + 3));
1180 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1181 * @inode - pointer to inode
1182 * @fattr - attributes
1184 * Attempt to divine whether or not an RPC call reply carrying stale
1185 * attributes got scheduled after another call carrying updated ones.
1187 * To do so, the function first assumes that a more recent ctime means
1188 * that the attributes in fattr are newer, however it also attempt to
1189 * catch the case where ctime either didn't change, or went backwards
1190 * (if someone reset the clock on the server) by looking at whether
1191 * or not this RPC call was started after the inode was last updated.
1192 * Note also the check for wraparound of 'attr_gencount'
1194 * The function returns 'true' if it thinks the attributes in 'fattr' are
1195 * more recent than the ones cached in the inode.
1198 static int nfs_inode_attrs_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1200 const struct nfs_inode *nfsi = NFS_I(inode);
1202 return ((long)fattr->gencount - (long)nfsi->attr_gencount) > 0 ||
1203 nfs_ctime_need_update(inode, fattr) ||
1204 nfs_size_need_update(inode, fattr) ||
1205 ((long)nfsi->attr_gencount - (long)nfs_read_attr_generation_counter() > 0);
1208 static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr)
1210 if (nfs_inode_attrs_need_update(inode, fattr))
1211 return nfs_update_inode(inode, fattr);
1212 return nfs_check_inode_attributes(inode, fattr);
1216 * nfs_refresh_inode - try to update the inode attribute cache
1217 * @inode - pointer to inode
1218 * @fattr - updated attributes
1220 * Check that an RPC call that returned attributes has not overlapped with
1221 * other recent updates of the inode metadata, then decide whether it is
1222 * safe to do a full update of the inode attributes, or whether just to
1223 * call nfs_check_inode_attributes.
1225 int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1229 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1231 spin_lock(&inode->i_lock);
1232 status = nfs_refresh_inode_locked(inode, fattr);
1233 spin_unlock(&inode->i_lock);
1237 EXPORT_SYMBOL_GPL(nfs_refresh_inode);
1239 static int nfs_post_op_update_inode_locked(struct inode *inode, struct nfs_fattr *fattr)
1241 struct nfs_inode *nfsi = NFS_I(inode);
1243 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1244 if (S_ISDIR(inode->i_mode)) {
1245 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1246 nfs_fscache_invalidate(inode);
1248 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1250 return nfs_refresh_inode_locked(inode, fattr);
1254 * nfs_post_op_update_inode - try to update the inode attribute cache
1255 * @inode - pointer to inode
1256 * @fattr - updated attributes
1258 * After an operation that has changed the inode metadata, mark the
1259 * attribute cache as being invalid, then try to update it.
1261 * NB: if the server didn't return any post op attributes, this
1262 * function will force the retrieval of attributes before the next
1263 * NFS request. Thus it should be used only for operations that
1264 * are expected to change one or more attributes, to avoid
1265 * unnecessary NFS requests and trips through nfs_update_inode().
1267 int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1271 spin_lock(&inode->i_lock);
1272 status = nfs_post_op_update_inode_locked(inode, fattr);
1273 spin_unlock(&inode->i_lock);
1276 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode);
1279 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1280 * @inode - pointer to inode
1281 * @fattr - updated attributes
1283 * After an operation that has changed the inode metadata, mark the
1284 * attribute cache as being invalid, then try to update it. Fake up
1285 * weak cache consistency data, if none exist.
1287 * This function is mainly designed to be used by the ->write_done() functions.
1289 int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fattr)
1293 spin_lock(&inode->i_lock);
1294 /* Don't do a WCC update if these attributes are already stale */
1295 if ((fattr->valid & NFS_ATTR_FATTR) == 0 ||
1296 !nfs_inode_attrs_need_update(inode, fattr)) {
1297 fattr->valid &= ~(NFS_ATTR_FATTR_PRECHANGE
1298 | NFS_ATTR_FATTR_PRESIZE
1299 | NFS_ATTR_FATTR_PREMTIME
1300 | NFS_ATTR_FATTR_PRECTIME);
1303 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1304 (fattr->valid & NFS_ATTR_FATTR_PRECHANGE) == 0) {
1305 fattr->pre_change_attr = inode->i_version;
1306 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
1308 if ((fattr->valid & NFS_ATTR_FATTR_CTIME) != 0 &&
1309 (fattr->valid & NFS_ATTR_FATTR_PRECTIME) == 0) {
1310 memcpy(&fattr->pre_ctime, &inode->i_ctime, sizeof(fattr->pre_ctime));
1311 fattr->valid |= NFS_ATTR_FATTR_PRECTIME;
1313 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) != 0 &&
1314 (fattr->valid & NFS_ATTR_FATTR_PREMTIME) == 0) {
1315 memcpy(&fattr->pre_mtime, &inode->i_mtime, sizeof(fattr->pre_mtime));
1316 fattr->valid |= NFS_ATTR_FATTR_PREMTIME;
1318 if ((fattr->valid & NFS_ATTR_FATTR_SIZE) != 0 &&
1319 (fattr->valid & NFS_ATTR_FATTR_PRESIZE) == 0) {
1320 fattr->pre_size = i_size_read(inode);
1321 fattr->valid |= NFS_ATTR_FATTR_PRESIZE;
1324 status = nfs_post_op_update_inode_locked(inode, fattr);
1325 spin_unlock(&inode->i_lock);
1328 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc);
1331 * Many nfs protocol calls return the new file attributes after
1332 * an operation. Here we update the inode to reflect the state
1333 * of the server's inode.
1335 * This is a bit tricky because we have to make sure all dirty pages
1336 * have been sent off to the server before calling invalidate_inode_pages.
1337 * To make sure no other process adds more write requests while we try
1338 * our best to flush them, we make them sleep during the attribute refresh.
1340 * A very similar scenario holds for the dir cache.
1342 static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1344 struct nfs_server *server;
1345 struct nfs_inode *nfsi = NFS_I(inode);
1346 loff_t cur_isize, new_isize;
1347 unsigned long invalid = 0;
1348 unsigned long now = jiffies;
1349 unsigned long save_cache_validity;
1351 dfprintk(VFS, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1352 __func__, inode->i_sb->s_id, inode->i_ino,
1353 nfs_display_fhandle_hash(NFS_FH(inode)),
1354 atomic_read(&inode->i_count), fattr->valid);
1356 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid) {
1357 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1358 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1359 NFS_SERVER(inode)->nfs_client->cl_hostname,
1360 inode->i_sb->s_id, (long long)nfsi->fileid,
1361 (long long)fattr->fileid);
1366 * Make sure the inode's type hasn't changed.
1368 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1370 * Big trouble! The inode has become a different object.
1372 printk(KERN_DEBUG "NFS: %s: inode %ld mode changed, %07o to %07o\n",
1373 __func__, inode->i_ino, inode->i_mode, fattr->mode);
1377 server = NFS_SERVER(inode);
1378 /* Update the fsid? */
1379 if (S_ISDIR(inode->i_mode) && (fattr->valid & NFS_ATTR_FATTR_FSID) &&
1380 !nfs_fsid_equal(&server->fsid, &fattr->fsid) &&
1381 !IS_AUTOMOUNT(inode))
1382 server->fsid = fattr->fsid;
1385 * Update the read time so we don't revalidate too often.
1387 nfsi->read_cache_jiffies = fattr->time_start;
1389 save_cache_validity = nfsi->cache_validity;
1390 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR
1391 | NFS_INO_INVALID_ATIME
1392 | NFS_INO_REVAL_FORCED
1393 | NFS_INO_REVAL_PAGECACHE);
1395 /* Do atomic weak cache consistency updates */
1396 invalid |= nfs_wcc_update_inode(inode, fattr);
1398 /* More cache consistency checks */
1399 if (fattr->valid & NFS_ATTR_FATTR_CHANGE) {
1400 if (inode->i_version != fattr->change_attr) {
1401 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1402 inode->i_sb->s_id, inode->i_ino);
1403 invalid |= NFS_INO_INVALID_ATTR
1404 | NFS_INO_INVALID_DATA
1405 | NFS_INO_INVALID_ACCESS
1406 | NFS_INO_INVALID_ACL
1407 | NFS_INO_REVAL_PAGECACHE;
1408 if (S_ISDIR(inode->i_mode))
1409 nfs_force_lookup_revalidate(inode);
1410 inode->i_version = fattr->change_attr;
1412 } else if (server->caps & NFS_CAP_CHANGE_ATTR)
1413 invalid |= save_cache_validity;
1415 if (fattr->valid & NFS_ATTR_FATTR_MTIME) {
1416 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1417 } else if (server->caps & NFS_CAP_MTIME)
1418 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1419 | NFS_INO_REVAL_FORCED);
1421 if (fattr->valid & NFS_ATTR_FATTR_CTIME) {
1422 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1423 } else if (server->caps & NFS_CAP_CTIME)
1424 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1425 | NFS_INO_REVAL_FORCED);
1427 /* Check if our cached file size is stale */
1428 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1429 new_isize = nfs_size_to_loff_t(fattr->size);
1430 cur_isize = i_size_read(inode);
1431 if (new_isize != cur_isize) {
1432 /* Do we perhaps have any outstanding writes, or has
1433 * the file grown beyond our last write? */
1434 if ((nfsi->npages == 0 && !test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) ||
1435 new_isize > cur_isize) {
1436 i_size_write(inode, new_isize);
1437 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1439 dprintk("NFS: isize change on server for file %s/%ld "
1443 (long long)cur_isize,
1444 (long long)new_isize);
1447 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1448 | NFS_INO_REVAL_PAGECACHE
1449 | NFS_INO_REVAL_FORCED);
1452 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
1453 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1454 else if (server->caps & NFS_CAP_ATIME)
1455 invalid |= save_cache_validity & (NFS_INO_INVALID_ATIME
1456 | NFS_INO_REVAL_FORCED);
1458 if (fattr->valid & NFS_ATTR_FATTR_MODE) {
1459 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) {
1460 umode_t newmode = inode->i_mode & S_IFMT;
1461 newmode |= fattr->mode & S_IALLUGO;
1462 inode->i_mode = newmode;
1463 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1465 } else if (server->caps & NFS_CAP_MODE)
1466 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1467 | NFS_INO_INVALID_ACCESS
1468 | NFS_INO_INVALID_ACL
1469 | NFS_INO_REVAL_FORCED);
1471 if (fattr->valid & NFS_ATTR_FATTR_OWNER) {
1472 if (!uid_eq(inode->i_uid, fattr->uid)) {
1473 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1474 inode->i_uid = fattr->uid;
1476 } else if (server->caps & NFS_CAP_OWNER)
1477 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1478 | NFS_INO_INVALID_ACCESS
1479 | NFS_INO_INVALID_ACL
1480 | NFS_INO_REVAL_FORCED);
1482 if (fattr->valid & NFS_ATTR_FATTR_GROUP) {
1483 if (!gid_eq(inode->i_gid, fattr->gid)) {
1484 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1485 inode->i_gid = fattr->gid;
1487 } else if (server->caps & NFS_CAP_OWNER_GROUP)
1488 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1489 | NFS_INO_INVALID_ACCESS
1490 | NFS_INO_INVALID_ACL
1491 | NFS_INO_REVAL_FORCED);
1493 if (fattr->valid & NFS_ATTR_FATTR_NLINK) {
1494 if (inode->i_nlink != fattr->nlink) {
1495 invalid |= NFS_INO_INVALID_ATTR;
1496 if (S_ISDIR(inode->i_mode))
1497 invalid |= NFS_INO_INVALID_DATA;
1498 set_nlink(inode, fattr->nlink);
1500 } else if (server->caps & NFS_CAP_NLINK)
1501 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1502 | NFS_INO_REVAL_FORCED);
1504 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
1506 * report the blocks in 512byte units
1508 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1510 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
1511 inode->i_blocks = fattr->du.nfs2.blocks;
1513 /* Update attrtimeo value if we're out of the unstable period */
1514 if (invalid & NFS_INO_INVALID_ATTR) {
1515 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
1516 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1517 nfsi->attrtimeo_timestamp = now;
1518 nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1520 if (!time_in_range_open(now, nfsi->attrtimeo_timestamp, nfsi->attrtimeo_timestamp + nfsi->attrtimeo)) {
1521 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1522 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1523 nfsi->attrtimeo_timestamp = now;
1526 invalid &= ~NFS_INO_INVALID_ATTR;
1527 /* Don't invalidate the data if we were to blame */
1528 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1529 || S_ISLNK(inode->i_mode)))
1530 invalid &= ~NFS_INO_INVALID_DATA;
1531 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ) ||
1532 (save_cache_validity & NFS_INO_REVAL_FORCED))
1533 nfsi->cache_validity |= invalid;
1535 if (invalid & NFS_INO_INVALID_DATA)
1536 nfs_fscache_invalidate(inode);
1541 * No need to worry about unhashing the dentry, as the
1542 * lookup validation will know that the inode is bad.
1543 * (But we fall through to invalidate the caches.)
1545 nfs_invalidate_inode(inode);
1549 struct inode *nfs_alloc_inode(struct super_block *sb)
1551 struct nfs_inode *nfsi;
1552 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, GFP_KERNEL);
1556 nfsi->cache_validity = 0UL;
1557 #ifdef CONFIG_NFS_V3_ACL
1558 nfsi->acl_access = ERR_PTR(-EAGAIN);
1559 nfsi->acl_default = ERR_PTR(-EAGAIN);
1561 #if IS_ENABLED(CONFIG_NFS_V4)
1562 nfsi->nfs4_acl = NULL;
1563 #endif /* CONFIG_NFS_V4 */
1564 return &nfsi->vfs_inode;
1566 EXPORT_SYMBOL_GPL(nfs_alloc_inode);
1568 static void nfs_i_callback(struct rcu_head *head)
1570 struct inode *inode = container_of(head, struct inode, i_rcu);
1571 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1574 void nfs_destroy_inode(struct inode *inode)
1576 call_rcu(&inode->i_rcu, nfs_i_callback);
1578 EXPORT_SYMBOL_GPL(nfs_destroy_inode);
1580 static inline void nfs4_init_once(struct nfs_inode *nfsi)
1582 #if IS_ENABLED(CONFIG_NFS_V4)
1583 INIT_LIST_HEAD(&nfsi->open_states);
1584 nfsi->delegation = NULL;
1585 nfsi->delegation_state = 0;
1586 init_rwsem(&nfsi->rwsem);
1587 nfsi->layout = NULL;
1591 static void init_once(void *foo)
1593 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1595 inode_init_once(&nfsi->vfs_inode);
1596 INIT_LIST_HEAD(&nfsi->open_files);
1597 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
1598 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
1599 INIT_LIST_HEAD(&nfsi->commit_info.list);
1601 nfsi->commit_info.ncommit = 0;
1602 atomic_set(&nfsi->commit_info.rpcs_out, 0);
1603 atomic_set(&nfsi->silly_count, 1);
1604 INIT_HLIST_HEAD(&nfsi->silly_list);
1605 init_waitqueue_head(&nfsi->waitqueue);
1606 nfs4_init_once(nfsi);
1609 static int __init nfs_init_inodecache(void)
1611 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
1612 sizeof(struct nfs_inode),
1613 0, (SLAB_RECLAIM_ACCOUNT|
1616 if (nfs_inode_cachep == NULL)
1622 static void nfs_destroy_inodecache(void)
1625 * Make sure all delayed rcu free inodes are flushed before we
1629 kmem_cache_destroy(nfs_inode_cachep);
1632 struct workqueue_struct *nfsiod_workqueue;
1633 EXPORT_SYMBOL_GPL(nfsiod_workqueue);
1636 * start up the nfsiod workqueue
1638 static int nfsiod_start(void)
1640 struct workqueue_struct *wq;
1641 dprintk("RPC: creating workqueue nfsiod\n");
1642 wq = alloc_workqueue("nfsiod", WQ_MEM_RECLAIM, 0);
1645 nfsiod_workqueue = wq;
1650 * Destroy the nfsiod workqueue
1652 static void nfsiod_stop(void)
1654 struct workqueue_struct *wq;
1656 wq = nfsiod_workqueue;
1659 nfsiod_workqueue = NULL;
1660 destroy_workqueue(wq);
1664 EXPORT_SYMBOL_GPL(nfs_net_id);
1666 static int nfs_net_init(struct net *net)
1668 nfs_clients_init(net);
1669 return nfs_dns_resolver_cache_init(net);
1672 static void nfs_net_exit(struct net *net)
1674 nfs_dns_resolver_cache_destroy(net);
1675 nfs_cleanup_cb_ident_idr(net);
1678 static struct pernet_operations nfs_net_ops = {
1679 .init = nfs_net_init,
1680 .exit = nfs_net_exit,
1682 .size = sizeof(struct nfs_net),
1688 static int __init init_nfs_fs(void)
1692 err = nfs_dns_resolver_init();
1696 err = register_pernet_subsys(&nfs_net_ops);
1700 err = nfs_fscache_register();
1704 err = nfsiod_start();
1708 err = nfs_fs_proc_init();
1712 err = nfs_init_nfspagecache();
1716 err = nfs_init_inodecache();
1720 err = nfs_init_readpagecache();
1724 err = nfs_init_writepagecache();
1728 err = nfs_init_directcache();
1732 #ifdef CONFIG_PROC_FS
1733 rpc_proc_register(&init_net, &nfs_rpcstat);
1735 if ((err = register_nfs_fs()) != 0)
1740 #ifdef CONFIG_PROC_FS
1741 rpc_proc_unregister(&init_net, "nfs");
1743 nfs_destroy_directcache();
1745 nfs_destroy_writepagecache();
1747 nfs_destroy_readpagecache();
1749 nfs_destroy_inodecache();
1751 nfs_destroy_nfspagecache();
1757 nfs_fscache_unregister();
1759 unregister_pernet_subsys(&nfs_net_ops);
1761 nfs_dns_resolver_destroy();
1766 static void __exit exit_nfs_fs(void)
1768 nfs_destroy_directcache();
1769 nfs_destroy_writepagecache();
1770 nfs_destroy_readpagecache();
1771 nfs_destroy_inodecache();
1772 nfs_destroy_nfspagecache();
1773 nfs_fscache_unregister();
1774 unregister_pernet_subsys(&nfs_net_ops);
1775 nfs_dns_resolver_destroy();
1776 #ifdef CONFIG_PROC_FS
1777 rpc_proc_unregister(&init_net, "nfs");
1779 unregister_nfs_fs();
1784 /* Not quite true; I just maintain it */
1785 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1786 MODULE_LICENSE("GPL");
1787 module_param(enable_ino64, bool, 0644);
1789 module_init(init_nfs_fs)
1790 module_exit(exit_nfs_fs)