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/signal.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/uaccess.h>
42 #include <linux/iversion.h>
46 #include "delegation.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);
74 static int nfs_wait_killable(int mode)
76 freezable_schedule_unsafe();
77 if (signal_pending_state(mode, current))
82 int nfs_wait_bit_killable(struct wait_bit_key *key, int mode)
84 return nfs_wait_killable(mode);
86 EXPORT_SYMBOL_GPL(nfs_wait_bit_killable);
89 * nfs_compat_user_ino64 - returns the user-visible inode number
90 * @fileid: 64-bit fileid
92 * This function returns a 32-bit inode number if the boot parameter
93 * nfs.enable_ino64 is zero.
95 u64 nfs_compat_user_ino64(u64 fileid)
106 if (sizeof(ino) < sizeof(fileid))
107 ino ^= fileid >> (sizeof(fileid)-sizeof(ino)) * 8;
111 int nfs_drop_inode(struct inode *inode)
113 return NFS_STALE(inode) || generic_drop_inode(inode);
115 EXPORT_SYMBOL_GPL(nfs_drop_inode);
117 void nfs_clear_inode(struct inode *inode)
120 * The following should never happen...
122 WARN_ON_ONCE(nfs_have_writebacks(inode));
123 WARN_ON_ONCE(!list_empty(&NFS_I(inode)->open_files));
124 nfs_zap_acl_cache(inode);
125 nfs_access_zap_cache(inode);
126 nfs_fscache_clear_inode(inode);
128 EXPORT_SYMBOL_GPL(nfs_clear_inode);
130 void nfs_evict_inode(struct inode *inode)
132 truncate_inode_pages_final(&inode->i_data);
134 nfs_clear_inode(inode);
137 int nfs_sync_inode(struct inode *inode)
139 inode_dio_wait(inode);
140 return nfs_wb_all(inode);
142 EXPORT_SYMBOL_GPL(nfs_sync_inode);
145 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
147 int nfs_sync_mapping(struct address_space *mapping)
151 if (mapping->nrpages != 0) {
152 unmap_mapping_range(mapping, 0, 0, 0);
153 ret = nfs_wb_all(mapping->host);
158 static int nfs_attribute_timeout(struct inode *inode)
160 struct nfs_inode *nfsi = NFS_I(inode);
162 return !time_in_range_open(jiffies, nfsi->read_cache_jiffies, nfsi->read_cache_jiffies + nfsi->attrtimeo);
165 static bool nfs_check_cache_invalid_delegated(struct inode *inode, unsigned long flags)
167 unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
169 /* Special case for the pagecache or access cache */
170 if (flags == NFS_INO_REVAL_PAGECACHE &&
171 !(cache_validity & NFS_INO_REVAL_FORCED))
173 return (cache_validity & flags) != 0;
176 static bool nfs_check_cache_invalid_not_delegated(struct inode *inode, unsigned long flags)
178 unsigned long cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
180 if ((cache_validity & flags) != 0)
182 if (nfs_attribute_timeout(inode))
187 bool nfs_check_cache_invalid(struct inode *inode, unsigned long flags)
189 if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
190 return nfs_check_cache_invalid_delegated(inode, flags);
192 return nfs_check_cache_invalid_not_delegated(inode, flags);
195 static void nfs_set_cache_invalid(struct inode *inode, unsigned long flags)
197 struct nfs_inode *nfsi = NFS_I(inode);
198 bool have_delegation = NFS_PROTO(inode)->have_delegation(inode, FMODE_READ);
200 if (have_delegation) {
201 if (!(flags & NFS_INO_REVAL_FORCED))
202 flags &= ~NFS_INO_INVALID_OTHER;
203 flags &= ~(NFS_INO_INVALID_CHANGE
204 | NFS_INO_INVALID_SIZE
205 | NFS_INO_REVAL_PAGECACHE);
208 if (inode->i_mapping->nrpages == 0)
209 flags &= ~NFS_INO_INVALID_DATA;
210 nfsi->cache_validity |= flags;
211 if (flags & NFS_INO_INVALID_DATA)
212 nfs_fscache_invalidate(inode);
216 * Invalidate the local caches
218 static void nfs_zap_caches_locked(struct inode *inode)
220 struct nfs_inode *nfsi = NFS_I(inode);
221 int mode = inode->i_mode;
223 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
225 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
226 nfsi->attrtimeo_timestamp = jiffies;
228 memset(NFS_I(inode)->cookieverf, 0, sizeof(NFS_I(inode)->cookieverf));
229 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)) {
230 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATTR
231 | NFS_INO_INVALID_DATA
232 | NFS_INO_INVALID_ACCESS
233 | NFS_INO_INVALID_ACL
234 | NFS_INO_REVAL_PAGECACHE);
236 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATTR
237 | NFS_INO_INVALID_ACCESS
238 | NFS_INO_INVALID_ACL
239 | NFS_INO_REVAL_PAGECACHE);
240 nfs_zap_label_cache_locked(nfsi);
243 void nfs_zap_caches(struct inode *inode)
245 spin_lock(&inode->i_lock);
246 nfs_zap_caches_locked(inode);
247 spin_unlock(&inode->i_lock);
250 void nfs_zap_mapping(struct inode *inode, struct address_space *mapping)
252 if (mapping->nrpages != 0) {
253 spin_lock(&inode->i_lock);
254 nfs_set_cache_invalid(inode, NFS_INO_INVALID_DATA);
255 spin_unlock(&inode->i_lock);
259 void nfs_zap_acl_cache(struct inode *inode)
261 void (*clear_acl_cache)(struct inode *);
263 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
264 if (clear_acl_cache != NULL)
265 clear_acl_cache(inode);
266 spin_lock(&inode->i_lock);
267 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
268 spin_unlock(&inode->i_lock);
270 EXPORT_SYMBOL_GPL(nfs_zap_acl_cache);
272 void nfs_invalidate_atime(struct inode *inode)
274 spin_lock(&inode->i_lock);
275 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATIME);
276 spin_unlock(&inode->i_lock);
278 EXPORT_SYMBOL_GPL(nfs_invalidate_atime);
281 * Invalidate, but do not unhash, the inode.
282 * NB: must be called with inode->i_lock held!
284 static void nfs_invalidate_inode(struct inode *inode)
286 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
287 nfs_zap_caches_locked(inode);
290 struct nfs_find_desc {
292 struct nfs_fattr *fattr;
296 * In NFSv3 we can have 64bit inode numbers. In order to support
297 * this, and re-exported directories (also seen in NFSv2)
298 * we are forced to allow 2 different inodes to have the same
302 nfs_find_actor(struct inode *inode, void *opaque)
304 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
305 struct nfs_fh *fh = desc->fh;
306 struct nfs_fattr *fattr = desc->fattr;
308 if (NFS_FILEID(inode) != fattr->fileid)
310 if ((S_IFMT & inode->i_mode) != (S_IFMT & fattr->mode))
312 if (nfs_compare_fh(NFS_FH(inode), fh))
314 if (is_bad_inode(inode) || NFS_STALE(inode))
320 nfs_init_locked(struct inode *inode, void *opaque)
322 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
323 struct nfs_fattr *fattr = desc->fattr;
325 set_nfs_fileid(inode, fattr->fileid);
326 inode->i_mode = fattr->mode;
327 nfs_copy_fh(NFS_FH(inode), desc->fh);
331 #ifdef CONFIG_NFS_V4_SECURITY_LABEL
332 static void nfs_clear_label_invalid(struct inode *inode)
334 spin_lock(&inode->i_lock);
335 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_LABEL;
336 spin_unlock(&inode->i_lock);
339 void nfs_setsecurity(struct inode *inode, struct nfs_fattr *fattr,
340 struct nfs4_label *label)
347 if ((fattr->valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL) && inode->i_security) {
348 error = security_inode_notifysecctx(inode, label->label,
351 printk(KERN_ERR "%s() %s %d "
352 "security_inode_notifysecctx() %d\n",
354 (char *)label->label,
356 nfs_clear_label_invalid(inode);
360 struct nfs4_label *nfs4_label_alloc(struct nfs_server *server, gfp_t flags)
362 struct nfs4_label *label = NULL;
363 int minor_version = server->nfs_client->cl_minorversion;
365 if (minor_version < 2)
368 if (!(server->caps & NFS_CAP_SECURITY_LABEL))
371 label = kzalloc(sizeof(struct nfs4_label), flags);
373 return ERR_PTR(-ENOMEM);
375 label->label = kzalloc(NFS4_MAXLABELLEN, flags);
376 if (label->label == NULL) {
378 return ERR_PTR(-ENOMEM);
380 label->len = NFS4_MAXLABELLEN;
384 EXPORT_SYMBOL_GPL(nfs4_label_alloc);
386 void nfs_setsecurity(struct inode *inode, struct nfs_fattr *fattr,
387 struct nfs4_label *label)
391 EXPORT_SYMBOL_GPL(nfs_setsecurity);
393 /* Search for inode identified by fh, fileid and i_mode in inode cache. */
395 nfs_ilookup(struct super_block *sb, struct nfs_fattr *fattr, struct nfs_fh *fh)
397 struct nfs_find_desc desc = {
404 if (!(fattr->valid & NFS_ATTR_FATTR_FILEID) ||
405 !(fattr->valid & NFS_ATTR_FATTR_TYPE))
408 hash = nfs_fattr_to_ino_t(fattr);
409 inode = ilookup5(sb, hash, nfs_find_actor, &desc);
411 dprintk("%s: returning %p\n", __func__, inode);
416 * This is our front-end to iget that looks up inodes by file handle
417 * instead of inode number.
420 nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr, struct nfs4_label *label)
422 struct nfs_find_desc desc = {
426 struct inode *inode = ERR_PTR(-ENOENT);
429 nfs_attr_check_mountpoint(sb, fattr);
431 if (nfs_attr_use_mounted_on_fileid(fattr))
432 fattr->fileid = fattr->mounted_on_fileid;
433 else if ((fattr->valid & NFS_ATTR_FATTR_FILEID) == 0)
435 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) == 0)
438 hash = nfs_fattr_to_ino_t(fattr);
440 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
442 inode = ERR_PTR(-ENOMEM);
446 if (inode->i_state & I_NEW) {
447 struct nfs_inode *nfsi = NFS_I(inode);
448 unsigned long now = jiffies;
450 /* We set i_ino for the few things that still rely on it,
454 /* We can't support update_atime(), since the server will reset it */
455 inode->i_flags |= S_NOATIME|S_NOCMTIME;
456 inode->i_mode = fattr->mode;
457 nfsi->cache_validity = 0;
458 if ((fattr->valid & NFS_ATTR_FATTR_MODE) == 0
459 && nfs_server_capable(inode, NFS_CAP_MODE))
460 nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER);
461 /* Why so? Because we want revalidate for devices/FIFOs, and
462 * that's precisely what we have in nfs_file_inode_operations.
464 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->file_inode_ops;
465 if (S_ISREG(inode->i_mode)) {
466 inode->i_fop = NFS_SB(sb)->nfs_client->rpc_ops->file_ops;
467 inode->i_data.a_ops = &nfs_file_aops;
468 } else if (S_ISDIR(inode->i_mode)) {
469 inode->i_op = NFS_SB(sb)->nfs_client->rpc_ops->dir_inode_ops;
470 inode->i_fop = &nfs_dir_operations;
471 inode->i_data.a_ops = &nfs_dir_aops;
472 /* Deal with crossing mountpoints */
473 if (fattr->valid & NFS_ATTR_FATTR_MOUNTPOINT ||
474 fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL) {
475 if (fattr->valid & NFS_ATTR_FATTR_V4_REFERRAL)
476 inode->i_op = &nfs_referral_inode_operations;
478 inode->i_op = &nfs_mountpoint_inode_operations;
480 inode->i_flags |= S_AUTOMOUNT;
482 } else if (S_ISLNK(inode->i_mode)) {
483 inode->i_op = &nfs_symlink_inode_operations;
484 inode_nohighmem(inode);
486 init_special_inode(inode, inode->i_mode, fattr->rdev);
488 memset(&inode->i_atime, 0, sizeof(inode->i_atime));
489 memset(&inode->i_mtime, 0, sizeof(inode->i_mtime));
490 memset(&inode->i_ctime, 0, sizeof(inode->i_ctime));
491 inode_set_iversion_raw(inode, 0);
494 inode->i_uid = make_kuid(&init_user_ns, -2);
495 inode->i_gid = make_kgid(&init_user_ns, -2);
497 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
501 nfsi->read_cache_jiffies = fattr->time_start;
502 nfsi->attr_gencount = fattr->gencount;
503 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
504 inode->i_atime = timespec_to_timespec64(fattr->atime);
505 else if (nfs_server_capable(inode, NFS_CAP_ATIME))
506 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATIME);
507 if (fattr->valid & NFS_ATTR_FATTR_MTIME)
508 inode->i_mtime = timespec_to_timespec64(fattr->mtime);
509 else if (nfs_server_capable(inode, NFS_CAP_MTIME))
510 nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME);
511 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
512 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
513 else if (nfs_server_capable(inode, NFS_CAP_CTIME))
514 nfs_set_cache_invalid(inode, NFS_INO_INVALID_CTIME);
515 if (fattr->valid & NFS_ATTR_FATTR_CHANGE)
516 inode_set_iversion_raw(inode, fattr->change_attr);
518 nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE);
519 if (fattr->valid & NFS_ATTR_FATTR_SIZE)
520 inode->i_size = nfs_size_to_loff_t(fattr->size);
522 nfs_set_cache_invalid(inode, NFS_INO_INVALID_SIZE);
523 if (fattr->valid & NFS_ATTR_FATTR_NLINK)
524 set_nlink(inode, fattr->nlink);
525 else if (nfs_server_capable(inode, NFS_CAP_NLINK))
526 nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER);
527 if (fattr->valid & NFS_ATTR_FATTR_OWNER)
528 inode->i_uid = fattr->uid;
529 else if (nfs_server_capable(inode, NFS_CAP_OWNER))
530 nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER);
531 if (fattr->valid & NFS_ATTR_FATTR_GROUP)
532 inode->i_gid = fattr->gid;
533 else if (nfs_server_capable(inode, NFS_CAP_OWNER_GROUP))
534 nfs_set_cache_invalid(inode, NFS_INO_INVALID_OTHER);
535 if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
536 inode->i_blocks = fattr->du.nfs2.blocks;
537 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
539 * report the blocks in 512byte units
541 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
544 if (nfsi->cache_validity != 0)
545 nfsi->cache_validity |= NFS_INO_REVAL_FORCED;
547 nfs_setsecurity(inode, fattr, label);
549 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
550 nfsi->attrtimeo_timestamp = now;
551 nfsi->access_cache = RB_ROOT;
553 nfs_fscache_init_inode(inode);
555 unlock_new_inode(inode);
557 int err = nfs_refresh_inode(inode, fattr);
560 inode = ERR_PTR(err);
564 dprintk("NFS: nfs_fhget(%s/%Lu fh_crc=0x%08x ct=%d)\n",
566 (unsigned long long)NFS_FILEID(inode),
567 nfs_display_fhandle_hash(fh),
568 atomic_read(&inode->i_count));
574 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
577 EXPORT_SYMBOL_GPL(nfs_fhget);
579 #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)
582 nfs_setattr(struct dentry *dentry, struct iattr *attr)
584 struct inode *inode = d_inode(dentry);
585 struct nfs_fattr *fattr;
588 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
590 /* skip mode change if it's just for clearing setuid/setgid */
591 if (attr->ia_valid & (ATTR_KILL_SUID | ATTR_KILL_SGID))
592 attr->ia_valid &= ~ATTR_MODE;
594 if (attr->ia_valid & ATTR_SIZE) {
595 BUG_ON(!S_ISREG(inode->i_mode));
597 error = inode_newsize_ok(inode, attr->ia_size);
601 if (attr->ia_size == i_size_read(inode))
602 attr->ia_valid &= ~ATTR_SIZE;
605 /* Optimization: if the end result is no change, don't RPC */
606 attr->ia_valid &= NFS_VALID_ATTRS;
607 if ((attr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
610 trace_nfs_setattr_enter(inode);
612 /* Write all dirty data */
613 if (S_ISREG(inode->i_mode))
614 nfs_sync_inode(inode);
616 fattr = nfs_alloc_fattr();
622 error = NFS_PROTO(inode)->setattr(dentry, fattr, attr);
624 error = nfs_refresh_inode(inode, fattr);
625 nfs_free_fattr(fattr);
627 trace_nfs_setattr_exit(inode, error);
630 EXPORT_SYMBOL_GPL(nfs_setattr);
633 * nfs_vmtruncate - unmap mappings "freed" by truncate() syscall
634 * @inode: inode of the file used
635 * @offset: file offset to start truncating
637 * This is a copy of the common vmtruncate, but with the locking
638 * corrected to take into account the fact that NFS requires
639 * inode->i_size to be updated under the inode->i_lock.
640 * Note: must be called with inode->i_lock held!
642 static int nfs_vmtruncate(struct inode * inode, loff_t offset)
646 err = inode_newsize_ok(inode, offset);
650 i_size_write(inode, offset);
653 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_DATA;
654 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_SIZE;
656 spin_unlock(&inode->i_lock);
657 truncate_pagecache(inode, offset);
658 spin_lock(&inode->i_lock);
664 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
665 * @inode: pointer to struct inode
666 * @attr: pointer to struct iattr
667 * @fattr: pointer to struct nfs_fattr
669 * Note: we do this in the *proc.c in order to ensure that
670 * it works for things like exclusive creates too.
672 void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
673 struct nfs_fattr *fattr)
675 /* Barrier: bump the attribute generation count. */
676 nfs_fattr_set_barrier(fattr);
678 spin_lock(&inode->i_lock);
679 NFS_I(inode)->attr_gencount = fattr->gencount;
680 if ((attr->ia_valid & ATTR_SIZE) != 0) {
681 nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME);
682 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
683 nfs_vmtruncate(inode, attr->ia_size);
685 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
686 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_CTIME;
687 if ((attr->ia_valid & ATTR_MODE) != 0) {
688 int mode = attr->ia_mode & S_IALLUGO;
689 mode |= inode->i_mode & ~S_IALLUGO;
690 inode->i_mode = mode;
692 if ((attr->ia_valid & ATTR_UID) != 0)
693 inode->i_uid = attr->ia_uid;
694 if ((attr->ia_valid & ATTR_GID) != 0)
695 inode->i_gid = attr->ia_gid;
696 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
697 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
699 nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE
700 | NFS_INO_INVALID_CTIME);
701 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ACCESS
702 | NFS_INO_INVALID_ACL);
704 if (attr->ia_valid & (ATTR_ATIME_SET|ATTR_ATIME)) {
705 NFS_I(inode)->cache_validity &= ~(NFS_INO_INVALID_ATIME
706 | NFS_INO_INVALID_CTIME);
707 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
708 inode->i_atime = timespec_to_timespec64(fattr->atime);
709 else if (attr->ia_valid & ATTR_ATIME_SET)
710 inode->i_atime = attr->ia_atime;
712 nfs_set_cache_invalid(inode, NFS_INO_INVALID_ATIME);
714 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
715 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
717 nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE
718 | NFS_INO_INVALID_CTIME);
720 if (attr->ia_valid & (ATTR_MTIME_SET|ATTR_MTIME)) {
721 NFS_I(inode)->cache_validity &= ~(NFS_INO_INVALID_MTIME
722 | NFS_INO_INVALID_CTIME);
723 if (fattr->valid & NFS_ATTR_FATTR_MTIME)
724 inode->i_mtime = timespec_to_timespec64(fattr->mtime);
725 else if (attr->ia_valid & ATTR_MTIME_SET)
726 inode->i_mtime = attr->ia_mtime;
728 nfs_set_cache_invalid(inode, NFS_INO_INVALID_MTIME);
730 if (fattr->valid & NFS_ATTR_FATTR_CTIME)
731 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
733 nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE
734 | NFS_INO_INVALID_CTIME);
737 nfs_update_inode(inode, fattr);
738 spin_unlock(&inode->i_lock);
740 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode);
742 static void nfs_readdirplus_parent_cache_miss(struct dentry *dentry)
744 struct dentry *parent;
746 if (!nfs_server_capable(d_inode(dentry), NFS_CAP_READDIRPLUS))
748 parent = dget_parent(dentry);
749 nfs_force_use_readdirplus(d_inode(parent));
753 static void nfs_readdirplus_parent_cache_hit(struct dentry *dentry)
755 struct dentry *parent;
757 if (!nfs_server_capable(d_inode(dentry), NFS_CAP_READDIRPLUS))
759 parent = dget_parent(dentry);
760 nfs_advise_use_readdirplus(d_inode(parent));
764 static bool nfs_need_revalidate_inode(struct inode *inode)
766 if (NFS_I(inode)->cache_validity &
767 (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_LABEL))
769 if (nfs_attribute_cache_expired(inode))
774 int nfs_getattr(const struct path *path, struct kstat *stat,
775 u32 request_mask, unsigned int query_flags)
777 struct inode *inode = d_inode(path->dentry);
778 struct nfs_server *server = NFS_SERVER(inode);
779 unsigned long cache_validity;
781 bool force_sync = query_flags & AT_STATX_FORCE_SYNC;
782 bool do_update = false;
784 trace_nfs_getattr_enter(inode);
786 if ((query_flags & AT_STATX_DONT_SYNC) && !force_sync)
789 /* Flush out writes to the server in order to update c/mtime. */
790 if ((request_mask & (STATX_CTIME|STATX_MTIME)) &&
791 S_ISREG(inode->i_mode)) {
792 err = filemap_write_and_wait(inode->i_mapping);
798 * We may force a getattr if the user cares about atime.
800 * Note that we only have to check the vfsmount flags here:
801 * - NFS always sets S_NOATIME by so checking it would give a
803 * - NFS never sets SB_NOATIME or SB_NODIRATIME so there is
804 * no point in checking those.
806 if ((path->mnt->mnt_flags & MNT_NOATIME) ||
807 ((path->mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
808 request_mask &= ~STATX_ATIME;
810 /* Is the user requesting attributes that might need revalidation? */
811 if (!(request_mask & (STATX_MODE|STATX_NLINK|STATX_ATIME|STATX_CTIME|
812 STATX_MTIME|STATX_UID|STATX_GID|
813 STATX_SIZE|STATX_BLOCKS)))
814 goto out_no_revalidate;
816 /* Check whether the cached attributes are stale */
817 do_update |= force_sync || nfs_attribute_cache_expired(inode);
818 cache_validity = READ_ONCE(NFS_I(inode)->cache_validity);
819 do_update |= cache_validity &
820 (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_LABEL);
821 if (request_mask & STATX_ATIME)
822 do_update |= cache_validity & NFS_INO_INVALID_ATIME;
823 if (request_mask & (STATX_CTIME|STATX_MTIME))
824 do_update |= cache_validity & NFS_INO_REVAL_PAGECACHE;
826 /* Update the attribute cache */
827 if (!(server->flags & NFS_MOUNT_NOAC))
828 nfs_readdirplus_parent_cache_miss(path->dentry);
830 nfs_readdirplus_parent_cache_hit(path->dentry);
831 err = __nfs_revalidate_inode(server, inode);
835 nfs_readdirplus_parent_cache_hit(path->dentry);
837 /* Only return attributes that were revalidated. */
838 stat->result_mask &= request_mask;
840 generic_fillattr(inode, stat);
841 stat->ino = nfs_compat_user_ino64(NFS_FILEID(inode));
842 if (S_ISDIR(inode->i_mode))
843 stat->blksize = NFS_SERVER(inode)->dtsize;
845 trace_nfs_getattr_exit(inode, err);
848 EXPORT_SYMBOL_GPL(nfs_getattr);
850 static void nfs_init_lock_context(struct nfs_lock_context *l_ctx)
852 refcount_set(&l_ctx->count, 1);
853 l_ctx->lockowner = current->files;
854 INIT_LIST_HEAD(&l_ctx->list);
855 atomic_set(&l_ctx->io_count, 0);
858 static struct nfs_lock_context *__nfs_find_lock_context(struct nfs_open_context *ctx)
860 struct nfs_lock_context *head = &ctx->lock_context;
861 struct nfs_lock_context *pos = head;
864 if (pos->lockowner != current->files)
866 refcount_inc(&pos->count);
868 } while ((pos = list_entry(pos->list.next, typeof(*pos), list)) != head);
872 struct nfs_lock_context *nfs_get_lock_context(struct nfs_open_context *ctx)
874 struct nfs_lock_context *res, *new = NULL;
875 struct inode *inode = d_inode(ctx->dentry);
877 spin_lock(&inode->i_lock);
878 res = __nfs_find_lock_context(ctx);
880 spin_unlock(&inode->i_lock);
881 new = kmalloc(sizeof(*new), GFP_KERNEL);
883 return ERR_PTR(-ENOMEM);
884 nfs_init_lock_context(new);
885 spin_lock(&inode->i_lock);
886 res = __nfs_find_lock_context(ctx);
888 list_add_tail(&new->list, &ctx->lock_context.list);
889 new->open_context = ctx;
894 spin_unlock(&inode->i_lock);
898 EXPORT_SYMBOL_GPL(nfs_get_lock_context);
900 void nfs_put_lock_context(struct nfs_lock_context *l_ctx)
902 struct nfs_open_context *ctx = l_ctx->open_context;
903 struct inode *inode = d_inode(ctx->dentry);
905 if (!refcount_dec_and_lock(&l_ctx->count, &inode->i_lock))
907 list_del(&l_ctx->list);
908 spin_unlock(&inode->i_lock);
911 EXPORT_SYMBOL_GPL(nfs_put_lock_context);
914 * nfs_close_context - Common close_context() routine NFSv2/v3
915 * @ctx: pointer to context
916 * @is_sync: is this a synchronous close
918 * Ensure that the attributes are up to date if we're mounted
919 * with close-to-open semantics and we have cached data that will
920 * need to be revalidated on open.
922 void nfs_close_context(struct nfs_open_context *ctx, int is_sync)
924 struct nfs_inode *nfsi;
926 struct nfs_server *server;
928 if (!(ctx->mode & FMODE_WRITE))
932 inode = d_inode(ctx->dentry);
933 if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
936 if (inode->i_mapping->nrpages == 0)
938 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
940 if (!list_empty(&nfsi->open_files))
942 server = NFS_SERVER(inode);
943 if (server->flags & NFS_MOUNT_NOCTO)
945 nfs_revalidate_inode(server, inode);
947 EXPORT_SYMBOL_GPL(nfs_close_context);
949 struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry,
953 struct nfs_open_context *ctx;
954 struct rpc_cred *cred = rpc_lookup_cred();
956 return ERR_CAST(cred);
958 ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
961 return ERR_PTR(-ENOMEM);
963 nfs_sb_active(dentry->d_sb);
964 ctx->dentry = dget(dentry);
970 ctx->flock_owner = (fl_owner_t)filp;
971 nfs_init_lock_context(&ctx->lock_context);
972 ctx->lock_context.open_context = ctx;
973 INIT_LIST_HEAD(&ctx->list);
974 ctx->mdsthreshold = NULL;
977 EXPORT_SYMBOL_GPL(alloc_nfs_open_context);
979 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
982 refcount_inc(&ctx->lock_context.count);
985 EXPORT_SYMBOL_GPL(get_nfs_open_context);
987 static void __put_nfs_open_context(struct nfs_open_context *ctx, int is_sync)
989 struct inode *inode = d_inode(ctx->dentry);
990 struct super_block *sb = ctx->dentry->d_sb;
992 if (!list_empty(&ctx->list)) {
993 if (!refcount_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
995 list_del(&ctx->list);
996 spin_unlock(&inode->i_lock);
997 } else if (!refcount_dec_and_test(&ctx->lock_context.count))
1000 NFS_PROTO(inode)->close_context(ctx, is_sync);
1001 if (ctx->cred != NULL)
1002 put_rpccred(ctx->cred);
1004 nfs_sb_deactive(sb);
1005 kfree(ctx->mdsthreshold);
1009 void put_nfs_open_context(struct nfs_open_context *ctx)
1011 __put_nfs_open_context(ctx, 0);
1013 EXPORT_SYMBOL_GPL(put_nfs_open_context);
1015 static void put_nfs_open_context_sync(struct nfs_open_context *ctx)
1017 __put_nfs_open_context(ctx, 1);
1021 * Ensure that mmap has a recent RPC credential for use when writing out
1024 void nfs_inode_attach_open_context(struct nfs_open_context *ctx)
1026 struct inode *inode = d_inode(ctx->dentry);
1027 struct nfs_inode *nfsi = NFS_I(inode);
1029 spin_lock(&inode->i_lock);
1030 if (ctx->mode & FMODE_WRITE)
1031 list_add(&ctx->list, &nfsi->open_files);
1033 list_add_tail(&ctx->list, &nfsi->open_files);
1034 spin_unlock(&inode->i_lock);
1036 EXPORT_SYMBOL_GPL(nfs_inode_attach_open_context);
1038 void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1040 filp->private_data = get_nfs_open_context(ctx);
1041 set_bit(NFS_CONTEXT_FILE_OPEN, &ctx->flags);
1042 if (list_empty(&ctx->list))
1043 nfs_inode_attach_open_context(ctx);
1045 EXPORT_SYMBOL_GPL(nfs_file_set_open_context);
1048 * Given an inode, search for an open context with the desired characteristics
1050 struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, fmode_t mode)
1052 struct nfs_inode *nfsi = NFS_I(inode);
1053 struct nfs_open_context *pos, *ctx = NULL;
1055 spin_lock(&inode->i_lock);
1056 list_for_each_entry(pos, &nfsi->open_files, list) {
1057 if (cred != NULL && pos->cred != cred)
1059 if ((pos->mode & (FMODE_READ|FMODE_WRITE)) != mode)
1061 if (!test_bit(NFS_CONTEXT_FILE_OPEN, &pos->flags))
1063 ctx = get_nfs_open_context(pos);
1066 spin_unlock(&inode->i_lock);
1070 void nfs_file_clear_open_context(struct file *filp)
1072 struct nfs_open_context *ctx = nfs_file_open_context(filp);
1075 struct inode *inode = d_inode(ctx->dentry);
1077 clear_bit(NFS_CONTEXT_FILE_OPEN, &ctx->flags);
1079 * We fatal error on write before. Try to writeback
1083 invalidate_inode_pages2(inode->i_mapping);
1084 filp->private_data = NULL;
1085 spin_lock(&inode->i_lock);
1086 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1087 spin_unlock(&inode->i_lock);
1088 put_nfs_open_context_sync(ctx);
1093 * These allocate and release file read/write context information.
1095 int nfs_open(struct inode *inode, struct file *filp)
1097 struct nfs_open_context *ctx;
1099 ctx = alloc_nfs_open_context(file_dentry(filp), filp->f_mode, filp);
1101 return PTR_ERR(ctx);
1102 nfs_file_set_open_context(filp, ctx);
1103 put_nfs_open_context(ctx);
1104 nfs_fscache_open_file(inode, filp);
1107 EXPORT_SYMBOL_GPL(nfs_open);
1110 * This function is called whenever some part of NFS notices that
1111 * the cached attributes have to be refreshed.
1114 __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1116 int status = -ESTALE;
1117 struct nfs4_label *label = NULL;
1118 struct nfs_fattr *fattr = NULL;
1119 struct nfs_inode *nfsi = NFS_I(inode);
1121 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Lu)\n",
1122 inode->i_sb->s_id, (unsigned long long)NFS_FILEID(inode));
1124 trace_nfs_revalidate_inode_enter(inode);
1126 if (is_bad_inode(inode))
1128 if (NFS_STALE(inode))
1131 /* pNFS: Attributes aren't updated until we layoutcommit */
1132 if (S_ISREG(inode->i_mode)) {
1133 status = pnfs_sync_inode(inode, false);
1139 fattr = nfs_alloc_fattr();
1143 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
1145 label = nfs4_label_alloc(NFS_SERVER(inode), GFP_KERNEL);
1146 if (IS_ERR(label)) {
1147 status = PTR_ERR(label);
1151 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), fattr,
1154 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Lu) getattr failed, error=%d\n",
1156 (unsigned long long)NFS_FILEID(inode), status);
1157 if (status == -ESTALE) {
1158 nfs_zap_caches(inode);
1159 if (!S_ISDIR(inode->i_mode))
1160 set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
1165 status = nfs_refresh_inode(inode, fattr);
1167 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Lu) refresh failed, error=%d\n",
1169 (unsigned long long)NFS_FILEID(inode), status);
1173 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
1174 nfs_zap_acl_cache(inode);
1176 nfs_setsecurity(inode, fattr, label);
1178 dfprintk(PAGECACHE, "NFS: (%s/%Lu) revalidation complete\n",
1180 (unsigned long long)NFS_FILEID(inode));
1183 nfs4_label_free(label);
1185 nfs_free_fattr(fattr);
1186 trace_nfs_revalidate_inode_exit(inode, status);
1190 int nfs_attribute_cache_expired(struct inode *inode)
1192 if (nfs_have_delegated_attributes(inode))
1194 return nfs_attribute_timeout(inode);
1198 * nfs_revalidate_inode - Revalidate the inode attributes
1199 * @server - pointer to nfs_server struct
1200 * @inode - pointer to inode struct
1202 * Updates inode attribute information by retrieving the data from the server.
1204 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1206 if (!nfs_need_revalidate_inode(inode))
1207 return NFS_STALE(inode) ? -ESTALE : 0;
1208 return __nfs_revalidate_inode(server, inode);
1210 EXPORT_SYMBOL_GPL(nfs_revalidate_inode);
1212 static int nfs_invalidate_mapping(struct inode *inode, struct address_space *mapping)
1214 struct nfs_inode *nfsi = NFS_I(inode);
1217 if (mapping->nrpages != 0) {
1218 if (S_ISREG(inode->i_mode)) {
1219 ret = nfs_sync_mapping(mapping);
1223 ret = invalidate_inode_pages2(mapping);
1227 if (S_ISDIR(inode->i_mode)) {
1228 spin_lock(&inode->i_lock);
1229 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1230 spin_unlock(&inode->i_lock);
1232 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
1233 nfs_fscache_wait_on_invalidate(inode);
1235 dfprintk(PAGECACHE, "NFS: (%s/%Lu) data cache invalidated\n",
1237 (unsigned long long)NFS_FILEID(inode));
1241 bool nfs_mapping_need_revalidate_inode(struct inode *inode)
1243 return nfs_check_cache_invalid(inode, NFS_INO_REVAL_PAGECACHE) ||
1247 int nfs_revalidate_mapping_rcu(struct inode *inode)
1249 struct nfs_inode *nfsi = NFS_I(inode);
1250 unsigned long *bitlock = &nfsi->flags;
1253 if (IS_SWAPFILE(inode))
1255 if (nfs_mapping_need_revalidate_inode(inode)) {
1259 spin_lock(&inode->i_lock);
1260 if (test_bit(NFS_INO_INVALIDATING, bitlock) ||
1261 (nfsi->cache_validity & NFS_INO_INVALID_DATA))
1263 spin_unlock(&inode->i_lock);
1269 * nfs_revalidate_mapping - Revalidate the pagecache
1270 * @inode - pointer to host inode
1271 * @mapping - pointer to mapping
1273 int nfs_revalidate_mapping(struct inode *inode,
1274 struct address_space *mapping)
1276 struct nfs_inode *nfsi = NFS_I(inode);
1277 unsigned long *bitlock = &nfsi->flags;
1280 /* swapfiles are not supposed to be shared. */
1281 if (IS_SWAPFILE(inode))
1284 if (nfs_mapping_need_revalidate_inode(inode)) {
1285 ret = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
1291 * We must clear NFS_INO_INVALID_DATA first to ensure that
1292 * invalidations that come in while we're shooting down the mappings
1293 * are respected. But, that leaves a race window where one revalidator
1294 * can clear the flag, and then another checks it before the mapping
1295 * gets invalidated. Fix that by serializing access to this part of
1298 * At the same time, we need to allow other tasks to see whether we
1299 * might be in the middle of invalidating the pages, so we only set
1300 * the bit lock here if it looks like we're going to be doing that.
1303 ret = wait_on_bit_action(bitlock, NFS_INO_INVALIDATING,
1304 nfs_wait_bit_killable, TASK_KILLABLE);
1307 spin_lock(&inode->i_lock);
1308 if (test_bit(NFS_INO_INVALIDATING, bitlock)) {
1309 spin_unlock(&inode->i_lock);
1312 if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
1314 spin_unlock(&inode->i_lock);
1318 set_bit(NFS_INO_INVALIDATING, bitlock);
1320 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
1321 spin_unlock(&inode->i_lock);
1322 trace_nfs_invalidate_mapping_enter(inode);
1323 ret = nfs_invalidate_mapping(inode, mapping);
1324 trace_nfs_invalidate_mapping_exit(inode, ret);
1326 clear_bit_unlock(NFS_INO_INVALIDATING, bitlock);
1327 smp_mb__after_atomic();
1328 wake_up_bit(bitlock, NFS_INO_INVALIDATING);
1333 static bool nfs_file_has_writers(struct nfs_inode *nfsi)
1335 struct inode *inode = &nfsi->vfs_inode;
1337 assert_spin_locked(&inode->i_lock);
1339 if (!S_ISREG(inode->i_mode))
1341 if (list_empty(&nfsi->open_files))
1343 /* Note: This relies on nfsi->open_files being ordered with writers
1344 * being placed at the head of the list.
1345 * See nfs_inode_attach_open_context()
1347 return (list_first_entry(&nfsi->open_files,
1348 struct nfs_open_context,
1349 list)->mode & FMODE_WRITE) == FMODE_WRITE;
1352 static bool nfs_file_has_buffered_writers(struct nfs_inode *nfsi)
1354 return nfs_file_has_writers(nfsi) && nfs_file_io_is_buffered(nfsi);
1357 static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1361 if ((fattr->valid & NFS_ATTR_FATTR_PRECHANGE)
1362 && (fattr->valid & NFS_ATTR_FATTR_CHANGE)
1363 && inode_eq_iversion_raw(inode, fattr->pre_change_attr)) {
1364 inode_set_iversion_raw(inode, fattr->change_attr);
1365 if (S_ISDIR(inode->i_mode))
1366 nfs_set_cache_invalid(inode, NFS_INO_INVALID_DATA);
1368 /* If we have atomic WCC data, we may update some attributes */
1369 ts = timespec64_to_timespec(inode->i_ctime);
1370 if ((fattr->valid & NFS_ATTR_FATTR_PRECTIME)
1371 && (fattr->valid & NFS_ATTR_FATTR_CTIME)
1372 && timespec_equal(&ts, &fattr->pre_ctime)) {
1373 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
1376 ts = timespec64_to_timespec(inode->i_mtime);
1377 if ((fattr->valid & NFS_ATTR_FATTR_PREMTIME)
1378 && (fattr->valid & NFS_ATTR_FATTR_MTIME)
1379 && timespec_equal(&ts, &fattr->pre_mtime)) {
1380 inode->i_mtime = timespec_to_timespec64(fattr->mtime);
1381 if (S_ISDIR(inode->i_mode))
1382 nfs_set_cache_invalid(inode, NFS_INO_INVALID_DATA);
1384 if ((fattr->valid & NFS_ATTR_FATTR_PRESIZE)
1385 && (fattr->valid & NFS_ATTR_FATTR_SIZE)
1386 && i_size_read(inode) == nfs_size_to_loff_t(fattr->pre_size)
1387 && !nfs_have_writebacks(inode)) {
1388 i_size_write(inode, nfs_size_to_loff_t(fattr->size));
1393 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1394 * @inode - pointer to inode
1395 * @fattr - updated attributes
1397 * Verifies the attribute cache. If we have just changed the attributes,
1398 * so that fattr carries weak cache consistency data, then it may
1399 * also update the ctime/mtime/change_attribute.
1401 static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1403 struct nfs_inode *nfsi = NFS_I(inode);
1404 loff_t cur_size, new_isize;
1405 unsigned long invalid = 0;
1408 if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1411 /* Has the inode gone and changed behind our back? */
1412 if ((fattr->valid & NFS_ATTR_FATTR_FILEID) && nfsi->fileid != fattr->fileid)
1414 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1417 if (!nfs_file_has_buffered_writers(nfsi)) {
1418 /* Verify a few of the more important attributes */
1419 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 && !inode_eq_iversion_raw(inode, fattr->change_attr))
1420 invalid |= NFS_INO_INVALID_CHANGE
1421 | NFS_INO_REVAL_PAGECACHE;
1423 ts = timespec64_to_timespec(inode->i_mtime);
1424 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) && !timespec_equal(&ts, &fattr->mtime))
1425 invalid |= NFS_INO_INVALID_MTIME;
1427 ts = timespec64_to_timespec(inode->i_ctime);
1428 if ((fattr->valid & NFS_ATTR_FATTR_CTIME) && !timespec_equal(&ts, &fattr->ctime))
1429 invalid |= NFS_INO_INVALID_CTIME;
1431 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1432 cur_size = i_size_read(inode);
1433 new_isize = nfs_size_to_loff_t(fattr->size);
1434 if (cur_size != new_isize)
1435 invalid |= NFS_INO_INVALID_SIZE
1436 | NFS_INO_REVAL_PAGECACHE;
1440 /* Have any file permissions changed? */
1441 if ((fattr->valid & NFS_ATTR_FATTR_MODE) && (inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO))
1442 invalid |= NFS_INO_INVALID_ACCESS
1443 | NFS_INO_INVALID_ACL
1444 | NFS_INO_INVALID_OTHER;
1445 if ((fattr->valid & NFS_ATTR_FATTR_OWNER) && !uid_eq(inode->i_uid, fattr->uid))
1446 invalid |= NFS_INO_INVALID_ACCESS
1447 | NFS_INO_INVALID_ACL
1448 | NFS_INO_INVALID_OTHER;
1449 if ((fattr->valid & NFS_ATTR_FATTR_GROUP) && !gid_eq(inode->i_gid, fattr->gid))
1450 invalid |= NFS_INO_INVALID_ACCESS
1451 | NFS_INO_INVALID_ACL
1452 | NFS_INO_INVALID_OTHER;
1454 /* Has the link count changed? */
1455 if ((fattr->valid & NFS_ATTR_FATTR_NLINK) && inode->i_nlink != fattr->nlink)
1456 invalid |= NFS_INO_INVALID_OTHER;
1458 ts = timespec64_to_timespec(inode->i_atime);
1459 if ((fattr->valid & NFS_ATTR_FATTR_ATIME) && !timespec_equal(&ts, &fattr->atime))
1460 invalid |= NFS_INO_INVALID_ATIME;
1463 nfs_set_cache_invalid(inode, invalid);
1465 nfsi->read_cache_jiffies = fattr->time_start;
1469 static atomic_long_t nfs_attr_generation_counter;
1471 static unsigned long nfs_read_attr_generation_counter(void)
1473 return atomic_long_read(&nfs_attr_generation_counter);
1476 unsigned long nfs_inc_attr_generation_counter(void)
1478 return atomic_long_inc_return(&nfs_attr_generation_counter);
1480 EXPORT_SYMBOL_GPL(nfs_inc_attr_generation_counter);
1482 void nfs_fattr_init(struct nfs_fattr *fattr)
1485 fattr->time_start = jiffies;
1486 fattr->gencount = nfs_inc_attr_generation_counter();
1487 fattr->owner_name = NULL;
1488 fattr->group_name = NULL;
1490 EXPORT_SYMBOL_GPL(nfs_fattr_init);
1493 * nfs_fattr_set_barrier
1494 * @fattr: attributes
1496 * Used to set a barrier after an attribute was updated. This
1497 * barrier ensures that older attributes from RPC calls that may
1498 * have raced with our update cannot clobber these new values.
1499 * Note that you are still responsible for ensuring that other
1500 * operations which change the attribute on the server do not
1503 void nfs_fattr_set_barrier(struct nfs_fattr *fattr)
1505 fattr->gencount = nfs_inc_attr_generation_counter();
1508 struct nfs_fattr *nfs_alloc_fattr(void)
1510 struct nfs_fattr *fattr;
1512 fattr = kmalloc(sizeof(*fattr), GFP_NOFS);
1514 nfs_fattr_init(fattr);
1517 EXPORT_SYMBOL_GPL(nfs_alloc_fattr);
1519 struct nfs_fh *nfs_alloc_fhandle(void)
1523 fh = kmalloc(sizeof(struct nfs_fh), GFP_NOFS);
1528 EXPORT_SYMBOL_GPL(nfs_alloc_fhandle);
1532 * _nfs_display_fhandle_hash - calculate the crc32 hash for the filehandle
1533 * in the same way that wireshark does
1537 * For debugging only.
1539 u32 _nfs_display_fhandle_hash(const struct nfs_fh *fh)
1541 /* wireshark uses 32-bit AUTODIN crc and does a bitwise
1542 * not on the result */
1543 return nfs_fhandle_hash(fh);
1545 EXPORT_SYMBOL_GPL(_nfs_display_fhandle_hash);
1548 * _nfs_display_fhandle - display an NFS file handle on the console
1550 * @fh: file handle to display
1551 * @caption: display caption
1553 * For debugging only.
1555 void _nfs_display_fhandle(const struct nfs_fh *fh, const char *caption)
1559 if (fh == NULL || fh->size == 0) {
1560 printk(KERN_DEFAULT "%s at %p is empty\n", caption, fh);
1564 printk(KERN_DEFAULT "%s at %p is %u bytes, crc: 0x%08x:\n",
1565 caption, fh, fh->size, _nfs_display_fhandle_hash(fh));
1566 for (i = 0; i < fh->size; i += 16) {
1567 __be32 *pos = (__be32 *)&fh->data[i];
1569 switch ((fh->size - i - 1) >> 2) {
1571 printk(KERN_DEFAULT " %08x\n",
1575 printk(KERN_DEFAULT " %08x %08x\n",
1576 be32_to_cpup(pos), be32_to_cpup(pos + 1));
1579 printk(KERN_DEFAULT " %08x %08x %08x\n",
1580 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1581 be32_to_cpup(pos + 2));
1584 printk(KERN_DEFAULT " %08x %08x %08x %08x\n",
1585 be32_to_cpup(pos), be32_to_cpup(pos + 1),
1586 be32_to_cpup(pos + 2), be32_to_cpup(pos + 3));
1590 EXPORT_SYMBOL_GPL(_nfs_display_fhandle);
1594 * nfs_inode_attrs_need_update - check if the inode attributes need updating
1595 * @inode - pointer to inode
1596 * @fattr - attributes
1598 * Attempt to divine whether or not an RPC call reply carrying stale
1599 * attributes got scheduled after another call carrying updated ones.
1601 * To do so, the function first assumes that a more recent ctime means
1602 * that the attributes in fattr are newer, however it also attempt to
1603 * catch the case where ctime either didn't change, or went backwards
1604 * (if someone reset the clock on the server) by looking at whether
1605 * or not this RPC call was started after the inode was last updated.
1606 * Note also the check for wraparound of 'attr_gencount'
1608 * The function returns 'true' if it thinks the attributes in 'fattr' are
1609 * more recent than the ones cached in the inode.
1612 static int nfs_inode_attrs_need_update(const struct inode *inode, const struct nfs_fattr *fattr)
1614 unsigned long attr_gencount = NFS_I(inode)->attr_gencount;
1616 return (long)(fattr->gencount - attr_gencount) > 0 ||
1617 (long)(attr_gencount - nfs_read_attr_generation_counter()) > 0;
1620 static int nfs_refresh_inode_locked(struct inode *inode, struct nfs_fattr *fattr)
1624 trace_nfs_refresh_inode_enter(inode);
1626 if (nfs_inode_attrs_need_update(inode, fattr))
1627 ret = nfs_update_inode(inode, fattr);
1629 ret = nfs_check_inode_attributes(inode, fattr);
1631 trace_nfs_refresh_inode_exit(inode, ret);
1636 * nfs_refresh_inode - try to update the inode attribute cache
1637 * @inode - pointer to inode
1638 * @fattr - updated attributes
1640 * Check that an RPC call that returned attributes has not overlapped with
1641 * other recent updates of the inode metadata, then decide whether it is
1642 * safe to do a full update of the inode attributes, or whether just to
1643 * call nfs_check_inode_attributes.
1645 int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1649 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1651 spin_lock(&inode->i_lock);
1652 status = nfs_refresh_inode_locked(inode, fattr);
1653 spin_unlock(&inode->i_lock);
1657 EXPORT_SYMBOL_GPL(nfs_refresh_inode);
1659 static int nfs_post_op_update_inode_locked(struct inode *inode,
1660 struct nfs_fattr *fattr, unsigned int invalid)
1662 if (S_ISDIR(inode->i_mode))
1663 invalid |= NFS_INO_INVALID_DATA;
1664 nfs_set_cache_invalid(inode, invalid);
1665 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1667 return nfs_refresh_inode_locked(inode, fattr);
1671 * nfs_post_op_update_inode - try to update the inode attribute cache
1672 * @inode - pointer to inode
1673 * @fattr - updated attributes
1675 * After an operation that has changed the inode metadata, mark the
1676 * attribute cache as being invalid, then try to update it.
1678 * NB: if the server didn't return any post op attributes, this
1679 * function will force the retrieval of attributes before the next
1680 * NFS request. Thus it should be used only for operations that
1681 * are expected to change one or more attributes, to avoid
1682 * unnecessary NFS requests and trips through nfs_update_inode().
1684 int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1688 spin_lock(&inode->i_lock);
1689 nfs_fattr_set_barrier(fattr);
1690 status = nfs_post_op_update_inode_locked(inode, fattr,
1691 NFS_INO_INVALID_CHANGE
1692 | NFS_INO_INVALID_CTIME
1693 | NFS_INO_REVAL_FORCED);
1694 spin_unlock(&inode->i_lock);
1698 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode);
1701 * nfs_post_op_update_inode_force_wcc_locked - update the inode attribute cache
1702 * @inode - pointer to inode
1703 * @fattr - updated attributes
1705 * After an operation that has changed the inode metadata, mark the
1706 * attribute cache as being invalid, then try to update it. Fake up
1707 * weak cache consistency data, if none exist.
1709 * This function is mainly designed to be used by the ->write_done() functions.
1711 int nfs_post_op_update_inode_force_wcc_locked(struct inode *inode, struct nfs_fattr *fattr)
1715 /* Don't do a WCC update if these attributes are already stale */
1716 if ((fattr->valid & NFS_ATTR_FATTR) == 0 ||
1717 !nfs_inode_attrs_need_update(inode, fattr)) {
1718 fattr->valid &= ~(NFS_ATTR_FATTR_PRECHANGE
1719 | NFS_ATTR_FATTR_PRESIZE
1720 | NFS_ATTR_FATTR_PREMTIME
1721 | NFS_ATTR_FATTR_PRECTIME);
1724 if ((fattr->valid & NFS_ATTR_FATTR_CHANGE) != 0 &&
1725 (fattr->valid & NFS_ATTR_FATTR_PRECHANGE) == 0) {
1726 fattr->pre_change_attr = inode_peek_iversion_raw(inode);
1727 fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
1729 if ((fattr->valid & NFS_ATTR_FATTR_CTIME) != 0 &&
1730 (fattr->valid & NFS_ATTR_FATTR_PRECTIME) == 0) {
1731 fattr->pre_ctime = timespec64_to_timespec(inode->i_ctime);
1732 fattr->valid |= NFS_ATTR_FATTR_PRECTIME;
1734 if ((fattr->valid & NFS_ATTR_FATTR_MTIME) != 0 &&
1735 (fattr->valid & NFS_ATTR_FATTR_PREMTIME) == 0) {
1736 fattr->pre_mtime = timespec64_to_timespec(inode->i_mtime);
1737 fattr->valid |= NFS_ATTR_FATTR_PREMTIME;
1739 if ((fattr->valid & NFS_ATTR_FATTR_SIZE) != 0 &&
1740 (fattr->valid & NFS_ATTR_FATTR_PRESIZE) == 0) {
1741 fattr->pre_size = i_size_read(inode);
1742 fattr->valid |= NFS_ATTR_FATTR_PRESIZE;
1745 status = nfs_post_op_update_inode_locked(inode, fattr,
1746 NFS_INO_INVALID_CHANGE
1747 | NFS_INO_INVALID_CTIME
1748 | NFS_INO_INVALID_MTIME);
1753 * nfs_post_op_update_inode_force_wcc - try to update the inode attribute cache
1754 * @inode - pointer to inode
1755 * @fattr - updated attributes
1757 * After an operation that has changed the inode metadata, mark the
1758 * attribute cache as being invalid, then try to update it. Fake up
1759 * weak cache consistency data, if none exist.
1761 * This function is mainly designed to be used by the ->write_done() functions.
1763 int nfs_post_op_update_inode_force_wcc(struct inode *inode, struct nfs_fattr *fattr)
1767 spin_lock(&inode->i_lock);
1768 nfs_fattr_set_barrier(fattr);
1769 status = nfs_post_op_update_inode_force_wcc_locked(inode, fattr);
1770 spin_unlock(&inode->i_lock);
1773 EXPORT_SYMBOL_GPL(nfs_post_op_update_inode_force_wcc);
1776 static inline bool nfs_fileid_valid(struct nfs_inode *nfsi,
1777 struct nfs_fattr *fattr)
1779 bool ret1 = true, ret2 = true;
1781 if (fattr->valid & NFS_ATTR_FATTR_FILEID)
1782 ret1 = (nfsi->fileid == fattr->fileid);
1783 if (fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID)
1784 ret2 = (nfsi->fileid == fattr->mounted_on_fileid);
1785 return ret1 || ret2;
1789 * Many nfs protocol calls return the new file attributes after
1790 * an operation. Here we update the inode to reflect the state
1791 * of the server's inode.
1793 * This is a bit tricky because we have to make sure all dirty pages
1794 * have been sent off to the server before calling invalidate_inode_pages.
1795 * To make sure no other process adds more write requests while we try
1796 * our best to flush them, we make them sleep during the attribute refresh.
1798 * A very similar scenario holds for the dir cache.
1800 static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1802 struct nfs_server *server;
1803 struct nfs_inode *nfsi = NFS_I(inode);
1804 loff_t cur_isize, new_isize;
1805 unsigned long invalid = 0;
1806 unsigned long now = jiffies;
1807 unsigned long save_cache_validity;
1808 bool have_writers = nfs_file_has_buffered_writers(nfsi);
1809 bool cache_revalidated = true;
1810 bool attr_changed = false;
1811 bool have_delegation;
1813 dfprintk(VFS, "NFS: %s(%s/%lu fh_crc=0x%08x ct=%d info=0x%x)\n",
1814 __func__, inode->i_sb->s_id, inode->i_ino,
1815 nfs_display_fhandle_hash(NFS_FH(inode)),
1816 atomic_read(&inode->i_count), fattr->valid);
1818 if (!nfs_fileid_valid(nfsi, fattr)) {
1819 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1820 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1821 NFS_SERVER(inode)->nfs_client->cl_hostname,
1822 inode->i_sb->s_id, (long long)nfsi->fileid,
1823 (long long)fattr->fileid);
1828 * Make sure the inode's type hasn't changed.
1830 if ((fattr->valid & NFS_ATTR_FATTR_TYPE) && (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1832 * Big trouble! The inode has become a different object.
1834 printk(KERN_DEBUG "NFS: %s: inode %lu mode changed, %07o to %07o\n",
1835 __func__, inode->i_ino, inode->i_mode, fattr->mode);
1839 server = NFS_SERVER(inode);
1840 /* Update the fsid? */
1841 if (S_ISDIR(inode->i_mode) && (fattr->valid & NFS_ATTR_FATTR_FSID) &&
1842 !nfs_fsid_equal(&server->fsid, &fattr->fsid) &&
1843 !IS_AUTOMOUNT(inode))
1844 server->fsid = fattr->fsid;
1846 /* Save the delegation state before clearing cache_validity */
1847 have_delegation = nfs_have_delegated_attributes(inode);
1850 * Update the read time so we don't revalidate too often.
1852 nfsi->read_cache_jiffies = fattr->time_start;
1854 save_cache_validity = nfsi->cache_validity;
1855 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR
1856 | NFS_INO_INVALID_ATIME
1857 | NFS_INO_REVAL_FORCED
1858 | NFS_INO_REVAL_PAGECACHE);
1860 /* Do atomic weak cache consistency updates */
1861 nfs_wcc_update_inode(inode, fattr);
1863 if (pnfs_layoutcommit_outstanding(inode)) {
1864 nfsi->cache_validity |= save_cache_validity & NFS_INO_INVALID_ATTR;
1865 cache_revalidated = false;
1868 /* More cache consistency checks */
1869 if (fattr->valid & NFS_ATTR_FATTR_CHANGE) {
1870 if (!inode_eq_iversion_raw(inode, fattr->change_attr)) {
1871 /* Could it be a race with writeback? */
1872 if (!(have_writers || have_delegation)) {
1873 invalid |= NFS_INO_INVALID_DATA
1874 | NFS_INO_INVALID_ACCESS
1875 | NFS_INO_INVALID_ACL;
1876 /* Force revalidate of all attributes */
1877 save_cache_validity |= NFS_INO_INVALID_CTIME
1878 | NFS_INO_INVALID_MTIME
1879 | NFS_INO_INVALID_SIZE
1880 | NFS_INO_INVALID_OTHER;
1881 if (S_ISDIR(inode->i_mode))
1882 nfs_force_lookup_revalidate(inode);
1883 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1887 inode_set_iversion_raw(inode, fattr->change_attr);
1888 attr_changed = true;
1891 nfsi->cache_validity |= save_cache_validity &
1892 (NFS_INO_INVALID_CHANGE
1893 | NFS_INO_REVAL_PAGECACHE
1894 | NFS_INO_REVAL_FORCED);
1895 cache_revalidated = false;
1898 if (fattr->valid & NFS_ATTR_FATTR_MTIME) {
1899 inode->i_mtime = timespec_to_timespec64(fattr->mtime);
1900 } else if (server->caps & NFS_CAP_MTIME) {
1901 nfsi->cache_validity |= save_cache_validity &
1902 (NFS_INO_INVALID_MTIME
1903 | NFS_INO_REVAL_FORCED);
1904 cache_revalidated = false;
1907 if (fattr->valid & NFS_ATTR_FATTR_CTIME) {
1908 inode->i_ctime = timespec_to_timespec64(fattr->ctime);
1909 } else if (server->caps & NFS_CAP_CTIME) {
1910 nfsi->cache_validity |= save_cache_validity &
1911 (NFS_INO_INVALID_CTIME
1912 | NFS_INO_REVAL_FORCED);
1913 cache_revalidated = false;
1916 /* Check if our cached file size is stale */
1917 if (fattr->valid & NFS_ATTR_FATTR_SIZE) {
1918 new_isize = nfs_size_to_loff_t(fattr->size);
1919 cur_isize = i_size_read(inode);
1920 if (new_isize != cur_isize && !have_delegation) {
1921 /* Do we perhaps have any outstanding writes, or has
1922 * the file grown beyond our last write? */
1923 if (!nfs_have_writebacks(inode) || new_isize > cur_isize) {
1924 i_size_write(inode, new_isize);
1926 invalid |= NFS_INO_INVALID_DATA;
1927 attr_changed = true;
1929 dprintk("NFS: isize change on server for file %s/%ld "
1933 (long long)cur_isize,
1934 (long long)new_isize);
1937 nfsi->cache_validity |= save_cache_validity &
1938 (NFS_INO_INVALID_SIZE
1939 | NFS_INO_REVAL_PAGECACHE
1940 | NFS_INO_REVAL_FORCED);
1941 cache_revalidated = false;
1945 if (fattr->valid & NFS_ATTR_FATTR_ATIME)
1946 inode->i_atime = timespec_to_timespec64(fattr->atime);
1947 else if (server->caps & NFS_CAP_ATIME) {
1948 nfsi->cache_validity |= save_cache_validity &
1949 (NFS_INO_INVALID_ATIME
1950 | NFS_INO_REVAL_FORCED);
1951 cache_revalidated = false;
1954 if (fattr->valid & NFS_ATTR_FATTR_MODE) {
1955 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)) {
1956 umode_t newmode = inode->i_mode & S_IFMT;
1957 newmode |= fattr->mode & S_IALLUGO;
1958 inode->i_mode = newmode;
1959 invalid |= NFS_INO_INVALID_ACCESS
1960 | NFS_INO_INVALID_ACL;
1961 attr_changed = true;
1963 } else if (server->caps & NFS_CAP_MODE) {
1964 nfsi->cache_validity |= save_cache_validity &
1965 (NFS_INO_INVALID_OTHER
1966 | NFS_INO_REVAL_FORCED);
1967 cache_revalidated = false;
1970 if (fattr->valid & NFS_ATTR_FATTR_OWNER) {
1971 if (!uid_eq(inode->i_uid, fattr->uid)) {
1972 invalid |= NFS_INO_INVALID_ACCESS
1973 | NFS_INO_INVALID_ACL;
1974 inode->i_uid = fattr->uid;
1975 attr_changed = true;
1977 } else if (server->caps & NFS_CAP_OWNER) {
1978 nfsi->cache_validity |= save_cache_validity &
1979 (NFS_INO_INVALID_OTHER
1980 | NFS_INO_REVAL_FORCED);
1981 cache_revalidated = false;
1984 if (fattr->valid & NFS_ATTR_FATTR_GROUP) {
1985 if (!gid_eq(inode->i_gid, fattr->gid)) {
1986 invalid |= NFS_INO_INVALID_ACCESS
1987 | NFS_INO_INVALID_ACL;
1988 inode->i_gid = fattr->gid;
1989 attr_changed = true;
1991 } else if (server->caps & NFS_CAP_OWNER_GROUP) {
1992 nfsi->cache_validity |= save_cache_validity &
1993 (NFS_INO_INVALID_OTHER
1994 | NFS_INO_REVAL_FORCED);
1995 cache_revalidated = false;
1998 if (fattr->valid & NFS_ATTR_FATTR_NLINK) {
1999 if (inode->i_nlink != fattr->nlink) {
2000 if (S_ISDIR(inode->i_mode))
2001 invalid |= NFS_INO_INVALID_DATA;
2002 set_nlink(inode, fattr->nlink);
2003 attr_changed = true;
2005 } else if (server->caps & NFS_CAP_NLINK) {
2006 nfsi->cache_validity |= save_cache_validity &
2007 (NFS_INO_INVALID_OTHER
2008 | NFS_INO_REVAL_FORCED);
2009 cache_revalidated = false;
2012 if (fattr->valid & NFS_ATTR_FATTR_SPACE_USED) {
2014 * report the blocks in 512byte units
2016 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
2017 } else if (fattr->valid & NFS_ATTR_FATTR_BLOCKS_USED)
2018 inode->i_blocks = fattr->du.nfs2.blocks;
2020 cache_revalidated = false;
2022 /* Update attrtimeo value if we're out of the unstable period */
2024 invalid &= ~NFS_INO_INVALID_ATTR;
2025 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
2026 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
2027 nfsi->attrtimeo_timestamp = now;
2028 /* Set barrier to be more recent than all outstanding updates */
2029 nfsi->attr_gencount = nfs_inc_attr_generation_counter();
2031 if (cache_revalidated) {
2032 if (!time_in_range_open(now, nfsi->attrtimeo_timestamp,
2033 nfsi->attrtimeo_timestamp + nfsi->attrtimeo)) {
2034 nfsi->attrtimeo <<= 1;
2035 if (nfsi->attrtimeo > NFS_MAXATTRTIMEO(inode))
2036 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
2038 nfsi->attrtimeo_timestamp = now;
2040 /* Set the barrier to be more recent than this fattr */
2041 if ((long)(fattr->gencount - nfsi->attr_gencount) > 0)
2042 nfsi->attr_gencount = fattr->gencount;
2045 /* Don't invalidate the data if we were to blame */
2046 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
2047 || S_ISLNK(inode->i_mode)))
2048 invalid &= ~NFS_INO_INVALID_DATA;
2049 nfs_set_cache_invalid(inode, invalid);
2054 * No need to worry about unhashing the dentry, as the
2055 * lookup validation will know that the inode is bad.
2056 * (But we fall through to invalidate the caches.)
2058 nfs_invalidate_inode(inode);
2062 struct inode *nfs_alloc_inode(struct super_block *sb)
2064 struct nfs_inode *nfsi;
2065 nfsi = kmem_cache_alloc(nfs_inode_cachep, GFP_KERNEL);
2069 nfsi->cache_validity = 0UL;
2070 #if IS_ENABLED(CONFIG_NFS_V4)
2071 nfsi->nfs4_acl = NULL;
2072 #endif /* CONFIG_NFS_V4 */
2073 return &nfsi->vfs_inode;
2075 EXPORT_SYMBOL_GPL(nfs_alloc_inode);
2077 static void nfs_i_callback(struct rcu_head *head)
2079 struct inode *inode = container_of(head, struct inode, i_rcu);
2080 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2083 void nfs_destroy_inode(struct inode *inode)
2085 call_rcu(&inode->i_rcu, nfs_i_callback);
2087 EXPORT_SYMBOL_GPL(nfs_destroy_inode);
2089 static inline void nfs4_init_once(struct nfs_inode *nfsi)
2091 #if IS_ENABLED(CONFIG_NFS_V4)
2092 INIT_LIST_HEAD(&nfsi->open_states);
2093 nfsi->delegation = NULL;
2094 init_rwsem(&nfsi->rwsem);
2095 nfsi->layout = NULL;
2099 static void init_once(void *foo)
2101 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2103 inode_init_once(&nfsi->vfs_inode);
2104 INIT_LIST_HEAD(&nfsi->open_files);
2105 INIT_LIST_HEAD(&nfsi->access_cache_entry_lru);
2106 INIT_LIST_HEAD(&nfsi->access_cache_inode_lru);
2107 INIT_LIST_HEAD(&nfsi->commit_info.list);
2108 atomic_long_set(&nfsi->nrequests, 0);
2109 atomic_long_set(&nfsi->commit_info.ncommit, 0);
2110 atomic_set(&nfsi->commit_info.rpcs_out, 0);
2111 init_rwsem(&nfsi->rmdir_sem);
2112 mutex_init(&nfsi->commit_mutex);
2113 nfs4_init_once(nfsi);
2116 static int __init nfs_init_inodecache(void)
2118 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2119 sizeof(struct nfs_inode),
2120 0, (SLAB_RECLAIM_ACCOUNT|
2121 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
2123 if (nfs_inode_cachep == NULL)
2129 static void nfs_destroy_inodecache(void)
2132 * Make sure all delayed rcu free inodes are flushed before we
2136 kmem_cache_destroy(nfs_inode_cachep);
2139 struct workqueue_struct *nfsiod_workqueue;
2140 EXPORT_SYMBOL_GPL(nfsiod_workqueue);
2143 * start up the nfsiod workqueue
2145 static int nfsiod_start(void)
2147 struct workqueue_struct *wq;
2148 dprintk("RPC: creating workqueue nfsiod\n");
2149 wq = alloc_workqueue("nfsiod", WQ_MEM_RECLAIM | WQ_UNBOUND, 0);
2152 nfsiod_workqueue = wq;
2157 * Destroy the nfsiod workqueue
2159 static void nfsiod_stop(void)
2161 struct workqueue_struct *wq;
2163 wq = nfsiod_workqueue;
2166 nfsiod_workqueue = NULL;
2167 destroy_workqueue(wq);
2170 unsigned int nfs_net_id;
2171 EXPORT_SYMBOL_GPL(nfs_net_id);
2173 static int nfs_net_init(struct net *net)
2175 nfs_clients_init(net);
2176 return nfs_fs_proc_net_init(net);
2179 static void nfs_net_exit(struct net *net)
2181 struct nfs_net *nn = net_generic(net, nfs_net_id);
2183 nfs_fs_proc_net_exit(net);
2184 nfs_cleanup_cb_ident_idr(net);
2185 WARN_ON_ONCE(!list_empty(&nn->nfs_client_list));
2186 WARN_ON_ONCE(!list_empty(&nn->nfs_volume_list));
2189 static struct pernet_operations nfs_net_ops = {
2190 .init = nfs_net_init,
2191 .exit = nfs_net_exit,
2193 .size = sizeof(struct nfs_net),
2199 static int __init init_nfs_fs(void)
2203 err = register_pernet_subsys(&nfs_net_ops);
2207 err = nfs_fscache_register();
2211 err = nfsiod_start();
2215 err = nfs_fs_proc_init();
2219 err = nfs_init_nfspagecache();
2223 err = nfs_init_inodecache();
2227 err = nfs_init_readpagecache();
2231 err = nfs_init_writepagecache();
2235 err = nfs_init_directcache();
2239 rpc_proc_register(&init_net, &nfs_rpcstat);
2241 err = register_nfs_fs();
2247 rpc_proc_unregister(&init_net, "nfs");
2248 nfs_destroy_directcache();
2250 nfs_destroy_writepagecache();
2252 nfs_destroy_readpagecache();
2254 nfs_destroy_inodecache();
2256 nfs_destroy_nfspagecache();
2262 nfs_fscache_unregister();
2264 unregister_pernet_subsys(&nfs_net_ops);
2269 static void __exit exit_nfs_fs(void)
2271 nfs_destroy_directcache();
2272 nfs_destroy_writepagecache();
2273 nfs_destroy_readpagecache();
2274 nfs_destroy_inodecache();
2275 nfs_destroy_nfspagecache();
2276 nfs_fscache_unregister();
2277 unregister_pernet_subsys(&nfs_net_ops);
2278 rpc_proc_unregister(&init_net, "nfs");
2279 unregister_nfs_fs();
2284 /* Not quite true; I just maintain it */
2285 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2286 MODULE_LICENSE("GPL");
2287 module_param(enable_ino64, bool, 0644);
2289 module_init(init_nfs_fs)
2290 module_exit(exit_nfs_fs)