Update bcachefs sources to 6afa1fcb13 bcachefs: Clean up error reporting in the startup path

12 files changed, 182 insertions, 286 deletions

diff --git a/libbcachefs/acl.c b/libbcachefs/acl.c
index d0c6878b..0cde2638 100644
--- a/libbcachefs/acl.c
+++ b/libbcachefs/acl.c
@@ -331,8 +331,7 @@ retry:
 	inode_u.bi_mode		= mode;
 
 	ret =   bch2_inode_write(&trans, &inode_iter, &inode_u) ?:
-		bch2_trans_commit(&trans, NULL,
-				  &inode->ei_journal_seq, 0);
+		bch2_trans_commit(&trans, NULL, NULL, 0);
 btree_err:
 	bch2_trans_iter_exit(&trans, &inode_iter);
 
diff --git a/libbcachefs/buckets.c b/libbcachefs/buckets.c
index a762a31b..c3542d3c 100644
--- a/libbcachefs/buckets.c
+++ b/libbcachefs/buckets.c
@@ -1203,11 +1203,15 @@ static int bch2_mark_reflink_p(struct btree_trans *trans,
 	struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
 	struct reflink_gc *ref;
 	size_t l, r, m;
-	u64 idx = le64_to_cpu(p.v->idx) - le32_to_cpu(p.v->front_pad);
-	u64 end_idx = le64_to_cpu(p.v->idx) + p.k->size +
-		le32_to_cpu(p.v->back_pad);
+	u64 idx = le64_to_cpu(p.v->idx);
+	u64 end = le64_to_cpu(p.v->idx) + p.k->size;
 	int ret = 0;
 
+	if (c->sb.version >= bcachefs_metadata_version_reflink_p_fix) {
+		idx -= le32_to_cpu(p.v->front_pad);
+		end += le32_to_cpu(p.v->back_pad);
+	}
+
 	l = 0;
 	r = c->reflink_gc_nr;
 	while (l < r) {
@@ -1220,7 +1224,7 @@ static int bch2_mark_reflink_p(struct btree_trans *trans,
 			r = m;
 	}
 
-	while (idx < end_idx && !ret)
+	while (idx < end && !ret)
 		ret = __bch2_mark_reflink_p(c, p, &idx, flags, l++);
 
 	return ret;
diff --git a/libbcachefs/fs-io.c b/libbcachefs/fs-io.c
index 684488ef..a2367b3e 100644
--- a/libbcachefs/fs-io.c
+++ b/libbcachefs/fs-io.c
@@ -1126,7 +1126,6 @@ static void bch2_writepage_io_alloc(struct bch_fs *c,
 	op			= &w->io->op;
 	bch2_write_op_init(op, c, w->opts);
 	op->target		= w->opts.foreground_target;
-	op_journal_seq_set(op, &inode->ei_journal_seq);
 	op->nr_replicas		= nr_replicas;
 	op->res.nr_replicas	= nr_replicas;
 	op->write_point		= writepoint_hashed(inode->ei_last_dirtied);
@@ -1936,7 +1935,6 @@ static long bch2_dio_write_loop(struct dio_write *dio)
 		bch2_write_op_init(&dio->op, c, io_opts(c, &inode->ei_inode));
 		dio->op.end_io		= bch2_dio_write_loop_async;
 		dio->op.target		= dio->op.opts.foreground_target;
-		op_journal_seq_set(&dio->op, &inode->ei_journal_seq);
 		dio->op.write_point	= writepoint_hashed((unsigned long) current);
 		dio->op.nr_replicas	= dio->op.opts.data_replicas;
 		dio->op.subvol		= inode->ei_subvol;
@@ -2168,27 +2166,33 @@ unlock:
 
 /* fsync: */
 
-int bch2_fsync(struct file *file, loff_t start, loff_t end, int datasync)
+/*
+ * inode->ei_inode.bi_journal_seq won't be up to date since it's set in an
+ * insert trigger: look up the btree inode instead
+ */
+static int bch2_flush_inode(struct bch_fs *c, subvol_inum inum)
 {
-	struct bch_inode_info *inode = file_bch_inode(file);
-	struct bch_fs *c = inode->v.i_sb->s_fs_info;
-	int ret, ret2;
+	struct bch_inode_unpacked inode;
+	int ret;
 
-	ret = file_write_and_wait_range(file, start, end);
+	if (c->opts.journal_flush_disabled)
+		return 0;
+
+	ret = bch2_inode_find_by_inum(c, inum, &inode);
 	if (ret)
 		return ret;
 
-	if (datasync && !(inode->v.i_state & I_DIRTY_DATASYNC))
-		goto out;
+	return bch2_journal_flush_seq(&c->journal, inode.bi_journal_seq);
+}
 
-	ret = sync_inode_metadata(&inode->v, 1);
-	if (ret)
-		return ret;
-out:
-	if (!c->opts.journal_flush_disabled)
-		ret = bch2_journal_flush_seq(&c->journal,
-					     inode->ei_journal_seq);
-	ret2 = file_check_and_advance_wb_err(file);
+int bch2_fsync(struct file *file, loff_t start, loff_t end, int datasync)
+{
+	struct bch_inode_info *inode = file_bch_inode(file);
+	struct bch_fs *c = inode->v.i_sb->s_fs_info;
+	int ret, ret2 = 0;
+
+	ret = file_write_and_wait_range(file, start, end);
+	ret2 = bch2_flush_inode(c, inode_inum(inode));
 
 	return ret ?: ret2;
 }
@@ -2452,7 +2456,7 @@ int bch2_truncate(struct user_namespace *mnt_userns,
 
 	ret = bch2_fpunch(c, inode_inum(inode),
 			round_up(iattr->ia_size, block_bytes(c)) >> 9,
-			U64_MAX, &inode->ei_journal_seq, &i_sectors_delta);
+			U64_MAX, &i_sectors_delta);
 	i_sectors_acct(c, inode, NULL, i_sectors_delta);
 
 	if (unlikely(ret))
@@ -2512,7 +2516,6 @@ static long bchfs_fpunch(struct bch_inode_info *inode, loff_t offset, loff_t len
 
 		ret = bch2_fpunch(c, inode_inum(inode),
 				  discard_start, discard_end,
-				  &inode->ei_journal_seq,
 				  &i_sectors_delta);
 		i_sectors_acct(c, inode, NULL, i_sectors_delta);
 	}
@@ -2591,7 +2594,6 @@ static long bchfs_fcollapse_finsert(struct bch_inode_info *inode,
 
 		ret = bch2_fpunch(c, inode_inum(inode),
 				  offset >> 9, (offset + len) >> 9,
-				  &inode->ei_journal_seq,
 				  &i_sectors_delta);
 		i_sectors_acct(c, inode, NULL, i_sectors_delta);
 
@@ -2695,8 +2697,7 @@ reassemble:
 		ret =   bch2_btree_iter_traverse(&del) ?:
 			bch2_trans_update(&trans, &del, &delete, trigger_flags) ?:
 			bch2_trans_update(&trans, &dst, copy.k, trigger_flags) ?:
-			bch2_trans_commit(&trans, &disk_res,
-					  &inode->ei_journal_seq,
+			bch2_trans_commit(&trans, &disk_res, NULL,
 					  BTREE_INSERT_NOFAIL);
 		bch2_disk_reservation_put(c, &disk_res);
 
@@ -2807,7 +2808,7 @@ static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
 
 		ret = bch2_extent_update(&trans, inode_inum(inode), &iter,
 					 &reservation.k_i,
-				&disk_res, &inode->ei_journal_seq,
+				&disk_res, NULL,
 				0, &i_sectors_delta, true);
 		i_sectors_acct(c, inode, &quota_res, i_sectors_delta);
 bkey_err:
@@ -3011,7 +3012,6 @@ loff_t bch2_remap_file_range(struct file *file_src, loff_t pos_src,
 			       inode_inum(dst), pos_dst >> 9,
 			       inode_inum(src), pos_src >> 9,
 			       aligned_len >> 9,
-			       &dst->ei_journal_seq,
 			       pos_dst + len, &i_sectors_delta);
 	if (ret < 0)
 		goto err;
@@ -3029,10 +3029,9 @@ loff_t bch2_remap_file_range(struct file *file_src, loff_t pos_src,
 		i_size_write(&dst->v, pos_dst + ret);
 	spin_unlock(&dst->v.i_lock);
 
-	if (((file_dst->f_flags & (__O_SYNC | O_DSYNC)) ||
-	     IS_SYNC(file_inode(file_dst))) &&
-	    !c->opts.journal_flush_disabled)
-		ret = bch2_journal_flush_seq(&c->journal, dst->ei_journal_seq);
+	if ((file_dst->f_flags & (__O_SYNC | O_DSYNC)) ||
+	    IS_SYNC(file_inode(file_dst)))
+		ret = bch2_flush_inode(c, inode_inum(dst));
 err:
 	bch2_unlock_inodes(INODE_LOCK|INODE_PAGECACHE_BLOCK, src, dst);
 
diff --git a/libbcachefs/fs.c b/libbcachefs/fs.c
index 9c5d16c6..74de7bc1 100644
--- a/libbcachefs/fs.c
+++ b/libbcachefs/fs.c
@@ -40,25 +40,6 @@ static void bch2_vfs_inode_init(struct bch_fs *, subvol_inum,
 				struct bch_inode_info *,
 				struct bch_inode_unpacked *);
 
-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);
-}
-
 static void __pagecache_lock_put(struct pagecache_lock *lock, long i)
 {
 	BUG_ON(atomic_long_read(&lock->v) == 0);
@@ -151,9 +132,7 @@ retry:
 				BTREE_ITER_INTENT) ?:
 		(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_NOFAIL);
+		bch2_trans_commit(&trans, NULL, NULL, BTREE_INSERT_NOFAIL);
 
 	/*
 	 * the btree node lock protects inode->ei_inode, not ei_update_lock;
@@ -328,7 +307,6 @@ err_before_quota:
 	if (!(flags & BCH_CREATE_TMPFILE)) {
 		bch2_inode_update_after_write(c, dir, &dir_u,
 					      ATTR_MTIME|ATTR_CTIME);
-		journal_seq_copy(c, dir, journal_seq);
 		mutex_unlock(&dir->ei_update_lock);
 	}
 
@@ -336,7 +314,6 @@ err_before_quota:
 	inum.inum = inode_u.bi_inum;
 
 	bch2_vfs_inode_init(c, inum, inode, &inode_u);
-	journal_seq_copy(c, inode, journal_seq);
 
 	set_cached_acl(&inode->v, ACL_TYPE_ACCESS, acl);
 	set_cached_acl(&inode->v, ACL_TYPE_DEFAULT, default_acl);
@@ -361,7 +338,6 @@ err_before_quota:
 		 * 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);
 
@@ -445,7 +421,7 @@ static int __bch2_link(struct bch_fs *c,
 	mutex_lock(&inode->ei_update_lock);
 	bch2_trans_init(&trans, c, 4, 1024);
 
-	ret = __bch2_trans_do(&trans, NULL, &inode->ei_journal_seq, 0,
+	ret = __bch2_trans_do(&trans, NULL, NULL, 0,
 			bch2_link_trans(&trans,
 					inode_inum(dir),   &dir_u,
 					inode_inum(inode), &inode_u,
@@ -454,7 +430,6 @@ static int __bch2_link(struct bch_fs *c,
 	if (likely(!ret)) {
 		BUG_ON(inode_u.bi_inum != inode->v.i_ino);
 
-		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_CTIME);
@@ -497,7 +472,7 @@ int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
 	bch2_lock_inodes(INODE_UPDATE_LOCK, dir, inode);
 	bch2_trans_init(&trans, c, 4, 1024);
 
-	ret = __bch2_trans_do(&trans, NULL, &dir->ei_journal_seq,
+	ret = __bch2_trans_do(&trans, NULL, NULL,
 			      BTREE_INSERT_NOFAIL,
 			bch2_unlink_trans(&trans,
 					  inode_inum(dir), &dir_u,
@@ -507,7 +482,6 @@ int __bch2_unlink(struct inode *vdir, struct dentry *dentry,
 	if (likely(!ret)) {
 		BUG_ON(inode_u.bi_inum != inode->v.i_ino);
 
-		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,
@@ -549,8 +523,6 @@ static int bch2_symlink(struct user_namespace *mnt_userns,
 	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;
@@ -585,7 +557,6 @@ static int bch2_rename2(struct user_namespace *mnt_userns,
 		? BCH_RENAME_EXCHANGE
 		: dst_dentry->d_inode
 		? BCH_RENAME_OVERWRITE : BCH_RENAME;
-	u64 journal_seq = 0;
 	int ret;
 
 	if (flags & ~(RENAME_NOREPLACE|RENAME_EXCHANGE))
@@ -625,7 +596,7 @@ static int bch2_rename2(struct user_namespace *mnt_userns,
 			goto err;
 	}
 
-	ret = __bch2_trans_do(&trans, NULL, &journal_seq, 0,
+	ret = __bch2_trans_do(&trans, NULL, NULL, 0,
 			bch2_rename_trans(&trans,
 					  inode_inum(src_dir), &src_dir_u,
 					  inode_inum(dst_dir), &dst_dir_u,
@@ -643,23 +614,17 @@ static int bch2_rename2(struct user_namespace *mnt_userns,
 
 	bch2_inode_update_after_write(c, src_dir, &src_dir_u,
 				      ATTR_MTIME|ATTR_CTIME);
-	journal_seq_copy(c, src_dir, journal_seq);
 
-	if (src_dir != dst_dir) {
+	if (src_dir != dst_dir)
 		bch2_inode_update_after_write(c, 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,
 				      ATTR_CTIME);
-	journal_seq_copy(c, src_inode, journal_seq);
 
-	if (dst_inode) {
+	if (dst_inode)
 		bch2_inode_update_after_write(c, dst_inode, &dst_inode_u,
 					      ATTR_CTIME);
-		journal_seq_copy(c, dst_inode, journal_seq);
-	}
 err:
 	bch2_trans_exit(&trans);
 
@@ -766,8 +731,7 @@ retry:
 	}
 
 	ret =   bch2_inode_write(&trans, &inode_iter, &inode_u) ?:
-		bch2_trans_commit(&trans, NULL,
-				  &inode->ei_journal_seq,
+		bch2_trans_commit(&trans, NULL, NULL,
 				  BTREE_INSERT_NOFAIL);
 btree_err:
 	bch2_trans_iter_exit(&trans, &inode_iter);
@@ -1201,7 +1165,6 @@ static void bch2_vfs_inode_init(struct bch_fs *c, subvol_inum inum,
 	inode->v.i_size		= bi->bi_size;
 
 	inode->ei_flags		= 0;
-	inode->ei_journal_seq	= bi->bi_journal_seq;
 	inode->ei_quota_reserved = 0;
 	inode->ei_qid		= bch_qid(bi);
 	inode->ei_subvol	= inum.subvol;
@@ -1240,7 +1203,6 @@ static struct inode *bch2_alloc_inode(struct super_block *sb)
 	mutex_init(&inode->ei_update_lock);
 	pagecache_lock_init(&inode->ei_pagecache_lock);
 	mutex_init(&inode->ei_quota_lock);
-	inode->ei_journal_seq = 0;
 
 	return &inode->v;
 }
diff --git a/libbcachefs/fs.h b/libbcachefs/fs.h
index bf62e80f..40212b3d 100644
--- a/libbcachefs/fs.h
+++ b/libbcachefs/fs.h
@@ -36,7 +36,6 @@ struct bch_inode_info {
 	unsigned long		ei_flags;
 
 	struct mutex		ei_update_lock;
-	u64			ei_journal_seq;
 	u64			ei_quota_reserved;
 	unsigned long		ei_last_dirtied;
 
diff --git a/libbcachefs/io.c b/libbcachefs/io.c
index 701e9d0e..7c9ea91d 100644
--- a/libbcachefs/io.c
+++ b/libbcachefs/io.c
@@ -376,7 +376,7 @@ int bch2_extent_update(struct btree_trans *trans,
  */
 int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
 		   subvol_inum inum, u64 end,
-		   u64 *journal_seq, s64 *i_sectors_delta)
+		   s64 *i_sectors_delta)
 {
 	struct bch_fs *c	= trans->c;
 	unsigned max_sectors	= KEY_SIZE_MAX & (~0 << c->block_bits);
@@ -414,7 +414,7 @@ int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
 		bch2_cut_back(end_pos, &delete);
 
 		ret = bch2_extent_update(trans, inum, iter, &delete,
-				&disk_res, journal_seq,
+				&disk_res, NULL,
 				0, i_sectors_delta, false);
 		bch2_disk_reservation_put(c, &disk_res);
 btree_err:
@@ -433,7 +433,7 @@ btree_err:
 }
 
 int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
-		u64 *journal_seq, s64 *i_sectors_delta)
+		s64 *i_sectors_delta)
 {
 	struct btree_trans trans;
 	struct btree_iter iter;
@@ -444,8 +444,7 @@ int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
 			     POS(inum.inum, start),
 			     BTREE_ITER_INTENT);
 
-	ret = bch2_fpunch_at(&trans, &iter, inum, end,
-			     journal_seq, i_sectors_delta);
+	ret = bch2_fpunch_at(&trans, &iter, inum, end, i_sectors_delta);
 
 	bch2_trans_iter_exit(&trans, &iter);
 	bch2_trans_exit(&trans);
diff --git a/libbcachefs/io.h b/libbcachefs/io.h
index 38efd39c..fbe46660 100644
--- a/libbcachefs/io.h
+++ b/libbcachefs/io.h
@@ -48,12 +48,6 @@ static inline u64 *op_journal_seq(struct bch_write_op *op)
 		? op->journal_seq_p : &op->journal_seq;
 }
 
-static inline void op_journal_seq_set(struct bch_write_op *op, u64 *journal_seq)
-{
-	op->journal_seq_p = journal_seq;
-	op->flags |= BCH_WRITE_JOURNAL_SEQ_PTR;
-}
-
 static inline struct workqueue_struct *index_update_wq(struct bch_write_op *op)
 {
 	return op->alloc_reserve == RESERVE_MOVINGGC
@@ -68,8 +62,8 @@ int bch2_extent_update(struct btree_trans *, subvol_inum,
 		       struct disk_reservation *, u64 *, u64, s64 *, bool);
 
 int bch2_fpunch_at(struct btree_trans *, struct btree_iter *,
-		   subvol_inum, u64, u64 *, s64 *);
-int bch2_fpunch(struct bch_fs *c, subvol_inum, u64, u64, u64 *, s64 *);
+		   subvol_inum, u64, s64 *);
+int bch2_fpunch(struct bch_fs *c, subvol_inum, u64, u64, s64 *);
 
 int bch2_write_index_default(struct bch_write_op *);
 
diff --git a/libbcachefs/reflink.c b/libbcachefs/reflink.c
index 8e66e639..d003f408 100644
--- a/libbcachefs/reflink.c
+++ b/libbcachefs/reflink.c
@@ -210,7 +210,7 @@ static struct bkey_s_c get_next_src(struct btree_iter *iter, struct bpos end)
 s64 bch2_remap_range(struct bch_fs *c,
 		     subvol_inum dst_inum, u64 dst_offset,
 		     subvol_inum src_inum, u64 src_offset,
-		     u64 remap_sectors, u64 *journal_seq,
+		     u64 remap_sectors,
 		     u64 new_i_size, s64 *i_sectors_delta)
 {
 	struct btree_trans trans;
@@ -281,7 +281,7 @@ s64 bch2_remap_range(struct bch_fs *c,
 					min(dst_end.offset,
 					    dst_iter.pos.offset +
 					    src_iter.pos.offset - src_want.offset),
-					journal_seq, i_sectors_delta);
+					i_sectors_delta);
 			continue;
 		}
 
@@ -320,7 +320,7 @@ s64 bch2_remap_range(struct bch_fs *c,
 				    dst_end.offset - dst_iter.pos.offset));
 
 		ret = bch2_extent_update(&trans, dst_inum, &dst_iter,
-					 new_dst.k, &disk_res, journal_seq,
+					 new_dst.k, &disk_res, NULL,
 					 new_i_size, i_sectors_delta,
 					 true);
 		bch2_disk_reservation_put(c, &disk_res);
@@ -347,7 +347,7 @@ s64 bch2_remap_range(struct bch_fs *c,
 		    inode_u.bi_size < new_i_size) {
 			inode_u.bi_size = new_i_size;
 			ret2  = bch2_inode_write(&trans, &inode_iter, &inode_u) ?:
-				bch2_trans_commit(&trans, NULL, journal_seq, 0);
+				bch2_trans_commit(&trans, NULL, NULL, 0);
 		}
 
 		bch2_trans_iter_exit(&trans, &inode_iter);
diff --git a/libbcachefs/reflink.h b/libbcachefs/reflink.h
index 4c1b8286..3745873f 100644
--- a/libbcachefs/reflink.h
+++ b/libbcachefs/reflink.h
@@ -58,6 +58,6 @@ static inline __le64 *bkey_refcount(struct bkey_i *k)
 }
 
 s64 bch2_remap_range(struct bch_fs *, subvol_inum, u64,
-		     subvol_inum, u64, u64, u64 *, u64, s64 *);
+		     subvol_inum, u64, u64, u64, s64 *);
 
 #endif /* _BCACHEFS_REFLINK_H */
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index dc8f6415..3744b6d5 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -588,48 +588,53 @@ void bch2_fs_stop(struct bch_fs *c)
 	bch2_fs_free(c);
 }
 
-static const char *bch2_fs_online(struct bch_fs *c)
+static int bch2_fs_online(struct bch_fs *c)
 {
 	struct bch_dev *ca;
-	const char *err = NULL;
 	unsigned i;
-	int ret;
+	int ret = 0;
 
 	lockdep_assert_held(&bch_fs_list_lock);
 
-	if (!list_empty(&c->list))
-		return NULL;
-
-	if (__bch2_uuid_to_fs(c->sb.uuid))
-		return "filesystem UUID already open";
+	if (__bch2_uuid_to_fs(c->sb.uuid)) {
+		bch_err(c, "filesystem UUID already open");
+		return -EINVAL;
+	}
 
 	ret = bch2_fs_chardev_init(c);
-	if (ret)
-		return "error creating character device";
+	if (ret) {
+		bch_err(c, "error creating character device");
+		return ret;
+	}
 
 	bch2_fs_debug_init(c);
 
-	if (kobject_add(&c->kobj, NULL, "%pU", c->sb.user_uuid.b) ||
-	    kobject_add(&c->internal, &c->kobj, "internal") ||
-	    kobject_add(&c->opts_dir, &c->kobj, "options") ||
-	    kobject_add(&c->time_stats, &c->kobj, "time_stats") ||
-	    bch2_opts_create_sysfs_files(&c->opts_dir))
-		return "error creating sysfs objects";
+	ret = kobject_add(&c->kobj, NULL, "%pU", c->sb.user_uuid.b) ?:
+	    kobject_add(&c->internal, &c->kobj, "internal") ?:
+	    kobject_add(&c->opts_dir, &c->kobj, "options") ?:
+	    kobject_add(&c->time_stats, &c->kobj, "time_stats") ?:
+	    bch2_opts_create_sysfs_files(&c->opts_dir);
+	if (ret) {
+		bch_err(c, "error creating sysfs objects");
+		return ret;
+	}
 
 	down_write(&c->state_lock);
 
-	err = "error creating sysfs objects";
-	for_each_member_device(ca, c, i)
-		if (bch2_dev_sysfs_online(c, ca)) {
+	for_each_member_device(ca, c, i) {
+		ret = bch2_dev_sysfs_online(c, ca);
+		if (ret) {
+			bch_err(c, "error creating sysfs objects");
 			percpu_ref_put(&ca->ref);
 			goto err;
 		}
+	}
 
+	BUG_ON(!list_empty(&c->list));
 	list_add(&c->list, &bch_fs_list);
-	err = NULL;
 err:
 	up_write(&c->state_lock);
-	return err;
+	return ret;
 }
 
 static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
@@ -637,13 +642,15 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	struct bch_sb_field_members *mi;
 	struct bch_fs *c;
 	unsigned i, iter_size;
-	const char *err;
+	int ret = 0;
 
 	pr_verbose_init(opts, "");
 
 	c = kvpmalloc(sizeof(struct bch_fs), GFP_KERNEL|__GFP_ZERO);
-	if (!c)
+	if (!c) {
+		c = ERR_PTR(-ENOMEM);
 		goto out;
+	}
 
 	__module_get(THIS_MODULE);
 
@@ -724,17 +731,16 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 
 	mutex_init(&c->sectors_available_lock);
 
-	if (percpu_init_rwsem(&c->mark_lock))
+	ret = percpu_init_rwsem(&c->mark_lock);
+	if (ret)
 		goto err;
 
 	mutex_lock(&c->sb_lock);
+	ret = bch2_sb_to_fs(c, sb);
+	mutex_unlock(&c->sb_lock);
 
-	if (bch2_sb_to_fs(c, sb)) {
-		mutex_unlock(&c->sb_lock);
+	if (ret)
 		goto err;
-	}
-
-	mutex_unlock(&c->sb_lock);
 
 	scnprintf(c->name, sizeof(c->name), "%pU", &c->sb.user_uuid);
 
@@ -745,8 +751,11 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	c->block_bits		= ilog2(c->opts.block_size);
 	c->btree_foreground_merge_threshold = BTREE_FOREGROUND_MERGE_THRESHOLD(c);
 
-	if (bch2_fs_init_fault("fs_alloc"))
+	if (bch2_fs_init_fault("fs_alloc")) {
+		bch_err(c, "fs_alloc fault injected");
+		ret = -EFAULT;
 		goto err;
+	}
 
 	iter_size = sizeof(struct sort_iter) +
 		(btree_blocks(c) + 1) * 2 *
@@ -776,21 +785,26 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 					btree_bytes(c)) ||
 	    mempool_init_kmalloc_pool(&c->large_bkey_pool, 1, 2048) ||
 	    !(c->unused_inode_hints = kcalloc(1U << c->inode_shard_bits,
-					      sizeof(u64), GFP_KERNEL)) ||
-	    bch2_io_clock_init(&c->io_clock[READ]) ||
-	    bch2_io_clock_init(&c->io_clock[WRITE]) ||
-	    bch2_fs_journal_init(&c->journal) ||
-	    bch2_fs_replicas_init(c) ||
-	    bch2_fs_btree_cache_init(c) ||
-	    bch2_fs_btree_key_cache_init(&c->btree_key_cache) ||
-	    bch2_fs_btree_iter_init(c) ||
-	    bch2_fs_btree_interior_update_init(c) ||
-	    bch2_fs_subvolumes_init(c) ||
-	    bch2_fs_io_init(c) ||
-	    bch2_fs_encryption_init(c) ||
-	    bch2_fs_compress_init(c) ||
-	    bch2_fs_ec_init(c) ||
-	    bch2_fs_fsio_init(c))
+					      sizeof(u64), GFP_KERNEL))) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	ret = bch2_io_clock_init(&c->io_clock[READ]) ?:
+	    bch2_io_clock_init(&c->io_clock[WRITE]) ?:
+	    bch2_fs_journal_init(&c->journal) ?:
+	    bch2_fs_replicas_init(c) ?:
+	    bch2_fs_btree_cache_init(c) ?:
+	    bch2_fs_btree_key_cache_init(&c->btree_key_cache) ?:
+	    bch2_fs_btree_iter_init(c) ?:
+	    bch2_fs_btree_interior_update_init(c) ?:
+	    bch2_fs_subvolumes_init(c) ?:
+	    bch2_fs_io_init(c) ?:
+	    bch2_fs_encryption_init(c) ?:
+	    bch2_fs_compress_init(c) ?:
+	    bch2_fs_ec_init(c) ?:
+	    bch2_fs_fsio_init(c);
+	if (ret)
 		goto err;
 
 	if (c->opts.nochanges)
@@ -799,8 +813,10 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	mi = bch2_sb_get_members(c->disk_sb.sb);
 	for (i = 0; i < c->sb.nr_devices; i++)
 		if (bch2_dev_exists(c->disk_sb.sb, mi, i) &&
-		    bch2_dev_alloc(c, i))
+		    bch2_dev_alloc(c, i)) {
+			ret = -EEXIST;
 			goto err;
+		}
 
 	bch2_journal_entry_res_resize(&c->journal,
 			&c->btree_root_journal_res,
@@ -811,18 +827,17 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 			(sizeof(struct jset_entry_clock) / sizeof(u64)) * 2);
 
 	mutex_lock(&bch_fs_list_lock);
-	err = bch2_fs_online(c);
+	ret = bch2_fs_online(c);
 	mutex_unlock(&bch_fs_list_lock);
-	if (err) {
-		bch_err(c, "bch2_fs_online() error: %s", err);
+
+	if (ret)
 		goto err;
-	}
 out:
-	pr_verbose_init(opts, "ret %i", c ? 0 : -ENOMEM);
+	pr_verbose_init(opts, "ret %i", PTR_ERR_OR_ZERO(c));
 	return c;
 err:
 	bch2_fs_free(c);
-	c = NULL;
+	c = ERR_PTR(ret);
 	goto out;
 }
 
@@ -862,7 +877,6 @@ static void print_mount_opts(struct bch_fs *c)
 
 int bch2_fs_start(struct bch_fs *c)
 {
-	const char *err = "cannot allocate memory";
 	struct bch_sb_field_members *mi;
 	struct bch_dev *ca;
 	time64_t now = ktime_get_real_seconds();
@@ -898,10 +912,11 @@ int bch2_fs_start(struct bch_fs *c)
 	if (ret)
 		goto err;
 
-	err = "dynamic fault";
 	ret = -EINVAL;
-	if (bch2_fs_init_fault("fs_start"))
+	if (bch2_fs_init_fault("fs_start")) {
+		bch_err(c, "fs_start fault injected");
 		goto err;
+	}
 
 	set_bit(BCH_FS_STARTED, &c->flags);
 
@@ -922,7 +937,6 @@ int bch2_fs_start(struct bch_fs *c)
 	if (c->opts.read_only || c->opts.nochanges) {
 		bch2_fs_read_only(c);
 	} else {
-		err = "error going read write";
 		ret = !test_bit(BCH_FS_RW, &c->flags)
 			? bch2_fs_read_write(c)
 			: bch2_fs_read_write_late(c);
@@ -940,25 +954,22 @@ err:
 	case BCH_FSCK_ERRORS_NOT_FIXED:
 		bch_err(c, "filesystem contains errors: please report this to the developers");
 		pr_cont("mount with -o fix_errors to repair\n");
-		err = "fsck error";
 		break;
 	case BCH_FSCK_REPAIR_UNIMPLEMENTED:
 		bch_err(c, "filesystem contains errors: please report this to the developers");
 		pr_cont("repair unimplemented: inform the developers so that it can be added\n");
-		err = "fsck error";
 		break;
 	case BCH_FSCK_REPAIR_IMPOSSIBLE:
 		bch_err(c, "filesystem contains errors, but repair impossible");
-		err = "fsck error";
 		break;
 	case BCH_FSCK_UNKNOWN_VERSION:
-		err = "unknown metadata version";;
+		bch_err(c, "unknown metadata version");
 		break;
 	case -ENOMEM:
-		err = "cannot allocate memory";
+		bch_err(c, "cannot allocate memory");
 		break;
 	case -EIO:
-		err = "IO error";
+		bch_err(c, "IO error");
 		break;
 	}
 
@@ -1378,7 +1389,7 @@ static void __bch2_dev_read_only(struct bch_fs *c, struct bch_dev *ca)
 	bch2_copygc_start(c);
 }
 
-static const char *__bch2_dev_read_write(struct bch_fs *c, struct bch_dev *ca)
+static int __bch2_dev_read_write(struct bch_fs *c, struct bch_dev *ca)
 {
 	lockdep_assert_held(&c->state_lock);
 
@@ -1387,10 +1398,7 @@ static const char *__bch2_dev_read_write(struct bch_fs *c, struct bch_dev *ca)
 	bch2_dev_allocator_add(c, ca);
 	bch2_recalc_capacity(c);
 
-	if (bch2_dev_allocator_start(ca))
-		return "error starting allocator thread";
-
-	return NULL;
+	return bch2_dev_allocator_start(ca);
 }
 
 int __bch2_dev_set_state(struct bch_fs *c, struct bch_dev *ca,
@@ -1416,9 +1424,8 @@ int __bch2_dev_set_state(struct bch_fs *c, struct bch_dev *ca,
 	bch2_write_super(c);
 	mutex_unlock(&c->sb_lock);
 
-	if (new_state == BCH_MEMBER_STATE_rw &&
-	    __bch2_dev_read_write(c, ca))
-		ret = -ENOMEM;
+	if (new_state == BCH_MEMBER_STATE_rw)
+		ret = __bch2_dev_read_write(c, ca);
 
 	rebalance_wakeup(c);
 
@@ -1702,8 +1709,8 @@ have_slot:
 		goto err_late;
 
 	if (ca->mi.state == BCH_MEMBER_STATE_rw) {
-		err = __bch2_dev_read_write(c, ca);
-		if (err)
+		ret = __bch2_dev_read_write(c, ca);
+		if (ret)
 			goto err_late;
 	}
 
@@ -1747,24 +1754,27 @@ int bch2_dev_online(struct bch_fs *c, const char *path)
 	dev_idx = sb.sb->dev_idx;
 
 	err = bch2_dev_in_fs(c->disk_sb.sb, sb.sb);
-	if (err)
+	if (err) {
+		bch_err(c, "error bringing %s online: %s", path, err);
 		goto err;
+	}
 
-	if (bch2_dev_attach_bdev(c, &sb)) {
-		err = "bch2_dev_attach_bdev() error";
+	ret = bch2_dev_attach_bdev(c, &sb);
+	if (ret)
 		goto err;
-	}
 
 	ca = bch_dev_locked(c, dev_idx);
 
-	if (bch2_trans_mark_dev_sb(c, ca)) {
-		err = "bch2_trans_mark_dev_sb() error";
+	ret = bch2_trans_mark_dev_sb(c, ca);
+	if (ret) {
+		bch_err(c, "error bringing %s online: error %i from bch2_trans_mark_dev_sb",
+			path, ret);
 		goto err;
 	}
 
 	if (ca->mi.state == BCH_MEMBER_STATE_rw) {
-		err = __bch2_dev_read_write(c, ca);
-		if (err)
+		ret = __bch2_dev_read_write(c, ca);
+		if (ret)
 			goto err;
 	}
 
@@ -1782,7 +1792,6 @@ int bch2_dev_online(struct bch_fs *c, const char *path)
 err:
 	up_write(&c->state_lock);
 	bch2_free_super(&sb);
-	bch_err(c, "error bringing %s online: %s", path, err);
 	return -EINVAL;
 }
 
@@ -1886,7 +1895,7 @@ struct bch_fs *bch2_fs_open(char * const *devices, unsigned nr_devices,
 	struct bch_sb_field_members *mi;
 	unsigned i, best_sb = 0;
 	const char *err;
-	int ret = -ENOMEM;
+	int ret = 0;
 
 	pr_verbose_init(opts, "");
 
@@ -1901,8 +1910,10 @@ struct bch_fs *bch2_fs_open(char * const *devices, unsigned nr_devices,
 	}
 
 	sb = kcalloc(nr_devices, sizeof(*sb), GFP_KERNEL);
-	if (!sb)
+	if (!sb) {
+		ret = -ENOMEM;
 		goto err;
+	}
 
 	for (i = 0; i < nr_devices; i++) {
 		ret = bch2_read_super(devices[i], &opts, &sb[i]);
@@ -1939,18 +1950,20 @@ struct bch_fs *bch2_fs_open(char * const *devices, unsigned nr_devices,
 		i++;
 	}
 
-	ret = -ENOMEM;
 	c = bch2_fs_alloc(sb[best_sb].sb, opts);
-	if (!c)
+	if (IS_ERR(c)) {
+		ret = PTR_ERR(c);
 		goto err;
+	}
 
-	err = "bch2_dev_online() error";
 	down_write(&c->state_lock);
-	for (i = 0; i < nr_devices; i++)
-		if (bch2_dev_attach_bdev(c, &sb[i])) {
+	for (i = 0; i < nr_devices; i++) {
+		ret = bch2_dev_attach_bdev(c, &sb[i]);
+		if (ret) {
 			up_write(&c->state_lock);
-			goto err_print;
+			goto err;
 		}
+	}
 	up_write(&c->state_lock);
 
 	err = "insufficient devices";
@@ -1973,10 +1986,11 @@ err_print:
 	       devices[0], err);
 	ret = -EINVAL;
 err:
-	if (c)
+	if (!IS_ERR_OR_NULL(c))
 		bch2_fs_stop(c);
-	for (i = 0; i < nr_devices; i++)
-		bch2_free_super(&sb[i]);
+	if (sb)
+		for (i = 0; i < nr_devices; i++)
+			bch2_free_super(&sb[i]);
 	c = ERR_PTR(ret);
 	goto out;
 }
@@ -2002,12 +2016,12 @@ static const char *__bch2_fs_open_incremental(struct bch_sb_handle *sb,
 		if (err)
 			goto err;
 	} else {
+		allocated_fs = true;
 		c = bch2_fs_alloc(sb->sb, opts);
-		err = "cannot allocate memory";
-		if (!c)
-			goto err;
 
-		allocated_fs = true;
+		err = "bch2_fs_alloc() error";
+		if (IS_ERR(c))
+			goto err;
 	}
 
 	err = "bch2_dev_online() error";
@@ -2033,7 +2047,7 @@ static const char *__bch2_fs_open_incremental(struct bch_sb_handle *sb,
 err:
 	mutex_unlock(&bch_fs_list_lock);
 
-	if (allocated_fs)
+	if (allocated_fs && !IS_ERR(c))
 		bch2_fs_stop(c);
 	else if (c)
 		closure_put(&c->cl);
diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c
index 51eb19b8..864be860 100644
--- a/libbcachefs/sysfs.c
+++ b/libbcachefs/sysfs.c
@@ -155,11 +155,6 @@ read_attribute(congested);
 
 read_attribute(btree_avg_write_size);
 
-read_attribute(bucket_quantiles_last_read);
-read_attribute(bucket_quantiles_last_write);
-read_attribute(bucket_quantiles_fragmentation);
-read_attribute(bucket_quantiles_oldest_gen);
-
 read_attribute(reserve_stats);
 read_attribute(btree_cache_size);
 read_attribute(compression_stats);
@@ -751,76 +746,6 @@ struct attribute *bch2_fs_time_stats_files[] = {
 	NULL
 };
 
-typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
-				 size_t, void *);
-
-static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
-				  size_t b, void *private)
-{
-	int rw = (private ? 1 : 0);
-
-	return atomic64_read(&c->io_clock[rw].now) - bucket(ca, b)->io_time[rw];
-}
-
-static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
-				       size_t b, void *private)
-{
-	struct bucket *g = bucket(ca, b);
-	return bucket_sectors_used(g->mark);
-}
-
-static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
-				     size_t b, void *private)
-{
-	return bucket_gc_gen(bucket(ca, b));
-}
-
-static int unsigned_cmp(const void *_l, const void *_r)
-{
-	const unsigned *l = _l;
-	const unsigned *r = _r;
-
-	return cmp_int(*l, *r);
-}
-
-static int quantiles_to_text(struct printbuf *out,
-			     struct bch_fs *c, struct bch_dev *ca,
-			     bucket_map_fn *fn, void *private)
-{
-	size_t i, n;
-	/* Compute 31 quantiles */
-	unsigned q[31], *p;
-
-	down_read(&ca->bucket_lock);
-	n = ca->mi.nbuckets;
-
-	p = vzalloc(n * sizeof(unsigned));
-	if (!p) {
-		up_read(&ca->bucket_lock);
-		return -ENOMEM;
-	}
-
-	for (i = ca->mi.first_bucket; i < n; i++)
-		p[i] = fn(c, ca, i, private);
-
-	sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
-	up_read(&ca->bucket_lock);
-
-	while (n &&
-	       !p[n - 1])
-		--n;
-
-	for (i = 0; i < ARRAY_SIZE(q); i++)
-		q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
-
-	vfree(p);
-
-	for (i = 0; i < ARRAY_SIZE(q); i++)
-		pr_buf(out, "%u ", q[i]);
-	pr_buf(out, "\n");
-	return 0;
-}
-
 static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca)
 {
 	enum alloc_reserve i;
@@ -982,15 +907,6 @@ SHOW(bch2_dev)
 		     clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 		     * 100 / CONGESTED_MAX);
 
-	if (attr == &sysfs_bucket_quantiles_last_read)
-		return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 0) ?: out.pos - buf;
-	if (attr == &sysfs_bucket_quantiles_last_write)
-		return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 1) ?: out.pos - buf;
-	if (attr == &sysfs_bucket_quantiles_fragmentation)
-		return quantiles_to_text(&out, c, ca, bucket_sectors_used_fn, NULL)  ?: out.pos - buf;
-	if (attr == &sysfs_bucket_quantiles_oldest_gen)
-		return quantiles_to_text(&out, c, ca, bucket_oldest_gen_fn, NULL)    ?: out.pos - buf;
-
 	if (attr == &sysfs_reserve_stats) {
 		reserve_stats_to_text(&out, ca);
 		return out.pos - buf;
@@ -1082,12 +998,6 @@ struct attribute *bch2_dev_files[] = {
 	&sysfs_io_latency_stats_write,
 	&sysfs_congested,
 
-	/* alloc info - other stats: */
-	&sysfs_bucket_quantiles_last_read,
-	&sysfs_bucket_quantiles_last_write,
-	&sysfs_bucket_quantiles_fragmentation,
-	&sysfs_bucket_quantiles_oldest_gen,
-
 	&sysfs_reserve_stats,
 
 	/* debug: */
diff --git a/libbcachefs/xattr.c b/libbcachefs/xattr.c
index bb5da310..464ed683 100644
--- a/libbcachefs/xattr.c
+++ b/libbcachefs/xattr.c
@@ -165,8 +165,24 @@ int bch2_xattr_set(struct btree_trans *trans, subvol_inum inum,
 		   const char *name, const void *value, size_t size,
 		   int type, int flags)
 {
+	struct btree_iter inode_iter = { NULL };
+	struct bch_inode_unpacked inode_u;
 	int ret;
 
+	/*
+	 * We need to do an inode update so that bi_journal_sync gets updated
+	 * and fsync works:
+	 *
+	 * Perhaps we should be updating bi_mtime too?
+	 */
+
+	ret   = bch2_inode_peek(trans, &inode_iter, &inode_u, inum, BTREE_ITER_INTENT) ?:
+		bch2_inode_write(trans, &inode_iter, &inode_u);
+	bch2_trans_iter_exit(trans, &inode_iter);
+
+	if (ret)
+		return ret;
+
 	if (value) {
 		struct bkey_i_xattr *xattr;
 		unsigned namelen = strlen(name);
@@ -352,7 +368,7 @@ static int bch2_xattr_set_handler(const struct xattr_handler *handler,
 	struct bch_fs *c = inode->v.i_sb->s_fs_info;
 	struct bch_hash_info hash = bch2_hash_info_init(c, &inode->ei_inode);
 
-	return bch2_trans_do(c, NULL, &inode->ei_journal_seq, 0,
+	return bch2_trans_do(c, NULL, NULL, 0,
 			bch2_xattr_set(&trans, inode_inum(inode), &hash,
 				       name, value, size,
 				       handler->flags, flags));