X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Ffs.c;h=093f5404a655a0b5c7a445f4e99cab851efe5fb8;hb=HEAD;hp=e3edca4d265b60fe094905e3f74b0d415d4fd6a0;hpb=db931a4571817d7d61be6bce306f1d42f7cd3398;p=bcachefs-tools-debian

diff --git a/libbcachefs/fs.c b/libbcachefs/fs.c
index e3edca4..093f540 100644
--- a/libbcachefs/fs.c
+++ b/libbcachefs/fs.c
@@ -3,22 +3,27 @@
 
 #include "bcachefs.h"
 #include "acl.h"
-#include "bkey_on_stack.h"
+#include "bkey_buf.h"
 #include "btree_update.h"
 #include "buckets.h"
 #include "chardev.h"
 #include "dirent.h"
+#include "errcode.h"
 #include "extents.h"
 #include "fs.h"
 #include "fs-common.h"
 #include "fs-io.h"
 #include "fs-ioctl.h"
+#include "fs-io-buffered.h"
+#include "fs-io-direct.h"
+#include "fs-io-pagecache.h"
 #include "fsck.h"
 #include "inode.h"
-#include "io.h"
+#include "io_read.h"
 #include "journal.h"
 #include "keylist.h"
 #include "quota.h"
+#include "snapshot.h"
 #include "super.h"
 #include "xattr.h"
 
@@ -27,106 +32,45 @@
 #include <linux/exportfs.h>
 #include <linux/fiemap.h>
 #include <linux/module.h>
+#include <linux/pagemap.h>
 #include <linux/posix_acl.h>
 #include <linux/random.h>
+#include <linux/seq_file.h>
 #include <linux/statfs.h>
+#include <linux/string.h>
 #include <linux/xattr.h>
 
 static struct kmem_cache *bch2_inode_cache;
 
-static void bch2_vfs_inode_init(struct bch_fs *,
+static void bch2_vfs_inode_init(struct btree_trans *, subvol_inum,
 				struct bch_inode_info *,
-				struct bch_inode_unpacked *);
+				struct bch_inode_unpacked *,
+				struct bch_subvolume *);
 
-static void journal_seq_copy(struct bch_fs *c,
-			     struct bch_inode_info *dst,
-			     u64 journal_seq)
-{
-	/*
-	 * atomic64_cmpxchg has a fallback for archs that don't support it,
-	 * cmpxchg does not:
-	 */
-	atomic64_t *dst_seq = (void *) &dst->ei_journal_seq;
-	u64 old, v = READ_ONCE(dst->ei_journal_seq);
-
-	do {
-		old = v;
-
-		if (old >= journal_seq)
-			break;
-	} while ((v = atomic64_cmpxchg(dst_seq, old, journal_seq)) != old);
-
-	bch2_journal_set_has_inum(&c->journal, dst->v.i_ino, journal_seq);
-}
-
-static void __pagecache_lock_put(struct pagecache_lock *lock, long i)
-{
-	BUG_ON(atomic_long_read(&lock->v) == 0);
-
-	if (atomic_long_sub_return_release(i, &lock->v) == 0)
-		wake_up_all(&lock->wait);
-}
-
-static bool __pagecache_lock_tryget(struct pagecache_lock *lock, long i)
-{
-	long v = atomic_long_read(&lock->v), old;
-
-	do {
-		old = v;
-
-		if (i > 0 ? v < 0 : v > 0)
-			return false;
-	} while ((v = atomic_long_cmpxchg_acquire(&lock->v,
-					old, old + i)) != old);
-	return true;
-}
-
-static void __pagecache_lock_get(struct pagecache_lock *lock, long i)
-{
-	wait_event(lock->wait, __pagecache_lock_tryget(lock, i));
-}
-
-void bch2_pagecache_add_put(struct pagecache_lock *lock)
-{
-	__pagecache_lock_put(lock, 1);
-}
-
-bool bch2_pagecache_add_tryget(struct pagecache_lock *lock)
-{
-	return __pagecache_lock_tryget(lock, 1);
-}
-
-void bch2_pagecache_add_get(struct pagecache_lock *lock)
-{
-	__pagecache_lock_get(lock, 1);
-}
-
-void bch2_pagecache_block_put(struct pagecache_lock *lock)
-{
-	__pagecache_lock_put(lock, -1);
-}
-
-void bch2_pagecache_block_get(struct pagecache_lock *lock)
-{
-	__pagecache_lock_get(lock, -1);
-}
-
-void bch2_inode_update_after_write(struct bch_fs *c,
+void bch2_inode_update_after_write(struct btree_trans *trans,
 				   struct bch_inode_info *inode,
 				   struct bch_inode_unpacked *bi,
 				   unsigned fields)
 {
+	struct bch_fs *c = trans->c;
+
+	BUG_ON(bi->bi_inum != inode->v.i_ino);
+
+	bch2_assert_pos_locked(trans, BTREE_ID_inodes,
+			       POS(0, bi->bi_inum),
+			       c->opts.inodes_use_key_cache);
+
 	set_nlink(&inode->v, bch2_inode_nlink_get(bi));
 	i_uid_write(&inode->v, bi->bi_uid);
 	i_gid_write(&inode->v, bi->bi_gid);
 	inode->v.i_mode	= bi->bi_mode;
 
 	if (fields & ATTR_ATIME)
-		inode->v.i_atime = bch2_time_to_timespec(c, bi->bi_atime);
+		inode_set_atime_to_ts(&inode->v, bch2_time_to_timespec(c, bi->bi_atime));
 	if (fields & ATTR_MTIME)
-		inode->v.i_mtime = bch2_time_to_timespec(c, bi->bi_mtime);
+		inode_set_mtime_to_ts(&inode->v, bch2_time_to_timespec(c, bi->bi_mtime));
 	if (fields & ATTR_CTIME)
-		inode->v.i_ctime = bch2_time_to_timespec(c, bi->bi_ctime);
+		inode_set_ctime_to_ts(&inode->v, bch2_time_to_timespec(c, bi->bi_ctime));
 
 	inode->ei_inode		= *bi;
 
@@ -138,38 +82,37 @@ int __must_check bch2_write_inode(struct bch_fs *c,
 				  inode_set_fn set,
 				  void *p, unsigned fields)
 {
-	struct btree_trans trans;
-	struct btree_iter *iter;
+	struct btree_trans *trans = bch2_trans_get(c);
+	struct btree_iter iter = { NULL };
 	struct bch_inode_unpacked inode_u;
 	int ret;
-
-	bch2_trans_init(&trans, c, 0, 0);
 retry:
-	bch2_trans_begin(&trans);
-
-	iter = bch2_inode_peek(&trans, &inode_u, inode->v.i_ino,
-			       BTREE_ITER_INTENT);
-	ret   = PTR_ERR_OR_ZERO(iter) ?:
-		(set ? set(inode, &inode_u, p) : 0) ?:
-		bch2_inode_write(&trans, iter, &inode_u) ?:
-		bch2_trans_commit(&trans, NULL,
-				  &inode->ei_journal_seq,
-				  BTREE_INSERT_NOUNLOCK|
-				  BTREE_INSERT_NOFAIL);
+	bch2_trans_begin(trans);
+
+	ret   = bch2_inode_peek(trans, &iter, &inode_u, inode_inum(inode),
+				BTREE_ITER_INTENT) ?:
+		(set ? set(trans, inode, &inode_u, p) : 0) ?:
+		bch2_inode_write(trans, &iter, &inode_u) ?:
+		bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
 
 	/*
 	 * the btree node lock protects inode->ei_inode, not ei_update_lock;
 	 * this is important for inode updates via bchfs_write_index_update
 	 */
 	if (!ret)
-		bch2_inode_update_after_write(c, inode, &inode_u, fields);
+		bch2_inode_update_after_write(trans, inode, &inode_u, fields);
 
-	bch2_trans_iter_put(&trans, iter);
+	bch2_trans_iter_exit(trans, &iter);
 
-	if (ret == -EINTR)
+	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		goto retry;
 
-	bch2_trans_exit(&trans);
+	bch2_fs_fatal_err_on(bch2_err_matches(ret, ENOENT), c,
+			     "inode %u:%llu not found when updating",
+			     inode_inum(inode).subvol,
+			     inode_inum(inode).inum);
+
+	bch2_trans_put(trans);
 	return ret < 0 ? ret : 0;
 }
 
@@ -208,51 +151,129 @@ int bch2_fs_quota_transfer(struct bch_fs *c,
 	return ret;
 }
 
-struct inode *bch2_vfs_inode_get(struct bch_fs *c, u64 inum)
+static int bch2_iget5_test(struct inode *vinode, void *p)
 {
-	struct bch_inode_unpacked inode_u;
-	struct bch_inode_info *inode;
-	int ret;
+	struct bch_inode_info *inode = to_bch_ei(vinode);
+	subvol_inum *inum = p;
 
-	inode = to_bch_ei(iget_locked(c->vfs_sb, inum));
-	if (unlikely(!inode))
-		return ERR_PTR(-ENOMEM);
-	if (!(inode->v.i_state & I_NEW))
-		return &inode->v;
+	return inode->ei_subvol == inum->subvol &&
+		inode->ei_inode.bi_inum == inum->inum;
+}
 
-	ret = bch2_inode_find_by_inum(c, inum, &inode_u);
-	if (ret) {
-		iget_failed(&inode->v);
-		return ERR_PTR(ret);
+static int bch2_iget5_set(struct inode *vinode, void *p)
+{
+	struct bch_inode_info *inode = to_bch_ei(vinode);
+	subvol_inum *inum = p;
+
+	inode->v.i_ino		= inum->inum;
+	inode->ei_subvol	= inum->subvol;
+	inode->ei_inode.bi_inum	= inum->inum;
+	return 0;
+}
+
+static unsigned bch2_inode_hash(subvol_inum inum)
+{
+	return jhash_3words(inum.subvol, inum.inum >> 32, inum.inum, JHASH_INITVAL);
+}
+
+static struct bch_inode_info *bch2_inode_insert(struct bch_fs *c, struct bch_inode_info *inode)
+{
+	subvol_inum inum = inode_inum(inode);
+	struct bch_inode_info *old = to_bch_ei(inode_insert5(&inode->v,
+				      bch2_inode_hash(inum),
+				      bch2_iget5_test,
+				      bch2_iget5_set,
+				      &inum));
+	BUG_ON(!old);
+
+	if (unlikely(old != inode)) {
+		discard_new_inode(&inode->v);
+		inode = old;
+	} else {
+		mutex_lock(&c->vfs_inodes_lock);
+		list_add(&inode->ei_vfs_inode_list, &c->vfs_inodes_list);
+		mutex_unlock(&c->vfs_inodes_lock);
+		/*
+		 * we really don't want insert_inode_locked2() to be setting
+		 * I_NEW...
+		 */
+		unlock_new_inode(&inode->v);
 	}
 
-	bch2_vfs_inode_init(c, inode, &inode_u);
+	return inode;
+}
+
+#define memalloc_flags_do(_flags, _do)						\
+({										\
+	unsigned _saved_flags = memalloc_flags_save(_flags);			\
+	typeof(_do) _ret = _do;							\
+	memalloc_noreclaim_restore(_saved_flags);				\
+	_ret;									\
+})
+
+/*
+ * Allocate a new inode, dropping/retaking btree locks if necessary:
+ */
+static struct bch_inode_info *bch2_new_inode(struct btree_trans *trans)
+{
+	struct bch_fs *c = trans->c;
 
-	inode->ei_journal_seq = bch2_inode_journal_seq(&c->journal, inum);
+	struct bch_inode_info *inode =
+		memalloc_flags_do(PF_MEMALLOC_NORECLAIM|PF_MEMALLOC_NOWARN,
+				  to_bch_ei(new_inode(c->vfs_sb)));
 
-	unlock_new_inode(&inode->v);
+	if (unlikely(!inode)) {
+		int ret = drop_locks_do(trans, (inode = to_bch_ei(new_inode(c->vfs_sb))) ? 0 : -ENOMEM);
+		if (ret && inode)
+			discard_new_inode(&inode->v);
+		if (ret)
+			return ERR_PTR(ret);
+	}
 
-	return &inode->v;
+	return inode;
 }
 
-static int inum_test(struct inode *inode, void *p)
+struct inode *bch2_vfs_inode_get(struct bch_fs *c, subvol_inum inum)
 {
-	unsigned long *ino = p;
+	struct bch_inode_info *inode =
+		to_bch_ei(ilookup5_nowait(c->vfs_sb,
+					  bch2_inode_hash(inum),
+					  bch2_iget5_test,
+					  &inum));
+	if (inode)
+		return &inode->v;
+
+	struct btree_trans *trans = bch2_trans_get(c);
+
+	struct bch_inode_unpacked inode_u;
+	struct bch_subvolume subvol;
+	int ret = lockrestart_do(trans,
+		bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
+		bch2_inode_find_by_inum_trans(trans, inum, &inode_u)) ?:
+		PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
+	if (!ret) {
+		bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
+		inode = bch2_inode_insert(c, inode);
+	}
+	bch2_trans_put(trans);
 
-	return *ino == inode->i_ino;
+	return ret ? ERR_PTR(ret) : &inode->v;
 }
 
-static struct bch_inode_info *
-__bch2_create(struct bch_inode_info *dir, struct dentry *dentry,
-	      umode_t mode, dev_t rdev, bool tmpfile)
+struct bch_inode_info *
+__bch2_create(struct mnt_idmap *idmap,
+	      struct bch_inode_info *dir, struct dentry *dentry,
+	      umode_t mode, dev_t rdev, subvol_inum snapshot_src,
+	      unsigned flags)
 {
 	struct bch_fs *c = dir->v.i_sb->s_fs_info;
-	struct user_namespace *ns = dir->v.i_sb->s_user_ns;
-	struct btree_trans trans;
+	struct btree_trans *trans;
 	struct bch_inode_unpacked dir_u;
-	struct bch_inode_info *inode, *old;
+	struct bch_inode_info *inode;
 	struct bch_inode_unpacked inode_u;
 	struct posix_acl *default_acl = NULL, *acl = NULL;
+	subvol_inum inum;
+	struct bch_subvolume subvol;
 	u64 journal_seq = 0;
 	int ret;
 
@@ -273,45 +294,49 @@ __bch2_create(struct bch_inode_info *dir, struct dentry *dentry,
 
 	bch2_inode_init_early(c, &inode_u);
 
-	if (!tmpfile)
+	if (!(flags & BCH_CREATE_TMPFILE))
 		mutex_lock(&dir->ei_update_lock);
 
-	bch2_trans_init(&trans, c, 8,
-			2048 + (!tmpfile ? dentry->d_name.len : 0));
+	trans = bch2_trans_get(c);
 retry:
-	bch2_trans_begin(&trans);
-
-	ret   = bch2_create_trans(&trans, dir->v.i_ino, &dir_u, &inode_u,
-				  !tmpfile ? &dentry->d_name : NULL,
-				  from_kuid(ns, current_fsuid()),
-				  from_kgid(ns, current_fsgid()),
+	bch2_trans_begin(trans);
+
+	ret   = bch2_subvol_is_ro_trans(trans, dir->ei_subvol) ?:
+		bch2_create_trans(trans,
+				  inode_inum(dir), &dir_u, &inode_u,
+				  !(flags & BCH_CREATE_TMPFILE)
+				  ? &dentry->d_name : NULL,
+				  from_kuid(i_user_ns(&dir->v), current_fsuid()),
+				  from_kgid(i_user_ns(&dir->v), current_fsgid()),
 				  mode, rdev,
-				  default_acl, acl) ?:
+				  default_acl, acl, snapshot_src, flags) ?:
 		bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, 1,
 				KEY_TYPE_QUOTA_PREALLOC);
 	if (unlikely(ret))
 		goto err_before_quota;
 
-	ret   = bch2_trans_commit(&trans, NULL, &journal_seq,
-				  BTREE_INSERT_NOUNLOCK);
+	inum.subvol = inode_u.bi_subvol ?: dir->ei_subvol;
+	inum.inum = inode_u.bi_inum;
+
+	ret   = bch2_subvolume_get(trans, inum.subvol, true,
+				   BTREE_ITER_WITH_UPDATES, &subvol) ?:
+		bch2_trans_commit(trans, NULL, &journal_seq, 0);
 	if (unlikely(ret)) {
 		bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
 				KEY_TYPE_QUOTA_WARN);
 err_before_quota:
-		if (ret == -EINTR)
+		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 			goto retry;
 		goto err_trans;
 	}
 
-	if (!tmpfile) {
-		bch2_inode_update_after_write(c, dir, &dir_u,
+	if (!(flags & BCH_CREATE_TMPFILE)) {
+		bch2_inode_update_after_write(trans, dir, &dir_u,
 					      ATTR_MTIME|ATTR_CTIME);
-		journal_seq_copy(c, dir, journal_seq);
 		mutex_unlock(&dir->ei_update_lock);
 	}
 
-	bch2_vfs_inode_init(c, inode, &inode_u);
-	journal_seq_copy(c, inode, journal_seq);
+	bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
 
 	set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
 	set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
@@ -321,40 +346,17 @@ err_before_quota:
 	 * bch2_trans_exit() and dropping locks, else we could race with another
 	 * thread pulling the inode in and modifying it:
 	 */
-
-	inode->v.i_state |= I_CREATING;
-	old = to_bch_ei(inode_insert5(&inode->v, inode->v.i_ino,
-				      inum_test, NULL, &inode->v.i_ino));
-	BUG_ON(!old);
-
-	if (unlikely(old != inode)) {
-		/*
-		 * We raced, another process pulled the new inode into cache
-		 * before us:
-		 */
-		journal_seq_copy(c, old, journal_seq);
-		make_bad_inode(&inode->v);
-		iput(&inode->v);
-
-		inode = old;
-	} else {
-		/*
-		 * we really don't want insert_inode_locked2() to be setting
-		 * I_NEW...
-		 */
-		unlock_new_inode(&inode->v);
-	}
-
-	bch2_trans_exit(&trans);
+	inode = bch2_inode_insert(c, inode);
+	bch2_trans_put(trans);
 err:
 	posix_acl_release(default_acl);
 	posix_acl_release(acl);
 	return inode;
 err_trans:
-	if (!tmpfile)
+	if (!(flags & BCH_CREATE_TMPFILE))
 		mutex_unlock(&dir->ei_update_lock);
 
-	bch2_trans_exit(&trans);
+	bch2_trans_put(trans);
 	make_bad_inode(&inode->v);
 	iput(&inode->v);
 	inode = ERR_PTR(ret);
@@ -363,41 +365,100 @@ err_trans:
 
 /* methods */
 
+static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
+			subvol_inum dir, struct bch_hash_info *dir_hash_info,
+			const struct qstr *name)
+{
+	struct bch_fs *c = trans->c;
+	struct btree_iter dirent_iter = {};
+	subvol_inum inum = {};
+
+	int ret = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc,
+				   dir_hash_info, dir, name, 0);
+	if (ret)
+		return ERR_PTR(ret);
+
+	struct bkey_s_c k = bch2_btree_iter_peek_slot(&dirent_iter);
+	ret = bkey_err(k);
+	if (ret)
+		goto err;
+
+	ret = bch2_dirent_read_target(trans, dir, bkey_s_c_to_dirent(k), &inum);
+	if (ret > 0)
+		ret = -ENOENT;
+	if (ret)
+		goto err;
+
+	struct bch_inode_info *inode =
+		to_bch_ei(ilookup5_nowait(c->vfs_sb,
+					  bch2_inode_hash(inum),
+					  bch2_iget5_test,
+					  &inum));
+	if (inode)
+		goto out;
+
+	struct bch_subvolume subvol;
+	struct bch_inode_unpacked inode_u;
+	ret =   bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
+		bch2_inode_find_by_inum_nowarn_trans(trans, inum, &inode_u) ?:
+		PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
+	if (bch2_err_matches(ret, ENOENT)) {
+		struct printbuf buf = PRINTBUF;
+
+		bch2_bkey_val_to_text(&buf, c, k);
+		bch_err(c, "%s points to missing inode", buf.buf);
+		printbuf_exit(&buf);
+	}
+	if (ret)
+		goto err;
+
+	bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
+	inode = bch2_inode_insert(c, inode);
+out:
+	bch2_trans_iter_exit(trans, &dirent_iter);
+	return inode;
+err:
+	inode = ERR_PTR(ret);
+	goto out;
+}
+
 static struct dentry *bch2_lookup(struct inode *vdir, struct dentry *dentry,
 				  unsigned int flags)
 {
 	struct bch_fs *c = vdir->i_sb->s_fs_info;
 	struct bch_inode_info *dir = to_bch_ei(vdir);
-	struct inode *vinode = NULL;
-	u64 inum;
-
-	inum = bch2_dirent_lookup(c, dir->v.i_ino,
-				  &dir->ei_str_hash,
-				  &dentry->d_name);
+	struct bch_hash_info hash = bch2_hash_info_init(c, &dir->ei_inode);
 
-	if (inum)
-		vinode = bch2_vfs_inode_get(c, inum);
+	struct bch_inode_info *inode;
+	bch2_trans_do(c, NULL, NULL, 0,
+		PTR_ERR_OR_ZERO(inode = bch2_lookup_trans(trans, inode_inum(dir),
+							  &hash, &dentry->d_name)));
+	if (IS_ERR(inode))
+		inode = NULL;
 
-	return d_splice_alias(vinode, dentry);
+	return d_splice_alias(&inode->v, dentry);
 }
 
-static int bch2_mknod(struct inode *vdir, struct dentry *dentry,
+static int bch2_mknod(struct mnt_idmap *idmap,
+		      struct inode *vdir, struct dentry *dentry,
 		      umode_t mode, dev_t rdev)
 {
 	struct bch_inode_info *inode =
-		__bch2_create(to_bch_ei(vdir), dentry, mode, rdev, false);
+		__bch2_create(idmap, to_bch_ei(vdir), dentry, mode, rdev,
+			      (subvol_inum) { 0 }, 0);
 
 	if (IS_ERR(inode))
-		return PTR_ERR(inode);
+		return bch2_err_class(PTR_ERR(inode));
 
 	d_instantiate(dentry, &inode->v);
 	return 0;
 }
 
-static int bch2_create(struct inode *vdir, struct dentry *dentry,
+static int bch2_create(struct mnt_idmap *idmap,
+		       struct inode *vdir, struct dentry *dentry,
 		       umode_t mode, bool excl)
 {
-	return bch2_mknod(vdir, dentry, mode|S_IFREG, 0);
+	return bch2_mknod(idmap, vdir, dentry, mode|S_IFREG, 0);
 }
 
 static int __bch2_link(struct bch_fs *c,
@@ -405,34 +466,25 @@ static int __bch2_link(struct bch_fs *c,
 		       struct bch_inode_info *dir,
 		       struct dentry *dentry)
 {
-	struct btree_trans trans;
+	struct btree_trans *trans = bch2_trans_get(c);
 	struct bch_inode_unpacked dir_u, inode_u;
 	int ret;
 
 	mutex_lock(&inode->ei_update_lock);
-	bch2_trans_init(&trans, c, 4, 1024);
-
-	do {
-		bch2_trans_begin(&trans);
-		ret   = bch2_link_trans(&trans,
-					dir->v.i_ino,
-					inode->v.i_ino, &dir_u, &inode_u,
-					&dentry->d_name) ?:
-			bch2_trans_commit(&trans, NULL,
-					&inode->ei_journal_seq,
-					BTREE_INSERT_NOUNLOCK);
-	} while (ret == -EINTR);
 
-	if (likely(!ret)) {
-		BUG_ON(inode_u.bi_inum != inode->v.i_ino);
+	ret = commit_do(trans, NULL, NULL, 0,
+			bch2_link_trans(trans,
+					inode_inum(dir),   &dir_u,
+					inode_inum(inode), &inode_u,
+					&dentry->d_name));
 
-		journal_seq_copy(c, inode, dir->ei_journal_seq);
-		bch2_inode_update_after_write(c, dir, &dir_u,
+	if (likely(!ret)) {
+		bch2_inode_update_after_write(trans, dir, &dir_u,
 					      ATTR_MTIME|ATTR_CTIME);
-		bch2_inode_update_after_write(c, inode, &inode_u, ATTR_CTIME);
+		bch2_inode_update_after_write(trans, inode, &inode_u, ATTR_CTIME);
 	}
 
-	bch2_trans_exit(&trans);
+	bch2_trans_put(trans);
 	mutex_unlock(&inode->ei_update_lock);
 	return ret;
 }
@@ -447,65 +499,79 @@ static int bch2_link(struct dentry *old_dentry, struct inode *vdir,
 
 	lockdep_assert_held(&inode->v.i_rwsem);
 
-	ret = __bch2_link(c, inode, dir, dentry);
+	ret   = bch2_subvol_is_ro(c, dir->ei_subvol) ?:
+		bch2_subvol_is_ro(c, inode->ei_subvol) ?:
+		__bch2_link(c, inode, dir, dentry);
 	if (unlikely(ret))
-		return ret;
+		return bch2_err_class(ret);
 
 	ihold(&inode->v);
 	d_instantiate(dentry, &inode->v);
 	return 0;
 }
 
-static int bch2_unlink(struct inode *vdir, struct dentry *dentry)
+int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
+		  bool deleting_snapshot)
 {
 	struct bch_fs *c = vdir->i_sb->s_fs_info;
 	struct bch_inode_info *dir = to_bch_ei(vdir);
 	struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
 	struct bch_inode_unpacked dir_u, inode_u;
-	struct btree_trans trans;
+	struct btree_trans *trans = bch2_trans_get(c);
 	int ret;
 
 	bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
-	bch2_trans_init(&trans, c, 4, 1024);
-
-	do {
-		bch2_trans_begin(&trans);
 
-		ret   = bch2_unlink_trans(&trans,
-					  dir->v.i_ino, &dir_u,
-					  &inode_u, &dentry->d_name) ?:
-			bch2_trans_commit(&trans, NULL,
-					  &dir->ei_journal_seq,
-					  BTREE_INSERT_NOUNLOCK|
-					  BTREE_INSERT_NOFAIL);
-	} while (ret == -EINTR);
+	ret = commit_do(trans, NULL, NULL,
+			BCH_TRANS_COMMIT_no_enospc,
+		bch2_unlink_trans(trans,
+				  inode_inum(dir), &dir_u,
+				  &inode_u, &dentry->d_name,
+				  deleting_snapshot));
+	if (unlikely(ret))
+		goto err;
 
-	if (likely(!ret)) {
-		BUG_ON(inode_u.bi_inum != inode->v.i_ino);
+	bch2_inode_update_after_write(trans, dir, &dir_u,
+				      ATTR_MTIME|ATTR_CTIME);
+	bch2_inode_update_after_write(trans, inode, &inode_u,
+				      ATTR_MTIME);
 
-		journal_seq_copy(c, inode, dir->ei_journal_seq);
-		bch2_inode_update_after_write(c, dir, &dir_u,
-					      ATTR_MTIME|ATTR_CTIME);
-		bch2_inode_update_after_write(c, inode, &inode_u,
-					      ATTR_MTIME);
+	if (inode_u.bi_subvol) {
+		/*
+		 * Subvolume deletion is asynchronous, but we still want to tell
+		 * the VFS that it's been deleted here:
+		 */
+		set_nlink(&inode->v, 0);
 	}
-
-	bch2_trans_exit(&trans);
+err:
 	bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode);
+	bch2_trans_put(trans);
 
 	return ret;
 }
 
-static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
+static int bch2_unlink(struct inode *vdir, struct dentry *dentry)
+{
+	struct bch_inode_info *dir= to_bch_ei(vdir);
+	struct bch_fs *c = dir->v.i_sb->s_fs_info;
+
+	int ret = bch2_subvol_is_ro(c, dir->ei_subvol) ?:
+		__bch2_unlink(vdir, dentry, false);
+	return bch2_err_class(ret);
+}
+
+static int bch2_symlink(struct mnt_idmap *idmap,
+			struct inode *vdir, struct dentry *dentry,
 			const char *symname)
 {
 	struct bch_fs *c = vdir->i_sb->s_fs_info;
 	struct bch_inode_info *dir = to_bch_ei(vdir), *inode;
 	int ret;
 
-	inode = __bch2_create(dir, dentry, S_IFLNK|S_IRWXUGO, 0, true);
-	if (unlikely(IS_ERR(inode)))
-		return PTR_ERR(inode);
+	inode = __bch2_create(idmap, dir, dentry, S_IFLNK|S_IRWXUGO, 0,
+			      (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
+	if (IS_ERR(inode))
+		return bch2_err_class(PTR_ERR(inode));
 
 	inode_lock(&inode->v);
 	ret = page_symlink(&inode->v, symname, strlen(symname) + 1);
@@ -518,8 +584,6 @@ static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
 	if (unlikely(ret))
 		goto err;
 
-	journal_seq_copy(c, dir, inode->ei_journal_seq);
-
 	ret = __bch2_link(c, inode, dir, dentry);
 	if (unlikely(ret))
 		goto err;
@@ -528,15 +592,17 @@ static int bch2_symlink(struct inode *vdir, struct dentry *dentry,
 	return 0;
 err:
 	iput(&inode->v);
-	return ret;
+	return bch2_err_class(ret);
 }
 
-static int bch2_mkdir(struct inode *vdir, struct dentry *dentry, umode_t mode)
+static int bch2_mkdir(struct mnt_idmap *idmap,
+		      struct inode *vdir, struct dentry *dentry, umode_t mode)
 {
-	return bch2_mknod(vdir, dentry, mode|S_IFDIR, 0);
+	return bch2_mknod(idmap, vdir, dentry, mode|S_IFDIR, 0);
 }
 
-static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
+static int bch2_rename2(struct mnt_idmap *idmap,
+			struct inode *src_vdir, struct dentry *src_dentry,
 			struct inode *dst_vdir, struct dentry *dst_dentry,
 			unsigned flags)
 {
@@ -547,12 +613,11 @@ static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
 	struct bch_inode_info *dst_inode = to_bch_ei(dst_dentry->d_inode);
 	struct bch_inode_unpacked dst_dir_u, src_dir_u;
 	struct bch_inode_unpacked src_inode_u, dst_inode_u;
-	struct btree_trans trans;
+	struct btree_trans *trans;
 	enum bch_rename_mode mode = flags & RENAME_EXCHANGE
 		? BCH_RENAME_EXCHANGE
 		: dst_dentry->d_inode
 		? BCH_RENAME_OVERWRITE : BCH_RENAME;
-	u64 journal_seq = 0;
 	int ret;
 
 	if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
@@ -565,7 +630,7 @@ static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
 			return ret;
 	}
 
-	bch2_trans_init(&trans, c, 8, 2048);
+	trans = bch2_trans_get(c);
 
 	bch2_lock_inodes(INODE_UPDATE_LOCK,
 			 src_dir,
@@ -573,6 +638,11 @@ static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
 			 src_inode,
 			 dst_inode);
 
+	ret   = bch2_subvol_is_ro_trans(trans, src_dir->ei_subvol) ?:
+		bch2_subvol_is_ro_trans(trans, dst_dir->ei_subvol);
+	if (ret)
+		goto err;
+
 	if (inode_attr_changing(dst_dir, src_inode, Inode_opt_project)) {
 		ret = bch2_fs_quota_transfer(c, src_inode,
 					     dst_dir->ei_qid,
@@ -592,21 +662,15 @@ static int bch2_rename2(struct inode *src_vdir, struct dentry *src_dentry,
 			goto err;
 	}
 
-retry:
-	bch2_trans_begin(&trans);
-	ret   = bch2_rename_trans(&trans,
-				  src_dir->v.i_ino, &src_dir_u,
-				  dst_dir->v.i_ino, &dst_dir_u,
-				  &src_inode_u,
-				  &dst_inode_u,
-				  &src_dentry->d_name,
-				  &dst_dentry->d_name,
-				  mode) ?:
-		bch2_trans_commit(&trans, NULL,
-				  &journal_seq,
-				  BTREE_INSERT_NOUNLOCK);
-	if (ret == -EINTR)
-		goto retry;
+	ret = commit_do(trans, NULL, NULL, 0,
+			bch2_rename_trans(trans,
+					  inode_inum(src_dir), &src_dir_u,
+					  inode_inum(dst_dir), &dst_dir_u,
+					  &src_inode_u,
+					  &dst_inode_u,
+					  &src_dentry->d_name,
+					  &dst_dentry->d_name,
+					  mode));
 	if (unlikely(ret))
 		goto err;
 
@@ -614,27 +678,21 @@ retry:
 	BUG_ON(dst_inode &&
 	       dst_inode->v.i_ino != dst_inode_u.bi_inum);
 
-	bch2_inode_update_after_write(c, src_dir, &src_dir_u,
+	bch2_inode_update_after_write(trans, src_dir, &src_dir_u,
 				      ATTR_MTIME|ATTR_CTIME);
-	journal_seq_copy(c, src_dir, journal_seq);
 
-	if (src_dir != dst_dir) {
-		bch2_inode_update_after_write(c, dst_dir, &dst_dir_u,
+	if (src_dir != dst_dir)
+		bch2_inode_update_after_write(trans, dst_dir, &dst_dir_u,
 					      ATTR_MTIME|ATTR_CTIME);
-		journal_seq_copy(c, dst_dir, journal_seq);
-	}
 
-	bch2_inode_update_after_write(c, src_inode, &src_inode_u,
+	bch2_inode_update_after_write(trans, src_inode, &src_inode_u,
 				      ATTR_CTIME);
-	journal_seq_copy(c, src_inode, journal_seq);
 
-	if (dst_inode) {
-		bch2_inode_update_after_write(c, dst_inode, &dst_inode_u,
+	if (dst_inode)
+		bch2_inode_update_after_write(trans, dst_inode, &dst_inode_u,
 					      ATTR_CTIME);
-		journal_seq_copy(c, dst_inode, journal_seq);
-	}
 err:
-	bch2_trans_exit(&trans);
+	bch2_trans_put(trans);
 
 	bch2_fs_quota_transfer(c, src_inode,
 			       bch_qid(&src_inode->ei_inode),
@@ -652,20 +710,24 @@ err:
 			   src_inode,
 			   dst_inode);
 
-	return ret;
+	return bch2_err_class(ret);
 }
 
-void bch2_setattr_copy(struct bch_inode_info *inode,
-		       struct bch_inode_unpacked *bi,
-		       struct iattr *attr)
+static void bch2_setattr_copy(struct mnt_idmap *idmap,
+			      struct bch_inode_info *inode,
+			      struct bch_inode_unpacked *bi,
+			      struct iattr *attr)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	unsigned int ia_valid = attr->ia_valid;
 
 	if (ia_valid & ATTR_UID)
-		bi->bi_uid = from_kuid(c->vfs_sb->s_user_ns, attr->ia_uid);
+		bi->bi_uid = from_kuid(i_user_ns(&inode->v), attr->ia_uid);
 	if (ia_valid & ATTR_GID)
-		bi->bi_gid = from_kgid(c->vfs_sb->s_user_ns, attr->ia_gid);
+		bi->bi_gid = from_kgid(i_user_ns(&inode->v), attr->ia_gid);
+
+	if (ia_valid & ATTR_SIZE)
+		bi->bi_size = attr->ia_size;
 
 	if (ia_valid & ATTR_ATIME)
 		bi->bi_atime = timespec_to_bch2_time(c, attr->ia_atime);
@@ -681,19 +743,20 @@ void bch2_setattr_copy(struct bch_inode_info *inode,
 			: inode->v.i_gid;
 
 		if (!in_group_p(gid) &&
-		    !capable_wrt_inode_uidgid(&inode->v, CAP_FSETID))
+		    !capable_wrt_inode_uidgid(idmap, &inode->v, CAP_FSETID))
 			mode &= ~S_ISGID;
 		bi->bi_mode = mode;
 	}
 }
 
-static int bch2_setattr_nonsize(struct bch_inode_info *inode,
-				struct iattr *attr)
+int bch2_setattr_nonsize(struct mnt_idmap *idmap,
+			 struct bch_inode_info *inode,
+			 struct iattr *attr)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	struct bch_qid qid;
-	struct btree_trans trans;
-	struct btree_iter *inode_iter;
+	struct btree_trans *trans;
+	struct btree_iter inode_iter = { NULL };
 	struct bch_inode_unpacked inode_u;
 	struct posix_acl *acl = NULL;
 	int ret;
@@ -703,60 +766,61 @@ static int bch2_setattr_nonsize(struct bch_inode_info *inode,
 	qid = inode->ei_qid;
 
 	if (attr->ia_valid & ATTR_UID)
-		qid.q[QTYP_USR] = from_kuid(&init_user_ns, attr->ia_uid);
+		qid.q[QTYP_USR] = from_kuid(i_user_ns(&inode->v), attr->ia_uid);
 
 	if (attr->ia_valid & ATTR_GID)
-		qid.q[QTYP_GRP] = from_kgid(&init_user_ns, attr->ia_gid);
+		qid.q[QTYP_GRP] = from_kgid(i_user_ns(&inode->v), attr->ia_gid);
 
 	ret = bch2_fs_quota_transfer(c, inode, qid, ~0,
 				     KEY_TYPE_QUOTA_PREALLOC);
 	if (ret)
 		goto err;
 
-	bch2_trans_init(&trans, c, 0, 0);
+	trans = bch2_trans_get(c);
 retry:
-	bch2_trans_begin(&trans);
+	bch2_trans_begin(trans);
 	kfree(acl);
 	acl = NULL;
 
-	inode_iter = bch2_inode_peek(&trans, &inode_u, inode->v.i_ino,
-				     BTREE_ITER_INTENT);
-	ret = PTR_ERR_OR_ZERO(inode_iter);
+	ret = bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
+			      BTREE_ITER_INTENT);
 	if (ret)
 		goto btree_err;
 
-	bch2_setattr_copy(inode, &inode_u, attr);
+	bch2_setattr_copy(idmap, inode, &inode_u, attr);
 
 	if (attr->ia_valid & ATTR_MODE) {
-		ret = bch2_acl_chmod(&trans, inode, inode_u.bi_mode, &acl);
+		ret = bch2_acl_chmod(trans, inode_inum(inode), &inode_u,
+				     inode_u.bi_mode, &acl);
 		if (ret)
 			goto btree_err;
 	}
 
-	ret =   bch2_inode_write(&trans, inode_iter, &inode_u) ?:
-		bch2_trans_commit(&trans, NULL,
-				  &inode->ei_journal_seq,
-				  BTREE_INSERT_NOUNLOCK|
-				  BTREE_INSERT_NOFAIL);
+	ret =   bch2_inode_write(trans, &inode_iter, &inode_u) ?:
+		bch2_trans_commit(trans, NULL, NULL,
+				  BCH_TRANS_COMMIT_no_enospc);
 btree_err:
-	if (ret == -EINTR)
+	bch2_trans_iter_exit(trans, &inode_iter);
+
+	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		goto retry;
 	if (unlikely(ret))
 		goto err_trans;
 
-	bch2_inode_update_after_write(c, inode, &inode_u, attr->ia_valid);
+	bch2_inode_update_after_write(trans, inode, &inode_u, attr->ia_valid);
 
 	if (acl)
 		set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
 err_trans:
-	bch2_trans_exit(&trans);
+	bch2_trans_put(trans);
 err:
 	mutex_unlock(&inode->ei_update_lock);
 
-	return ret;
+	return bch2_err_class(ret);
 }
 
-static int bch2_getattr(const struct path *path, struct kstat *stat,
+static int bch2_getattr(struct mnt_idmap *idmap,
+			const struct path *path, struct kstat *stat,
 			u32 request_mask, unsigned query_flags)
 {
 	struct bch_inode_info *inode = to_bch_ei(d_inode(path->dentry));
@@ -770,9 +834,9 @@ static int bch2_getattr(const struct path *path, struct kstat *stat,
 	stat->gid	= inode->v.i_gid;
 	stat->rdev	= inode->v.i_rdev;
 	stat->size	= i_size_read(&inode->v);
-	stat->atime	= inode->v.i_atime;
-	stat->mtime	= inode->v.i_mtime;
-	stat->ctime	= inode->v.i_ctime;
+	stat->atime	= inode_get_atime(&inode->v);
+	stat->mtime	= inode_get_mtime(&inode->v);
+	stat->ctime	= inode_get_ctime(&inode->v);
 	stat->blksize	= block_bytes(c);
 	stat->blocks	= inode->v.i_blocks;
 
@@ -781,48 +845,54 @@ static int bch2_getattr(const struct path *path, struct kstat *stat,
 		stat->btime = bch2_time_to_timespec(c, inode->ei_inode.bi_otime);
 	}
 
-	if (inode->ei_inode.bi_flags & BCH_INODE_IMMUTABLE)
+	if (inode->ei_inode.bi_flags & BCH_INODE_immutable)
 		stat->attributes |= STATX_ATTR_IMMUTABLE;
 	stat->attributes_mask	 |= STATX_ATTR_IMMUTABLE;
 
-	if (inode->ei_inode.bi_flags & BCH_INODE_APPEND)
+	if (inode->ei_inode.bi_flags & BCH_INODE_append)
 		stat->attributes |= STATX_ATTR_APPEND;
 	stat->attributes_mask	 |= STATX_ATTR_APPEND;
 
-	if (inode->ei_inode.bi_flags & BCH_INODE_NODUMP)
+	if (inode->ei_inode.bi_flags & BCH_INODE_nodump)
 		stat->attributes |= STATX_ATTR_NODUMP;
 	stat->attributes_mask	 |= STATX_ATTR_NODUMP;
 
 	return 0;
 }
 
-static int bch2_setattr(struct dentry *dentry, struct iattr *iattr)
+static int bch2_setattr(struct mnt_idmap *idmap,
+			struct dentry *dentry, struct iattr *iattr)
 {
 	struct bch_inode_info *inode = to_bch_ei(dentry->d_inode);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	int ret;
 
 	lockdep_assert_held(&inode->v.i_rwsem);
 
-	ret = setattr_prepare(dentry, iattr);
+	ret   = bch2_subvol_is_ro(c, inode->ei_subvol) ?:
+		setattr_prepare(idmap, dentry, iattr);
 	if (ret)
 		return ret;
 
 	return iattr->ia_valid & ATTR_SIZE
-		? bch2_truncate(inode, iattr)
-		: bch2_setattr_nonsize(inode, iattr);
+		? bchfs_truncate(idmap, inode, iattr)
+		: bch2_setattr_nonsize(idmap, inode, iattr);
 }
 
-static int bch2_tmpfile(struct inode *vdir, struct dentry *dentry, umode_t mode)
+static int bch2_tmpfile(struct mnt_idmap *idmap,
+			struct inode *vdir, struct file *file, umode_t mode)
 {
 	struct bch_inode_info *inode =
-		__bch2_create(to_bch_ei(vdir), dentry, mode, 0, true);
+		__bch2_create(idmap, to_bch_ei(vdir),
+			      file->f_path.dentry, mode, 0,
+			      (subvol_inum) { 0 }, BCH_CREATE_TMPFILE);
 
 	if (IS_ERR(inode))
-		return PTR_ERR(inode);
+		return bch2_err_class(PTR_ERR(inode));
 
-	d_mark_tmpfile(dentry, &inode->v);
-	d_instantiate(dentry, &inode->v);
-	return 0;
+	d_mark_tmpfile(file, &inode->v);
+	d_instantiate(file->f_path.dentry, &inode->v);
+	return finish_open_simple(file, 0);
 }
 
 static int bch2_fill_extent(struct bch_fs *c,
@@ -842,13 +912,16 @@ static int bch2_fill_extent(struct bch_fs *c,
 			int flags2 = 0;
 			u64 offset = p.ptr.offset;
 
+			if (p.ptr.unwritten)
+				flags2 |= FIEMAP_EXTENT_UNWRITTEN;
+
 			if (p.crc.compression_type)
 				flags2 |= FIEMAP_EXTENT_ENCODED;
 			else
 				offset += p.crc.offset;
 
-			if ((offset & (c->opts.block_size - 1)) ||
-			    (k.k->size & (c->opts.block_size - 1)))
+			if ((offset & (block_sectors(c) - 1)) ||
+			    (k.k->size & (block_sectors(c) - 1)))
 				flags2 |= FIEMAP_EXTENT_NOT_ALIGNED;
 
 			ret = fiemap_fill_next_extent(info,
@@ -883,13 +956,14 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 {
 	struct bch_fs *c = vinode->i_sb->s_fs_info;
 	struct bch_inode_info *ei = to_bch_ei(vinode);
-	struct btree_trans trans;
-	struct btree_iter *iter;
+	struct btree_trans *trans;
+	struct btree_iter iter;
 	struct bkey_s_c k;
-	struct bkey_on_stack cur, prev;
+	struct bkey_buf cur, prev;
 	struct bpos end = POS(ei->v.i_ino, (start + len) >> 9);
 	unsigned offset_into_extent, sectors;
 	bool have_extent = false;
+	u32 snapshot;
 	int ret = 0;
 
 	ret = fiemap_prep(&ei->v, info, start, &len, FIEMAP_FLAG_SYNC);
@@ -899,35 +973,45 @@ static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info,
 	if (start + len < start)
 		return -EINVAL;
 
-	bkey_on_stack_init(&cur);
-	bkey_on_stack_init(&prev);
-	bch2_trans_init(&trans, c, 0, 0);
+	start >>= 9;
 
-	iter = bch2_trans_get_iter(&trans, BTREE_ID_EXTENTS,
-				   POS(ei->v.i_ino, start >> 9), 0);
+	bch2_bkey_buf_init(&cur);
+	bch2_bkey_buf_init(&prev);
+	trans = bch2_trans_get(c);
 retry:
-	while ((k = bch2_btree_iter_peek(iter)).k &&
-	       !(ret = bkey_err(k)) &&
-	       bkey_cmp(iter->pos, end) < 0) {
+	bch2_trans_begin(trans);
+
+	ret = bch2_subvolume_get_snapshot(trans, ei->ei_subvol, &snapshot);
+	if (ret)
+		goto err;
+
+	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
+			     SPOS(ei->v.i_ino, start, snapshot), 0);
+
+	while (!(ret = btree_trans_too_many_iters(trans)) &&
+	       (k = bch2_btree_iter_peek_upto(&iter, end)).k &&
+	       !(ret = bkey_err(k))) {
+		enum btree_id data_btree = BTREE_ID_extents;
+
 		if (!bkey_extent_is_data(k.k) &&
 		    k.k->type != KEY_TYPE_reservation) {
-			bch2_btree_iter_next(iter);
+			bch2_btree_iter_advance(&iter);
 			continue;
 		}
 
-		offset_into_extent	= iter->pos.offset -
+		offset_into_extent	= iter.pos.offset -
 			bkey_start_offset(k.k);
 		sectors			= k.k->size - offset_into_extent;
 
-		bkey_on_stack_reassemble(&cur, c, k);
+		bch2_bkey_buf_reassemble(&cur, c, k);
 
-		ret = bch2_read_indirect_extent(&trans,
+		ret = bch2_read_indirect_extent(trans, &data_btree,
 					&offset_into_extent, &cur);
 		if (ret)
 			break;
 
 		k = bkey_i_to_s_c(cur.k);
-		bkey_on_stack_realloc(&prev, c, k.k->u64s);
+		bch2_bkey_buf_realloc(&prev, c, k.k->u64s);
 
 		sectors = min(sectors, k.k->size - offset_into_extent);
 
@@ -936,10 +1020,11 @@ retry:
 				   offset_into_extent),
 			       cur.k);
 		bch2_key_resize(&cur.k->k, sectors);
-		cur.k->k.p = iter->pos;
+		cur.k->k.p = iter.pos;
 		cur.k->k.p.offset += cur.k->k.size;
 
 		if (have_extent) {
+			bch2_trans_unlock(trans);
 			ret = bch2_fill_extent(c, info,
 					bkey_i_to_s_c(prev.k), 0);
 			if (ret)
@@ -949,20 +1034,24 @@ retry:
 		bkey_copy(prev.k, cur.k);
 		have_extent = true;
 
-		bch2_btree_iter_set_pos(iter,
-			POS(iter->pos.inode, iter->pos.offset + sectors));
+		bch2_btree_iter_set_pos(&iter,
+			POS(iter.pos.inode, iter.pos.offset + sectors));
 	}
-
-	if (ret == -EINTR)
+	start = iter.pos.offset;
+	bch2_trans_iter_exit(trans, &iter);
+err:
+	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		goto retry;
 
-	if (!ret && have_extent)
+	if (!ret && have_extent) {
+		bch2_trans_unlock(trans);
 		ret = bch2_fill_extent(c, info, bkey_i_to_s_c(prev.k),
 				       FIEMAP_EXTENT_LAST);
+	}
 
-	ret = bch2_trans_exit(&trans) ?: ret;
-	bkey_on_stack_exit(&cur, c);
-	bkey_on_stack_exit(&prev, c);
+	bch2_trans_put(trans);
+	bch2_bkey_buf_exit(&cur, c);
+	bch2_bkey_buf_exit(&prev, c);
 	return ret < 0 ? ret : 0;
 }
 
@@ -996,21 +1085,35 @@ static int bch2_vfs_readdir(struct file *file, struct dir_context *ctx)
 	if (!dir_emit_dots(file, ctx))
 		return 0;
 
-	return bch2_readdir(c, inode->v.i_ino, ctx);
+	int ret = bch2_readdir(c, inode_inum(inode), ctx);
+
+	bch_err_fn(c, ret);
+	return bch2_err_class(ret);
+}
+
+static int bch2_open(struct inode *vinode, struct file *file)
+{
+	if (file->f_flags & (O_WRONLY|O_RDWR)) {
+		struct bch_inode_info *inode = to_bch_ei(vinode);
+		struct bch_fs *c = inode->v.i_sb->s_fs_info;
+
+		int ret = bch2_subvol_is_ro(c, inode->ei_subvol);
+		if (ret)
+			return ret;
+	}
+
+	return generic_file_open(vinode, file);
 }
 
 static const struct file_operations bch_file_operations = {
+	.open		= bch2_open,
 	.llseek		= bch2_llseek,
 	.read_iter	= bch2_read_iter,
 	.write_iter	= bch2_write_iter,
 	.mmap		= bch2_mmap,
-	.open		= generic_file_open,
 	.fsync		= bch2_fsync,
-	.splice_read	= generic_file_splice_read,
-	/*
-	 * Broken, on v5.3:
+	.splice_read	= filemap_splice_read,
 	.splice_write	= iter_file_splice_write,
-	*/
 	.fallocate	= bch2_fallocate_dispatch,
 	.unlocked_ioctl = bch2_fs_file_ioctl,
 #ifdef CONFIG_COMPAT
@@ -1083,67 +1186,267 @@ static const struct inode_operations bch_special_inode_operations = {
 };
 
 static const struct address_space_operations bch_address_space_operations = {
-	.writepage	= bch2_writepage,
-	.readpage	= bch2_readpage,
+	.read_folio	= bch2_read_folio,
 	.writepages	= bch2_writepages,
 	.readahead	= bch2_readahead,
-	.set_page_dirty	= __set_page_dirty_nobuffers,
+	.dirty_folio	= filemap_dirty_folio,
 	.write_begin	= bch2_write_begin,
 	.write_end	= bch2_write_end,
-	.invalidatepage	= bch2_invalidatepage,
-	.releasepage	= bch2_releasepage,
+	.invalidate_folio = bch2_invalidate_folio,
+	.release_folio	= bch2_release_folio,
 	.direct_IO	= noop_direct_IO,
 #ifdef CONFIG_MIGRATION
-	.migratepage	= bch2_migrate_page,
+	.migrate_folio	= filemap_migrate_folio,
 #endif
-	.error_remove_page = generic_error_remove_page,
+	.error_remove_folio = generic_error_remove_folio,
 };
 
-static struct inode *bch2_nfs_get_inode(struct super_block *sb,
-		u64 ino, u32 generation)
+struct bcachefs_fid {
+	u64		inum;
+	u32		subvol;
+	u32		gen;
+} __packed;
+
+struct bcachefs_fid_with_parent {
+	struct bcachefs_fid	fid;
+	struct bcachefs_fid	dir;
+} __packed;
+
+static int bcachefs_fid_valid(int fh_len, int fh_type)
 {
-	struct bch_fs *c = sb->s_fs_info;
-	struct inode *vinode;
+	switch (fh_type) {
+	case FILEID_BCACHEFS_WITHOUT_PARENT:
+		return fh_len == sizeof(struct bcachefs_fid) / sizeof(u32);
+	case FILEID_BCACHEFS_WITH_PARENT:
+		return fh_len == sizeof(struct bcachefs_fid_with_parent) / sizeof(u32);
+	default:
+		return false;
+	}
+}
+
+static struct bcachefs_fid bch2_inode_to_fid(struct bch_inode_info *inode)
+{
+	return (struct bcachefs_fid) {
+		.inum	= inode->ei_inode.bi_inum,
+		.subvol	= inode->ei_subvol,
+		.gen	= inode->ei_inode.bi_generation,
+	};
+}
+
+static int bch2_encode_fh(struct inode *vinode, u32 *fh, int *len,
+			  struct inode *vdir)
+{
+	struct bch_inode_info *inode	= to_bch_ei(vinode);
+	struct bch_inode_info *dir	= to_bch_ei(vdir);
+	int min_len;
 
-	if (ino < BCACHEFS_ROOT_INO)
-		return ERR_PTR(-ESTALE);
+	if (!S_ISDIR(inode->v.i_mode) && dir) {
+		struct bcachefs_fid_with_parent *fid = (void *) fh;
 
-	vinode = bch2_vfs_inode_get(c, ino);
-	if (IS_ERR(vinode))
-		return ERR_CAST(vinode);
-	if (generation && vinode->i_generation != generation) {
-		/* we didn't find the right inode.. */
+		min_len = sizeof(*fid) / sizeof(u32);
+		if (*len < min_len) {
+			*len = min_len;
+			return FILEID_INVALID;
+		}
+
+		fid->fid = bch2_inode_to_fid(inode);
+		fid->dir = bch2_inode_to_fid(dir);
+
+		*len = min_len;
+		return FILEID_BCACHEFS_WITH_PARENT;
+	} else {
+		struct bcachefs_fid *fid = (void *) fh;
+
+		min_len = sizeof(*fid) / sizeof(u32);
+		if (*len < min_len) {
+			*len = min_len;
+			return FILEID_INVALID;
+		}
+		*fid = bch2_inode_to_fid(inode);
+
+		*len = min_len;
+		return FILEID_BCACHEFS_WITHOUT_PARENT;
+	}
+}
+
+static struct inode *bch2_nfs_get_inode(struct super_block *sb,
+					struct bcachefs_fid fid)
+{
+	struct bch_fs *c = sb->s_fs_info;
+	struct inode *vinode = bch2_vfs_inode_get(c, (subvol_inum) {
+				    .subvol = fid.subvol,
+				    .inum = fid.inum,
+	});
+	if (!IS_ERR(vinode) && vinode->i_generation != fid.gen) {
 		iput(vinode);
-		return ERR_PTR(-ESTALE);
+		vinode = ERR_PTR(-ESTALE);
 	}
 	return vinode;
 }
 
-static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *fid,
+static struct dentry *bch2_fh_to_dentry(struct super_block *sb, struct fid *_fid,
 		int fh_len, int fh_type)
 {
-	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
-				    bch2_nfs_get_inode);
+	struct bcachefs_fid *fid = (void *) _fid;
+
+	if (!bcachefs_fid_valid(fh_len, fh_type))
+		return NULL;
+
+	return d_obtain_alias(bch2_nfs_get_inode(sb, *fid));
 }
 
-static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *fid,
+static struct dentry *bch2_fh_to_parent(struct super_block *sb, struct fid *_fid,
 		int fh_len, int fh_type)
 {
-	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
-				    bch2_nfs_get_inode);
+	struct bcachefs_fid_with_parent *fid = (void *) _fid;
+
+	if (!bcachefs_fid_valid(fh_len, fh_type) ||
+	    fh_type != FILEID_BCACHEFS_WITH_PARENT)
+		return NULL;
+
+	return d_obtain_alias(bch2_nfs_get_inode(sb, fid->dir));
+}
+
+static struct dentry *bch2_get_parent(struct dentry *child)
+{
+	struct bch_inode_info *inode = to_bch_ei(child->d_inode);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+	subvol_inum parent_inum = {
+		.subvol = inode->ei_inode.bi_parent_subvol ?:
+			inode->ei_subvol,
+		.inum = inode->ei_inode.bi_dir,
+	};
+
+	return d_obtain_alias(bch2_vfs_inode_get(c, parent_inum));
+}
+
+static int bch2_get_name(struct dentry *parent, char *name, struct dentry *child)
+{
+	struct bch_inode_info *inode	= to_bch_ei(child->d_inode);
+	struct bch_inode_info *dir	= to_bch_ei(parent->d_inode);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+	struct btree_trans *trans;
+	struct btree_iter iter1;
+	struct btree_iter iter2;
+	struct bkey_s_c k;
+	struct bkey_s_c_dirent d;
+	struct bch_inode_unpacked inode_u;
+	subvol_inum target;
+	u32 snapshot;
+	struct qstr dirent_name;
+	unsigned name_len = 0;
+	int ret;
+
+	if (!S_ISDIR(dir->v.i_mode))
+		return -EINVAL;
+
+	trans = bch2_trans_get(c);
+
+	bch2_trans_iter_init(trans, &iter1, BTREE_ID_dirents,
+			     POS(dir->ei_inode.bi_inum, 0), 0);
+	bch2_trans_iter_init(trans, &iter2, BTREE_ID_dirents,
+			     POS(dir->ei_inode.bi_inum, 0), 0);
+retry:
+	bch2_trans_begin(trans);
+
+	ret = bch2_subvolume_get_snapshot(trans, dir->ei_subvol, &snapshot);
+	if (ret)
+		goto err;
+
+	bch2_btree_iter_set_snapshot(&iter1, snapshot);
+	bch2_btree_iter_set_snapshot(&iter2, snapshot);
+
+	ret = bch2_inode_find_by_inum_trans(trans, inode_inum(inode), &inode_u);
+	if (ret)
+		goto err;
+
+	if (inode_u.bi_dir == dir->ei_inode.bi_inum) {
+		bch2_btree_iter_set_pos(&iter1, POS(inode_u.bi_dir, inode_u.bi_dir_offset));
+
+		k = bch2_btree_iter_peek_slot(&iter1);
+		ret = bkey_err(k);
+		if (ret)
+			goto err;
+
+		if (k.k->type != KEY_TYPE_dirent) {
+			ret = -BCH_ERR_ENOENT_dirent_doesnt_match_inode;
+			goto err;
+		}
+
+		d = bkey_s_c_to_dirent(k);
+		ret = bch2_dirent_read_target(trans, inode_inum(dir), d, &target);
+		if (ret > 0)
+			ret = -BCH_ERR_ENOENT_dirent_doesnt_match_inode;
+		if (ret)
+			goto err;
+
+		if (target.subvol	== inode->ei_subvol &&
+		    target.inum		== inode->ei_inode.bi_inum)
+			goto found;
+	} else {
+		/*
+		 * File with multiple hardlinks and our backref is to the wrong
+		 * directory - linear search:
+		 */
+		for_each_btree_key_continue_norestart(iter2, 0, k, ret) {
+			if (k.k->p.inode > dir->ei_inode.bi_inum)
+				break;
+
+			if (k.k->type != KEY_TYPE_dirent)
+				continue;
+
+			d = bkey_s_c_to_dirent(k);
+			ret = bch2_dirent_read_target(trans, inode_inum(dir), d, &target);
+			if (ret < 0)
+				break;
+			if (ret)
+				continue;
+
+			if (target.subvol	== inode->ei_subvol &&
+			    target.inum		== inode->ei_inode.bi_inum)
+				goto found;
+		}
+	}
+
+	ret = -ENOENT;
+	goto err;
+found:
+	dirent_name = bch2_dirent_get_name(d);
+
+	name_len = min_t(unsigned, dirent_name.len, NAME_MAX);
+	memcpy(name, dirent_name.name, name_len);
+	name[name_len] = '\0';
+err:
+	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+		goto retry;
+
+	bch2_trans_iter_exit(trans, &iter1);
+	bch2_trans_iter_exit(trans, &iter2);
+	bch2_trans_put(trans);
+
+	return ret;
 }
 
 static const struct export_operations bch_export_ops = {
+	.encode_fh	= bch2_encode_fh,
 	.fh_to_dentry	= bch2_fh_to_dentry,
 	.fh_to_parent	= bch2_fh_to_parent,
-	//.get_parent	= bch2_get_parent,
+	.get_parent	= bch2_get_parent,
+	.get_name	= bch2_get_name,
 };
 
-static void bch2_vfs_inode_init(struct bch_fs *c,
+static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
 				struct bch_inode_info *inode,
-				struct bch_inode_unpacked *bi)
+				struct bch_inode_unpacked *bi,
+				struct bch_subvolume *subvol)
 {
-	bch2_inode_update_after_write(c, inode, bi, ~0);
+	bch2_iget5_set(&inode->v, &inum);
+	bch2_inode_update_after_write(trans, inode, bi, ~0);
+
+	if (BCH_SUBVOLUME_SNAP(subvol))
+		set_bit(EI_INODE_SNAPSHOT, &inode->ei_flags);
+	else
+		clear_bit(EI_INODE_SNAPSHOT, &inode->ei_flags);
 
 	inode->v.i_blocks	= bi->bi_sectors;
 	inode->v.i_ino		= bi->bi_inum;
@@ -1152,10 +1455,9 @@ static void bch2_vfs_inode_init(struct bch_fs *c,
 	inode->v.i_size		= bi->bi_size;
 
 	inode->ei_flags		= 0;
-	inode->ei_journal_seq	= 0;
 	inode->ei_quota_reserved = 0;
-	inode->ei_str_hash	= bch2_hash_info_init(c, bi);
 	inode->ei_qid		= bch_qid(bi);
+	inode->ei_subvol	= inum.subvol;
 
 	inode->v.i_mapping->a_ops = &bch_address_space_operations;
 
@@ -1177,6 +1479,8 @@ static void bch2_vfs_inode_init(struct bch_fs *c,
 		inode->v.i_op	= &bch_special_inode_operations;
 		break;
 	}
+
+	mapping_set_large_folios(inode->v.i_mapping);
 }
 
 static struct inode *bch2_alloc_inode(struct super_block *sb)
@@ -1189,9 +1493,9 @@ static struct inode *bch2_alloc_inode(struct super_block *sb)
 
 	inode_init_once(&inode->v);
 	mutex_init(&inode->ei_update_lock);
-	pagecache_lock_init(&inode->ei_pagecache_lock);
+	two_state_lock_init(&inode->ei_pagecache_lock);
+	INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
 	mutex_init(&inode->ei_quota_lock);
-	inode->ei_journal_seq = 0;
 
 	return &inode->v;
 }
@@ -1209,15 +1513,16 @@ static void bch2_destroy_inode(struct inode *vinode)
 	call_rcu(&vinode->i_rcu, bch2_i_callback);
 }
 
-static int inode_update_times_fn(struct bch_inode_info *inode,
+static int inode_update_times_fn(struct btree_trans *trans,
+				 struct bch_inode_info *inode,
 				 struct bch_inode_unpacked *bi,
 				 void *p)
 {
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 
-	bi->bi_atime	= timespec_to_bch2_time(c, inode->v.i_atime);
-	bi->bi_mtime	= timespec_to_bch2_time(c, inode->v.i_mtime);
-	bi->bi_ctime	= timespec_to_bch2_time(c, inode->v.i_ctime);
+	bi->bi_atime	= timespec_to_bch2_time(c, inode_get_atime(&inode->v));
+	bi->bi_mtime	= timespec_to_bch2_time(c, inode_get_mtime(&inode->v));
+	bi->bi_ctime	= timespec_to_bch2_time(c, inode_get_ctime(&inode->v));
 
 	return 0;
 }
@@ -1234,7 +1539,7 @@ static int bch2_vfs_write_inode(struct inode *vinode,
 			       ATTR_ATIME|ATTR_MTIME|ATTR_CTIME);
 	mutex_unlock(&inode->ei_update_lock);
 
-	return ret;
+	return bch2_err_class(ret);
 }
 
 static void bch2_evict_inode(struct inode *vinode)
@@ -1253,8 +1558,76 @@ static void bch2_evict_inode(struct inode *vinode)
 				KEY_TYPE_QUOTA_WARN);
 		bch2_quota_acct(c, inode->ei_qid, Q_INO, -1,
 				KEY_TYPE_QUOTA_WARN);
-		bch2_inode_rm(c, inode->v.i_ino, true);
+		bch2_inode_rm(c, inode_inum(inode));
+	}
+
+	mutex_lock(&c->vfs_inodes_lock);
+	list_del_init(&inode->ei_vfs_inode_list);
+	mutex_unlock(&c->vfs_inodes_lock);
+}
+
+void bch2_evict_subvolume_inodes(struct bch_fs *c, snapshot_id_list *s)
+{
+	struct bch_inode_info *inode;
+	DARRAY(struct bch_inode_info *) grabbed;
+	bool clean_pass = false, this_pass_clean;
+
+	/*
+	 * Initially, we scan for inodes without I_DONTCACHE, then mark them to
+	 * be pruned with d_mark_dontcache().
+	 *
+	 * Once we've had a clean pass where we didn't find any inodes without
+	 * I_DONTCACHE, we wait for them to be freed:
+	 */
+
+	darray_init(&grabbed);
+	darray_make_room(&grabbed, 1024);
+again:
+	cond_resched();
+	this_pass_clean = true;
+
+	mutex_lock(&c->vfs_inodes_lock);
+	list_for_each_entry(inode, &c->vfs_inodes_list, ei_vfs_inode_list) {
+		if (!snapshot_list_has_id(s, inode->ei_subvol))
+			continue;
+
+		if (!(inode->v.i_state & I_DONTCACHE) &&
+		    !(inode->v.i_state & I_FREEING) &&
+		    igrab(&inode->v)) {
+			this_pass_clean = false;
+
+			if (darray_push_gfp(&grabbed, inode, GFP_ATOMIC|__GFP_NOWARN)) {
+				iput(&inode->v);
+				break;
+			}
+		} else if (clean_pass && this_pass_clean) {
+			wait_queue_head_t *wq = bit_waitqueue(&inode->v.i_state, __I_NEW);
+			DEFINE_WAIT_BIT(wait, &inode->v.i_state, __I_NEW);
+
+			prepare_to_wait(wq, &wait.wq_entry, TASK_UNINTERRUPTIBLE);
+			mutex_unlock(&c->vfs_inodes_lock);
+
+			schedule();
+			finish_wait(wq, &wait.wq_entry);
+			goto again;
+		}
+	}
+	mutex_unlock(&c->vfs_inodes_lock);
+
+	darray_for_each(grabbed, i) {
+		inode = *i;
+		d_mark_dontcache(&inode->v);
+		d_prune_aliases(&inode->v);
+		iput(&inode->v);
 	}
+	grabbed.nr = 0;
+
+	if (!clean_pass || !this_pass_clean) {
+		clean_pass = this_pass_clean;
+		goto again;
+	}
+
+	darray_exit(&grabbed);
 }
 
 static int bch2_statfs(struct dentry *dentry, struct kstatfs *buf)
@@ -1268,21 +1641,17 @@ static int bch2_statfs(struct dentry *dentry, struct kstatfs *buf)
 	 * number:
 	 */
 	u64 avail_inodes = ((usage.capacity - usage.used) << 3);
-	u64 fsid;
 
 	buf->f_type	= BCACHEFS_STATFS_MAGIC;
 	buf->f_bsize	= sb->s_blocksize;
 	buf->f_blocks	= usage.capacity >> shift;
-	buf->f_bfree	= (usage.capacity - usage.used) >> shift;
-	buf->f_bavail	= buf->f_bfree;
+	buf->f_bfree	= usage.free >> shift;
+	buf->f_bavail	= avail_factor(usage.free) >> shift;
 
 	buf->f_files	= usage.nr_inodes + avail_inodes;
 	buf->f_ffree	= avail_inodes;
 
-	fsid = le64_to_cpup((void *) c->sb.user_uuid.b) ^
-	       le64_to_cpup((void *) c->sb.user_uuid.b + sizeof(u64));
-	buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
-	buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
+	buf->f_fsid	= uuid_to_fsid(c->sb.user_uuid.b);
 	buf->f_namelen	= BCH_NAME_MAX;
 
 	return 0;
@@ -1291,6 +1660,7 @@ static int bch2_statfs(struct dentry *dentry, struct kstatfs *buf)
 static int bch2_sync_fs(struct super_block *sb, int wait)
 {
 	struct bch_fs *c = sb->s_fs_info;
+	int ret;
 
 	if (c->opts.journal_flush_disabled)
 		return 0;
@@ -1300,62 +1670,37 @@ static int bch2_sync_fs(struct super_block *sb, int wait)
 		return 0;
 	}
 
-	return bch2_journal_flush(&c->journal);
+	ret = bch2_journal_flush(&c->journal);
+	return bch2_err_class(ret);
 }
 
-static struct bch_fs *bch2_path_to_fs(const char *dev)
+static struct bch_fs *bch2_path_to_fs(const char *path)
 {
 	struct bch_fs *c;
-	struct block_device *bdev = lookup_bdev(dev);
+	dev_t dev;
+	int ret;
 
-	if (IS_ERR(bdev))
-		return ERR_CAST(bdev);
+	ret = lookup_bdev(path, &dev);
+	if (ret)
+		return ERR_PTR(ret);
 
-	c = bch2_bdev_to_fs(bdev);
-	bdput(bdev);
+	c = bch2_dev_to_fs(dev);
 	if (c)
 		closure_put(&c->cl);
 	return c ?: ERR_PTR(-ENOENT);
 }
 
-static char **split_devs(const char *_dev_name, unsigned *nr)
-{
-	char *dev_name = NULL, **devs = NULL, *s;
-	size_t i, nr_devs = 0;
-
-	dev_name = kstrdup(_dev_name, GFP_KERNEL);
-	if (!dev_name)
-		return NULL;
-
-	for (s = dev_name; s; s = strchr(s + 1, ':'))
-		nr_devs++;
-
-	devs = kcalloc(nr_devs + 1, sizeof(const char *), GFP_KERNEL);
-	if (!devs) {
-		kfree(dev_name);
-		return NULL;
-	}
-
-	for (i = 0, s = dev_name;
-	     s;
-	     (s = strchr(s, ':')) && (*s++ = '\0'))
-		devs[i++] = s;
-
-	*nr = nr_devs;
-	return devs;
-}
-
 static int bch2_remount(struct super_block *sb, int *flags, char *data)
 {
 	struct bch_fs *c = sb->s_fs_info;
 	struct bch_opts opts = bch2_opts_empty();
 	int ret;
 
-	opt_set(opts, read_only, (*flags & SB_RDONLY) != 0);
-
 	ret = bch2_parse_mount_opts(c, &opts, data);
 	if (ret)
-		return ret;
+		goto err;
+
+	opt_set(opts, read_only, (*flags & SB_RDONLY) != 0);
 
 	if (opts.read_only != c->opts.read_only) {
 		down_write(&c->state_lock);
@@ -1369,7 +1714,8 @@ static int bch2_remount(struct super_block *sb, int *flags, char *data)
 			if (ret) {
 				bch_err(c, "error going rw: %i", ret);
 				up_write(&c->state_lock);
-				return -EINVAL;
+				ret = -EINVAL;
+				goto err;
 			}
 
 			sb->s_flags &= ~SB_RDONLY;
@@ -1380,25 +1726,22 @@ static int bch2_remount(struct super_block *sb, int *flags, char *data)
 		up_write(&c->state_lock);
 	}
 
-	if (opts.errors >= 0)
+	if (opt_defined(opts, errors))
 		c->opts.errors = opts.errors;
-
-	return ret;
+err:
+	return bch2_err_class(ret);
 }
 
 static int bch2_show_devname(struct seq_file *seq, struct dentry *root)
 {
 	struct bch_fs *c = root->d_sb->s_fs_info;
-	struct bch_dev *ca;
-	unsigned i;
 	bool first = true;
 
-	for_each_online_member(ca, c, i) {
+	for_each_online_member(c, ca) {
 		if (!first)
 			seq_putc(seq, ':');
 		first = false;
-		seq_puts(seq, "/dev/");
-		seq_puts(seq, ca->name);
+		seq_puts(seq, ca->disk_sb.sb_name);
 	}
 
 	return 0;
@@ -1408,25 +1751,30 @@ static int bch2_show_options(struct seq_file *seq, struct dentry *root)
 {
 	struct bch_fs *c = root->d_sb->s_fs_info;
 	enum bch_opt_id i;
-	char buf[512];
+	struct printbuf buf = PRINTBUF;
+	int ret = 0;
 
 	for (i = 0; i < bch2_opts_nr; i++) {
 		const struct bch_option *opt = &bch2_opt_table[i];
 		u64 v = bch2_opt_get_by_id(&c->opts, i);
 
-		if (!(opt->mode & OPT_MOUNT))
+		if (!(opt->flags & OPT_MOUNT))
 			continue;
 
 		if (v == bch2_opt_get_by_id(&bch2_opts_default, i))
 			continue;
 
-		bch2_opt_to_text(&PBUF(buf), c, opt, v,
+		printbuf_reset(&buf);
+		bch2_opt_to_text(&buf, c, c->disk_sb.sb, opt, v,
 				 OPT_SHOW_MOUNT_STYLE);
 		seq_putc(seq, ',');
-		seq_puts(seq, buf);
+		seq_puts(seq, buf.buf);
 	}
 
-	return 0;
+	if (buf.allocation_failure)
+		ret = -ENOMEM;
+	printbuf_exit(&buf);
+	return ret;
 }
 
 static void bch2_put_super(struct super_block *sb)
@@ -1436,6 +1784,38 @@ static void bch2_put_super(struct super_block *sb)
 	__bch2_fs_stop(c);
 }
 
+/*
+ * bcachefs doesn't currently integrate intwrite freeze protection but the
+ * internal write references serve the same purpose. Therefore reuse the
+ * read-only transition code to perform the quiesce. The caveat is that we don't
+ * currently have the ability to block tasks that want a write reference while
+ * the superblock is frozen. This is fine for now, but we should either add
+ * blocking support or find a way to integrate sb_start_intwrite() and friends.
+ */
+static int bch2_freeze(struct super_block *sb)
+{
+	struct bch_fs *c = sb->s_fs_info;
+
+	down_write(&c->state_lock);
+	bch2_fs_read_only(c);
+	up_write(&c->state_lock);
+	return 0;
+}
+
+static int bch2_unfreeze(struct super_block *sb)
+{
+	struct bch_fs *c = sb->s_fs_info;
+	int ret;
+
+	if (test_bit(BCH_FS_emergency_ro, &c->flags))
+		return 0;
+
+	down_write(&c->state_lock);
+	ret = bch2_fs_read_write(c);
+	up_write(&c->state_lock);
+	return ret;
+}
+
 static const struct super_operations bch_super_operations = {
 	.alloc_inode	= bch2_alloc_inode,
 	.destroy_inode	= bch2_destroy_inode,
@@ -1447,10 +1827,8 @@ static const struct super_operations bch_super_operations = {
 	.show_options	= bch2_show_options,
 	.remount_fs	= bch2_remount,
 	.put_super	= bch2_put_super,
-#if 0
 	.freeze_fs	= bch2_freeze,
 	.unfreeze_fs	= bch2_unfreeze,
-#endif
 };
 
 static int bch2_set_super(struct super_block *s, void *data)
@@ -1464,17 +1842,18 @@ static int bch2_noset_super(struct super_block *s, void *data)
 	return -EBUSY;
 }
 
+typedef DARRAY(struct bch_fs *) darray_fs;
+
 static int bch2_test_super(struct super_block *s, void *data)
 {
 	struct bch_fs *c = s->s_fs_info;
-	struct bch_fs **devs = data;
-	unsigned i;
+	darray_fs *d = data;
 
 	if (!c)
 		return false;
 
-	for (i = 0; devs[i]; i++)
-		if (c != devs[i])
+	darray_for_each(*d, i)
+		if (c != *i)
 			return false;
 	return true;
 }
@@ -1483,13 +1862,9 @@ static struct dentry *bch2_mount(struct file_system_type *fs_type,
 				 int flags, const char *dev_name, void *data)
 {
 	struct bch_fs *c;
-	struct bch_dev *ca;
 	struct super_block *sb;
 	struct inode *vinode;
 	struct bch_opts opts = bch2_opts_empty();
-	char **devs;
-	struct bch_fs **devs_to_fs = NULL;
-	unsigned i, nr_devs;
 	int ret;
 
 	opt_set(opts, read_only, (flags & SB_RDONLY) != 0);
@@ -1498,25 +1873,28 @@ static struct dentry *bch2_mount(struct file_system_type *fs_type,
 	if (ret)
 		return ERR_PTR(ret);
 
-	devs = split_devs(dev_name, &nr_devs);
-	if (!devs)
-		return ERR_PTR(-ENOMEM);
+	if (!dev_name || strlen(dev_name) == 0)
+		return ERR_PTR(-EINVAL);
 
-	devs_to_fs = kcalloc(nr_devs + 1, sizeof(void *), GFP_KERNEL);
-	if (!devs_to_fs) {
-		sb = ERR_PTR(-ENOMEM);
-		goto got_sb;
-	}
+	darray_str devs;
+	ret = bch2_split_devs(dev_name, &devs);
+	if (ret)
+		return ERR_PTR(ret);
 
-	for (i = 0; i < nr_devs; i++)
-		devs_to_fs[i] = bch2_path_to_fs(devs[i]);
+	darray_fs devs_to_fs = {};
+	darray_for_each(devs, i) {
+		ret = darray_push(&devs_to_fs, bch2_path_to_fs(*i));
+		if (ret) {
+			sb = ERR_PTR(ret);
+			goto got_sb;
+		}
+	}
 
-	sb = sget(fs_type, bch2_test_super, bch2_noset_super,
-		  flags|SB_NOSEC, devs_to_fs);
+	sb = sget(fs_type, bch2_test_super, bch2_noset_super, flags|SB_NOSEC, &devs_to_fs);
 	if (!IS_ERR(sb))
 		goto got_sb;
 
-	c = bch2_fs_open(devs, nr_devs, opts);
+	c = bch2_fs_open(devs.data, devs.nr, opts);
 	if (IS_ERR(c)) {
 		sb = ERR_CAST(c);
 		goto got_sb;
@@ -1536,12 +1914,14 @@ static struct dentry *bch2_mount(struct file_system_type *fs_type,
 	if (IS_ERR(sb))
 		bch2_fs_stop(c);
 got_sb:
-	kfree(devs_to_fs);
-	kfree(devs[0]);
-	kfree(devs);
+	darray_exit(&devs_to_fs);
+	bch2_darray_str_exit(&devs);
 
-	if (IS_ERR(sb))
-		return ERR_CAST(sb);
+	if (IS_ERR(sb)) {
+		ret = PTR_ERR(sb);
+		ret = bch2_err_class(ret);
+		return ERR_PTR(ret);
+	}
 
 	c = sb->s_fs_info;
 
@@ -1564,9 +1944,12 @@ got_sb:
 #endif
 	sb->s_xattr		= bch2_xattr_handlers;
 	sb->s_magic		= BCACHEFS_STATFS_MAGIC;
-	sb->s_time_gran		= c->sb.time_precision;
+	sb->s_time_gran		= c->sb.nsec_per_time_unit;
+	sb->s_time_min		= div_s64(S64_MIN, c->sb.time_units_per_sec) + 1;
+	sb->s_time_max		= div_s64(S64_MAX, c->sb.time_units_per_sec);
+	sb->s_uuid		= c->sb.user_uuid;
 	c->vfs_sb		= sb;
-	strlcpy(sb->s_id, c->name, sizeof(sb->s_id));
+	strscpy(sb->s_id, c->name, sizeof(sb->s_id));
 
 	ret = super_setup_bdi(sb);
 	if (ret)
@@ -1574,7 +1957,7 @@ got_sb:
 
 	sb->s_bdi->ra_pages		= VM_READAHEAD_PAGES;
 
-	for_each_online_member(ca, c, i) {
+	for_each_online_member(c, ca) {
 		struct block_device *bdev = ca->disk_sb.bdev;
 
 		/* XXX: create an anonymous device for multi device filesystems */
@@ -1584,18 +1967,20 @@ got_sb:
 		break;
 	}
 
+	c->dev = sb->s_dev;
+
 #ifdef CONFIG_BCACHEFS_POSIX_ACL
 	if (c->opts.acl)
 		sb->s_flags	|= SB_POSIXACL;
 #endif
 
-	vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_INO);
-	if (IS_ERR(vinode)) {
-		bch_err(c, "error mounting: error getting root inode %i",
-			(int) PTR_ERR(vinode));
-		ret = PTR_ERR(vinode);
+	sb->s_shrink->seeks = 0;
+
+	vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_SUBVOL_INUM);
+	ret = PTR_ERR_OR_ZERO(vinode);
+	bch_err_msg(c, ret, "mounting: error getting root inode");
+	if (ret)
 		goto err_put_super;
-	}
 
 	sb->s_root = d_make_root(vinode);
 	if (!sb->s_root) {
@@ -1610,7 +1995,7 @@ out:
 
 err_put_super:
 	deactivate_locked_super(sb);
-	return ERR_PTR(ret);
+	return ERR_PTR(bch2_err_class(ret));
 }
 
 static void bch2_kill_sb(struct super_block *sb)
@@ -1634,15 +2019,14 @@ MODULE_ALIAS_FS("bcachefs");
 void bch2_vfs_exit(void)
 {
 	unregister_filesystem(&bcache_fs_type);
-	if (bch2_inode_cache)
-		kmem_cache_destroy(bch2_inode_cache);
+	kmem_cache_destroy(bch2_inode_cache);
 }
 
 int __init bch2_vfs_init(void)
 {
 	int ret = -ENOMEM;
 
-	bch2_inode_cache = KMEM_CACHE(bch_inode_info, 0);
+	bch2_inode_cache = KMEM_CACHE(bch_inode_info, SLAB_RECLAIM_ACCOUNT);
 	if (!bch2_inode_cache)
 		goto err;