1 // SPDX-License-Identifier: GPL-2.0
7 #include "btree_update.h"
14 #include "fs-common.h"
17 #include "fs-io-buffered.h"
18 #include "fs-io-pagecache.h"
28 #include <linux/aio.h>
29 #include <linux/backing-dev.h>
30 #include <linux/exportfs.h>
31 #include <linux/fiemap.h>
32 #include <linux/module.h>
33 #include <linux/pagemap.h>
34 #include <linux/posix_acl.h>
35 #include <linux/random.h>
36 #include <linux/seq_file.h>
37 #include <linux/statfs.h>
38 #include <linux/string.h>
39 #include <linux/xattr.h>
41 static struct kmem_cache *bch2_inode_cache;
43 static void bch2_vfs_inode_init(struct btree_trans *, subvol_inum,
44 struct bch_inode_info *,
45 struct bch_inode_unpacked *,
46 struct bch_subvolume *);
48 void bch2_inode_update_after_write(struct btree_trans *trans,
49 struct bch_inode_info *inode,
50 struct bch_inode_unpacked *bi,
53 struct bch_fs *c = trans->c;
55 BUG_ON(bi->bi_inum != inode->v.i_ino);
57 bch2_assert_pos_locked(trans, BTREE_ID_inodes,
59 c->opts.inodes_use_key_cache);
61 set_nlink(&inode->v, bch2_inode_nlink_get(bi));
62 i_uid_write(&inode->v, bi->bi_uid);
63 i_gid_write(&inode->v, bi->bi_gid);
64 inode->v.i_mode = bi->bi_mode;
66 if (fields & ATTR_ATIME)
67 inode->v.i_atime = bch2_time_to_timespec(c, bi->bi_atime);
68 if (fields & ATTR_MTIME)
69 inode->v.i_mtime = bch2_time_to_timespec(c, bi->bi_mtime);
70 if (fields & ATTR_CTIME)
71 inode->v.i_ctime = bch2_time_to_timespec(c, bi->bi_ctime);
73 inode->ei_inode = *bi;
75 bch2_inode_flags_to_vfs(inode);
78 int __must_check bch2_write_inode(struct bch_fs *c,
79 struct bch_inode_info *inode,
81 void *p, unsigned fields)
83 struct btree_trans trans;
84 struct btree_iter iter = { NULL };
85 struct bch_inode_unpacked inode_u;
88 bch2_trans_init(&trans, c, 0, 512);
90 bch2_trans_begin(&trans);
92 ret = bch2_inode_peek(&trans, &iter, &inode_u, inode_inum(inode),
94 (set ? set(inode, &inode_u, p) : 0) ?:
95 bch2_inode_write(&trans, &iter, &inode_u) ?:
96 bch2_trans_commit(&trans, NULL, NULL, BTREE_INSERT_NOFAIL);
99 * the btree node lock protects inode->ei_inode, not ei_update_lock;
100 * this is important for inode updates via bchfs_write_index_update
103 bch2_inode_update_after_write(&trans, inode, &inode_u, fields);
105 bch2_trans_iter_exit(&trans, &iter);
107 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
110 bch2_fs_fatal_err_on(bch2_err_matches(ret, ENOENT), c,
111 "inode %u:%llu not found when updating",
112 inode_inum(inode).subvol,
113 inode_inum(inode).inum);
115 bch2_trans_exit(&trans);
116 return ret < 0 ? ret : 0;
119 int bch2_fs_quota_transfer(struct bch_fs *c,
120 struct bch_inode_info *inode,
121 struct bch_qid new_qid,
123 enum quota_acct_mode mode)
128 qtypes &= enabled_qtypes(c);
130 for (i = 0; i < QTYP_NR; i++)
131 if (new_qid.q[i] == inode->ei_qid.q[i])
132 qtypes &= ~(1U << i);
137 mutex_lock(&inode->ei_quota_lock);
139 ret = bch2_quota_transfer(c, qtypes, new_qid,
142 inode->ei_quota_reserved,
145 for (i = 0; i < QTYP_NR; i++)
146 if (qtypes & (1 << i))
147 inode->ei_qid.q[i] = new_qid.q[i];
149 mutex_unlock(&inode->ei_quota_lock);
154 static int bch2_iget5_test(struct inode *vinode, void *p)
156 struct bch_inode_info *inode = to_bch_ei(vinode);
157 subvol_inum *inum = p;
159 return inode->ei_subvol == inum->subvol &&
160 inode->ei_inode.bi_inum == inum->inum;
163 static int bch2_iget5_set(struct inode *vinode, void *p)
165 struct bch_inode_info *inode = to_bch_ei(vinode);
166 subvol_inum *inum = p;
168 inode->v.i_ino = inum->inum;
169 inode->ei_subvol = inum->subvol;
170 inode->ei_inode.bi_inum = inum->inum;
174 static unsigned bch2_inode_hash(subvol_inum inum)
176 return jhash_3words(inum.subvol, inum.inum >> 32, inum.inum, JHASH_INITVAL);
179 struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
181 struct bch_inode_unpacked inode_u;
182 struct bch_inode_info *inode;
183 struct btree_trans trans;
184 struct bch_subvolume subvol;
187 inode = to_bch_ei(iget5_locked(c->vfs_sb,
188 bch2_inode_hash(inum),
192 if (unlikely(!inode))
193 return ERR_PTR(-ENOMEM);
194 if (!(inode->v.i_state & I_NEW))
197 bch2_trans_init(&trans, c, 8, 0);
198 ret = lockrestart_do(&trans,
199 bch2_subvolume_get(&trans, inum.subvol, true, 0, &subvol) ?:
200 bch2_inode_find_by_inum_trans(&trans, inum, &inode_u));
203 bch2_vfs_inode_init(&trans, inum, inode, &inode_u, &subvol);
204 bch2_trans_exit(&trans);
207 iget_failed(&inode->v);
208 return ERR_PTR(bch2_err_class(ret));
211 mutex_lock(&c->vfs_inodes_lock);
212 list_add(&inode->ei_vfs_inode_list, &c->vfs_inodes_list);
213 mutex_unlock(&c->vfs_inodes_lock);
215 unlock_new_inode(&inode->v);
220 struct bch_inode_info *
221 __bch2_create(struct mnt_idmap *idmap,
222 struct bch_inode_info *dir, struct dentry *dentry,
223 umode_t mode, dev_t rdev, subvol_inum snapshot_src,
226 struct bch_fs *c = dir->v.i_sb->s_fs_info;
227 struct btree_trans trans;
228 struct bch_inode_unpacked dir_u;
229 struct bch_inode_info *inode, *old;
230 struct bch_inode_unpacked inode_u;
231 struct posix_acl *default_acl = NULL, *acl = NULL;
233 struct bch_subvolume subvol;
238 * preallocate acls + vfs inode before btree transaction, so that
239 * nothing can fail after the transaction succeeds:
241 #ifdef CONFIG_BCACHEFS_POSIX_ACL
242 ret = posix_acl_create(&dir->v, &mode, &default_acl, &acl);
246 inode = to_bch_ei(new_inode(c->vfs_sb));
247 if (unlikely(!inode)) {
248 inode = ERR_PTR(-ENOMEM);
252 bch2_inode_init_early(c, &inode_u);
254 if (!(flags & BCH_CREATE_TMPFILE))
255 mutex_lock(&dir->ei_update_lock);
257 bch2_trans_init(&trans, c, 8,
258 2048 + (!(flags & BCH_CREATE_TMPFILE)
259 ? dentry->d_name.len : 0));
261 bch2_trans_begin(&trans);
263 ret = bch2_create_trans(&trans,
264 inode_inum(dir), &dir_u, &inode_u,
265 !(flags & BCH_CREATE_TMPFILE)
266 ? &dentry->d_name : NULL,
267 from_kuid(i_user_ns(&dir->v), current_fsuid()),
268 from_kgid(i_user_ns(&dir->v), current_fsgid()),
270 default_acl, acl, snapshot_src, flags) ?:
271 bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
272 KEY_TYPE_QUOTA_PREALLOC);
274 goto err_before_quota;
276 inum.subvol = inode_u.bi_subvol ?: dir->ei_subvol;
277 inum.inum = inode_u.bi_inum;
279 ret = bch2_subvolume_get(&trans, inum.subvol, true,
280 BTREE_ITER_WITH_UPDATES, &subvol) ?:
281 bch2_trans_commit(&trans, NULL, &journal_seq, 0);
283 bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
284 KEY_TYPE_QUOTA_WARN);
286 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
291 if (!(flags & BCH_CREATE_TMPFILE)) {
292 bch2_inode_update_after_write(&trans, dir, &dir_u,
293 ATTR_MTIME|ATTR_CTIME);
294 mutex_unlock(&dir->ei_update_lock);
297 bch2_iget5_set(&inode->v, &inum);
298 bch2_vfs_inode_init(&trans, inum, inode, &inode_u, &subvol);
300 set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
301 set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
304 * we must insert the new inode into the inode cache before calling
305 * bch2_trans_exit() and dropping locks, else we could race with another
306 * thread pulling the inode in and modifying it:
309 inode->v.i_state |= I_CREATING;
311 old = to_bch_ei(inode_insert5(&inode->v,
312 bch2_inode_hash(inum),
318 if (unlikely(old != inode)) {
320 * We raced, another process pulled the new inode into cache
323 make_bad_inode(&inode->v);
328 mutex_lock(&c->vfs_inodes_lock);
329 list_add(&inode->ei_vfs_inode_list, &c->vfs_inodes_list);
330 mutex_unlock(&c->vfs_inodes_lock);
332 * we really don't want insert_inode_locked2() to be setting
335 unlock_new_inode(&inode->v);
338 bch2_trans_exit(&trans);
340 posix_acl_release(default_acl);
341 posix_acl_release(acl);
344 if (!(flags & BCH_CREATE_TMPFILE))
345 mutex_unlock(&dir->ei_update_lock);
347 bch2_trans_exit(&trans);
348 make_bad_inode(&inode->v);
350 inode = ERR_PTR(ret);
356 static struct dentry *bch2_lookup(struct inode *vdir, struct dentry *dentry,
359 struct bch_fs *c = vdir->i_sb->s_fs_info;
360 struct bch_inode_info *dir = to_bch_ei(vdir);
361 struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
362 struct inode *vinode = NULL;
363 subvol_inum inum = { .subvol = 1 };
366 ret = bch2_dirent_lookup(c, inode_inum(dir), &hash,
367 &dentry->d_name, &inum);
370 vinode = bch2_vfs_inode_get(c, inum);
372 return d_splice_alias(vinode, dentry);
375 static int bch2_mknod(struct mnt_idmap *idmap,
376 struct inode *vdir, struct dentry *dentry,
377 umode_t mode, dev_t rdev)
379 struct bch_inode_info *inode =
380 __bch2_create(idmap, to_bch_ei(vdir), dentry, mode, rdev,
381 (subvol_inum) { 0 }, 0);
384 return bch2_err_class(PTR_ERR(inode));
386 d_instantiate(dentry, &inode->v);
390 static int bch2_create(struct mnt_idmap *idmap,
391 struct inode *vdir, struct dentry *dentry,
392 umode_t mode, bool excl)
394 return bch2_mknod(idmap, vdir, dentry, mode|S_IFREG, 0);
397 static int __bch2_link(struct bch_fs *c,
398 struct bch_inode_info *inode,
399 struct bch_inode_info *dir,
400 struct dentry *dentry)
402 struct btree_trans trans;
403 struct bch_inode_unpacked dir_u, inode_u;
406 mutex_lock(&inode->ei_update_lock);
407 bch2_trans_init(&trans, c, 4, 1024);
409 ret = commit_do(&trans, NULL, NULL, 0,
410 bch2_link_trans(&trans,
411 inode_inum(dir), &dir_u,
412 inode_inum(inode), &inode_u,
416 bch2_inode_update_after_write(&trans, dir, &dir_u,
417 ATTR_MTIME|ATTR_CTIME);
418 bch2_inode_update_after_write(&trans, inode, &inode_u, ATTR_CTIME);
421 bch2_trans_exit(&trans);
422 mutex_unlock(&inode->ei_update_lock);
426 static int bch2_link(struct dentry *old_dentry, struct inode *vdir,
427 struct dentry *dentry)
429 struct bch_fs *c = vdir->i_sb->s_fs_info;
430 struct bch_inode_info *dir = to_bch_ei(vdir);
431 struct bch_inode_info *inode = to_bch_ei(old_dentry->d_inode);
434 lockdep_assert_held(&inode->v.i_rwsem);
436 ret = __bch2_link(c, inode, dir, dentry);
441 d_instantiate(dentry, &inode->v);
445 int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
446 bool deleting_snapshot)
448 struct bch_fs *c = vdir->i_sb->s_fs_info;
449 struct bch_inode_info *dir = to_bch_ei(vdir);
450 struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
451 struct bch_inode_unpacked dir_u, inode_u;
452 struct btree_trans trans;
455 bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
456 bch2_trans_init(&trans, c, 4, 1024);
458 ret = commit_do(&trans, NULL, NULL,
460 bch2_unlink_trans(&trans,
461 inode_inum(dir), &dir_u,
462 &inode_u, &dentry->d_name,
467 bch2_inode_update_after_write(&trans, dir, &dir_u,
468 ATTR_MTIME|ATTR_CTIME);
469 bch2_inode_update_after_write(&trans, inode, &inode_u,
472 if (inode_u.bi_subvol) {
474 * Subvolume deletion is asynchronous, but we still want to tell
475 * the VFS that it's been deleted here:
477 set_nlink(&inode->v, 0);
480 bch2_trans_exit(&trans);
481 bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
486 static int bch2_unlink(struct inode *vdir, struct dentry *dentry)
488 return __bch2_unlink(vdir, dentry, false);
491 static int bch2_symlink(struct mnt_idmap *idmap,
492 struct inode *vdir, struct dentry *dentry,
495 struct bch_fs *c = vdir->i_sb->s_fs_info;
496 struct bch_inode_info *dir = to_bch_ei(vdir), *inode;
499 inode = __bch2_create(idmap, dir, dentry, S_IFLNK|S_IRWXUGO, 0,
500 (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
502 return bch2_err_class(PTR_ERR(inode));
504 inode_lock(&inode->v);
505 ret = page_symlink(&inode->v, symname, strlen(symname) + 1);
506 inode_unlock(&inode->v);
511 ret = filemap_write_and_wait_range(inode->v.i_mapping, 0, LLONG_MAX);
515 ret = __bch2_link(c, inode, dir, dentry);
519 d_instantiate(dentry, &inode->v);
526 static int bch2_mkdir(struct mnt_idmap *idmap,
527 struct inode *vdir, struct dentry *dentry, umode_t mode)
529 return bch2_mknod(idmap, vdir, dentry, mode|S_IFDIR, 0);
532 static int bch2_rename2(struct mnt_idmap *idmap,
533 struct inode *src_vdir, struct dentry *src_dentry,
534 struct inode *dst_vdir, struct dentry *dst_dentry,
537 struct bch_fs *c = src_vdir->i_sb->s_fs_info;
538 struct bch_inode_info *src_dir = to_bch_ei(src_vdir);
539 struct bch_inode_info *dst_dir = to_bch_ei(dst_vdir);
540 struct bch_inode_info *src_inode = to_bch_ei(src_dentry->d_inode);
541 struct bch_inode_info *dst_inode = to_bch_ei(dst_dentry->d_inode);
542 struct bch_inode_unpacked dst_dir_u, src_dir_u;
543 struct bch_inode_unpacked src_inode_u, dst_inode_u;
544 struct btree_trans trans;
545 enum bch_rename_mode mode = flags & RENAME_EXCHANGE
546 ? BCH_RENAME_EXCHANGE
547 : dst_dentry->d_inode
548 ? BCH_RENAME_OVERWRITE : BCH_RENAME;
551 if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
554 if (mode == BCH_RENAME_OVERWRITE) {
555 ret = filemap_write_and_wait_range(src_inode->v.i_mapping,
561 bch2_trans_init(&trans, c, 8, 2048);
563 bch2_lock_inodes(INODE_UPDATE_LOCK,
569 if (inode_attr_changing(dst_dir, src_inode, Inode_opt_project)) {
570 ret = bch2_fs_quota_transfer(c, src_inode,
573 KEY_TYPE_QUOTA_PREALLOC);
578 if (mode == BCH_RENAME_EXCHANGE &&
579 inode_attr_changing(src_dir, dst_inode, Inode_opt_project)) {
580 ret = bch2_fs_quota_transfer(c, dst_inode,
583 KEY_TYPE_QUOTA_PREALLOC);
588 ret = commit_do(&trans, NULL, NULL, 0,
589 bch2_rename_trans(&trans,
590 inode_inum(src_dir), &src_dir_u,
591 inode_inum(dst_dir), &dst_dir_u,
600 BUG_ON(src_inode->v.i_ino != src_inode_u.bi_inum);
602 dst_inode->v.i_ino != dst_inode_u.bi_inum);
604 bch2_inode_update_after_write(&trans, src_dir, &src_dir_u,
605 ATTR_MTIME|ATTR_CTIME);
607 if (src_dir != dst_dir)
608 bch2_inode_update_after_write(&trans, dst_dir, &dst_dir_u,
609 ATTR_MTIME|ATTR_CTIME);
611 bch2_inode_update_after_write(&trans, src_inode, &src_inode_u,
615 bch2_inode_update_after_write(&trans, dst_inode, &dst_inode_u,
618 bch2_trans_exit(&trans);
620 bch2_fs_quota_transfer(c, src_inode,
621 bch_qid(&src_inode->ei_inode),
623 KEY_TYPE_QUOTA_NOCHECK);
625 bch2_fs_quota_transfer(c, dst_inode,
626 bch_qid(&dst_inode->ei_inode),
628 KEY_TYPE_QUOTA_NOCHECK);
630 bch2_unlock_inodes(INODE_UPDATE_LOCK,
639 static void bch2_setattr_copy(struct mnt_idmap *idmap,
640 struct bch_inode_info *inode,
641 struct bch_inode_unpacked *bi,
644 struct bch_fs *c = inode->v.i_sb->s_fs_info;
645 unsigned int ia_valid = attr->ia_valid;
647 if (ia_valid & ATTR_UID)
648 bi->bi_uid = from_kuid(i_user_ns(&inode->v), attr->ia_uid);
649 if (ia_valid & ATTR_GID)
650 bi->bi_gid = from_kgid(i_user_ns(&inode->v), attr->ia_gid);
652 if (ia_valid & ATTR_SIZE)
653 bi->bi_size = attr->ia_size;
655 if (ia_valid & ATTR_ATIME)
656 bi->bi_atime = timespec_to_bch2_time(c, attr->ia_atime);
657 if (ia_valid & ATTR_MTIME)
658 bi->bi_mtime = timespec_to_bch2_time(c, attr->ia_mtime);
659 if (ia_valid & ATTR_CTIME)
660 bi->bi_ctime = timespec_to_bch2_time(c, attr->ia_ctime);
662 if (ia_valid & ATTR_MODE) {
663 umode_t mode = attr->ia_mode;
664 kgid_t gid = ia_valid & ATTR_GID
668 if (!in_group_p(gid) &&
669 !capable_wrt_inode_uidgid(idmap, &inode->v, CAP_FSETID))
675 int bch2_setattr_nonsize(struct mnt_idmap *idmap,
676 struct bch_inode_info *inode,
679 struct bch_fs *c = inode->v.i_sb->s_fs_info;
681 struct btree_trans trans;
682 struct btree_iter inode_iter = { NULL };
683 struct bch_inode_unpacked inode_u;
684 struct posix_acl *acl = NULL;
687 mutex_lock(&inode->ei_update_lock);
691 if (attr->ia_valid & ATTR_UID)
692 qid.q[QTYP_USR] = from_kuid(i_user_ns(&inode->v), attr->ia_uid);
694 if (attr->ia_valid & ATTR_GID)
695 qid.q[QTYP_GRP] = from_kgid(i_user_ns(&inode->v), attr->ia_gid);
697 ret = bch2_fs_quota_transfer(c, inode, qid, ~0,
698 KEY_TYPE_QUOTA_PREALLOC);
702 bch2_trans_init(&trans, c, 0, 0);
704 bch2_trans_begin(&trans);
708 ret = bch2_inode_peek(&trans, &inode_iter, &inode_u, inode_inum(inode),
713 bch2_setattr_copy(idmap, inode, &inode_u, attr);
715 if (attr->ia_valid & ATTR_MODE) {
716 ret = bch2_acl_chmod(&trans, inode_inum(inode), &inode_u,
717 inode_u.bi_mode, &acl);
722 ret = bch2_inode_write(&trans, &inode_iter, &inode_u) ?:
723 bch2_trans_commit(&trans, NULL, NULL,
724 BTREE_INSERT_NOFAIL);
726 bch2_trans_iter_exit(&trans, &inode_iter);
728 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
733 bch2_inode_update_after_write(&trans, inode, &inode_u, attr->ia_valid);
736 set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
738 bch2_trans_exit(&trans);
740 mutex_unlock(&inode->ei_update_lock);
742 return bch2_err_class(ret);
745 static int bch2_getattr(struct mnt_idmap *idmap,
746 const struct path *path, struct kstat *stat,
747 u32 request_mask, unsigned query_flags)
749 struct bch_inode_info *inode = to_bch_ei(d_inode(path->dentry));
750 struct bch_fs *c = inode->v.i_sb->s_fs_info;
752 stat->dev = inode->v.i_sb->s_dev;
753 stat->ino = inode->v.i_ino;
754 stat->mode = inode->v.i_mode;
755 stat->nlink = inode->v.i_nlink;
756 stat->uid = inode->v.i_uid;
757 stat->gid = inode->v.i_gid;
758 stat->rdev = inode->v.i_rdev;
759 stat->size = i_size_read(&inode->v);
760 stat->atime = inode->v.i_atime;
761 stat->mtime = inode->v.i_mtime;
762 stat->ctime = inode->v.i_ctime;
763 stat->blksize = block_bytes(c);
764 stat->blocks = inode->v.i_blocks;
766 if (request_mask & STATX_BTIME) {
767 stat->result_mask |= STATX_BTIME;
768 stat->btime = bch2_time_to_timespec(c, inode->ei_inode.bi_otime);
771 if (inode->ei_inode.bi_flags & BCH_INODE_IMMUTABLE)
772 stat->attributes |= STATX_ATTR_IMMUTABLE;
773 stat->attributes_mask |= STATX_ATTR_IMMUTABLE;
775 if (inode->ei_inode.bi_flags & BCH_INODE_APPEND)
776 stat->attributes |= STATX_ATTR_APPEND;
777 stat->attributes_mask |= STATX_ATTR_APPEND;
779 if (inode->ei_inode.bi_flags & BCH_INODE_NODUMP)
780 stat->attributes |= STATX_ATTR_NODUMP;
781 stat->attributes_mask |= STATX_ATTR_NODUMP;
786 static int bch2_setattr(struct mnt_idmap *idmap,
787 struct dentry *dentry, struct iattr *iattr)
789 struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
792 lockdep_assert_held(&inode->v.i_rwsem);
794 ret = setattr_prepare(idmap, dentry, iattr);
798 return iattr->ia_valid & ATTR_SIZE
799 ? bch2_truncate(idmap, inode, iattr)
800 : bch2_setattr_nonsize(idmap, inode, iattr);
803 static int bch2_tmpfile(struct mnt_idmap *idmap,
804 struct inode *vdir, struct file *file, umode_t mode)
806 struct bch_inode_info *inode =
807 __bch2_create(idmap, to_bch_ei(vdir),
808 file->f_path.dentry, mode, 0,
809 (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
812 return bch2_err_class(PTR_ERR(inode));
814 d_mark_tmpfile(file, &inode->v);
815 d_instantiate(file->f_path.dentry, &inode->v);
816 return finish_open_simple(file, 0);
819 static int bch2_fill_extent(struct bch_fs *c,
820 struct fiemap_extent_info *info,
821 struct bkey_s_c k, unsigned flags)
823 if (bkey_extent_is_direct_data(k.k)) {
824 struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
825 const union bch_extent_entry *entry;
826 struct extent_ptr_decoded p;
829 if (k.k->type == KEY_TYPE_reflink_v)
830 flags |= FIEMAP_EXTENT_SHARED;
832 bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
834 u64 offset = p.ptr.offset;
837 flags2 |= FIEMAP_EXTENT_UNWRITTEN;
839 if (p.crc.compression_type)
840 flags2 |= FIEMAP_EXTENT_ENCODED;
842 offset += p.crc.offset;
844 if ((offset & (block_sectors(c) - 1)) ||
845 (k.k->size & (block_sectors(c) - 1)))
846 flags2 |= FIEMAP_EXTENT_NOT_ALIGNED;
848 ret = fiemap_fill_next_extent(info,
849 bkey_start_offset(k.k) << 9,
851 k.k->size << 9, flags|flags2);
857 } else if (bkey_extent_is_inline_data(k.k)) {
858 return fiemap_fill_next_extent(info,
859 bkey_start_offset(k.k) << 9,
862 FIEMAP_EXTENT_DATA_INLINE);
863 } else if (k.k->type == KEY_TYPE_reservation) {
864 return fiemap_fill_next_extent(info,
865 bkey_start_offset(k.k) << 9,
868 FIEMAP_EXTENT_DELALLOC|
869 FIEMAP_EXTENT_UNWRITTEN);
875 static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
878 struct bch_fs *c = vinode->i_sb->s_fs_info;
879 struct bch_inode_info *ei = to_bch_ei(vinode);
880 struct btree_trans trans;
881 struct btree_iter iter;
883 struct bkey_buf cur, prev;
884 struct bpos end = POS(ei->v.i_ino, (start + len) >> 9);
885 unsigned offset_into_extent, sectors;
886 bool have_extent = false;
890 ret = fiemap_prep(&ei->v, info, start, &len, FIEMAP_FLAG_SYNC);
894 if (start + len < start)
899 bch2_bkey_buf_init(&cur);
900 bch2_bkey_buf_init(&prev);
901 bch2_trans_init(&trans, c, 0, 0);
903 bch2_trans_begin(&trans);
905 ret = bch2_subvolume_get_snapshot(&trans, ei->ei_subvol, &snapshot);
909 bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
910 SPOS(ei->v.i_ino, start, snapshot), 0);
912 while (!(ret = btree_trans_too_many_iters(&trans)) &&
913 (k = bch2_btree_iter_peek_upto(&iter, end)).k &&
914 !(ret = bkey_err(k))) {
915 enum btree_id data_btree = BTREE_ID_extents;
917 if (!bkey_extent_is_data(k.k) &&
918 k.k->type != KEY_TYPE_reservation) {
919 bch2_btree_iter_advance(&iter);
923 offset_into_extent = iter.pos.offset -
924 bkey_start_offset(k.k);
925 sectors = k.k->size - offset_into_extent;
927 bch2_bkey_buf_reassemble(&cur, c, k);
929 ret = bch2_read_indirect_extent(&trans, &data_btree,
930 &offset_into_extent, &cur);
934 k = bkey_i_to_s_c(cur.k);
935 bch2_bkey_buf_realloc(&prev, c, k.k->u64s);
937 sectors = min(sectors, k.k->size - offset_into_extent);
939 bch2_cut_front(POS(k.k->p.inode,
940 bkey_start_offset(k.k) +
943 bch2_key_resize(&cur.k->k, sectors);
944 cur.k->k.p = iter.pos;
945 cur.k->k.p.offset += cur.k->k.size;
948 bch2_trans_unlock(&trans);
949 ret = bch2_fill_extent(c, info,
950 bkey_i_to_s_c(prev.k), 0);
955 bkey_copy(prev.k, cur.k);
958 bch2_btree_iter_set_pos(&iter,
959 POS(iter.pos.inode, iter.pos.offset + sectors));
961 start = iter.pos.offset;
962 bch2_trans_iter_exit(&trans, &iter);
964 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
967 if (!ret && have_extent) {
968 bch2_trans_unlock(&trans);
969 ret = bch2_fill_extent(c, info, bkey_i_to_s_c(prev.k),
973 bch2_trans_exit(&trans);
974 bch2_bkey_buf_exit(&cur, c);
975 bch2_bkey_buf_exit(&prev, c);
976 return ret < 0 ? ret : 0;
979 static const struct vm_operations_struct bch_vm_ops = {
980 .fault = bch2_page_fault,
981 .map_pages = filemap_map_pages,
982 .page_mkwrite = bch2_page_mkwrite,
985 static int bch2_mmap(struct file *file, struct vm_area_struct *vma)
989 vma->vm_ops = &bch_vm_ops;
995 static loff_t bch2_dir_llseek(struct file *file, loff_t offset, int whence)
997 return generic_file_llseek_size(file, offset, whence,
1001 static int bch2_vfs_readdir(struct file *file, struct dir_context *ctx)
1003 struct bch_inode_info *inode = file_bch_inode(file);
1004 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1007 if (!dir_emit_dots(file, ctx))
1010 ret = bch2_readdir(c, inode_inum(inode), ctx);
1014 return bch2_err_class(ret);
1017 static const struct file_operations bch_file_operations = {
1018 .llseek = bch2_llseek,
1019 .read_iter = bch2_read_iter,
1020 .write_iter = bch2_write_iter,
1022 .open = generic_file_open,
1023 .fsync = bch2_fsync,
1024 .splice_read = filemap_splice_read,
1025 .splice_write = iter_file_splice_write,
1026 .fallocate = bch2_fallocate_dispatch,
1027 .unlocked_ioctl = bch2_fs_file_ioctl,
1028 #ifdef CONFIG_COMPAT
1029 .compat_ioctl = bch2_compat_fs_ioctl,
1031 .remap_file_range = bch2_remap_file_range,
1034 static const struct inode_operations bch_file_inode_operations = {
1035 .getattr = bch2_getattr,
1036 .setattr = bch2_setattr,
1037 .fiemap = bch2_fiemap,
1038 .listxattr = bch2_xattr_list,
1039 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1040 .get_acl = bch2_get_acl,
1041 .set_acl = bch2_set_acl,
1045 static const struct inode_operations bch_dir_inode_operations = {
1046 .lookup = bch2_lookup,
1047 .create = bch2_create,
1049 .unlink = bch2_unlink,
1050 .symlink = bch2_symlink,
1051 .mkdir = bch2_mkdir,
1052 .rmdir = bch2_unlink,
1053 .mknod = bch2_mknod,
1054 .rename = bch2_rename2,
1055 .getattr = bch2_getattr,
1056 .setattr = bch2_setattr,
1057 .tmpfile = bch2_tmpfile,
1058 .listxattr = bch2_xattr_list,
1059 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1060 .get_acl = bch2_get_acl,
1061 .set_acl = bch2_set_acl,
1065 static const struct file_operations bch_dir_file_operations = {
1066 .llseek = bch2_dir_llseek,
1067 .read = generic_read_dir,
1068 .iterate_shared = bch2_vfs_readdir,
1069 .fsync = bch2_fsync,
1070 .unlocked_ioctl = bch2_fs_file_ioctl,
1071 #ifdef CONFIG_COMPAT
1072 .compat_ioctl = bch2_compat_fs_ioctl,
1076 static const struct inode_operations bch_symlink_inode_operations = {
1077 .get_link = page_get_link,
1078 .getattr = bch2_getattr,
1079 .setattr = bch2_setattr,
1080 .listxattr = bch2_xattr_list,
1081 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1082 .get_acl = bch2_get_acl,
1083 .set_acl = bch2_set_acl,
1087 static const struct inode_operations bch_special_inode_operations = {
1088 .getattr = bch2_getattr,
1089 .setattr = bch2_setattr,
1090 .listxattr = bch2_xattr_list,
1091 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1092 .get_acl = bch2_get_acl,
1093 .set_acl = bch2_set_acl,
1097 static const struct address_space_operations bch_address_space_operations = {
1098 .read_folio = bch2_read_folio,
1099 .writepages = bch2_writepages,
1100 .readahead = bch2_readahead,
1101 .dirty_folio = filemap_dirty_folio,
1102 .write_begin = bch2_write_begin,
1103 .write_end = bch2_write_end,
1104 .invalidate_folio = bch2_invalidate_folio,
1105 .release_folio = bch2_release_folio,
1106 .direct_IO = noop_direct_IO,
1107 #ifdef CONFIG_MIGRATION
1108 .migrate_folio = filemap_migrate_folio,
1110 .error_remove_page = generic_error_remove_page,
1113 struct bcachefs_fid {
1119 struct bcachefs_fid_with_parent {
1120 struct bcachefs_fid fid;
1121 struct bcachefs_fid dir;
1124 static int bcachefs_fid_valid(int fh_len, int fh_type)
1127 case FILEID_BCACHEFS_WITHOUT_PARENT:
1128 return fh_len == sizeof(struct bcachefs_fid) / sizeof(u32);
1129 case FILEID_BCACHEFS_WITH_PARENT:
1130 return fh_len == sizeof(struct bcachefs_fid_with_parent) / sizeof(u32);
1136 static struct bcachefs_fid bch2_inode_to_fid(struct bch_inode_info *inode)
1138 return (struct bcachefs_fid) {
1139 .inum = inode->ei_inode.bi_inum,
1140 .subvol = inode->ei_subvol,
1141 .gen = inode->ei_inode.bi_generation,
1145 static int bch2_encode_fh(struct inode *vinode, u32 *fh, int *len,
1148 struct bch_inode_info *inode = to_bch_ei(vinode);
1149 struct bch_inode_info *dir = to_bch_ei(vdir);
1151 if (*len < sizeof(struct bcachefs_fid_with_parent) / sizeof(u32))
1152 return FILEID_INVALID;
1154 if (!S_ISDIR(inode->v.i_mode) && dir) {
1155 struct bcachefs_fid_with_parent *fid = (void *) fh;
1157 fid->fid = bch2_inode_to_fid(inode);
1158 fid->dir = bch2_inode_to_fid(dir);
1160 *len = sizeof(*fid) / sizeof(u32);
1161 return FILEID_BCACHEFS_WITH_PARENT;
1163 struct bcachefs_fid *fid = (void *) fh;
1165 *fid = bch2_inode_to_fid(inode);
1167 *len = sizeof(*fid) / sizeof(u32);
1168 return FILEID_BCACHEFS_WITHOUT_PARENT;
1172 static struct inode *bch2_nfs_get_inode(struct super_block *sb,
1173 struct bcachefs_fid fid)
1175 struct bch_fs *c = sb->s_fs_info;
1176 struct inode *vinode = bch2_vfs_inode_get(c, (subvol_inum) {
1177 .subvol = fid.subvol,
1180 if (!IS_ERR(vinode) && vinode->i_generation != fid.gen) {
1182 vinode = ERR_PTR(-ESTALE);
1187 static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *_fid,
1188 int fh_len, int fh_type)
1190 struct bcachefs_fid *fid = (void *) _fid;
1192 if (!bcachefs_fid_valid(fh_len, fh_type))
1195 return d_obtain_alias(bch2_nfs_get_inode(sb, *fid));
1198 static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *_fid,
1199 int fh_len, int fh_type)
1201 struct bcachefs_fid_with_parent *fid = (void *) _fid;
1203 if (!bcachefs_fid_valid(fh_len, fh_type) ||
1204 fh_type != FILEID_BCACHEFS_WITH_PARENT)
1207 return d_obtain_alias(bch2_nfs_get_inode(sb, fid->dir));
1210 static struct dentry *bch2_get_parent(struct dentry *child)
1212 struct bch_inode_info *inode = to_bch_ei(child->d_inode);
1213 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1214 subvol_inum parent_inum = {
1215 .subvol = inode->ei_inode.bi_parent_subvol ?:
1217 .inum = inode->ei_inode.bi_dir,
1220 if (!parent_inum.inum)
1223 return d_obtain_alias(bch2_vfs_inode_get(c, parent_inum));
1226 static int bch2_get_name(struct dentry *parent, char *name, struct dentry *child)
1228 struct bch_inode_info *inode = to_bch_ei(child->d_inode);
1229 struct bch_inode_info *dir = to_bch_ei(parent->d_inode);
1230 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1231 struct btree_trans trans;
1232 struct btree_iter iter1;
1233 struct btree_iter iter2;
1235 struct bkey_s_c_dirent d;
1236 struct bch_inode_unpacked inode_u;
1242 if (!S_ISDIR(dir->v.i_mode))
1245 bch2_trans_init(&trans, c, 0, 0);
1247 bch2_trans_iter_init(&trans, &iter1, BTREE_ID_dirents,
1248 POS(dir->ei_inode.bi_inum, 0), 0);
1249 bch2_trans_iter_init(&trans, &iter2, BTREE_ID_dirents,
1250 POS(dir->ei_inode.bi_inum, 0), 0);
1252 bch2_trans_begin(&trans);
1254 ret = bch2_subvolume_get_snapshot(&trans, dir->ei_subvol, &snapshot);
1258 bch2_btree_iter_set_snapshot(&iter1, snapshot);
1259 bch2_btree_iter_set_snapshot(&iter2, snapshot);
1261 ret = bch2_inode_find_by_inum_trans(&trans, inode_inum(inode), &inode_u);
1265 if (inode_u.bi_dir == dir->ei_inode.bi_inum) {
1266 bch2_btree_iter_set_pos(&iter1, POS(inode_u.bi_dir, inode_u.bi_dir_offset));
1268 k = bch2_btree_iter_peek_slot(&iter1);
1273 if (k.k->type != KEY_TYPE_dirent) {
1274 ret = -BCH_ERR_ENOENT_dirent_doesnt_match_inode;
1278 d = bkey_s_c_to_dirent(k);
1279 ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
1281 ret = -BCH_ERR_ENOENT_dirent_doesnt_match_inode;
1285 if (target.subvol == inode->ei_subvol &&
1286 target.inum == inode->ei_inode.bi_inum)
1290 * File with multiple hardlinks and our backref is to the wrong
1291 * directory - linear search:
1293 for_each_btree_key_continue_norestart(iter2, 0, k, ret) {
1294 if (k.k->p.inode > dir->ei_inode.bi_inum)
1297 if (k.k->type != KEY_TYPE_dirent)
1300 d = bkey_s_c_to_dirent(k);
1301 ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
1307 if (target.subvol == inode->ei_subvol &&
1308 target.inum == inode->ei_inode.bi_inum)
1316 name_len = min_t(unsigned, bch2_dirent_name_bytes(d), NAME_MAX);
1318 memcpy(name, d.v->d_name, name_len);
1319 name[name_len] = '\0';
1321 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
1324 bch2_trans_iter_exit(&trans, &iter1);
1325 bch2_trans_iter_exit(&trans, &iter2);
1326 bch2_trans_exit(&trans);
1331 static const struct export_operations bch_export_ops = {
1332 .encode_fh = bch2_encode_fh,
1333 .fh_to_dentry = bch2_fh_to_dentry,
1334 .fh_to_parent = bch2_fh_to_parent,
1335 .get_parent = bch2_get_parent,
1336 .get_name = bch2_get_name,
1339 static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
1340 struct bch_inode_info *inode,
1341 struct bch_inode_unpacked *bi,
1342 struct bch_subvolume *subvol)
1344 bch2_inode_update_after_write(trans, inode, bi, ~0);
1346 if (BCH_SUBVOLUME_SNAP(subvol))
1347 set_bit(EI_INODE_SNAPSHOT, &inode->ei_flags);
1349 clear_bit(EI_INODE_SNAPSHOT, &inode->ei_flags);
1351 inode->v.i_blocks = bi->bi_sectors;
1352 inode->v.i_ino = bi->bi_inum;
1353 inode->v.i_rdev = bi->bi_dev;
1354 inode->v.i_generation = bi->bi_generation;
1355 inode->v.i_size = bi->bi_size;
1357 inode->ei_flags = 0;
1358 inode->ei_quota_reserved = 0;
1359 inode->ei_qid = bch_qid(bi);
1360 inode->ei_subvol = inum.subvol;
1362 inode->v.i_mapping->a_ops = &bch_address_space_operations;
1364 switch (inode->v.i_mode & S_IFMT) {
1366 inode->v.i_op = &bch_file_inode_operations;
1367 inode->v.i_fop = &bch_file_operations;
1370 inode->v.i_op = &bch_dir_inode_operations;
1371 inode->v.i_fop = &bch_dir_file_operations;
1374 inode_nohighmem(&inode->v);
1375 inode->v.i_op = &bch_symlink_inode_operations;
1378 init_special_inode(&inode->v, inode->v.i_mode, inode->v.i_rdev);
1379 inode->v.i_op = &bch_special_inode_operations;
1383 mapping_set_large_folios(inode->v.i_mapping);
1386 static struct inode *bch2_alloc_inode(struct super_block *sb)
1388 struct bch_inode_info *inode;
1390 inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
1394 inode_init_once(&inode->v);
1395 mutex_init(&inode->ei_update_lock);
1396 two_state_lock_init(&inode->ei_pagecache_lock);
1397 INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
1398 mutex_init(&inode->ei_quota_lock);
1403 static void bch2_i_callback(struct rcu_head *head)
1405 struct inode *vinode = container_of(head, struct inode, i_rcu);
1406 struct bch_inode_info *inode = to_bch_ei(vinode);
1408 kmem_cache_free(bch2_inode_cache, inode);
1411 static void bch2_destroy_inode(struct inode *vinode)
1413 call_rcu(&vinode->i_rcu, bch2_i_callback);
1416 static int inode_update_times_fn(struct bch_inode_info *inode,
1417 struct bch_inode_unpacked *bi,
1420 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1422 bi->bi_atime = timespec_to_bch2_time(c, inode->v.i_atime);
1423 bi->bi_mtime = timespec_to_bch2_time(c, inode->v.i_mtime);
1424 bi->bi_ctime = timespec_to_bch2_time(c, inode->v.i_ctime);
1429 static int bch2_vfs_write_inode(struct inode *vinode,
1430 struct writeback_control *wbc)
1432 struct bch_fs *c = vinode->i_sb->s_fs_info;
1433 struct bch_inode_info *inode = to_bch_ei(vinode);
1436 mutex_lock(&inode->ei_update_lock);
1437 ret = bch2_write_inode(c, inode, inode_update_times_fn, NULL,
1438 ATTR_ATIME|ATTR_MTIME|ATTR_CTIME);
1439 mutex_unlock(&inode->ei_update_lock);
1441 return bch2_err_class(ret);
1444 static void bch2_evict_inode(struct inode *vinode)
1446 struct bch_fs *c = vinode->i_sb->s_fs_info;
1447 struct bch_inode_info *inode = to_bch_ei(vinode);
1449 truncate_inode_pages_final(&inode->v.i_data);
1451 clear_inode(&inode->v);
1453 BUG_ON(!is_bad_inode(&inode->v) && inode->ei_quota_reserved);
1455 if (!inode->v.i_nlink && !is_bad_inode(&inode->v)) {
1456 bch2_quota_acct(c, inode->ei_qid, Q_SPC, -((s64) inode->v.i_blocks),
1457 KEY_TYPE_QUOTA_WARN);
1458 bch2_quota_acct(c, inode->ei_qid, Q_INO, -1,
1459 KEY_TYPE_QUOTA_WARN);
1460 bch2_inode_rm(c, inode_inum(inode));
1463 mutex_lock(&c->vfs_inodes_lock);
1464 list_del_init(&inode->ei_vfs_inode_list);
1465 mutex_unlock(&c->vfs_inodes_lock);
1468 void bch2_evict_subvolume_inodes(struct bch_fs *c, snapshot_id_list *s)
1470 struct bch_inode_info *inode, **i;
1471 DARRAY(struct bch_inode_info *) grabbed;
1472 bool clean_pass = false, this_pass_clean;
1475 * Initially, we scan for inodes without I_DONTCACHE, then mark them to
1476 * be pruned with d_mark_dontcache().
1478 * Once we've had a clean pass where we didn't find any inodes without
1479 * I_DONTCACHE, we wait for them to be freed:
1482 darray_init(&grabbed);
1483 darray_make_room(&grabbed, 1024);
1486 this_pass_clean = true;
1488 mutex_lock(&c->vfs_inodes_lock);
1489 list_for_each_entry(inode, &c->vfs_inodes_list, ei_vfs_inode_list) {
1490 if (!snapshot_list_has_id(s, inode->ei_subvol))
1493 if (!(inode->v.i_state & I_DONTCACHE) &&
1494 !(inode->v.i_state & I_FREEING) &&
1496 this_pass_clean = false;
1498 if (darray_push_gfp(&grabbed, inode, GFP_ATOMIC|__GFP_NOWARN)) {
1502 } else if (clean_pass && this_pass_clean) {
1503 wait_queue_head_t *wq = bit_waitqueue(&inode->v.i_state, __I_NEW);
1504 DEFINE_WAIT_BIT(wait, &inode->v.i_state, __I_NEW);
1506 prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
1507 mutex_unlock(&c->vfs_inodes_lock);
1510 finish_wait(wq, &wait.wq_entry);
1514 mutex_unlock(&c->vfs_inodes_lock);
1516 darray_for_each(grabbed, i) {
1518 d_mark_dontcache(&inode->v);
1519 d_prune_aliases(&inode->v);
1524 if (!clean_pass || !this_pass_clean) {
1525 clean_pass = this_pass_clean;
1529 darray_exit(&grabbed);
1532 static int bch2_statfs(struct dentry *dentry, struct kstatfs *buf)
1534 struct super_block *sb = dentry->d_sb;
1535 struct bch_fs *c = sb->s_fs_info;
1536 struct bch_fs_usage_short usage = bch2_fs_usage_read_short(c);
1537 unsigned shift = sb->s_blocksize_bits - 9;
1539 * this assumes inodes take up 64 bytes, which is a decent average
1542 u64 avail_inodes = ((usage.capacity - usage.used) << 3);
1545 buf->f_type = BCACHEFS_STATFS_MAGIC;
1546 buf->f_bsize = sb->s_blocksize;
1547 buf->f_blocks = usage.capacity >> shift;
1548 buf->f_bfree = usage.free >> shift;
1549 buf->f_bavail = avail_factor(usage.free) >> shift;
1551 buf->f_files = usage.nr_inodes + avail_inodes;
1552 buf->f_ffree = avail_inodes;
1554 fsid = le64_to_cpup((void *) c->sb.user_uuid.b) ^
1555 le64_to_cpup((void *) c->sb.user_uuid.b + sizeof(u64));
1556 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1557 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1558 buf->f_namelen = BCH_NAME_MAX;
1563 static int bch2_sync_fs(struct super_block *sb, int wait)
1565 struct bch_fs *c = sb->s_fs_info;
1568 if (c->opts.journal_flush_disabled)
1572 bch2_journal_flush_async(&c->journal, NULL);
1576 ret = bch2_journal_flush(&c->journal);
1577 return bch2_err_class(ret);
1580 static struct bch_fs *bch2_path_to_fs(const char *path)
1586 ret = lookup_bdev(path, &dev);
1588 return ERR_PTR(ret);
1590 c = bch2_dev_to_fs(dev);
1592 closure_put(&c->cl);
1593 return c ?: ERR_PTR(-ENOENT);
1596 static char **split_devs(const char *_dev_name, unsigned *nr)
1598 char *dev_name = NULL, **devs = NULL, *s;
1599 size_t i, nr_devs = 0;
1601 dev_name = kstrdup(_dev_name, GFP_KERNEL);
1605 for (s = dev_name; s; s = strchr(s + 1, ':'))
1608 devs = kcalloc(nr_devs + 1, sizeof(const char *), GFP_KERNEL);
1614 for (i = 0, s = dev_name;
1616 (s = strchr(s, ':')) && (*s++ = '\0'))
1623 static int bch2_remount(struct super_block *sb, int *flags, char *data)
1625 struct bch_fs *c = sb->s_fs_info;
1626 struct bch_opts opts = bch2_opts_empty();
1629 opt_set(opts, read_only, (*flags & SB_RDONLY) != 0);
1631 ret = bch2_parse_mount_opts(c, &opts, data);
1635 if (opts.read_only != c->opts.read_only) {
1636 down_write(&c->state_lock);
1638 if (opts.read_only) {
1639 bch2_fs_read_only(c);
1641 sb->s_flags |= SB_RDONLY;
1643 ret = bch2_fs_read_write(c);
1645 bch_err(c, "error going rw: %i", ret);
1646 up_write(&c->state_lock);
1651 sb->s_flags &= ~SB_RDONLY;
1654 c->opts.read_only = opts.read_only;
1656 up_write(&c->state_lock);
1659 if (opts.errors >= 0)
1660 c->opts.errors = opts.errors;
1662 return bch2_err_class(ret);
1665 static int bch2_show_devname(struct seq_file *seq, struct dentry *root)
1667 struct bch_fs *c = root->d_sb->s_fs_info;
1672 for_each_online_member(ca, c, i) {
1676 seq_puts(seq, "/dev/");
1677 seq_puts(seq, ca->name);
1683 static int bch2_show_options(struct seq_file *seq, struct dentry *root)
1685 struct bch_fs *c = root->d_sb->s_fs_info;
1687 struct printbuf buf = PRINTBUF;
1690 for (i = 0; i < bch2_opts_nr; i++) {
1691 const struct bch_option *opt = &bch2_opt_table[i];
1692 u64 v = bch2_opt_get_by_id(&c->opts, i);
1694 if (!(opt->flags & OPT_MOUNT))
1697 if (v == bch2_opt_get_by_id(&bch2_opts_default, i))
1700 printbuf_reset(&buf);
1701 bch2_opt_to_text(&buf, c, c->disk_sb.sb, opt, v,
1702 OPT_SHOW_MOUNT_STYLE);
1704 seq_puts(seq, buf.buf);
1707 if (buf.allocation_failure)
1709 printbuf_exit(&buf);
1713 static void bch2_put_super(struct super_block *sb)
1715 struct bch_fs *c = sb->s_fs_info;
1720 static const struct super_operations bch_super_operations = {
1721 .alloc_inode = bch2_alloc_inode,
1722 .destroy_inode = bch2_destroy_inode,
1723 .write_inode = bch2_vfs_write_inode,
1724 .evict_inode = bch2_evict_inode,
1725 .sync_fs = bch2_sync_fs,
1726 .statfs = bch2_statfs,
1727 .show_devname = bch2_show_devname,
1728 .show_options = bch2_show_options,
1729 .remount_fs = bch2_remount,
1730 .put_super = bch2_put_super,
1732 .freeze_fs = bch2_freeze,
1733 .unfreeze_fs = bch2_unfreeze,
1737 static int bch2_set_super(struct super_block *s, void *data)
1739 s->s_fs_info = data;
1743 static int bch2_noset_super(struct super_block *s, void *data)
1748 static int bch2_test_super(struct super_block *s, void *data)
1750 struct bch_fs *c = s->s_fs_info;
1751 struct bch_fs **devs = data;
1757 for (i = 0; devs[i]; i++)
1763 static struct dentry *bch2_mount(struct file_system_type *fs_type,
1764 int flags, const char *dev_name, void *data)
1768 struct super_block *sb;
1769 struct inode *vinode;
1770 struct bch_opts opts = bch2_opts_empty();
1772 struct bch_fs **devs_to_fs = NULL;
1773 unsigned i, nr_devs;
1776 opt_set(opts, read_only, (flags & SB_RDONLY) != 0);
1778 ret = bch2_parse_mount_opts(NULL, &opts, data);
1780 return ERR_PTR(ret);
1782 if (!dev_name || strlen(dev_name) == 0)
1783 return ERR_PTR(-EINVAL);
1785 devs = split_devs(dev_name, &nr_devs);
1787 return ERR_PTR(-ENOMEM);
1789 devs_to_fs = kcalloc(nr_devs + 1, sizeof(void *), GFP_KERNEL);
1791 sb = ERR_PTR(-ENOMEM);
1795 for (i = 0; i < nr_devs; i++)
1796 devs_to_fs[i] = bch2_path_to_fs(devs[i]);
1798 sb = sget(fs_type, bch2_test_super, bch2_noset_super,
1799 flags|SB_NOSEC, devs_to_fs);
1803 c = bch2_fs_open(devs, nr_devs, opts);
1809 /* Some options can't be parsed until after the fs is started: */
1810 ret = bch2_parse_mount_opts(c, &opts, data);
1817 bch2_opts_apply(&c->opts, opts);
1819 sb = sget(fs_type, NULL, bch2_set_super, flags|SB_NOSEC, c);
1829 ret = bch2_err_class(ret);
1830 return ERR_PTR(ret);
1836 if ((flags ^ sb->s_flags) & SB_RDONLY) {
1843 sb->s_blocksize = block_bytes(c);
1844 sb->s_blocksize_bits = ilog2(block_bytes(c));
1845 sb->s_maxbytes = MAX_LFS_FILESIZE;
1846 sb->s_op = &bch_super_operations;
1847 sb->s_export_op = &bch_export_ops;
1848 #ifdef CONFIG_BCACHEFS_QUOTA
1849 sb->s_qcop = &bch2_quotactl_operations;
1850 sb->s_quota_types = QTYPE_MASK_USR|QTYPE_MASK_GRP|QTYPE_MASK_PRJ;
1852 sb->s_xattr = bch2_xattr_handlers;
1853 sb->s_magic = BCACHEFS_STATFS_MAGIC;
1854 sb->s_time_gran = c->sb.nsec_per_time_unit;
1855 sb->s_time_min = div_s64(S64_MIN, c->sb.time_units_per_sec) + 1;
1856 sb->s_time_max = div_s64(S64_MAX, c->sb.time_units_per_sec);
1858 strscpy(sb->s_id, c->name, sizeof(sb->s_id));
1860 ret = super_setup_bdi(sb);
1864 sb->s_bdi->ra_pages = VM_READAHEAD_PAGES;
1866 for_each_online_member(ca, c, i) {
1867 struct block_device *bdev = ca->disk_sb.bdev;
1869 /* XXX: create an anonymous device for multi device filesystems */
1871 sb->s_dev = bdev->bd_dev;
1872 percpu_ref_put(&ca->io_ref);
1878 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1880 sb->s_flags |= SB_POSIXACL;
1883 sb->s_shrink.seeks = 0;
1885 vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_SUBVOL_INUM);
1886 ret = PTR_ERR_OR_ZERO(vinode);
1888 bch_err(c, "error mounting: error getting root inode: %s", bch2_err_str(ret));
1892 sb->s_root = d_make_root(vinode);
1894 bch_err(c, "error mounting: error allocating root dentry");
1899 sb->s_flags |= SB_ACTIVE;
1901 return dget(sb->s_root);
1904 deactivate_locked_super(sb);
1905 return ERR_PTR(bch2_err_class(ret));
1908 static void bch2_kill_sb(struct super_block *sb)
1910 struct bch_fs *c = sb->s_fs_info;
1912 generic_shutdown_super(sb);
1916 static struct file_system_type bcache_fs_type = {
1917 .owner = THIS_MODULE,
1919 .mount = bch2_mount,
1920 .kill_sb = bch2_kill_sb,
1921 .fs_flags = FS_REQUIRES_DEV,
1924 MODULE_ALIAS_FS("bcachefs");
1926 void bch2_vfs_exit(void)
1928 unregister_filesystem(&bcache_fs_type);
1929 kmem_cache_destroy(bch2_inode_cache);
1932 int __init bch2_vfs_init(void)
1936 bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
1937 if (!bch2_inode_cache)
1940 ret = register_filesystem(&bcache_fs_type);
1950 #endif /* NO_BCACHEFS_FS */