1 // SPDX-License-Identifier: GPL-2.0
7 #include "btree_update.h"
13 #include "fs-common.h"
25 #include <linux/aio.h>
26 #include <linux/backing-dev.h>
27 #include <linux/exportfs.h>
28 #include <linux/fiemap.h>
29 #include <linux/module.h>
30 #include <linux/pagemap.h>
31 #include <linux/posix_acl.h>
32 #include <linux/random.h>
33 #include <linux/statfs.h>
34 #include <linux/string.h>
35 #include <linux/xattr.h>
37 static struct kmem_cache *bch2_inode_cache;
39 static void bch2_vfs_inode_init(struct btree_trans *, subvol_inum,
40 struct bch_inode_info *,
41 struct bch_inode_unpacked *);
43 static void __pagecache_lock_put(struct pagecache_lock *lock, long i)
45 BUG_ON(atomic_long_read(&lock->v) == 0);
47 if (atomic_long_sub_return_release(i, &lock->v) == 0)
48 wake_up_all(&lock->wait);
51 static bool __pagecache_lock_tryget(struct pagecache_lock *lock, long i)
53 long v = atomic_long_read(&lock->v), old;
58 if (i > 0 ? v < 0 : v > 0)
60 } while ((v = atomic_long_cmpxchg_acquire(&lock->v,
61 old, old + i)) != old);
65 static void __pagecache_lock_get(struct pagecache_lock *lock, long i)
67 wait_event(lock->wait, __pagecache_lock_tryget(lock, i));
70 void bch2_pagecache_add_put(struct pagecache_lock *lock)
72 __pagecache_lock_put(lock, 1);
75 bool bch2_pagecache_add_tryget(struct pagecache_lock *lock)
77 return __pagecache_lock_tryget(lock, 1);
80 void bch2_pagecache_add_get(struct pagecache_lock *lock)
82 __pagecache_lock_get(lock, 1);
85 void bch2_pagecache_block_put(struct pagecache_lock *lock)
87 __pagecache_lock_put(lock, -1);
90 void bch2_pagecache_block_get(struct pagecache_lock *lock)
92 __pagecache_lock_get(lock, -1);
95 void bch2_inode_update_after_write(struct btree_trans *trans,
96 struct bch_inode_info *inode,
97 struct bch_inode_unpacked *bi,
100 struct bch_fs *c = trans->c;
102 BUG_ON(bi->bi_inum != inode->v.i_ino);
104 bch2_assert_pos_locked(trans, BTREE_ID_inodes,
106 0 && c->opts.inodes_use_key_cache);
108 set_nlink(&inode->v, bch2_inode_nlink_get(bi));
109 i_uid_write(&inode->v, bi->bi_uid);
110 i_gid_write(&inode->v, bi->bi_gid);
111 inode->v.i_mode = bi->bi_mode;
113 if (fields & ATTR_ATIME)
114 inode->v.i_atime = bch2_time_to_timespec(c, bi->bi_atime);
115 if (fields & ATTR_MTIME)
116 inode->v.i_mtime = bch2_time_to_timespec(c, bi->bi_mtime);
117 if (fields & ATTR_CTIME)
118 inode->v.i_ctime = bch2_time_to_timespec(c, bi->bi_ctime);
120 inode->ei_inode = *bi;
122 bch2_inode_flags_to_vfs(inode);
125 int __must_check bch2_write_inode(struct bch_fs *c,
126 struct bch_inode_info *inode,
128 void *p, unsigned fields)
130 struct btree_trans trans;
131 struct btree_iter iter = { NULL };
132 struct bch_inode_unpacked inode_u;
135 bch2_trans_init(&trans, c, 0, 512);
138 bch2_trans_begin(&trans);
140 ret = bch2_inode_peek(&trans, &iter, &inode_u, inode_inum(inode),
141 BTREE_ITER_INTENT) ?:
142 (set ? set(inode, &inode_u, p) : 0) ?:
143 bch2_inode_write(&trans, &iter, &inode_u) ?:
144 bch2_trans_commit(&trans, NULL, NULL, BTREE_INSERT_NOFAIL);
147 * the btree node lock protects inode->ei_inode, not ei_update_lock;
148 * this is important for inode updates via bchfs_write_index_update
151 bch2_inode_update_after_write(&trans, inode, &inode_u, fields);
153 bch2_trans_iter_exit(&trans, &iter);
158 bch2_trans_exit(&trans);
159 return ret < 0 ? ret : 0;
162 int bch2_fs_quota_transfer(struct bch_fs *c,
163 struct bch_inode_info *inode,
164 struct bch_qid new_qid,
166 enum quota_acct_mode mode)
171 qtypes &= enabled_qtypes(c);
173 for (i = 0; i < QTYP_NR; i++)
174 if (new_qid.q[i] == inode->ei_qid.q[i])
175 qtypes &= ~(1U << i);
180 mutex_lock(&inode->ei_quota_lock);
182 ret = bch2_quota_transfer(c, qtypes, new_qid,
185 inode->ei_quota_reserved,
188 for (i = 0; i < QTYP_NR; i++)
189 if (qtypes & (1 << i))
190 inode->ei_qid.q[i] = new_qid.q[i];
192 mutex_unlock(&inode->ei_quota_lock);
197 static int bch2_iget5_test(struct inode *vinode, void *p)
199 struct bch_inode_info *inode = to_bch_ei(vinode);
200 subvol_inum *inum = p;
202 return inode->ei_subvol == inum->subvol &&
203 inode->ei_inode.bi_inum == inum->inum;
206 static int bch2_iget5_set(struct inode *vinode, void *p)
208 struct bch_inode_info *inode = to_bch_ei(vinode);
209 subvol_inum *inum = p;
211 inode->v.i_ino = inum->inum;
212 inode->ei_subvol = inum->subvol;
213 inode->ei_inode.bi_inum = inum->inum;
217 static unsigned bch2_inode_hash(subvol_inum inum)
219 return jhash_3words(inum.subvol, inum.inum >> 32, inum.inum, JHASH_INITVAL);
222 struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
224 struct bch_inode_unpacked inode_u;
225 struct bch_inode_info *inode;
226 struct btree_trans trans;
229 inode = to_bch_ei(iget5_locked(c->vfs_sb,
230 bch2_inode_hash(inum),
234 if (unlikely(!inode))
235 return ERR_PTR(-ENOMEM);
236 if (!(inode->v.i_state & I_NEW))
239 bch2_trans_init(&trans, c, 8, 0);
240 ret = lockrestart_do(&trans,
241 bch2_inode_find_by_inum_trans(&trans, inum, &inode_u));
244 bch2_vfs_inode_init(&trans, inum, inode, &inode_u);
245 bch2_trans_exit(&trans);
248 iget_failed(&inode->v);
252 unlock_new_inode(&inode->v);
257 struct bch_inode_info *
258 __bch2_create(struct user_namespace *mnt_userns,
259 struct bch_inode_info *dir, struct dentry *dentry,
260 umode_t mode, dev_t rdev, subvol_inum snapshot_src,
263 struct bch_fs *c = dir->v.i_sb->s_fs_info;
264 struct btree_trans trans;
265 struct bch_inode_unpacked dir_u;
266 struct bch_inode_info *inode, *old;
267 struct bch_inode_unpacked inode_u;
268 struct posix_acl *default_acl = NULL, *acl = NULL;
274 * preallocate acls + vfs inode before btree transaction, so that
275 * nothing can fail after the transaction succeeds:
277 #ifdef CONFIG_BCACHEFS_POSIX_ACL
278 ret = posix_acl_create(&dir->v, &mode, &default_acl, &acl);
282 inode = to_bch_ei(new_inode(c->vfs_sb));
283 if (unlikely(!inode)) {
284 inode = ERR_PTR(-ENOMEM);
288 bch2_inode_init_early(c, &inode_u);
290 if (!(flags & BCH_CREATE_TMPFILE))
291 mutex_lock(&dir->ei_update_lock);
293 bch2_trans_init(&trans, c, 8,
294 2048 + (!(flags & BCH_CREATE_TMPFILE)
295 ? dentry->d_name.len : 0));
297 bch2_trans_begin(&trans);
299 ret = bch2_create_trans(&trans,
300 inode_inum(dir), &dir_u, &inode_u,
301 !(flags & BCH_CREATE_TMPFILE)
302 ? &dentry->d_name : NULL,
303 from_kuid(mnt_userns, current_fsuid()),
304 from_kgid(mnt_userns, current_fsgid()),
306 default_acl, acl, snapshot_src, flags) ?:
307 bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
308 KEY_TYPE_QUOTA_PREALLOC);
310 goto err_before_quota;
312 ret = bch2_trans_commit(&trans, NULL, &journal_seq, 0);
314 bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
315 KEY_TYPE_QUOTA_WARN);
322 if (!(flags & BCH_CREATE_TMPFILE)) {
323 bch2_inode_update_after_write(&trans, dir, &dir_u,
324 ATTR_MTIME|ATTR_CTIME);
325 mutex_unlock(&dir->ei_update_lock);
328 inum.subvol = inode_u.bi_subvol ?: dir->ei_subvol;
329 inum.inum = inode_u.bi_inum;
331 bch2_iget5_set(&inode->v, &inum);
332 bch2_vfs_inode_init(&trans, inum, inode, &inode_u);
334 set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
335 set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
338 * we must insert the new inode into the inode cache before calling
339 * bch2_trans_exit() and dropping locks, else we could race with another
340 * thread pulling the inode in and modifying it:
343 inode->v.i_state |= I_CREATING;
345 old = to_bch_ei(inode_insert5(&inode->v,
346 bch2_inode_hash(inum),
352 if (unlikely(old != inode)) {
354 * We raced, another process pulled the new inode into cache
357 make_bad_inode(&inode->v);
363 * we really don't want insert_inode_locked2() to be setting
366 unlock_new_inode(&inode->v);
369 bch2_trans_exit(&trans);
371 posix_acl_release(default_acl);
372 posix_acl_release(acl);
375 if (!(flags & BCH_CREATE_TMPFILE))
376 mutex_unlock(&dir->ei_update_lock);
378 bch2_trans_exit(&trans);
379 make_bad_inode(&inode->v);
381 inode = ERR_PTR(ret);
387 static struct dentry *bch2_lookup(struct inode *vdir, struct dentry *dentry,
390 struct bch_fs *c = vdir->i_sb->s_fs_info;
391 struct bch_inode_info *dir = to_bch_ei(vdir);
392 struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
393 struct inode *vinode = NULL;
394 subvol_inum inum = { .subvol = 1 };
397 ret = bch2_dirent_lookup(c, inode_inum(dir), &hash,
398 &dentry->d_name, &inum);
401 vinode = bch2_vfs_inode_get(c, inum);
403 return d_splice_alias(vinode, dentry);
406 static int bch2_mknod(struct user_namespace *mnt_userns,
407 struct inode *vdir, struct dentry *dentry,
408 umode_t mode, dev_t rdev)
410 struct bch_inode_info *inode =
411 __bch2_create(mnt_userns, to_bch_ei(vdir), dentry, mode, rdev,
412 (subvol_inum) { 0 }, 0);
415 return PTR_ERR(inode);
417 d_instantiate(dentry, &inode->v);
421 static int bch2_create(struct user_namespace *mnt_userns,
422 struct inode *vdir, struct dentry *dentry,
423 umode_t mode, bool excl)
425 return bch2_mknod(mnt_userns, vdir, dentry, mode|S_IFREG, 0);
428 static int __bch2_link(struct bch_fs *c,
429 struct bch_inode_info *inode,
430 struct bch_inode_info *dir,
431 struct dentry *dentry)
433 struct btree_trans trans;
434 struct bch_inode_unpacked dir_u, inode_u;
437 mutex_lock(&inode->ei_update_lock);
438 bch2_trans_init(&trans, c, 4, 1024);
440 ret = __bch2_trans_do(&trans, NULL, NULL, 0,
441 bch2_link_trans(&trans,
442 inode_inum(dir), &dir_u,
443 inode_inum(inode), &inode_u,
447 bch2_inode_update_after_write(&trans, dir, &dir_u,
448 ATTR_MTIME|ATTR_CTIME);
449 bch2_inode_update_after_write(&trans, inode, &inode_u, ATTR_CTIME);
452 bch2_trans_exit(&trans);
453 mutex_unlock(&inode->ei_update_lock);
457 static int bch2_link(struct dentry *old_dentry, struct inode *vdir,
458 struct dentry *dentry)
460 struct bch_fs *c = vdir->i_sb->s_fs_info;
461 struct bch_inode_info *dir = to_bch_ei(vdir);
462 struct bch_inode_info *inode = to_bch_ei(old_dentry->d_inode);
465 lockdep_assert_held(&inode->v.i_rwsem);
467 ret = __bch2_link(c, inode, dir, dentry);
472 d_instantiate(dentry, &inode->v);
476 int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
477 bool deleting_snapshot)
479 struct bch_fs *c = vdir->i_sb->s_fs_info;
480 struct bch_inode_info *dir = to_bch_ei(vdir);
481 struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
482 struct bch_inode_unpacked dir_u, inode_u;
483 struct btree_trans trans;
486 bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
487 bch2_trans_init(&trans, c, 4, 1024);
489 ret = __bch2_trans_do(&trans, NULL, NULL,
491 bch2_unlink_trans(&trans,
492 inode_inum(dir), &dir_u,
493 &inode_u, &dentry->d_name,
497 bch2_inode_update_after_write(&trans, dir, &dir_u,
498 ATTR_MTIME|ATTR_CTIME);
499 bch2_inode_update_after_write(&trans, inode, &inode_u,
503 bch2_trans_exit(&trans);
504 bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
509 static int bch2_unlink(struct inode *vdir, struct dentry *dentry)
511 return __bch2_unlink(vdir, dentry, false);
514 static int bch2_symlink(struct user_namespace *mnt_userns,
515 struct inode *vdir, struct dentry *dentry,
518 struct bch_fs *c = vdir->i_sb->s_fs_info;
519 struct bch_inode_info *dir = to_bch_ei(vdir), *inode;
522 inode = __bch2_create(mnt_userns, dir, dentry, S_IFLNK|S_IRWXUGO, 0,
523 (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
524 if (unlikely(IS_ERR(inode)))
525 return PTR_ERR(inode);
527 inode_lock(&inode->v);
528 ret = page_symlink(&inode->v, symname, strlen(symname) + 1);
529 inode_unlock(&inode->v);
534 ret = filemap_write_and_wait_range(inode->v.i_mapping, 0, LLONG_MAX);
538 ret = __bch2_link(c, inode, dir, dentry);
542 d_instantiate(dentry, &inode->v);
549 static int bch2_mkdir(struct user_namespace *mnt_userns,
550 struct inode *vdir, struct dentry *dentry, umode_t mode)
552 return bch2_mknod(mnt_userns, vdir, dentry, mode|S_IFDIR, 0);
555 static int bch2_rename2(struct user_namespace *mnt_userns,
556 struct inode *src_vdir, struct dentry *src_dentry,
557 struct inode *dst_vdir, struct dentry *dst_dentry,
560 struct bch_fs *c = src_vdir->i_sb->s_fs_info;
561 struct bch_inode_info *src_dir = to_bch_ei(src_vdir);
562 struct bch_inode_info *dst_dir = to_bch_ei(dst_vdir);
563 struct bch_inode_info *src_inode = to_bch_ei(src_dentry->d_inode);
564 struct bch_inode_info *dst_inode = to_bch_ei(dst_dentry->d_inode);
565 struct bch_inode_unpacked dst_dir_u, src_dir_u;
566 struct bch_inode_unpacked src_inode_u, dst_inode_u;
567 struct btree_trans trans;
568 enum bch_rename_mode mode = flags & RENAME_EXCHANGE
569 ? BCH_RENAME_EXCHANGE
570 : dst_dentry->d_inode
571 ? BCH_RENAME_OVERWRITE : BCH_RENAME;
574 if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
577 if (mode == BCH_RENAME_OVERWRITE) {
578 ret = filemap_write_and_wait_range(src_inode->v.i_mapping,
584 bch2_trans_init(&trans, c, 8, 2048);
586 bch2_lock_inodes(INODE_UPDATE_LOCK,
592 if (inode_attr_changing(dst_dir, src_inode, Inode_opt_project)) {
593 ret = bch2_fs_quota_transfer(c, src_inode,
596 KEY_TYPE_QUOTA_PREALLOC);
601 if (mode == BCH_RENAME_EXCHANGE &&
602 inode_attr_changing(src_dir, dst_inode, Inode_opt_project)) {
603 ret = bch2_fs_quota_transfer(c, dst_inode,
606 KEY_TYPE_QUOTA_PREALLOC);
611 ret = __bch2_trans_do(&trans, NULL, NULL, 0,
612 bch2_rename_trans(&trans,
613 inode_inum(src_dir), &src_dir_u,
614 inode_inum(dst_dir), &dst_dir_u,
623 BUG_ON(src_inode->v.i_ino != src_inode_u.bi_inum);
625 dst_inode->v.i_ino != dst_inode_u.bi_inum);
627 bch2_inode_update_after_write(&trans, src_dir, &src_dir_u,
628 ATTR_MTIME|ATTR_CTIME);
630 if (src_dir != dst_dir)
631 bch2_inode_update_after_write(&trans, dst_dir, &dst_dir_u,
632 ATTR_MTIME|ATTR_CTIME);
634 bch2_inode_update_after_write(&trans, src_inode, &src_inode_u,
638 bch2_inode_update_after_write(&trans, dst_inode, &dst_inode_u,
641 bch2_trans_exit(&trans);
643 bch2_fs_quota_transfer(c, src_inode,
644 bch_qid(&src_inode->ei_inode),
646 KEY_TYPE_QUOTA_NOCHECK);
648 bch2_fs_quota_transfer(c, dst_inode,
649 bch_qid(&dst_inode->ei_inode),
651 KEY_TYPE_QUOTA_NOCHECK);
653 bch2_unlock_inodes(INODE_UPDATE_LOCK,
662 static void bch2_setattr_copy(struct user_namespace *mnt_userns,
663 struct bch_inode_info *inode,
664 struct bch_inode_unpacked *bi,
667 struct bch_fs *c = inode->v.i_sb->s_fs_info;
668 unsigned int ia_valid = attr->ia_valid;
670 if (ia_valid & ATTR_UID)
671 bi->bi_uid = from_kuid(mnt_userns, attr->ia_uid);
672 if (ia_valid & ATTR_GID)
673 bi->bi_gid = from_kgid(mnt_userns, attr->ia_gid);
675 if (ia_valid & ATTR_SIZE)
676 bi->bi_size = attr->ia_size;
678 if (ia_valid & ATTR_ATIME)
679 bi->bi_atime = timespec_to_bch2_time(c, attr->ia_atime);
680 if (ia_valid & ATTR_MTIME)
681 bi->bi_mtime = timespec_to_bch2_time(c, attr->ia_mtime);
682 if (ia_valid & ATTR_CTIME)
683 bi->bi_ctime = timespec_to_bch2_time(c, attr->ia_ctime);
685 if (ia_valid & ATTR_MODE) {
686 umode_t mode = attr->ia_mode;
687 kgid_t gid = ia_valid & ATTR_GID
691 if (!in_group_p(gid) &&
692 !capable_wrt_inode_uidgid(mnt_userns, &inode->v, CAP_FSETID))
698 int bch2_setattr_nonsize(struct user_namespace *mnt_userns,
699 struct bch_inode_info *inode,
702 struct bch_fs *c = inode->v.i_sb->s_fs_info;
704 struct btree_trans trans;
705 struct btree_iter inode_iter = { NULL };
706 struct bch_inode_unpacked inode_u;
707 struct posix_acl *acl = NULL;
710 mutex_lock(&inode->ei_update_lock);
714 if (attr->ia_valid & ATTR_UID)
715 qid.q[QTYP_USR] = from_kuid(&init_user_ns, attr->ia_uid);
717 if (attr->ia_valid & ATTR_GID)
718 qid.q[QTYP_GRP] = from_kgid(&init_user_ns, attr->ia_gid);
720 ret = bch2_fs_quota_transfer(c, inode, qid, ~0,
721 KEY_TYPE_QUOTA_PREALLOC);
725 bch2_trans_init(&trans, c, 0, 0);
727 bch2_trans_begin(&trans);
731 ret = bch2_inode_peek(&trans, &inode_iter, &inode_u, inode_inum(inode),
736 bch2_setattr_copy(mnt_userns, inode, &inode_u, attr);
738 if (attr->ia_valid & ATTR_MODE) {
739 ret = bch2_acl_chmod(&trans, inode_inum(inode), &inode_u,
740 inode_u.bi_mode, &acl);
745 ret = bch2_inode_write(&trans, &inode_iter, &inode_u) ?:
746 bch2_trans_commit(&trans, NULL, NULL,
747 BTREE_INSERT_NOFAIL);
749 bch2_trans_iter_exit(&trans, &inode_iter);
756 bch2_inode_update_after_write(&trans, inode, &inode_u, attr->ia_valid);
759 set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
761 bch2_trans_exit(&trans);
763 mutex_unlock(&inode->ei_update_lock);
768 static int bch2_getattr(struct user_namespace *mnt_userns,
769 const struct path *path, struct kstat *stat,
770 u32 request_mask, unsigned query_flags)
772 struct bch_inode_info *inode = to_bch_ei(d_inode(path->dentry));
773 struct bch_fs *c = inode->v.i_sb->s_fs_info;
775 stat->dev = inode->v.i_sb->s_dev;
776 stat->ino = inode->v.i_ino;
777 stat->mode = inode->v.i_mode;
778 stat->nlink = inode->v.i_nlink;
779 stat->uid = inode->v.i_uid;
780 stat->gid = inode->v.i_gid;
781 stat->rdev = inode->v.i_rdev;
782 stat->size = i_size_read(&inode->v);
783 stat->atime = inode->v.i_atime;
784 stat->mtime = inode->v.i_mtime;
785 stat->ctime = inode->v.i_ctime;
786 stat->blksize = block_bytes(c);
787 stat->blocks = inode->v.i_blocks;
789 if (request_mask & STATX_BTIME) {
790 stat->result_mask |= STATX_BTIME;
791 stat->btime = bch2_time_to_timespec(c, inode->ei_inode.bi_otime);
794 if (inode->ei_inode.bi_flags & BCH_INODE_IMMUTABLE)
795 stat->attributes |= STATX_ATTR_IMMUTABLE;
796 stat->attributes_mask |= STATX_ATTR_IMMUTABLE;
798 if (inode->ei_inode.bi_flags & BCH_INODE_APPEND)
799 stat->attributes |= STATX_ATTR_APPEND;
800 stat->attributes_mask |= STATX_ATTR_APPEND;
802 if (inode->ei_inode.bi_flags & BCH_INODE_NODUMP)
803 stat->attributes |= STATX_ATTR_NODUMP;
804 stat->attributes_mask |= STATX_ATTR_NODUMP;
809 static int bch2_setattr(struct user_namespace *mnt_userns,
810 struct dentry *dentry, struct iattr *iattr)
812 struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
815 lockdep_assert_held(&inode->v.i_rwsem);
817 ret = setattr_prepare(mnt_userns, dentry, iattr);
821 return iattr->ia_valid & ATTR_SIZE
822 ? bch2_truncate(mnt_userns, inode, iattr)
823 : bch2_setattr_nonsize(mnt_userns, inode, iattr);
826 static int bch2_tmpfile(struct user_namespace *mnt_userns,
827 struct inode *vdir, struct dentry *dentry, umode_t mode)
829 struct bch_inode_info *inode =
830 __bch2_create(mnt_userns, to_bch_ei(vdir), dentry, mode, 0,
831 (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
834 return PTR_ERR(inode);
836 d_mark_tmpfile(dentry, &inode->v);
837 d_instantiate(dentry, &inode->v);
841 static int bch2_fill_extent(struct bch_fs *c,
842 struct fiemap_extent_info *info,
843 struct bkey_s_c k, unsigned flags)
845 if (bkey_extent_is_direct_data(k.k)) {
846 struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
847 const union bch_extent_entry *entry;
848 struct extent_ptr_decoded p;
851 if (k.k->type == KEY_TYPE_reflink_v)
852 flags |= FIEMAP_EXTENT_SHARED;
854 bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
856 u64 offset = p.ptr.offset;
858 if (p.crc.compression_type)
859 flags2 |= FIEMAP_EXTENT_ENCODED;
861 offset += p.crc.offset;
863 if ((offset & (c->opts.block_size - 1)) ||
864 (k.k->size & (c->opts.block_size - 1)))
865 flags2 |= FIEMAP_EXTENT_NOT_ALIGNED;
867 ret = fiemap_fill_next_extent(info,
868 bkey_start_offset(k.k) << 9,
870 k.k->size << 9, flags|flags2);
876 } else if (bkey_extent_is_inline_data(k.k)) {
877 return fiemap_fill_next_extent(info,
878 bkey_start_offset(k.k) << 9,
881 FIEMAP_EXTENT_DATA_INLINE);
882 } else if (k.k->type == KEY_TYPE_reservation) {
883 return fiemap_fill_next_extent(info,
884 bkey_start_offset(k.k) << 9,
887 FIEMAP_EXTENT_DELALLOC|
888 FIEMAP_EXTENT_UNWRITTEN);
894 static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
897 struct bch_fs *c = vinode->i_sb->s_fs_info;
898 struct bch_inode_info *ei = to_bch_ei(vinode);
899 struct btree_trans trans;
900 struct btree_iter iter;
902 struct bkey_buf cur, prev;
903 struct bpos end = POS(ei->v.i_ino, (start + len) >> 9);
904 unsigned offset_into_extent, sectors;
905 bool have_extent = false;
909 ret = fiemap_prep(&ei->v, info, start, &len, FIEMAP_FLAG_SYNC);
913 if (start + len < start)
918 bch2_bkey_buf_init(&cur);
919 bch2_bkey_buf_init(&prev);
920 bch2_trans_init(&trans, c, 0, 0);
922 bch2_trans_begin(&trans);
924 ret = bch2_subvolume_get_snapshot(&trans, ei->ei_subvol, &snapshot);
928 bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
929 SPOS(ei->v.i_ino, start, snapshot), 0);
931 while ((k = bch2_btree_iter_peek(&iter)).k &&
932 !(ret = bkey_err(k)) &&
933 bkey_cmp(iter.pos, end) < 0) {
934 enum btree_id data_btree = BTREE_ID_extents;
936 if (!bkey_extent_is_data(k.k) &&
937 k.k->type != KEY_TYPE_reservation) {
938 bch2_btree_iter_advance(&iter);
942 offset_into_extent = iter.pos.offset -
943 bkey_start_offset(k.k);
944 sectors = k.k->size - offset_into_extent;
946 bch2_bkey_buf_reassemble(&cur, c, k);
948 ret = bch2_read_indirect_extent(&trans, &data_btree,
949 &offset_into_extent, &cur);
953 k = bkey_i_to_s_c(cur.k);
954 bch2_bkey_buf_realloc(&prev, c, k.k->u64s);
956 sectors = min(sectors, k.k->size - offset_into_extent);
958 bch2_cut_front(POS(k.k->p.inode,
959 bkey_start_offset(k.k) +
962 bch2_key_resize(&cur.k->k, sectors);
963 cur.k->k.p = iter.pos;
964 cur.k->k.p.offset += cur.k->k.size;
967 ret = bch2_fill_extent(c, info,
968 bkey_i_to_s_c(prev.k), 0);
973 bkey_copy(prev.k, cur.k);
976 bch2_btree_iter_set_pos(&iter,
977 POS(iter.pos.inode, iter.pos.offset + sectors));
979 start = iter.pos.offset;
980 bch2_trans_iter_exit(&trans, &iter);
985 if (!ret && have_extent)
986 ret = bch2_fill_extent(c, info, bkey_i_to_s_c(prev.k),
989 bch2_trans_exit(&trans);
990 bch2_bkey_buf_exit(&cur, c);
991 bch2_bkey_buf_exit(&prev, c);
992 return ret < 0 ? ret : 0;
995 static const struct vm_operations_struct bch_vm_ops = {
996 .fault = bch2_page_fault,
997 .map_pages = filemap_map_pages,
998 .page_mkwrite = bch2_page_mkwrite,
1001 static int bch2_mmap(struct file *file, struct vm_area_struct *vma)
1003 file_accessed(file);
1005 vma->vm_ops = &bch_vm_ops;
1011 static loff_t bch2_dir_llseek(struct file *file, loff_t offset, int whence)
1013 return generic_file_llseek_size(file, offset, whence,
1017 static int bch2_vfs_readdir(struct file *file, struct dir_context *ctx)
1019 struct bch_inode_info *inode = file_bch_inode(file);
1020 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1022 if (!dir_emit_dots(file, ctx))
1025 return bch2_readdir(c, inode_inum(inode), ctx);
1028 static const struct file_operations bch_file_operations = {
1029 .llseek = bch2_llseek,
1030 .read_iter = bch2_read_iter,
1031 .write_iter = bch2_write_iter,
1033 .open = generic_file_open,
1034 .fsync = bch2_fsync,
1035 .splice_read = generic_file_splice_read,
1036 .splice_write = iter_file_splice_write,
1037 .fallocate = bch2_fallocate_dispatch,
1038 .unlocked_ioctl = bch2_fs_file_ioctl,
1039 #ifdef CONFIG_COMPAT
1040 .compat_ioctl = bch2_compat_fs_ioctl,
1042 .remap_file_range = bch2_remap_file_range,
1045 static const struct inode_operations bch_file_inode_operations = {
1046 .getattr = bch2_getattr,
1047 .setattr = bch2_setattr,
1048 .fiemap = bch2_fiemap,
1049 .listxattr = bch2_xattr_list,
1050 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1051 .get_acl = bch2_get_acl,
1052 .set_acl = bch2_set_acl,
1056 static const struct inode_operations bch_dir_inode_operations = {
1057 .lookup = bch2_lookup,
1058 .create = bch2_create,
1060 .unlink = bch2_unlink,
1061 .symlink = bch2_symlink,
1062 .mkdir = bch2_mkdir,
1063 .rmdir = bch2_unlink,
1064 .mknod = bch2_mknod,
1065 .rename = bch2_rename2,
1066 .getattr = bch2_getattr,
1067 .setattr = bch2_setattr,
1068 .tmpfile = bch2_tmpfile,
1069 .listxattr = bch2_xattr_list,
1070 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1071 .get_acl = bch2_get_acl,
1072 .set_acl = bch2_set_acl,
1076 static const struct file_operations bch_dir_file_operations = {
1077 .llseek = bch2_dir_llseek,
1078 .read = generic_read_dir,
1079 .iterate_shared = bch2_vfs_readdir,
1080 .fsync = bch2_fsync,
1081 .unlocked_ioctl = bch2_fs_file_ioctl,
1082 #ifdef CONFIG_COMPAT
1083 .compat_ioctl = bch2_compat_fs_ioctl,
1087 static const struct inode_operations bch_symlink_inode_operations = {
1088 .get_link = page_get_link,
1089 .getattr = bch2_getattr,
1090 .setattr = bch2_setattr,
1091 .listxattr = bch2_xattr_list,
1092 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1093 .get_acl = bch2_get_acl,
1094 .set_acl = bch2_set_acl,
1098 static const struct inode_operations bch_special_inode_operations = {
1099 .getattr = bch2_getattr,
1100 .setattr = bch2_setattr,
1101 .listxattr = bch2_xattr_list,
1102 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1103 .get_acl = bch2_get_acl,
1104 .set_acl = bch2_set_acl,
1108 static const struct address_space_operations bch_address_space_operations = {
1109 .writepage = bch2_writepage,
1110 .readpage = bch2_readpage,
1111 .writepages = bch2_writepages,
1112 .readahead = bch2_readahead,
1113 .set_page_dirty = __set_page_dirty_nobuffers,
1114 .write_begin = bch2_write_begin,
1115 .write_end = bch2_write_end,
1116 .invalidatepage = bch2_invalidatepage,
1117 .releasepage = bch2_releasepage,
1118 .direct_IO = noop_direct_IO,
1119 #ifdef CONFIG_MIGRATION
1120 .migratepage = bch2_migrate_page,
1122 .error_remove_page = generic_error_remove_page,
1125 struct bcachefs_fid {
1131 struct bcachefs_fid_with_parent {
1132 struct bcachefs_fid fid;
1133 struct bcachefs_fid dir;
1136 static int bcachefs_fid_valid(int fh_len, int fh_type)
1139 case FILEID_BCACHEFS_WITHOUT_PARENT:
1140 return fh_len == sizeof(struct bcachefs_fid) / sizeof(u32);
1141 case FILEID_BCACHEFS_WITH_PARENT:
1142 return fh_len == sizeof(struct bcachefs_fid_with_parent) / sizeof(u32);
1148 static struct bcachefs_fid bch2_inode_to_fid(struct bch_inode_info *inode)
1150 return (struct bcachefs_fid) {
1151 .inum = inode->ei_inode.bi_inum,
1152 .subvol = inode->ei_subvol,
1153 .gen = inode->ei_inode.bi_generation,
1157 static int bch2_encode_fh(struct inode *vinode, u32 *fh, int *len,
1160 struct bch_inode_info *inode = to_bch_ei(vinode);
1161 struct bch_inode_info *dir = to_bch_ei(vdir);
1163 if (*len < sizeof(struct bcachefs_fid_with_parent) / sizeof(u32))
1164 return FILEID_INVALID;
1166 if (!S_ISDIR(inode->v.i_mode) && dir) {
1167 struct bcachefs_fid_with_parent *fid = (void *) fh;
1169 fid->fid = bch2_inode_to_fid(inode);
1170 fid->dir = bch2_inode_to_fid(dir);
1172 *len = sizeof(*fid) / sizeof(u32);
1173 return FILEID_BCACHEFS_WITH_PARENT;
1175 struct bcachefs_fid *fid = (void *) fh;
1177 *fid = bch2_inode_to_fid(inode);
1179 *len = sizeof(*fid) / sizeof(u32);
1180 return FILEID_BCACHEFS_WITHOUT_PARENT;
1184 static struct inode *bch2_nfs_get_inode(struct super_block *sb,
1185 struct bcachefs_fid fid)
1187 struct bch_fs *c = sb->s_fs_info;
1188 struct inode *vinode = bch2_vfs_inode_get(c, (subvol_inum) {
1189 .subvol = fid.subvol,
1192 if (!IS_ERR(vinode) && vinode->i_generation != fid.gen) {
1194 vinode = ERR_PTR(-ESTALE);
1199 static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *_fid,
1200 int fh_len, int fh_type)
1202 struct bcachefs_fid *fid = (void *) _fid;
1204 if (!bcachefs_fid_valid(fh_len, fh_type))
1207 return d_obtain_alias(bch2_nfs_get_inode(sb, *fid));
1210 static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *_fid,
1211 int fh_len, int fh_type)
1213 struct bcachefs_fid_with_parent *fid = (void *) _fid;
1215 if (!bcachefs_fid_valid(fh_len, fh_type) ||
1216 fh_type != FILEID_BCACHEFS_WITH_PARENT)
1219 return d_obtain_alias(bch2_nfs_get_inode(sb, fid->dir));
1222 static struct dentry *bch2_get_parent(struct dentry *child)
1224 struct bch_inode_info *inode = to_bch_ei(child->d_inode);
1225 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1226 subvol_inum parent_inum = {
1227 .subvol = inode->ei_inode.bi_parent_subvol ?:
1229 .inum = inode->ei_inode.bi_dir,
1232 if (!parent_inum.inum)
1235 return d_obtain_alias(bch2_vfs_inode_get(c, parent_inum));
1238 static int bch2_get_name(struct dentry *parent, char *name, struct dentry *child)
1240 struct bch_inode_info *inode = to_bch_ei(child->d_inode);
1241 struct bch_inode_info *dir = to_bch_ei(parent->d_inode);
1242 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1243 struct btree_trans trans;
1244 struct btree_iter iter1;
1245 struct btree_iter iter2;
1247 struct bkey_s_c_dirent d;
1248 struct bch_inode_unpacked inode_u;
1254 if (!S_ISDIR(dir->v.i_mode))
1257 bch2_trans_init(&trans, c, 0, 0);
1259 bch2_trans_iter_init(&trans, &iter1, BTREE_ID_dirents,
1260 POS(dir->ei_inode.bi_inum, 0), 0);
1261 bch2_trans_iter_init(&trans, &iter2, BTREE_ID_dirents,
1262 POS(dir->ei_inode.bi_inum, 0), 0);
1264 bch2_trans_begin(&trans);
1266 ret = bch2_subvolume_get_snapshot(&trans, dir->ei_subvol, &snapshot);
1270 bch2_btree_iter_set_snapshot(&iter1, snapshot);
1271 bch2_btree_iter_set_snapshot(&iter2, snapshot);
1273 ret = bch2_inode_find_by_inum_trans(&trans, inode_inum(inode), &inode_u);
1277 if (inode_u.bi_dir == dir->ei_inode.bi_inum) {
1278 bch2_btree_iter_set_pos(&iter1, POS(inode_u.bi_dir, inode_u.bi_dir_offset));
1280 k = bch2_btree_iter_peek_slot(&iter1);
1285 if (k.k->type != KEY_TYPE_dirent) {
1290 d = bkey_s_c_to_dirent(k);
1291 ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
1297 if (target.subvol == inode->ei_subvol &&
1298 target.inum == inode->ei_inode.bi_inum)
1302 * File with multiple hardlinks and our backref is to the wrong
1303 * directory - linear search:
1305 for_each_btree_key_continue_norestart(iter2, 0, k, ret) {
1306 if (k.k->p.inode > dir->ei_inode.bi_inum)
1309 if (k.k->type != KEY_TYPE_dirent)
1312 d = bkey_s_c_to_dirent(k);
1313 ret = bch2_dirent_read_target(&trans, inode_inum(dir), d, &target);
1319 if (target.subvol == inode->ei_subvol &&
1320 target.inum == inode->ei_inode.bi_inum)
1328 name_len = min_t(unsigned, bch2_dirent_name_bytes(d), NAME_MAX);
1330 memcpy(name, d.v->d_name, name_len);
1331 name[name_len] = '\0';
1336 bch2_trans_iter_exit(&trans, &iter1);
1337 bch2_trans_iter_exit(&trans, &iter2);
1338 bch2_trans_exit(&trans);
1343 static const struct export_operations bch_export_ops = {
1344 .encode_fh = bch2_encode_fh,
1345 .fh_to_dentry = bch2_fh_to_dentry,
1346 .fh_to_parent = bch2_fh_to_parent,
1347 .get_parent = bch2_get_parent,
1348 .get_name = bch2_get_name,
1351 static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
1352 struct bch_inode_info *inode,
1353 struct bch_inode_unpacked *bi)
1355 bch2_inode_update_after_write(trans, inode, bi, ~0);
1357 inode->v.i_blocks = bi->bi_sectors;
1358 inode->v.i_ino = bi->bi_inum;
1359 inode->v.i_rdev = bi->bi_dev;
1360 inode->v.i_generation = bi->bi_generation;
1361 inode->v.i_size = bi->bi_size;
1363 inode->ei_flags = 0;
1364 inode->ei_quota_reserved = 0;
1365 inode->ei_qid = bch_qid(bi);
1366 inode->ei_subvol = inum.subvol;
1368 inode->v.i_mapping->a_ops = &bch_address_space_operations;
1370 switch (inode->v.i_mode & S_IFMT) {
1372 inode->v.i_op = &bch_file_inode_operations;
1373 inode->v.i_fop = &bch_file_operations;
1376 inode->v.i_op = &bch_dir_inode_operations;
1377 inode->v.i_fop = &bch_dir_file_operations;
1380 inode_nohighmem(&inode->v);
1381 inode->v.i_op = &bch_symlink_inode_operations;
1384 init_special_inode(&inode->v, inode->v.i_mode, inode->v.i_rdev);
1385 inode->v.i_op = &bch_special_inode_operations;
1390 static struct inode *bch2_alloc_inode(struct super_block *sb)
1392 struct bch_inode_info *inode;
1394 inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
1398 inode_init_once(&inode->v);
1399 mutex_init(&inode->ei_update_lock);
1400 pagecache_lock_init(&inode->ei_pagecache_lock);
1401 mutex_init(&inode->ei_quota_lock);
1406 static void bch2_i_callback(struct rcu_head *head)
1408 struct inode *vinode = container_of(head, struct inode, i_rcu);
1409 struct bch_inode_info *inode = to_bch_ei(vinode);
1411 kmem_cache_free(bch2_inode_cache, inode);
1414 static void bch2_destroy_inode(struct inode *vinode)
1416 call_rcu(&vinode->i_rcu, bch2_i_callback);
1419 static int inode_update_times_fn(struct bch_inode_info *inode,
1420 struct bch_inode_unpacked *bi,
1423 struct bch_fs *c = inode->v.i_sb->s_fs_info;
1425 bi->bi_atime = timespec_to_bch2_time(c, inode->v.i_atime);
1426 bi->bi_mtime = timespec_to_bch2_time(c, inode->v.i_mtime);
1427 bi->bi_ctime = timespec_to_bch2_time(c, inode->v.i_ctime);
1432 static int bch2_vfs_write_inode(struct inode *vinode,
1433 struct writeback_control *wbc)
1435 struct bch_fs *c = vinode->i_sb->s_fs_info;
1436 struct bch_inode_info *inode = to_bch_ei(vinode);
1439 mutex_lock(&inode->ei_update_lock);
1440 ret = bch2_write_inode(c, inode, inode_update_times_fn, NULL,
1441 ATTR_ATIME|ATTR_MTIME|ATTR_CTIME);
1442 mutex_unlock(&inode->ei_update_lock);
1447 static void bch2_evict_inode(struct inode *vinode)
1449 struct bch_fs *c = vinode->i_sb->s_fs_info;
1450 struct bch_inode_info *inode = to_bch_ei(vinode);
1452 truncate_inode_pages_final(&inode->v.i_data);
1454 clear_inode(&inode->v);
1456 BUG_ON(!is_bad_inode(&inode->v) && inode->ei_quota_reserved);
1458 if (!inode->v.i_nlink && !is_bad_inode(&inode->v)) {
1459 bch2_quota_acct(c, inode->ei_qid, Q_SPC, -((s64) inode->v.i_blocks),
1460 KEY_TYPE_QUOTA_WARN);
1461 bch2_quota_acct(c, inode->ei_qid, Q_INO, -1,
1462 KEY_TYPE_QUOTA_WARN);
1463 bch2_inode_rm(c, inode_inum(inode), true);
1467 void bch2_evict_subvolume_inodes(struct bch_fs *c,
1468 struct snapshot_id_list *s)
1470 struct super_block *sb = c->vfs_sb;
1471 struct inode *inode;
1473 spin_lock(&sb->s_inode_list_lock);
1474 list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
1475 if (!snapshot_list_has_id(s, to_bch_ei(inode)->ei_subvol) ||
1476 (inode->i_state & I_FREEING))
1479 d_mark_dontcache(inode);
1480 d_prune_aliases(inode);
1482 spin_unlock(&sb->s_inode_list_lock);
1485 spin_lock(&sb->s_inode_list_lock);
1486 list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
1487 if (!snapshot_list_has_id(s, to_bch_ei(inode)->ei_subvol) ||
1488 (inode->i_state & I_FREEING))
1491 if (!(inode->i_state & I_DONTCACHE)) {
1492 d_mark_dontcache(inode);
1493 d_prune_aliases(inode);
1496 spin_lock(&inode->i_lock);
1497 if (snapshot_list_has_id(s, to_bch_ei(inode)->ei_subvol) &&
1498 !(inode->i_state & I_FREEING)) {
1499 wait_queue_head_t *wq = bit_waitqueue(&inode->i_state, __I_NEW);
1500 DEFINE_WAIT_BIT(wait, &inode->i_state, __I_NEW);
1501 prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
1502 spin_unlock(&inode->i_lock);
1503 spin_unlock(&sb->s_inode_list_lock);
1505 finish_wait(wq, &wait.wq_entry);
1509 spin_unlock(&inode->i_lock);
1511 spin_unlock(&sb->s_inode_list_lock);
1514 static int bch2_statfs(struct dentry *dentry, struct kstatfs *buf)
1516 struct super_block *sb = dentry->d_sb;
1517 struct bch_fs *c = sb->s_fs_info;
1518 struct bch_fs_usage_short usage = bch2_fs_usage_read_short(c);
1519 unsigned shift = sb->s_blocksize_bits - 9;
1521 * this assumes inodes take up 64 bytes, which is a decent average
1524 u64 avail_inodes = ((usage.capacity - usage.used) << 3);
1527 buf->f_type = BCACHEFS_STATFS_MAGIC;
1528 buf->f_bsize = sb->s_blocksize;
1529 buf->f_blocks = usage.capacity >> shift;
1530 buf->f_bfree = usage.free >> shift;
1531 buf->f_bavail = avail_factor(usage.free) >> shift;
1533 buf->f_files = usage.nr_inodes + avail_inodes;
1534 buf->f_ffree = avail_inodes;
1536 fsid = le64_to_cpup((void *) c->sb.user_uuid.b) ^
1537 le64_to_cpup((void *) c->sb.user_uuid.b + sizeof(u64));
1538 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
1539 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
1540 buf->f_namelen = BCH_NAME_MAX;
1545 static int bch2_sync_fs(struct super_block *sb, int wait)
1547 struct bch_fs *c = sb->s_fs_info;
1549 if (c->opts.journal_flush_disabled)
1553 bch2_journal_flush_async(&c->journal, NULL);
1557 return bch2_journal_flush(&c->journal);
1560 static struct bch_fs *bch2_path_to_fs(const char *path)
1566 ret = lookup_bdev(path, &dev);
1568 return ERR_PTR(ret);
1570 c = bch2_dev_to_fs(dev);
1572 closure_put(&c->cl);
1573 return c ?: ERR_PTR(-ENOENT);
1576 static char **split_devs(const char *_dev_name, unsigned *nr)
1578 char *dev_name = NULL, **devs = NULL, *s;
1579 size_t i, nr_devs = 0;
1581 dev_name = kstrdup(_dev_name, GFP_KERNEL);
1585 for (s = dev_name; s; s = strchr(s + 1, ':'))
1588 devs = kcalloc(nr_devs + 1, sizeof(const char *), GFP_KERNEL);
1594 for (i = 0, s = dev_name;
1596 (s = strchr(s, ':')) && (*s++ = '\0'))
1603 static int bch2_remount(struct super_block *sb, int *flags, char *data)
1605 struct bch_fs *c = sb->s_fs_info;
1606 struct bch_opts opts = bch2_opts_empty();
1609 opt_set(opts, read_only, (*flags & SB_RDONLY) != 0);
1611 ret = bch2_parse_mount_opts(c, &opts, data);
1615 if (opts.read_only != c->opts.read_only) {
1616 down_write(&c->state_lock);
1618 if (opts.read_only) {
1619 bch2_fs_read_only(c);
1621 sb->s_flags |= SB_RDONLY;
1623 ret = bch2_fs_read_write(c);
1625 bch_err(c, "error going rw: %i", ret);
1626 up_write(&c->state_lock);
1630 sb->s_flags &= ~SB_RDONLY;
1633 c->opts.read_only = opts.read_only;
1635 up_write(&c->state_lock);
1638 if (opts.errors >= 0)
1639 c->opts.errors = opts.errors;
1644 static int bch2_show_devname(struct seq_file *seq, struct dentry *root)
1646 struct bch_fs *c = root->d_sb->s_fs_info;
1651 for_each_online_member(ca, c, i) {
1655 seq_puts(seq, "/dev/");
1656 seq_puts(seq, ca->name);
1662 static int bch2_show_options(struct seq_file *seq, struct dentry *root)
1664 struct bch_fs *c = root->d_sb->s_fs_info;
1668 for (i = 0; i < bch2_opts_nr; i++) {
1669 const struct bch_option *opt = &bch2_opt_table[i];
1670 u64 v = bch2_opt_get_by_id(&c->opts, i);
1672 if (!(opt->mode & OPT_MOUNT))
1675 if (v == bch2_opt_get_by_id(&bch2_opts_default, i))
1678 bch2_opt_to_text(&PBUF(buf), c, opt, v,
1679 OPT_SHOW_MOUNT_STYLE);
1687 static void bch2_put_super(struct super_block *sb)
1689 struct bch_fs *c = sb->s_fs_info;
1694 static const struct super_operations bch_super_operations = {
1695 .alloc_inode = bch2_alloc_inode,
1696 .destroy_inode = bch2_destroy_inode,
1697 .write_inode = bch2_vfs_write_inode,
1698 .evict_inode = bch2_evict_inode,
1699 .sync_fs = bch2_sync_fs,
1700 .statfs = bch2_statfs,
1701 .show_devname = bch2_show_devname,
1702 .show_options = bch2_show_options,
1703 .remount_fs = bch2_remount,
1704 .put_super = bch2_put_super,
1706 .freeze_fs = bch2_freeze,
1707 .unfreeze_fs = bch2_unfreeze,
1711 static int bch2_set_super(struct super_block *s, void *data)
1713 s->s_fs_info = data;
1717 static int bch2_noset_super(struct super_block *s, void *data)
1722 static int bch2_test_super(struct super_block *s, void *data)
1724 struct bch_fs *c = s->s_fs_info;
1725 struct bch_fs **devs = data;
1731 for (i = 0; devs[i]; i++)
1737 static struct dentry *bch2_mount(struct file_system_type *fs_type,
1738 int flags, const char *dev_name, void *data)
1742 struct super_block *sb;
1743 struct inode *vinode;
1744 struct bch_opts opts = bch2_opts_empty();
1746 struct bch_fs **devs_to_fs = NULL;
1747 unsigned i, nr_devs;
1750 opt_set(opts, read_only, (flags & SB_RDONLY) != 0);
1752 ret = bch2_parse_mount_opts(NULL, &opts, data);
1754 return ERR_PTR(ret);
1756 if (!dev_name || strlen(dev_name) == 0)
1757 return ERR_PTR(-EINVAL);
1759 devs = split_devs(dev_name, &nr_devs);
1761 return ERR_PTR(-ENOMEM);
1763 devs_to_fs = kcalloc(nr_devs + 1, sizeof(void *), GFP_KERNEL);
1765 sb = ERR_PTR(-ENOMEM);
1769 for (i = 0; i < nr_devs; i++)
1770 devs_to_fs[i] = bch2_path_to_fs(devs[i]);
1772 sb = sget(fs_type, bch2_test_super, bch2_noset_super,
1773 flags|SB_NOSEC, devs_to_fs);
1777 c = bch2_fs_open(devs, nr_devs, opts);
1783 /* Some options can't be parsed until after the fs is started: */
1784 ret = bch2_parse_mount_opts(c, &opts, data);
1791 bch2_opts_apply(&c->opts, opts);
1793 sb = sget(fs_type, NULL, bch2_set_super, flags|SB_NOSEC, c);
1802 return ERR_CAST(sb);
1807 if ((flags ^ sb->s_flags) & SB_RDONLY) {
1814 sb->s_blocksize = block_bytes(c);
1815 sb->s_blocksize_bits = ilog2(block_bytes(c));
1816 sb->s_maxbytes = MAX_LFS_FILESIZE;
1817 sb->s_op = &bch_super_operations;
1818 sb->s_export_op = &bch_export_ops;
1819 #ifdef CONFIG_BCACHEFS_QUOTA
1820 sb->s_qcop = &bch2_quotactl_operations;
1821 sb->s_quota_types = QTYPE_MASK_USR|QTYPE_MASK_GRP|QTYPE_MASK_PRJ;
1823 sb->s_xattr = bch2_xattr_handlers;
1824 sb->s_magic = BCACHEFS_STATFS_MAGIC;
1825 sb->s_time_gran = c->sb.nsec_per_time_unit;
1826 sb->s_time_min = div_s64(S64_MIN, c->sb.time_units_per_sec) + 1;
1827 sb->s_time_max = div_s64(S64_MAX, c->sb.time_units_per_sec);
1829 strlcpy(sb->s_id, c->name, sizeof(sb->s_id));
1831 ret = super_setup_bdi(sb);
1835 sb->s_bdi->ra_pages = VM_READAHEAD_PAGES;
1837 for_each_online_member(ca, c, i) {
1838 struct block_device *bdev = ca->disk_sb.bdev;
1840 /* XXX: create an anonymous device for multi device filesystems */
1842 sb->s_dev = bdev->bd_dev;
1843 percpu_ref_put(&ca->io_ref);
1849 #ifdef CONFIG_BCACHEFS_POSIX_ACL
1851 sb->s_flags |= SB_POSIXACL;
1854 sb->s_shrink.seeks = 0;
1856 vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_SUBVOL_INUM);
1857 if (IS_ERR(vinode)) {
1858 bch_err(c, "error mounting: error getting root inode %i",
1859 (int) PTR_ERR(vinode));
1860 ret = PTR_ERR(vinode);
1864 sb->s_root = d_make_root(vinode);
1866 bch_err(c, "error mounting: error allocating root dentry");
1871 sb->s_flags |= SB_ACTIVE;
1873 return dget(sb->s_root);
1876 deactivate_locked_super(sb);
1877 return ERR_PTR(ret);
1880 static void bch2_kill_sb(struct super_block *sb)
1882 struct bch_fs *c = sb->s_fs_info;
1884 generic_shutdown_super(sb);
1888 static struct file_system_type bcache_fs_type = {
1889 .owner = THIS_MODULE,
1891 .mount = bch2_mount,
1892 .kill_sb = bch2_kill_sb,
1893 .fs_flags = FS_REQUIRES_DEV,
1896 MODULE_ALIAS_FS("bcachefs");
1898 void bch2_vfs_exit(void)
1900 unregister_filesystem(&bcache_fs_type);
1901 if (bch2_inode_cache)
1902 kmem_cache_destroy(bch2_inode_cache);
1905 int __init bch2_vfs_init(void)
1909 bch2_inode_cache = KMEM_CACHE(bch_inode_info, 0);
1910 if (!bch2_inode_cache)
1913 ret = register_filesystem(&bcache_fs_type);
1923 #endif /* NO_BCACHEFS_FS */