2 * Copyright (c) 2002 Red Hat, Inc. All rights reserved.
4 * This software may be freely redistributed under the terms of the
5 * GNU General Public License.
7 * You should have received a copy of the GNU General Public License
8 * along with this program; if not, write to the Free Software
9 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
16 #include <linux/kernel.h>
17 #include <linux/module.h>
18 #include <linux/init.h>
20 #include <linux/pagemap.h>
21 #include <linux/sched.h>
22 #include <linux/mount.h>
23 #include <linux/namei.h>
24 #include <linux/iversion.h>
28 static const struct inode_operations afs_symlink_inode_operations = {
29 .get_link = page_get_link,
30 .listxattr = afs_listxattr,
33 static noinline void dump_vnode(struct afs_vnode *vnode, struct afs_vnode *parent_vnode)
35 static unsigned long once_only;
37 pr_warn("kAFS: AFS vnode with undefined type %u\n", vnode->status.type);
38 pr_warn("kAFS: A=%d m=%o s=%llx v=%llx\n",
39 vnode->status.abort_code,
42 vnode->status.data_version);
43 pr_warn("kAFS: vnode %llx:%llx:%x\n",
48 pr_warn("kAFS: dir %llx:%llx:%x\n",
49 parent_vnode->fid.vid,
50 parent_vnode->fid.vnode,
51 parent_vnode->fid.unique);
53 if (!test_and_set_bit(0, &once_only))
58 * Set the file size and block count. Estimate the number of 512 bytes blocks
59 * used, rounded up to nearest 1K for consistency with other AFS clients.
61 static void afs_set_i_size(struct afs_vnode *vnode, u64 size)
63 i_size_write(&vnode->vfs_inode, size);
64 vnode->vfs_inode.i_blocks = ((size + 1023) >> 10) << 1;
68 * Initialise an inode from the vnode status.
70 static int afs_inode_init_from_status(struct afs_operation *op,
71 struct afs_vnode_param *vp,
72 struct afs_vnode *vnode)
74 struct afs_file_status *status = &vp->scb.status;
75 struct inode *inode = AFS_VNODE_TO_I(vnode);
78 _enter("{%llx:%llu.%u} %s",
79 vp->fid.vid, vp->fid.vnode, vp->fid.unique,
80 op->type ? op->type->name : "???");
82 _debug("FS: ft=%d lk=%d sz=%llu ver=%Lu mod=%hu",
85 (unsigned long long) status->size,
89 write_seqlock(&vnode->cb_lock);
91 vnode->cb_v_break = op->cb_v_break;
92 vnode->cb_s_break = op->cb_s_break;
93 vnode->status = *status;
95 t = status->mtime_client;
99 inode->i_flags |= S_NOATIME;
100 inode->i_uid = make_kuid(&init_user_ns, status->owner);
101 inode->i_gid = make_kgid(&init_user_ns, status->group);
102 set_nlink(&vnode->vfs_inode, status->nlink);
104 switch (status->type) {
106 inode->i_mode = S_IFREG | status->mode;
107 inode->i_op = &afs_file_inode_operations;
108 inode->i_fop = &afs_file_operations;
109 inode->i_mapping->a_ops = &afs_fs_aops;
112 inode->i_mode = S_IFDIR | status->mode;
113 inode->i_op = &afs_dir_inode_operations;
114 inode->i_fop = &afs_dir_file_operations;
115 inode->i_mapping->a_ops = &afs_dir_aops;
117 case AFS_FTYPE_SYMLINK:
118 /* Symlinks with a mode of 0644 are actually mountpoints. */
119 if ((status->mode & 0777) == 0644) {
120 inode->i_flags |= S_AUTOMOUNT;
122 set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags);
124 inode->i_mode = S_IFDIR | 0555;
125 inode->i_op = &afs_mntpt_inode_operations;
126 inode->i_fop = &afs_mntpt_file_operations;
127 inode->i_mapping->a_ops = &afs_fs_aops;
129 inode->i_mode = S_IFLNK | status->mode;
130 inode->i_op = &afs_symlink_inode_operations;
131 inode->i_mapping->a_ops = &afs_fs_aops;
133 inode_nohighmem(inode);
136 dump_vnode(vnode, op->file[0].vnode != vnode ? op->file[0].vnode : NULL);
137 write_sequnlock(&vnode->cb_lock);
138 return afs_protocol_error(NULL, afs_eproto_file_type);
141 afs_set_i_size(vnode, status->size);
143 vnode->invalid_before = status->data_version;
144 inode_set_iversion_raw(&vnode->vfs_inode, status->data_version);
146 if (!vp->scb.have_cb) {
147 /* it's a symlink we just created (the fileserver
148 * didn't give us a callback) */
149 vnode->cb_expires_at = ktime_get_real_seconds();
151 vnode->cb_expires_at = vp->scb.callback.expires_at;
152 vnode->cb_server = op->server;
153 set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
156 write_sequnlock(&vnode->cb_lock);
161 * Update the core inode struct from a returned status record.
163 static void afs_apply_status(struct afs_operation *op,
164 struct afs_vnode_param *vp)
166 struct afs_file_status *status = &vp->scb.status;
167 struct afs_vnode *vnode = vp->vnode;
170 bool data_changed = false;
172 _enter("{%llx:%llu.%u} %s",
173 vp->fid.vid, vp->fid.vnode, vp->fid.unique,
174 op->type ? op->type->name : "???");
176 BUG_ON(test_bit(AFS_VNODE_UNSET, &vnode->flags));
178 if (status->type != vnode->status.type) {
179 pr_warn("Vnode %llx:%llx:%x changed type %u to %u\n",
183 status->type, vnode->status.type);
184 afs_protocol_error(NULL, afs_eproto_bad_status);
188 if (status->nlink != vnode->status.nlink)
189 set_nlink(&vnode->vfs_inode, status->nlink);
191 if (status->owner != vnode->status.owner)
192 vnode->vfs_inode.i_uid = make_kuid(&init_user_ns, status->owner);
194 if (status->group != vnode->status.group)
195 vnode->vfs_inode.i_gid = make_kgid(&init_user_ns, status->group);
197 if (status->mode != vnode->status.mode) {
198 mode = vnode->vfs_inode.i_mode;
200 mode |= status->mode;
201 WRITE_ONCE(vnode->vfs_inode.i_mode, mode);
204 t = status->mtime_client;
205 vnode->vfs_inode.i_ctime = t;
206 vnode->vfs_inode.i_mtime = t;
207 vnode->vfs_inode.i_atime = t;
209 if (vnode->status.data_version != status->data_version)
212 vnode->status = *status;
214 if (vp->dv_before + vp->dv_delta != status->data_version) {
215 if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags))
216 pr_warn("kAFS: vnode modified {%llx:%llu} %llx->%llx %s\n",
217 vnode->fid.vid, vnode->fid.vnode,
218 (unsigned long long)vp->dv_before + vp->dv_delta,
219 (unsigned long long)status->data_version,
220 op->type ? op->type->name : "???");
222 vnode->invalid_before = status->data_version;
223 if (vnode->status.type == AFS_FTYPE_DIR) {
224 if (test_and_clear_bit(AFS_VNODE_DIR_VALID, &vnode->flags))
225 afs_stat_v(vnode, n_inval);
227 set_bit(AFS_VNODE_ZAP_DATA, &vnode->flags);
229 } else if (vnode->status.type == AFS_FTYPE_DIR) {
230 /* Expected directory change is handled elsewhere so
231 * that we can locally edit the directory and save on a
234 if (test_bit(AFS_VNODE_DIR_VALID, &vnode->flags))
235 data_changed = false;
239 inode_set_iversion_raw(&vnode->vfs_inode, status->data_version);
240 afs_set_i_size(vnode, status->size);
245 * Apply a callback to a vnode.
247 static void afs_apply_callback(struct afs_operation *op,
248 struct afs_vnode_param *vp)
250 struct afs_callback *cb = &vp->scb.callback;
251 struct afs_vnode *vnode = vp->vnode;
253 if (!afs_cb_is_broken(vp->cb_break_before, vnode)) {
254 vnode->cb_expires_at = cb->expires_at;
255 vnode->cb_server = op->server;
256 set_bit(AFS_VNODE_CB_PROMISED, &vnode->flags);
261 * Apply the received status and callback to an inode all in the same critical
262 * section to avoid races with afs_validate().
264 void afs_vnode_commit_status(struct afs_operation *op, struct afs_vnode_param *vp)
266 struct afs_vnode *vnode = vp->vnode;
270 ASSERTCMP(op->error, ==, 0);
272 write_seqlock(&vnode->cb_lock);
274 if (vp->scb.have_error) {
275 if (vp->scb.status.abort_code == VNOVNODE) {
276 set_bit(AFS_VNODE_DELETED, &vnode->flags);
277 clear_nlink(&vnode->vfs_inode);
278 __afs_break_callback(vnode, afs_cb_break_for_deleted);
281 if (vp->scb.have_status)
282 afs_apply_status(op, vp);
284 afs_apply_callback(op, vp);
287 write_sequnlock(&vnode->cb_lock);
289 if (op->error == 0 && vp->scb.have_status)
290 afs_cache_permit(vnode, op->key, vp->cb_break_before, &vp->scb);
293 static void afs_fetch_status_success(struct afs_operation *op)
295 struct afs_vnode_param *vp = &op->file[0];
296 struct afs_vnode *vnode = vp->vnode;
299 if (vnode->vfs_inode.i_state & I_NEW) {
300 ret = afs_inode_init_from_status(op, vp, vnode);
303 afs_cache_permit(vnode, op->key, vp->cb_break_before, &vp->scb);
305 afs_vnode_commit_status(op, vp);
309 static const struct afs_operation_ops afs_fetch_status_operation = {
310 .issue_afs_rpc = afs_fs_fetch_status,
311 .issue_yfs_rpc = yfs_fs_fetch_status,
312 .success = afs_fetch_status_success,
316 * Fetch file status from the volume.
318 int afs_fetch_status(struct afs_vnode *vnode, struct key *key, bool is_new,
319 afs_access_t *_caller_access)
321 struct afs_operation *op;
323 _enter("%s,{%llx:%llu.%u,S=%lx}",
325 vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique,
328 op = afs_alloc_operation(key, vnode->volume);
332 afs_op_set_vnode(op, 0, vnode);
335 op->ops = &afs_fetch_status_operation;
336 afs_begin_vnode_operation(op);
337 afs_wait_for_operation(op);
340 *_caller_access = op->file[0].scb.status.caller_access;
341 return afs_put_operation(op);
345 * ilookup() comparator
347 int afs_ilookup5_test_by_fid(struct inode *inode, void *opaque)
349 struct afs_vnode *vnode = AFS_FS_I(inode);
350 struct afs_fid *fid = opaque;
352 return (fid->vnode == vnode->fid.vnode &&
353 fid->vnode_hi == vnode->fid.vnode_hi &&
354 fid->unique == vnode->fid.unique);
360 static int afs_iget5_test(struct inode *inode, void *opaque)
362 struct afs_vnode_param *vp = opaque;
363 //struct afs_vnode *vnode = AFS_FS_I(inode);
365 return afs_ilookup5_test_by_fid(inode, &vp->fid);
369 * iget5() inode initialiser
371 static int afs_iget5_set(struct inode *inode, void *opaque)
373 struct afs_vnode_param *vp = opaque;
374 struct afs_super_info *as = AFS_FS_S(inode->i_sb);
375 struct afs_vnode *vnode = AFS_FS_I(inode);
377 vnode->volume = as->volume;
378 vnode->fid = vp->fid;
380 /* YFS supports 96-bit vnode IDs, but Linux only supports
381 * 64-bit inode numbers.
383 inode->i_ino = vnode->fid.vnode;
384 inode->i_generation = vnode->fid.unique;
389 * Get a cache cookie for an inode.
391 static void afs_get_inode_cache(struct afs_vnode *vnode)
393 #ifdef CONFIG_AFS_FSCACHE
397 u32 vnode_id_ext[2]; /* Allow for a 96-bit key */
399 struct afs_vnode_cache_aux aux;
401 if (vnode->status.type == AFS_FTYPE_DIR) {
406 key.vnode_id = vnode->fid.vnode;
407 key.unique = vnode->fid.unique;
408 key.vnode_id_ext[0] = vnode->fid.vnode >> 32;
409 key.vnode_id_ext[1] = vnode->fid.vnode_hi;
410 aux.data_version = vnode->status.data_version;
412 vnode->cache = fscache_acquire_cookie(vnode->volume->cache,
413 &afs_vnode_cache_index_def,
416 vnode, vnode->status.size, true);
423 struct inode *afs_iget(struct afs_operation *op, struct afs_vnode_param *vp)
425 struct afs_vnode_param *dvp = &op->file[0];
426 struct super_block *sb = dvp->vnode->vfs_inode.i_sb;
427 struct afs_vnode *vnode;
431 _enter(",{%llx:%llu.%u},,", vp->fid.vid, vp->fid.vnode, vp->fid.unique);
433 inode = iget5_locked(sb, vp->fid.vnode, afs_iget5_test, afs_iget5_set, vp);
435 _leave(" = -ENOMEM");
436 return ERR_PTR(-ENOMEM);
439 vnode = AFS_FS_I(inode);
441 _debug("GOT INODE %p { vl=%llx vn=%llx, u=%x }",
442 inode, vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique);
444 /* deal with an existing inode */
445 if (!(inode->i_state & I_NEW)) {
446 _leave(" = %p", inode);
450 ret = afs_inode_init_from_status(op, vp, vnode);
454 afs_get_inode_cache(vnode);
457 clear_bit(AFS_VNODE_UNSET, &vnode->flags);
458 unlock_new_inode(inode);
459 _leave(" = %p", inode);
465 _leave(" = %d [bad]", ret);
469 static int afs_iget5_set_root(struct inode *inode, void *opaque)
471 struct afs_super_info *as = AFS_FS_S(inode->i_sb);
472 struct afs_vnode *vnode = AFS_FS_I(inode);
474 vnode->volume = as->volume;
475 vnode->fid.vid = as->volume->vid,
476 vnode->fid.vnode = 1;
477 vnode->fid.unique = 1;
479 inode->i_generation = 1;
484 * Set up the root inode for a volume. This is always vnode 1, unique 1 within
487 struct inode *afs_root_iget(struct super_block *sb, struct key *key)
489 struct afs_super_info *as = AFS_FS_S(sb);
490 struct afs_operation *op;
491 struct afs_vnode *vnode;
495 _enter(",{%llx},,", as->volume->vid);
497 inode = iget5_locked(sb, 1, NULL, afs_iget5_set_root, NULL);
499 _leave(" = -ENOMEM");
500 return ERR_PTR(-ENOMEM);
503 _debug("GOT ROOT INODE %p { vl=%llx }", inode, as->volume->vid);
505 BUG_ON(!(inode->i_state & I_NEW));
507 vnode = AFS_FS_I(inode);
508 vnode->cb_v_break = as->volume->cb_v_break,
510 op = afs_alloc_operation(key, as->volume);
516 afs_op_set_vnode(op, 0, vnode);
519 op->ops = &afs_fetch_status_operation;
520 ret = afs_do_sync_operation(op);
524 afs_get_inode_cache(vnode);
526 clear_bit(AFS_VNODE_UNSET, &vnode->flags);
527 unlock_new_inode(inode);
528 _leave(" = %p", inode);
533 _leave(" = %d [bad]", ret);
538 * mark the data attached to an inode as obsolete due to a write on the server
539 * - might also want to ditch all the outstanding writes and dirty pages
541 static void afs_zap_data(struct afs_vnode *vnode)
543 _enter("{%llx:%llu}", vnode->fid.vid, vnode->fid.vnode);
545 #ifdef CONFIG_AFS_FSCACHE
546 fscache_invalidate(vnode->cache);
549 /* nuke all the non-dirty pages that aren't locked, mapped or being
550 * written back in a regular file and completely discard the pages in a
551 * directory or symlink */
552 if (S_ISREG(vnode->vfs_inode.i_mode))
553 invalidate_remote_inode(&vnode->vfs_inode);
555 invalidate_inode_pages2(vnode->vfs_inode.i_mapping);
559 * Get the server reinit counter for a vnode's current server.
561 static bool afs_get_s_break_rcu(struct afs_vnode *vnode, unsigned int *_s_break)
563 struct afs_server_list *slist = rcu_dereference(vnode->volume->servers);
564 struct afs_server *server;
567 for (i = 0; i < slist->nr_servers; i++) {
568 server = slist->servers[i].server;
569 if (server == vnode->cb_server) {
570 *_s_break = READ_ONCE(server->cb_s_break);
579 * Check the validity of a vnode/inode.
581 bool afs_check_validity(struct afs_vnode *vnode)
583 struct afs_volume *volume = vnode->volume;
584 enum afs_cb_break_reason need_clear = afs_cb_break_no_break;
585 time64_t now = ktime_get_real_seconds();
587 unsigned int cb_break, cb_s_break, cb_v_break;
591 read_seqbegin_or_lock(&vnode->cb_lock, &seq);
592 cb_v_break = READ_ONCE(volume->cb_v_break);
593 cb_break = vnode->cb_break;
595 if (test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags) &&
596 afs_get_s_break_rcu(vnode, &cb_s_break)) {
597 if (vnode->cb_s_break != cb_s_break ||
598 vnode->cb_v_break != cb_v_break) {
599 vnode->cb_s_break = cb_s_break;
600 vnode->cb_v_break = cb_v_break;
601 need_clear = afs_cb_break_for_vsbreak;
603 } else if (test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) {
604 need_clear = afs_cb_break_for_zap;
606 } else if (vnode->cb_expires_at - 10 <= now) {
607 need_clear = afs_cb_break_for_lapsed;
612 } else if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
615 vnode->cb_v_break = cb_v_break;
619 } while (need_seqretry(&vnode->cb_lock, seq));
621 done_seqretry(&vnode->cb_lock, seq);
623 if (need_clear != afs_cb_break_no_break) {
624 write_seqlock(&vnode->cb_lock);
625 if (cb_break == vnode->cb_break)
626 __afs_break_callback(vnode, need_clear);
628 trace_afs_cb_miss(&vnode->fid, need_clear);
629 write_sequnlock(&vnode->cb_lock);
637 * validate a vnode/inode
638 * - there are several things we need to check
639 * - parent dir data changes (rm, rmdir, rename, mkdir, create, link,
641 * - parent dir metadata changed (security changes)
642 * - dentry data changed (write, truncate)
643 * - dentry metadata changed (security changes)
645 int afs_validate(struct afs_vnode *vnode, struct key *key)
650 _enter("{v={%llx:%llu} fl=%lx},%x",
651 vnode->fid.vid, vnode->fid.vnode, vnode->flags,
655 valid = afs_check_validity(vnode);
658 if (test_bit(AFS_VNODE_DELETED, &vnode->flags))
659 clear_nlink(&vnode->vfs_inode);
664 down_write(&vnode->validate_lock);
666 /* if the promise has expired, we need to check the server again to get
667 * a new promise - note that if the (parent) directory's metadata was
668 * changed then the security may be different and we may no longer have
670 if (!test_bit(AFS_VNODE_CB_PROMISED, &vnode->flags)) {
671 _debug("not promised");
672 ret = afs_fetch_status(vnode, key, false, NULL);
674 if (ret == -ENOENT) {
675 set_bit(AFS_VNODE_DELETED, &vnode->flags);
680 _debug("new promise [fl=%lx]", vnode->flags);
683 if (test_bit(AFS_VNODE_DELETED, &vnode->flags)) {
684 _debug("file already deleted");
689 /* if the vnode's data version number changed then its contents are
691 if (test_and_clear_bit(AFS_VNODE_ZAP_DATA, &vnode->flags))
693 up_write(&vnode->validate_lock);
699 up_write(&vnode->validate_lock);
700 _leave(" = %d", ret);
705 * read the attributes of an inode
707 int afs_getattr(const struct path *path, struct kstat *stat,
708 u32 request_mask, unsigned int query_flags)
710 struct inode *inode = d_inode(path->dentry);
711 struct afs_vnode *vnode = AFS_FS_I(inode);
714 _enter("{ ino=%lu v=%u }", inode->i_ino, inode->i_generation);
717 read_seqbegin_or_lock(&vnode->cb_lock, &seq);
718 generic_fillattr(inode, stat);
719 } while (need_seqretry(&vnode->cb_lock, seq));
721 done_seqretry(&vnode->cb_lock, seq);
726 * discard an AFS inode
728 int afs_drop_inode(struct inode *inode)
732 if (test_bit(AFS_VNODE_PSEUDODIR, &AFS_FS_I(inode)->flags))
733 return generic_delete_inode(inode);
735 return generic_drop_inode(inode);
741 void afs_evict_inode(struct inode *inode)
743 struct afs_vnode *vnode;
745 vnode = AFS_FS_I(inode);
747 _enter("{%llx:%llu.%d}",
752 _debug("CLEAR INODE %p", inode);
754 ASSERTCMP(inode->i_ino, ==, vnode->fid.vnode);
756 truncate_inode_pages_final(&inode->i_data);
759 while (!list_empty(&vnode->wb_keys)) {
760 struct afs_wb_key *wbk = list_entry(vnode->wb_keys.next,
761 struct afs_wb_key, vnode_link);
762 list_del(&wbk->vnode_link);
766 #ifdef CONFIG_AFS_FSCACHE
768 struct afs_vnode_cache_aux aux;
770 aux.data_version = vnode->status.data_version;
771 fscache_relinquish_cookie(vnode->cache, &aux,
772 test_bit(AFS_VNODE_DELETED, &vnode->flags));
777 afs_prune_wb_keys(vnode);
778 afs_put_permits(rcu_access_pointer(vnode->permit_cache));
779 key_put(vnode->silly_key);
780 vnode->silly_key = NULL;
781 key_put(vnode->lock_key);
782 vnode->lock_key = NULL;
786 static void afs_setattr_success(struct afs_operation *op)
788 afs_vnode_commit_status(op, &op->file[0]);
791 static const struct afs_operation_ops afs_setattr_operation = {
792 .issue_afs_rpc = afs_fs_setattr,
793 .issue_yfs_rpc = yfs_fs_setattr,
794 .success = afs_setattr_success,
798 * set the attributes of an inode
800 int afs_setattr(struct dentry *dentry, struct iattr *attr)
802 struct afs_operation *op;
803 struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
805 _enter("{%llx:%llu},{n=%pd},%x",
806 vnode->fid.vid, vnode->fid.vnode, dentry,
809 if (!(attr->ia_valid & (ATTR_SIZE | ATTR_MODE | ATTR_UID | ATTR_GID |
811 _leave(" = 0 [unsupported]");
815 /* flush any dirty data outstanding on a regular file */
816 if (S_ISREG(vnode->vfs_inode.i_mode))
817 filemap_write_and_wait(vnode->vfs_inode.i_mapping);
819 op = afs_alloc_operation(((attr->ia_valid & ATTR_FILE) ?
820 afs_file_key(attr->ia_file) : NULL),
825 afs_op_set_vnode(op, 0, vnode);
826 op->setattr.attr = attr;
828 if (attr->ia_valid & ATTR_SIZE)
829 op->file[0].dv_delta = 1;
831 op->ops = &afs_setattr_operation;
832 return afs_do_sync_operation(op);