4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/module.h>
17 #include <linux/init.h>
18 #include <linux/sched.h>
19 #include <linux/time.h>
20 #include <linux/kernel.h>
22 #include <linux/string.h>
23 #include <linux/stat.h>
24 #include <linux/errno.h>
25 #include <linux/unistd.h>
26 #include <linux/sunrpc/clnt.h>
27 #include <linux/sunrpc/stats.h>
28 #include <linux/sunrpc/metrics.h>
29 #include <linux/nfs_fs.h>
30 #include <linux/nfs_mount.h>
31 #include <linux/nfs4_mount.h>
32 #include <linux/lockd/bind.h>
33 #include <linux/seq_file.h>
34 #include <linux/mount.h>
35 #include <linux/vfs.h>
36 #include <linux/inet.h>
37 #include <linux/nfs_xdr.h>
38 #include <linux/slab.h>
39 #include <linux/compat.h>
40 #include <linux/freezer.h>
41 #include <linux/crc32.h>
43 #include <asm/uaccess.h>
47 #include "delegation.h"
55 #define NFSDBG_FACILITY NFSDBG_VFS
57 #define NFS_64_BIT_INODE_NUMBERS_ENABLED 1
59 /* Default is to see 64-bit inode numbers */
60 static bool enable_ino64 = NFS_64_BIT_INODE_NUMBERS_ENABLED;
62 static void nfs_invalidate_inode(struct inode *);
63 static int nfs_update_inode(struct inode *, struct nfs_fattr *);
65 static struct kmem_cache * nfs_inode_cachep;
67 static inline unsigned long
68 nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
70 return nfs_fileid_to_ino_t(fattr->fileid);
74 * nfs_wait_bit_killable - helper for functions that are sleeping on bit locks
75 * @word: long word containing the bit lock
77 int nfs_wait_bit_killable(void *word)
79 if (fatal_signal_pending(current))
84 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable);
87 * nfs_compat_user_ino64 - returns the user-visible inode number
88 * @fileid: 64-bit fileid
90 * This function returns a 32-bit inode number if the boot parameter
91 * nfs.enable_ino64 is zero.
93 u64 nfs_compat_user_ino64(u64 fileid)
104 if (sizeof(ino) < sizeof(fileid))
105 ino ^= fileid >> (sizeof(fileid)-sizeof(ino)) * 8;
109 int nfs_drop_inode(struct inode *inode)
111 return NFS_STALE(inode) || generic_drop_inode(inode);
113 EXPORT_SYMBOL_GPL(nfs_drop_inode);
115 void nfs_clear_inode(struct inode *inode)
118 * The following should never happen...
120 WARN_ON_ONCE(nfs_have_writebacks(inode));
121 WARN_ON_ONCE(!list_empty(&NFS_I(inode)->open_files));
122 nfs_zap_acl_cache(inode);
123 nfs_access_zap_cache(inode);
124 nfs_fscache_release_inode_cookie(inode);
126 EXPORT_SYMBOL_GPL(nfs_clear_inode);
128 void nfs_evict_inode(struct inode *inode)
130 truncate_inode_pages(&inode->i_data, 0);
132 nfs_clear_inode(inode);
136 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
138 int nfs_sync_mapping(struct address_space *mapping)
142 if (mapping->nrpages != 0) {
143 unmap_mapping_range(mapping, 0, 0, 0);
144 ret = nfs_wb_all(mapping->host);
150 * Invalidate the local caches
152 static void nfs_zap_caches_locked(struct inode *inode)
154 struct nfs_inode *nfsi = NFS_I(inode);
155 int mode = inode->i_mode;
157 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
159 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
160 nfsi->attrtimeo_timestamp = jiffies;
162 memset(NFS_I(inode)->cookieverf, 0, sizeof(NFS_I(inode)->cookieverf));
163 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)) {
164 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
165 nfs_fscache_invalidate(inode);
167 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
171 void nfs_zap_caches(struct inode *inode)
173 spin_lock(&inode->i_lock);
174 nfs_zap_caches_locked(inode);
175 spin_unlock(&inode->i_lock);
178 void nfs_zap_mapping(struct inode *inode, struct address_space *mapping)
180 if (mapping->nrpages != 0) {
181 spin_lock(&inode->i_lock);
182 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_DATA;
183 nfs_fscache_invalidate(inode);
184 spin_unlock(&inode->i_lock);
188 void nfs_zap_acl_cache(struct inode *inode)
190 void (*clear_acl_cache)(struct inode *);
192 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
193 if (clear_acl_cache != NULL)
194 clear_acl_cache(inode);
195 spin_lock(&inode->i_lock);
196 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
197 spin_unlock(&inode->i_lock);
199 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache);
201 void nfs_invalidate_atime(struct inode *inode)
203 spin_lock(&inode->i_lock);
204 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ATIME;
205 spin_unlock(&inode->i_lock);
207 EXPORT_SYMBOL_GPL(nfs_invalidate_atime);
210 * Invalidate, but do not unhash, the inode.
211 * NB: must be called with inode->i_lock held!
213 static void nfs_invalidate_inode(struct inode *inode)
215 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
216 nfs_zap_caches_locked(inode);
219 struct nfs_find_desc {
221 struct nfs_fattr *fattr;
225 * In NFSv3 we can have 64bit inode numbers. In order to support
226 * this, and re-exported directories (also seen in NFSv2)
227 * we are forced to allow 2 different inodes to have the same
231 nfs_find_actor(struct inode *inode, void *opaque)
233 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
234 struct nfs_fh *fh = desc->fh;
235 struct nfs_fattr *fattr = desc->fattr;
237 if (NFS_FILEID(inode) != fattr->fileid)
239 if ((S_IFMT & inode->i_mode) != (S_IFMT & fattr->mode))
241 if (nfs_compare_fh(NFS_FH(inode), fh))
243 if (is_bad_inode(inode) || NFS_STALE(inode))
249 nfs_init_locked(struct inode *inode, void *opaque)
251 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
252 struct nfs_fattr *fattr = desc->fattr;
254 set_nfs_fileid(inode, fattr->fileid);
255 nfs_copy_fh(NFS_FH(inode), desc->fh);
260 * This is our front-end to iget that looks up inodes by file handle
261 * instead of inode number.
264 nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
266 struct nfs_find_desc desc = {
270 struct inode *inode = ERR_PTR(-ENOENT);
273 nfs_attr_check_mountpoint(sb, fattr);
275 if (((fattr->valid & NFS_ATTR_FATTR_FILEID) == 0) &&
276 !nfs_attr_use_mounted_on_fileid(fattr))
278 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) == 0)
281 hash = nfs_fattr_to_ino_t(fattr);
283 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
285 inode = ERR_PTR(-ENOMEM);
289 if (inode->i_state & I_NEW) {
290 struct nfs_inode *nfsi = NFS_I(inode);
291 unsigned long now = jiffies;
293 /* We set i_ino for the few things that still rely on it,
297 /* We can't support update_atime(), since the server will reset it */
298 inode->i_flags |= S_NOATIME|S_NOCMTIME;
299 inode->i_mode = fattr->mode;
300 if ((fattr->valid & NFS_ATTR_FATTR_MODE) == 0
301 && nfs_server_capable(inode, NFS_CAP_MODE))
302 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
303 /* Why so? Because we want revalidate for devices/FIFOs, and
304 * that's precisely what we have in nfs_file_inode_operations.
306 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->file_inode_ops;
307 if (S_ISREG(inode->i_mode)) {
308 inode->i_fop = NFS_SB(sb)->nfs_client->rpc_ops->file_ops;
309 inode->i_data.a_ops = &nfs_file_aops;
310 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
311 } else if (S_ISDIR(inode->i_mode)) {
312 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->dir_inode_ops;
313 inode->i_fop = &nfs_dir_operations;
314 inode->i_data.a_ops = &nfs_dir_aops;
315 /* Deal with crossing mountpoints */
316 if (fattr->valid & NFS_ATTR_FATTR_MOUNTPOINT ||
317 fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL) {
318 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)
319 inode->i_op = &nfs_referral_inode_operations;
321 inode->i_op = &nfs_mountpoint_inode_operations;
323 inode->i_flags |= S_AUTOMOUNT;
325 } else if (S_ISLNK(inode->i_mode))
326 inode->i_op = &nfs_symlink_inode_operations;
328 init_special_inode(inode, inode->i_mode, fattr->rdev);
330 memset(&inode->i_atime, 0, sizeof(inode->i_atime));
331 memset(&inode->i_mtime, 0, sizeof(inode->i_mtime));
332 memset(&inode->i_ctime, 0, sizeof(inode->i_ctime));
333 inode->i_version = 0;
336 inode->i_uid = make_kuid(&init_user_ns, -2);
337 inode->i_gid = make_kgid(&init_user_ns, -2);
339 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
343 nfsi->read_cache_jiffies = fattr->time_start;
344 nfsi->attr_gencount = fattr->gencount;
345 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
346 inode->i_atime = fattr->atime;
347 else if (nfs_server_capable(inode, NFS_CAP_ATIME))
348 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
349 if (fattr->valid & NFS_ATTR_FATTR_MTIME)
350 inode->i_mtime = fattr->mtime;
351 else if (nfs_server_capable(inode, NFS_CAP_MTIME))
352 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
353 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
354 inode->i_ctime = fattr->ctime;
355 else if (nfs_server_capable(inode, NFS_CAP_CTIME))
356 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
357 if (fattr->valid & NFS_ATTR_FATTR_CHANGE)
358 inode->i_version = fattr->change_attr;
359 else if (nfs_server_capable(inode, NFS_CAP_CHANGE_ATTR))
360 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
361 if (fattr->valid & NFS_ATTR_FATTR_SIZE)
362 inode->i_size = nfs_size_to_loff_t(fattr->size);
364 nfsi->cache_validity |= NFS_INO_INVALID_ATTR
365 | NFS_INO_REVAL_PAGECACHE;
366 if (fattr->valid & NFS_ATTR_FATTR_NLINK)
367 set_nlink(inode, fattr->nlink);
368 else if (nfs_server_capable(inode, NFS_CAP_NLINK))
369 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
370 if (fattr->valid & NFS_ATTR_FATTR_OWNER)
371 inode->i_uid = fattr->uid;
372 else if (nfs_server_capable(inode, NFS_CAP_OWNER))
373 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
374 if (fattr->valid & NFS_ATTR_FATTR_GROUP)
375 inode->i_gid = fattr->gid;
376 else if (nfs_server_capable(inode, NFS_CAP_OWNER_GROUP))
377 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
378 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
379 inode->i_blocks = fattr->du.nfs2.blocks;
380 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
382 * report the blocks in 512byte units
384 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
386 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
387 nfsi->attrtimeo_timestamp = now;
388 nfsi->access_cache = RB_ROOT;
390 nfs_fscache_init_inode_cookie(inode);
392 unlock_new_inode(inode);
394 nfs_refresh_inode(inode, fattr);
395 dprintk("NFS: nfs_fhget(%s/%Ld fh_crc=0x%08x ct=%d)\n",
397 (long long)NFS_FILEID(inode),
398 nfs_display_fhandle_hash(fh),
399 atomic_read(&inode->i_count));
405 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
408 EXPORT_SYMBOL_GPL(nfs_fhget);
410 #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)
413 nfs_setattr(struct dentry *dentry, struct iattr *attr)
415 struct inode *inode = dentry->d_inode;
416 struct nfs_fattr *fattr;
419 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
421 /* skip mode change if it's just for clearing setuid/setgid */
422 if (attr->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
423 attr->ia_valid &= ~ATTR_MODE;
425 if (attr->ia_valid & ATTR_SIZE) {
426 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
427 attr->ia_valid &= ~ATTR_SIZE;
430 /* Optimization: if the end result is no change, don't RPC */
431 attr->ia_valid &= NFS_VALID_ATTRS;
432 if ((attr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
435 /* Write all dirty data */
436 if (S_ISREG(inode->i_mode)) {
437 nfs_inode_dio_wait(inode);
441 fattr = nfs_alloc_fattr();
445 * Return any delegations if we're going to change ACLs
447 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
448 NFS_PROTO(inode)->return_delegation(inode);
449 error = NFS_PROTO(inode)->setattr(dentry, fattr, attr);
451 nfs_refresh_inode(inode, fattr);
452 nfs_free_fattr(fattr);
456 EXPORT_SYMBOL_GPL(nfs_setattr);
459 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
460 * @inode: inode of the file used
461 * @offset: file offset to start truncating
463 * This is a copy of the common vmtruncate, but with the locking
464 * corrected to take into account the fact that NFS requires
465 * inode->i_size to be updated under the inode->i_lock.
467 static int nfs_vmtruncate(struct inode * inode, loff_t offset)
472 err = inode_newsize_ok(inode, offset);
476 spin_lock(&inode->i_lock);
477 oldsize = inode->i_size;
478 i_size_write(inode, offset);
479 spin_unlock(&inode->i_lock);
481 truncate_pagecache(inode, oldsize, offset);
487 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
488 * @inode: pointer to struct inode
489 * @attr: pointer to struct iattr
491 * Note: we do this in the *proc.c in order to ensure that
492 * it works for things like exclusive creates too.
494 void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
496 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
497 spin_lock(&inode->i_lock);
498 if ((attr->ia_valid & ATTR_MODE) != 0) {
499 int mode = attr->ia_mode & S_IALLUGO;
500 mode |= inode->i_mode & ~S_IALLUGO;
501 inode->i_mode = mode;
503 if ((attr->ia_valid & ATTR_UID) != 0)
504 inode->i_uid = attr->ia_uid;
505 if ((attr->ia_valid & ATTR_GID) != 0)
506 inode->i_gid = attr->ia_gid;
507 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
508 spin_unlock(&inode->i_lock);
510 if ((attr->ia_valid & ATTR_SIZE) != 0) {
511 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
512 nfs_vmtruncate(inode, attr->ia_size);
515 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode);
517 int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
519 struct inode *inode = dentry->d_inode;
520 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
523 /* Flush out writes to the server in order to update c/mtime. */
524 if (S_ISREG(inode->i_mode)) {
525 nfs_inode_dio_wait(inode);
526 err = filemap_write_and_wait(inode->i_mapping);
532 * We may force a getattr if the user cares about atime.
534 * Note that we only have to check the vfsmount flags here:
535 * - NFS always sets S_NOATIME by so checking it would give a
537 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
538 * no point in checking those.
540 if ((mnt->mnt_flags & MNT_NOATIME) ||
541 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
545 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
547 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
549 generic_fillattr(inode, stat);
550 stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode));
555 EXPORT_SYMBOL_GPL(nfs_getattr);
557 static void nfs_init_lock_context(struct nfs_lock_context *l_ctx)
559 atomic_set(&l_ctx->count, 1);
560 l_ctx->lockowner.l_owner = current->files;
561 l_ctx->lockowner.l_pid = current->tgid;
562 INIT_LIST_HEAD(&l_ctx->list);
563 nfs_iocounter_init(&l_ctx->io_count);
566 static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context *ctx)
568 struct nfs_lock_context *head = &ctx->lock_context;
569 struct nfs_lock_context *pos = head;
572 if (pos->lockowner.l_owner != current->files)
574 if (pos->lockowner.l_pid != current->tgid)
576 atomic_inc(&pos->count);
578 } while ((pos = list_entry(pos->list.next, typeof(*pos), list)) != head);
582 struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx)
584 struct nfs_lock_context *res, *new = NULL;
585 struct inode *inode = ctx->dentry->d_inode;
587 spin_lock(&inode->i_lock);
588 res = __nfs_find_lock_context(ctx);
590 spin_unlock(&inode->i_lock);
591 new = kmalloc(sizeof(*new), GFP_KERNEL);
593 return ERR_PTR(-ENOMEM);
594 nfs_init_lock_context(new);
595 spin_lock(&inode->i_lock);
596 res = __nfs_find_lock_context(ctx);
598 list_add_tail(&new->list, &ctx->lock_context.list);
599 new->open_context = ctx;
604 spin_unlock(&inode->i_lock);
609 void nfs_put_lock_context(struct nfs_lock_context *l_ctx)
611 struct nfs_open_context *ctx = l_ctx->open_context;
612 struct inode *inode = ctx->dentry->d_inode;
614 if (!atomic_dec_and_lock(&l_ctx->count, &inode->i_lock))
616 list_del(&l_ctx->list);
617 spin_unlock(&inode->i_lock);
622 * nfs_close_context - Common close_context() routine NFSv2/v3
623 * @ctx: pointer to context
624 * @is_sync: is this a synchronous close
626 * always ensure that the attributes are up to date if we're mounted
627 * with close-to-open semantics
629 void nfs_close_context(struct nfs_open_context *ctx, int is_sync)
632 struct nfs_server *server;
634 if (!(ctx->mode & FMODE_WRITE))
638 inode = ctx->dentry->d_inode;
639 if (!list_empty(&NFS_I(inode)->open_files))
641 server = NFS_SERVER(inode);
642 if (server->flags & NFS_MOUNT_NOCTO)
644 nfs_revalidate_inode(server, inode);
646 EXPORT_SYMBOL_GPL(nfs_close_context);
648 struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, fmode_t f_mode)
650 struct nfs_open_context *ctx;
651 struct rpc_cred *cred = rpc_lookup_cred();
653 return ERR_CAST(cred);
655 ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
658 return ERR_PTR(-ENOMEM);
660 nfs_sb_active(dentry->d_sb);
661 ctx->dentry = dget(dentry);
667 nfs_init_lock_context(&ctx->lock_context);
668 ctx->lock_context.open_context = ctx;
669 INIT_LIST_HEAD(&ctx->list);
670 ctx->mdsthreshold = NULL;
673 EXPORT_SYMBOL_GPL(alloc_nfs_open_context);
675 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
678 atomic_inc(&ctx->lock_context.count);
681 EXPORT_SYMBOL_GPL(get_nfs_open_context);
683 static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
685 struct inode *inode = ctx->dentry->d_inode;
686 struct super_block *sb = ctx->dentry->d_sb;
688 if (!list_empty(&ctx->list)) {
689 if (!atomic_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
691 list_del(&ctx->list);
692 spin_unlock(&inode->i_lock);
693 } else if (!atomic_dec_and_test(&ctx->lock_context.count))
696 NFS_PROTO(inode)->close_context(ctx, is_sync);
697 if (ctx->cred != NULL)
698 put_rpccred(ctx->cred);
701 kfree(ctx->mdsthreshold);
705 void put_nfs_open_context(struct nfs_open_context *ctx)
707 __put_nfs_open_context(ctx, 0);
709 EXPORT_SYMBOL_GPL(put_nfs_open_context);
712 * Ensure that mmap has a recent RPC credential for use when writing out
715 void nfs_inode_attach_open_context(struct nfs_open_context *ctx)
717 struct inode *inode = ctx->dentry->d_inode;
718 struct nfs_inode *nfsi = NFS_I(inode);
720 spin_lock(&inode->i_lock);
721 list_add(&ctx->list, &nfsi->open_files);
722 spin_unlock(&inode->i_lock);
724 EXPORT_SYMBOL_GPL(nfs_inode_attach_open_context);
726 void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
728 filp->private_data = get_nfs_open_context(ctx);
729 if (list_empty(&ctx->list))
730 nfs_inode_attach_open_context(ctx);
732 EXPORT_SYMBOL_GPL(nfs_file_set_open_context);
735 * Given an inode, search for an open context with the desired characteristics
737 struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, fmode_t mode)
739 struct nfs_inode *nfsi = NFS_I(inode);
740 struct nfs_open_context *pos, *ctx = NULL;
742 spin_lock(&inode->i_lock);
743 list_for_each_entry(pos, &nfsi->open_files, list) {
744 if (cred != NULL && pos->cred != cred)
746 if ((pos->mode & (FMODE_READ|FMODE_WRITE)) != mode)
748 ctx = get_nfs_open_context(pos);
751 spin_unlock(&inode->i_lock);
755 static void nfs_file_clear_open_context(struct file *filp)
757 struct nfs_open_context *ctx = nfs_file_open_context(filp);
760 struct inode *inode = ctx->dentry->d_inode;
762 filp->private_data = NULL;
763 spin_lock(&inode->i_lock);
764 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
765 spin_unlock(&inode->i_lock);
766 __put_nfs_open_context(ctx, filp->f_flags & O_DIRECT ? 0 : 1);
771 * These allocate and release file read/write context information.
773 int nfs_open(struct inode *inode, struct file *filp)
775 struct nfs_open_context *ctx;
777 ctx = alloc_nfs_open_context(filp->f_path.dentry, filp->f_mode);
780 nfs_file_set_open_context(filp, ctx);
781 put_nfs_open_context(ctx);
782 nfs_fscache_set_inode_cookie(inode, filp);
786 int nfs_release(struct inode *inode, struct file *filp)
788 nfs_file_clear_open_context(filp);
793 * This function is called whenever some part of NFS notices that
794 * the cached attributes have to be refreshed.
797 __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
799 int status = -ESTALE;
800 struct nfs_fattr *fattr = NULL;
801 struct nfs_inode *nfsi = NFS_I(inode);
803 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
804 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
806 if (is_bad_inode(inode))
808 if (NFS_STALE(inode))
812 fattr = nfs_alloc_fattr();
816 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
817 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), fattr);
819 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
821 (long long)NFS_FILEID(inode), status);
822 if (status == -ESTALE) {
823 nfs_zap_caches(inode);
824 if (!S_ISDIR(inode->i_mode))
825 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
830 status = nfs_refresh_inode(inode, fattr);
832 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
834 (long long)NFS_FILEID(inode), status);
838 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
839 nfs_zap_acl_cache(inode);
841 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
843 (long long)NFS_FILEID(inode));
846 nfs_free_fattr(fattr);
850 int nfs_attribute_timeout(struct inode *inode)
852 struct nfs_inode *nfsi = NFS_I(inode);
854 return !time_in_range_open(jiffies, nfsi->read_cache_jiffies, nfsi->read_cache_jiffies + nfsi->attrtimeo);
857 static int nfs_attribute_cache_expired(struct inode *inode)
859 if (nfs_have_delegated_attributes(inode))
861 return nfs_attribute_timeout(inode);
865 * nfs_revalidate_inode - Revalidate the inode attributes
866 * @server - pointer to nfs_server struct
867 * @inode - pointer to inode struct
869 * Updates inode attribute information by retrieving the data from the server.
871 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
873 if (!(NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATTR)
874 && !nfs_attribute_cache_expired(inode))
875 return NFS_STALE(inode) ? -ESTALE : 0;
876 return __nfs_revalidate_inode(server, inode);
878 EXPORT_SYMBOL_GPL(nfs_revalidate_inode);
880 static int nfs_invalidate_mapping(struct inode *inode, struct address_space *mapping)
882 struct nfs_inode *nfsi = NFS_I(inode);
884 if (mapping->nrpages != 0) {
885 int ret = invalidate_inode_pages2(mapping);
889 spin_lock(&inode->i_lock);
890 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
891 if (S_ISDIR(inode->i_mode))
892 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
893 spin_unlock(&inode->i_lock);
894 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
895 nfs_fscache_wait_on_invalidate(inode);
896 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
897 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
901 static bool nfs_mapping_need_revalidate_inode(struct inode *inode)
903 if (nfs_have_delegated_attributes(inode))
905 return (NFS_I(inode)->cache_validity & NFS_INO_REVAL_PAGECACHE)
906 || nfs_attribute_timeout(inode)
911 * nfs_revalidate_mapping - Revalidate the pagecache
912 * @inode - pointer to host inode
913 * @mapping - pointer to mapping
915 int nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
917 struct nfs_inode *nfsi = NFS_I(inode);
920 /* swapfiles are not supposed to be shared. */
921 if (IS_SWAPFILE(inode))
924 if (nfs_mapping_need_revalidate_inode(inode)) {
925 ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
929 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
930 ret = nfs_invalidate_mapping(inode, mapping);
935 static unsigned long nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
937 struct nfs_inode *nfsi = NFS_I(inode);
938 unsigned long ret = 0;
940 if ((fattr->valid & NFS_ATTR_FATTR_PRECHANGE)
941 && (fattr->valid & NFS_ATTR_FATTR_CHANGE)
942 && inode->i_version == fattr->pre_change_attr) {
943 inode->i_version = fattr->change_attr;
944 if (S_ISDIR(inode->i_mode))
945 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
946 ret |= NFS_INO_INVALID_ATTR;
948 /* If we have atomic WCC data, we may update some attributes */
949 if ((fattr->valid & NFS_ATTR_FATTR_PRECTIME)
950 && (fattr->valid & NFS_ATTR_FATTR_CTIME)
951 && timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
952 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
953 ret |= NFS_INO_INVALID_ATTR;
956 if ((fattr->valid & NFS_ATTR_FATTR_PREMTIME)
957 && (fattr->valid & NFS_ATTR_FATTR_MTIME)
958 && timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
959 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
960 if (S_ISDIR(inode->i_mode))
961 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
962 ret |= NFS_INO_INVALID_ATTR;
964 if ((fattr->valid & NFS_ATTR_FATTR_PRESIZE)
965 && (fattr->valid & NFS_ATTR_FATTR_SIZE)
966 && i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size)
967 && nfsi->npages == 0) {
968 i_size_write(inode, nfs_size_to_loff_t(fattr->size));
969 ret |= NFS_INO_INVALID_ATTR;
972 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
973 nfs_fscache_invalidate(inode);
979 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
980 * @inode - pointer to inode
981 * @fattr - updated attributes
983 * Verifies the attribute cache. If we have just changed the attributes,
984 * so that fattr carries weak cache consistency data, then it may
985 * also update the ctime/mtime/change_attribute.
987 static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
989 struct nfs_inode *nfsi = NFS_I(inode);
990 loff_t cur_size, new_isize;
991 unsigned long invalid = 0;
994 if (nfs_have_delegated_attributes(inode))
996 /* Has the inode gone and changed behind our back? */
997 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid)
999 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1002 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1003 inode->i_version != fattr->change_attr)
1004 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1006 /* Verify a few of the more important attributes */
1007 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) && !timespec_equal(&inode->i_mtime, &fattr->mtime))
1008 invalid |= NFS_INO_INVALID_ATTR;
1010 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1011 cur_size = i_size_read(inode);
1012 new_isize = nfs_size_to_loff_t(fattr->size);
1013 if (cur_size != new_isize && nfsi->npages == 0)
1014 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1017 /* Have any file permissions changed? */
1018 if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO))
1019 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1020 if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && !uid_eq(inode->i_uid, fattr->uid))
1021 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1022 if ((fattr->valid & NFS_ATTR_FATTR_GROUP) && !gid_eq(inode->i_gid, fattr->gid))
1023 invalid |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1025 /* Has the link count changed? */
1026 if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink)
1027 invalid |= NFS_INO_INVALID_ATTR;
1029 if ((fattr->valid & NFS_ATTR_FATTR_ATIME) && !timespec_equal(&inode->i_atime, &fattr->atime))
1030 invalid |= NFS_INO_INVALID_ATIME;
1033 nfsi->cache_validity |= invalid;
1035 nfsi->read_cache_jiffies = fattr->time_start;
1039 static int nfs_ctime_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1041 if (!(fattr->valid & NFS_ATTR_FATTR_CTIME))
1043 return timespec_compare(&fattr->ctime, &inode->i_ctime) > 0;
1046 static int nfs_size_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1048 if (!(fattr->valid & NFS_ATTR_FATTR_SIZE))
1050 return nfs_size_to_loff_t(fattr->size) > i_size_read(inode);
1053 static atomic_long_t nfs_attr_generation_counter;
1055 static unsigned long nfs_read_attr_generation_counter(void)
1057 return atomic_long_read(&nfs_attr_generation_counter);
1060 unsigned long nfs_inc_attr_generation_counter(void)
1062 return atomic_long_inc_return(&nfs_attr_generation_counter);
1065 void nfs_fattr_init(struct nfs_fattr *fattr)
1068 fattr->time_start = jiffies;
1069 fattr->gencount = nfs_inc_attr_generation_counter();
1070 fattr->owner_name = NULL;
1071 fattr->group_name = NULL;
1073 EXPORT_SYMBOL_GPL(nfs_fattr_init);
1075 struct nfs_fattr *nfs_alloc_fattr(void)
1077 struct nfs_fattr *fattr;
1079 fattr = kmalloc(sizeof(*fattr), GFP_NOFS);
1081 nfs_fattr_init(fattr);
1084 EXPORT_SYMBOL_GPL(nfs_alloc_fattr);
1086 struct nfs_fh *nfs_alloc_fhandle(void)
1090 fh = kmalloc(sizeof(struct nfs_fh), GFP_NOFS);
1095 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle);
1099 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1100 * in the same way that wireshark does
1104 * For debugging only.
1106 u32 _nfs_display_fhandle_hash(const struct nfs_fh *fh)
1108 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1109 * not on the result */
1110 return ~crc32(0xFFFFFFFF, &fh->data[0], fh->size);
1114 * _nfs_display_fhandle - display an NFS file handle on the console
1116 * @fh: file handle to display
1117 * @caption: display caption
1119 * For debugging only.
1121 void _nfs_display_fhandle(const struct nfs_fh *fh, const char *caption)
1125 if (fh == NULL || fh->size == 0) {
1126 printk(KERN_DEFAULT "%s at %p is empty\n", caption, fh);
1130 printk(KERN_DEFAULT "%s at %p is %u bytes, crc: 0x%08x:\n",
1131 caption, fh, fh->size, _nfs_display_fhandle_hash(fh));
1132 for (i = 0; i < fh->size; i += 16) {
1133 __be32 *pos = (__be32 *)&fh->data[i];
1135 switch ((fh->size - i - 1) >> 2) {
1137 printk(KERN_DEFAULT " %08x\n",
1141 printk(KERN_DEFAULT " %08x %08x\n",
1142 be32_to_cpup(pos), be32_to_cpup(pos + 1));
1145 printk(KERN_DEFAULT " %08x %08x %08x\n",
1146 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1147 be32_to_cpup(pos + 2));
1150 printk(KERN_DEFAULT " %08x %08x %08x %08x\n",
1151 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1152 be32_to_cpup(pos + 2), be32_to_cpup(pos + 3));
1159 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1160 * @inode - pointer to inode
1161 * @fattr - attributes
1163 * Attempt to divine whether or not an RPC call reply carrying stale
1164 * attributes got scheduled after another call carrying updated ones.
1166 * To do so, the function first assumes that a more recent ctime means
1167 * that the attributes in fattr are newer, however it also attempt to
1168 * catch the case where ctime either didn't change, or went backwards
1169 * (if someone reset the clock on the server) by looking at whether
1170 * or not this RPC call was started after the inode was last updated.
1171 * Note also the check for wraparound of 'attr_gencount'
1173 * The function returns 'true' if it thinks the attributes in 'fattr' are
1174 * more recent than the ones cached in the inode.
1177 static int nfs_inode_attrs_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1179 const struct nfs_inode *nfsi = NFS_I(inode);
1181 return ((long)fattr->gencount - (long)nfsi->attr_gencount) > 0 ||
1182 nfs_ctime_need_update(inode, fattr) ||
1183 nfs_size_need_update(inode, fattr) ||
1184 ((long)nfsi->attr_gencount - (long)nfs_read_attr_generation_counter() > 0);
1187 static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr)
1189 if (nfs_inode_attrs_need_update(inode, fattr))
1190 return nfs_update_inode(inode, fattr);
1191 return nfs_check_inode_attributes(inode, fattr);
1195 * nfs_refresh_inode - try to update the inode attribute cache
1196 * @inode - pointer to inode
1197 * @fattr - updated attributes
1199 * Check that an RPC call that returned attributes has not overlapped with
1200 * other recent updates of the inode metadata, then decide whether it is
1201 * safe to do a full update of the inode attributes, or whether just to
1202 * call nfs_check_inode_attributes.
1204 int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1208 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1210 spin_lock(&inode->i_lock);
1211 status = nfs_refresh_inode_locked(inode, fattr);
1212 spin_unlock(&inode->i_lock);
1216 EXPORT_SYMBOL_GPL(nfs_refresh_inode);
1218 static int nfs_post_op_update_inode_locked(struct inode *inode, struct nfs_fattr *fattr)
1220 struct nfs_inode *nfsi = NFS_I(inode);
1222 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1223 if (S_ISDIR(inode->i_mode)) {
1224 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1225 nfs_fscache_invalidate(inode);
1227 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1229 return nfs_refresh_inode_locked(inode, fattr);
1233 * nfs_post_op_update_inode - try to update the inode attribute cache
1234 * @inode - pointer to inode
1235 * @fattr - updated attributes
1237 * After an operation that has changed the inode metadata, mark the
1238 * attribute cache as being invalid, then try to update it.
1240 * NB: if the server didn't return any post op attributes, this
1241 * function will force the retrieval of attributes before the next
1242 * NFS request. Thus it should be used only for operations that
1243 * are expected to change one or more attributes, to avoid
1244 * unnecessary NFS requests and trips through nfs_update_inode().
1246 int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1250 spin_lock(&inode->i_lock);
1251 status = nfs_post_op_update_inode_locked(inode, fattr);
1252 spin_unlock(&inode->i_lock);
1255 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode);
1258 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1259 * @inode - pointer to inode
1260 * @fattr - updated attributes
1262 * After an operation that has changed the inode metadata, mark the
1263 * attribute cache as being invalid, then try to update it. Fake up
1264 * weak cache consistency data, if none exist.
1266 * This function is mainly designed to be used by the ->write_done() functions.
1268 int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fattr)
1272 spin_lock(&inode->i_lock);
1273 /* Don't do a WCC update if these attributes are already stale */
1274 if ((fattr->valid & NFS_ATTR_FATTR) == 0 ||
1275 !nfs_inode_attrs_need_update(inode, fattr)) {
1276 fattr->valid &= ~(NFS_ATTR_FATTR_PRECHANGE
1277 | NFS_ATTR_FATTR_PRESIZE
1278 | NFS_ATTR_FATTR_PREMTIME
1279 | NFS_ATTR_FATTR_PRECTIME);
1282 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1283 (fattr->valid & NFS_ATTR_FATTR_PRECHANGE) == 0) {
1284 fattr->pre_change_attr = inode->i_version;
1285 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
1287 if ((fattr->valid & NFS_ATTR_FATTR_CTIME) != 0 &&
1288 (fattr->valid & NFS_ATTR_FATTR_PRECTIME) == 0) {
1289 memcpy(&fattr->pre_ctime, &inode->i_ctime, sizeof(fattr->pre_ctime));
1290 fattr->valid |= NFS_ATTR_FATTR_PRECTIME;
1292 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) != 0 &&
1293 (fattr->valid & NFS_ATTR_FATTR_PREMTIME) == 0) {
1294 memcpy(&fattr->pre_mtime, &inode->i_mtime, sizeof(fattr->pre_mtime));
1295 fattr->valid |= NFS_ATTR_FATTR_PREMTIME;
1297 if ((fattr->valid & NFS_ATTR_FATTR_SIZE) != 0 &&
1298 (fattr->valid & NFS_ATTR_FATTR_PRESIZE) == 0) {
1299 fattr->pre_size = i_size_read(inode);
1300 fattr->valid |= NFS_ATTR_FATTR_PRESIZE;
1303 status = nfs_post_op_update_inode_locked(inode, fattr);
1304 spin_unlock(&inode->i_lock);
1307 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc);
1310 * Many nfs protocol calls return the new file attributes after
1311 * an operation. Here we update the inode to reflect the state
1312 * of the server's inode.
1314 * This is a bit tricky because we have to make sure all dirty pages
1315 * have been sent off to the server before calling invalidate_inode_pages.
1316 * To make sure no other process adds more write requests while we try
1317 * our best to flush them, we make them sleep during the attribute refresh.
1319 * A very similar scenario holds for the dir cache.
1321 static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1323 struct nfs_server *server;
1324 struct nfs_inode *nfsi = NFS_I(inode);
1325 loff_t cur_isize, new_isize;
1326 unsigned long invalid = 0;
1327 unsigned long now = jiffies;
1328 unsigned long save_cache_validity;
1330 dfprintk(VFS, "NFS: %s(%s/%ld fh_crc=0x%08x ct=%d info=0x%x)\n",
1331 __func__, inode->i_sb->s_id, inode->i_ino,
1332 nfs_display_fhandle_hash(NFS_FH(inode)),
1333 atomic_read(&inode->i_count), fattr->valid);
1335 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid) {
1336 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1337 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1338 NFS_SERVER(inode)->nfs_client->cl_hostname,
1339 inode->i_sb->s_id, (long long)nfsi->fileid,
1340 (long long)fattr->fileid);
1345 * Make sure the inode's type hasn't changed.
1347 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1349 * Big trouble! The inode has become a different object.
1351 printk(KERN_DEBUG "NFS: %s: inode %ld mode changed, %07o to %07o\n",
1352 __func__, inode->i_ino, inode->i_mode, fattr->mode);
1356 server = NFS_SERVER(inode);
1357 /* Update the fsid? */
1358 if (S_ISDIR(inode->i_mode) && (fattr->valid & NFS_ATTR_FATTR_FSID) &&
1359 !nfs_fsid_equal(&server->fsid, &fattr->fsid) &&
1360 !IS_AUTOMOUNT(inode))
1361 server->fsid = fattr->fsid;
1364 * Update the read time so we don't revalidate too often.
1366 nfsi->read_cache_jiffies = fattr->time_start;
1368 save_cache_validity = nfsi->cache_validity;
1369 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR
1370 | NFS_INO_INVALID_ATIME
1371 | NFS_INO_REVAL_FORCED
1372 | NFS_INO_REVAL_PAGECACHE);
1374 /* Do atomic weak cache consistency updates */
1375 invalid |= nfs_wcc_update_inode(inode, fattr);
1377 /* More cache consistency checks */
1378 if (fattr->valid & NFS_ATTR_FATTR_CHANGE) {
1379 if (inode->i_version != fattr->change_attr) {
1380 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1381 inode->i_sb->s_id, inode->i_ino);
1382 invalid |= NFS_INO_INVALID_ATTR
1383 | NFS_INO_INVALID_DATA
1384 | NFS_INO_INVALID_ACCESS
1385 | NFS_INO_INVALID_ACL
1386 | NFS_INO_REVAL_PAGECACHE;
1387 if (S_ISDIR(inode->i_mode))
1388 nfs_force_lookup_revalidate(inode);
1389 inode->i_version = fattr->change_attr;
1391 } else if (server->caps & NFS_CAP_CHANGE_ATTR)
1392 invalid |= save_cache_validity;
1394 if (fattr->valid & NFS_ATTR_FATTR_MTIME) {
1395 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1396 } else if (server->caps & NFS_CAP_MTIME)
1397 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1398 | NFS_INO_REVAL_FORCED);
1400 if (fattr->valid & NFS_ATTR_FATTR_CTIME) {
1401 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1402 } else if (server->caps & NFS_CAP_CTIME)
1403 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1404 | NFS_INO_REVAL_FORCED);
1406 /* Check if our cached file size is stale */
1407 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1408 new_isize = nfs_size_to_loff_t(fattr->size);
1409 cur_isize = i_size_read(inode);
1410 if (new_isize != cur_isize) {
1411 /* Do we perhaps have any outstanding writes, or has
1412 * the file grown beyond our last write? */
1413 if ((nfsi->npages == 0 && !test_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) ||
1414 new_isize > cur_isize) {
1415 i_size_write(inode, new_isize);
1416 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1418 dprintk("NFS: isize change on server for file %s/%ld "
1422 (long long)cur_isize,
1423 (long long)new_isize);
1426 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1427 | NFS_INO_REVAL_PAGECACHE
1428 | NFS_INO_REVAL_FORCED);
1431 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
1432 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1433 else if (server->caps & NFS_CAP_ATIME)
1434 invalid |= save_cache_validity & (NFS_INO_INVALID_ATIME
1435 | NFS_INO_REVAL_FORCED);
1437 if (fattr->valid & NFS_ATTR_FATTR_MODE) {
1438 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) {
1439 umode_t newmode = inode->i_mode & S_IFMT;
1440 newmode |= fattr->mode & S_IALLUGO;
1441 inode->i_mode = newmode;
1442 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1444 } else if (server->caps & NFS_CAP_MODE)
1445 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1446 | NFS_INO_INVALID_ACCESS
1447 | NFS_INO_INVALID_ACL
1448 | NFS_INO_REVAL_FORCED);
1450 if (fattr->valid & NFS_ATTR_FATTR_OWNER) {
1451 if (!uid_eq(inode->i_uid, fattr->uid)) {
1452 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1453 inode->i_uid = fattr->uid;
1455 } else if (server->caps & NFS_CAP_OWNER)
1456 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1457 | NFS_INO_INVALID_ACCESS
1458 | NFS_INO_INVALID_ACL
1459 | NFS_INO_REVAL_FORCED);
1461 if (fattr->valid & NFS_ATTR_FATTR_GROUP) {
1462 if (!gid_eq(inode->i_gid, fattr->gid)) {
1463 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1464 inode->i_gid = fattr->gid;
1466 } else if (server->caps & NFS_CAP_OWNER_GROUP)
1467 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1468 | NFS_INO_INVALID_ACCESS
1469 | NFS_INO_INVALID_ACL
1470 | NFS_INO_REVAL_FORCED);
1472 if (fattr->valid & NFS_ATTR_FATTR_NLINK) {
1473 if (inode->i_nlink != fattr->nlink) {
1474 invalid |= NFS_INO_INVALID_ATTR;
1475 if (S_ISDIR(inode->i_mode))
1476 invalid |= NFS_INO_INVALID_DATA;
1477 set_nlink(inode, fattr->nlink);
1479 } else if (server->caps & NFS_CAP_NLINK)
1480 invalid |= save_cache_validity & (NFS_INO_INVALID_ATTR
1481 | NFS_INO_REVAL_FORCED);
1483 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
1485 * report the blocks in 512byte units
1487 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1489 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
1490 inode->i_blocks = fattr->du.nfs2.blocks;
1492 /* Update attrtimeo value if we're out of the unstable period */
1493 if (invalid & NFS_INO_INVALID_ATTR) {
1494 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
1495 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1496 nfsi->attrtimeo_timestamp = now;
1497 nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1499 if (!time_in_range_open(now, nfsi->attrtimeo_timestamp, nfsi->attrtimeo_timestamp + nfsi->attrtimeo)) {
1500 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1501 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1502 nfsi->attrtimeo_timestamp = now;
1505 invalid &= ~NFS_INO_INVALID_ATTR;
1506 /* Don't invalidate the data if we were to blame */
1507 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1508 || S_ISLNK(inode->i_mode)))
1509 invalid &= ~NFS_INO_INVALID_DATA;
1510 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ) ||
1511 (save_cache_validity & NFS_INO_REVAL_FORCED))
1512 nfsi->cache_validity |= invalid;
1514 if (invalid & NFS_INO_INVALID_DATA)
1515 nfs_fscache_invalidate(inode);
1520 * No need to worry about unhashing the dentry, as the
1521 * lookup validation will know that the inode is bad.
1522 * (But we fall through to invalidate the caches.)
1524 nfs_invalidate_inode(inode);
1528 struct inode *nfs_alloc_inode(struct super_block *sb)
1530 struct nfs_inode *nfsi;
1531 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, GFP_KERNEL);
1535 nfsi->cache_validity = 0UL;
1536 #ifdef CONFIG_NFS_V3_ACL
1537 nfsi->acl_access = ERR_PTR(-EAGAIN);
1538 nfsi->acl_default = ERR_PTR(-EAGAIN);
1540 #if IS_ENABLED(CONFIG_NFS_V4)
1541 nfsi->nfs4_acl = NULL;
1542 #endif /* CONFIG_NFS_V4 */
1543 return &nfsi->vfs_inode;
1545 EXPORT_SYMBOL_GPL(nfs_alloc_inode);
1547 static void nfs_i_callback(struct rcu_head *head)
1549 struct inode *inode = container_of(head, struct inode, i_rcu);
1550 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1553 void nfs_destroy_inode(struct inode *inode)
1555 call_rcu(&inode->i_rcu, nfs_i_callback);
1557 EXPORT_SYMBOL_GPL(nfs_destroy_inode);
1559 static inline void nfs4_init_once(struct nfs_inode *nfsi)
1561 #if IS_ENABLED(CONFIG_NFS_V4)
1562 INIT_LIST_HEAD(&nfsi->open_states);
1563 nfsi->delegation = NULL;
1564 nfsi->delegation_state = 0;
1565 init_rwsem(&nfsi->rwsem);
1566 nfsi->layout = NULL;
1570 static void init_once(void *foo)
1572 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1574 inode_init_once(&nfsi->vfs_inode);
1575 INIT_LIST_HEAD(&nfsi->open_files);
1576 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
1577 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
1578 INIT_LIST_HEAD(&nfsi->commit_info.list);
1580 nfsi->commit_info.ncommit = 0;
1581 atomic_set(&nfsi->commit_info.rpcs_out, 0);
1582 atomic_set(&nfsi->silly_count, 1);
1583 INIT_HLIST_HEAD(&nfsi->silly_list);
1584 init_waitqueue_head(&nfsi->waitqueue);
1585 nfs4_init_once(nfsi);
1588 static int __init nfs_init_inodecache(void)
1590 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
1591 sizeof(struct nfs_inode),
1592 0, (SLAB_RECLAIM_ACCOUNT|
1595 if (nfs_inode_cachep == NULL)
1601 static void nfs_destroy_inodecache(void)
1604 * Make sure all delayed rcu free inodes are flushed before we
1608 kmem_cache_destroy(nfs_inode_cachep);
1611 struct workqueue_struct *nfsiod_workqueue;
1612 EXPORT_SYMBOL_GPL(nfsiod_workqueue);
1615 * start up the nfsiod workqueue
1617 static int nfsiod_start(void)
1619 struct workqueue_struct *wq;
1620 dprintk("RPC: creating workqueue nfsiod\n");
1621 wq = alloc_workqueue("nfsiod", WQ_MEM_RECLAIM, 0);
1624 nfsiod_workqueue = wq;
1629 * Destroy the nfsiod workqueue
1631 static void nfsiod_stop(void)
1633 struct workqueue_struct *wq;
1635 wq = nfsiod_workqueue;
1638 nfsiod_workqueue = NULL;
1639 destroy_workqueue(wq);
1643 EXPORT_SYMBOL_GPL(nfs_net_id);
1645 static int nfs_net_init(struct net *net)
1647 nfs_clients_init(net);
1651 static void nfs_net_exit(struct net *net)
1653 nfs_cleanup_cb_ident_idr(net);
1656 static struct pernet_operations nfs_net_ops = {
1657 .init = nfs_net_init,
1658 .exit = nfs_net_exit,
1660 .size = sizeof(struct nfs_net),
1666 static int __init init_nfs_fs(void)
1670 err = register_pernet_subsys(&nfs_net_ops);
1674 err = nfs_fscache_register();
1678 err = nfsiod_start();
1682 err = nfs_fs_proc_init();
1686 err = nfs_init_nfspagecache();
1690 err = nfs_init_inodecache();
1694 err = nfs_init_readpagecache();
1698 err = nfs_init_writepagecache();
1702 err = nfs_init_directcache();
1706 #ifdef CONFIG_PROC_FS
1707 rpc_proc_register(&init_net, &nfs_rpcstat);
1709 if ((err = register_nfs_fs()) != 0)
1714 #ifdef CONFIG_PROC_FS
1715 rpc_proc_unregister(&init_net, "nfs");
1717 nfs_destroy_directcache();
1719 nfs_destroy_writepagecache();
1721 nfs_destroy_readpagecache();
1723 nfs_destroy_inodecache();
1725 nfs_destroy_nfspagecache();
1731 nfs_fscache_unregister();
1733 unregister_pernet_subsys(&nfs_net_ops);
1738 static void __exit exit_nfs_fs(void)
1740 nfs_destroy_directcache();
1741 nfs_destroy_writepagecache();
1742 nfs_destroy_readpagecache();
1743 nfs_destroy_inodecache();
1744 nfs_destroy_nfspagecache();
1745 nfs_fscache_unregister();
1746 unregister_pernet_subsys(&nfs_net_ops);
1747 #ifdef CONFIG_PROC_FS
1748 rpc_proc_unregister(&init_net, "nfs");
1750 unregister_nfs_fs();
1755 /* Not quite true; I just maintain it */
1756 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1757 MODULE_LICENSE("GPL");
1758 module_param(enable_ino64, bool, 0644);
1760 module_init(init_nfs_fs)
1761 module_exit(exit_nfs_fs)