bcachefs: Kill for_each_btree_key()

for_each_btree_key() handles transaction restarts, like
for_each_btree_key2(), but only calls bch2_trans_begin() after a
transaction restart - for_each_btree_key2() wraps every loop iteration
in a transaction.

The for_each_btree_key() behaviour is problematic when it leads to
holding the SRCU lock that prevents key cache reclaim for an unbounded
amount of time - there's no real need to keep it around.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/alloc_background.c

@@ -544,8 +544,8 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 	u8 gen;
 	int ret;
 
-	for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-			   BTREE_ITER_PREFETCH, k, ret) {
+	ret = for_each_btree_key2(trans, iter, BTREE_ID_alloc, POS_MIN,
+				  BTREE_ITER_PREFETCH, k, ({
 		/*
 		 * Not a fsck error because this is checked/repaired by
 		 * bch2_check_alloc_key() which runs later:
@@ -572,8 +572,8 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 		}
 
 		g.v.gens[offset] = gen;
-	}
-	bch2_trans_iter_exit(trans, &iter);
+		0;
+	}));
 
 	if (have_bucket_gens_key && !ret)
 		ret = commit_do(trans, NULL, NULL,
@@ -582,7 +582,6 @@ int bch2_bucket_gens_init(struct bch_fs *c)
 
 	bch2_trans_put(trans);
 
-	if (ret)
 	bch_err_fn(c, ret);
 	return ret;
 }
@@ -601,8 +600,8 @@ int bch2_alloc_read(struct bch_fs *c)
 		const struct bch_bucket_gens *g;
 		u64 b;
 
-		for_each_btree_key(trans, iter, BTREE_ID_bucket_gens, POS_MIN,
-				   BTREE_ITER_PREFETCH, k, ret) {
+		ret = for_each_btree_key2(trans, iter, BTREE_ID_bucket_gens, POS_MIN,
+					  BTREE_ITER_PREFETCH, k, ({
 			u64 start = bucket_gens_pos_to_alloc(k.k->p, 0).offset;
 			u64 end = bucket_gens_pos_to_alloc(bpos_nosnap_successor(k.k->p), 0).offset;
 
@@ -624,13 +623,13 @@ int bch2_alloc_read(struct bch_fs *c)
 			     b < min_t(u64, ca->mi.nbuckets, end);
 			     b++)
 				*bucket_gen(ca, b) = g->gens[b & KEY_TYPE_BUCKET_GENS_MASK];
-		}
-		bch2_trans_iter_exit(trans, &iter);
+			0;
+		}));
 	} else {
 		struct bch_alloc_v4 a;
 
-		for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
-				   BTREE_ITER_PREFETCH, k, ret) {
+		ret = for_each_btree_key2(trans, iter, BTREE_ID_alloc, POS_MIN,
+					  BTREE_ITER_PREFETCH, k, ({
 			/*
 			 * Not a fsck error because this is checked/repaired by
 			 * bch2_check_alloc_key() which runs later:
@@ -641,16 +640,14 @@ int bch2_alloc_read(struct bch_fs *c)
 			ca = bch_dev_bkey_exists(c, k.k->p.inode);
 
 			*bucket_gen(ca, k.k->p.offset) = bch2_alloc_to_v4(k, &a)->gen;
-		}
-		bch2_trans_iter_exit(trans, &iter);
+			0;
+		}));
 	}
 
 	bch2_trans_put(trans);
 	up_read(&c->gc_lock);
 
-	if (ret)
 	bch_err_fn(c, ret);
-
 	return ret;
 }
 

fs/bcachefs/btree_gc.c

@@ -1665,7 +1665,6 @@ static int bch2_gc_reflink_done(struct bch_fs *c, bool metadata_only)
 static int bch2_gc_reflink_start(struct bch_fs *c,
 				 bool metadata_only)
 {
-	struct btree_trans *trans;
 	struct btree_iter iter;
 	struct bkey_s_c k;
 	struct reflink_gc *r;
@@ -1674,11 +1673,11 @@ static int bch2_gc_reflink_start(struct bch_fs *c,
 	if (metadata_only)
 		return 0;
 
-	trans = bch2_trans_get(c);
 	c->reflink_gc_nr = 0;
 
-	for_each_btree_key(trans, iter, BTREE_ID_reflink, POS_MIN,
-			   BTREE_ITER_PREFETCH, k, ret) {
+	ret = bch2_trans_run(c,
+		for_each_btree_key2(trans, iter, BTREE_ID_reflink, POS_MIN,
+				  BTREE_ITER_PREFETCH, k, ({
 			const __le64 *refcount = bkey_refcount_c(k);
 
 			if (!refcount)
@@ -1694,10 +1693,10 @@ static int bch2_gc_reflink_start(struct bch_fs *c,
 			r->offset	= k.k->p.offset;
 			r->size		= k.k->size;
 			r->refcount	= 0;
-	}
-	bch2_trans_iter_exit(trans, &iter);
+			0;
+		})));
 
-	bch2_trans_put(trans);
+	bch_err_fn(c, ret);
 	return ret;
 }
 

fs/bcachefs/btree_iter.h

@@ -777,7 +777,7 @@ transaction_restart:							\
 			    (_do) ?: bch2_trans_commit(_trans, (_disk_res),\
 					(_journal_seq), (_commit_flags)))
 
-#define for_each_btree_key(_trans, _iter, _btree_id,			\
+#define for_each_btree_key_old(_trans, _iter, _btree_id,			\
 			   _start, _flags, _k, _ret)			\
 	for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id),	\
 				  (_start), (_flags));			\

fs/bcachefs/ec.c

@@ -1833,7 +1833,6 @@ void bch2_fs_ec_flush(struct bch_fs *c)
 
 int bch2_stripes_read(struct bch_fs *c)
 {
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter iter;
 	struct bkey_s_c k;
 	const struct bch_stripe *s;
@@ -1841,8 +1840,9 @@ int bch2_stripes_read(struct bch_fs *c)
 	unsigned i;
 	int ret;
 
-	for_each_btree_key(trans, iter, BTREE_ID_stripes, POS_MIN,
-			   BTREE_ITER_PREFETCH, k, ret) {
+	ret = bch2_trans_run(c,
+		for_each_btree_key2(trans, iter, BTREE_ID_stripes, POS_MIN,
+				    BTREE_ITER_PREFETCH, k, ({
 			if (k.k->type != KEY_TYPE_stripe)
 				continue;
 
@@ -1863,14 +1863,10 @@ int bch2_stripes_read(struct bch_fs *c)
 				m->blocks_nonempty += !!stripe_blockcount_get(s, i);
 
 			bch2_stripes_heap_insert(c, m, k.k->p.offset);
-	}
-	bch2_trans_iter_exit(trans, &iter);
+			0;
+		})));
 
-	bch2_trans_put(trans);
-
-	if (ret)
 	bch_err_fn(c, ret);
-
 	return ret;
 }
 

fs/bcachefs/fsck.c

@@ -589,13 +589,12 @@ static int get_inodes_all_snapshots(struct btree_trans *trans,
 	struct bch_fs *c = trans->c;
 	struct btree_iter iter;
 	struct bkey_s_c k;
-	u32 restart_count = trans->restart_count;
 	int ret;
 
 	w->recalculate_sums = false;
 	w->inodes.nr = 0;
 
-	for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, inum),
+	for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
 				     BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
 		if (k.k->p.offset != inum)
 			break;
@@ -609,8 +608,7 @@ static int get_inodes_all_snapshots(struct btree_trans *trans,
 		return ret;
 
 	w->first_this_inode = true;
-
-	return trans_was_restarted(trans, restart_count);
+	return 0;
 }
 
 static struct inode_walker_entry *
@@ -2146,19 +2144,14 @@ int bch2_check_directory_structure(struct bch_fs *c)
 	pathbuf path = { 0, };
 	int ret;
 
-	for_each_btree_key(trans, iter, BTREE_ID_inodes, POS_MIN,
+	for_each_btree_key_old(trans, iter, BTREE_ID_inodes, POS_MIN,
 			       BTREE_ITER_INTENT|
 			       BTREE_ITER_PREFETCH|
 			       BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
 		if (!bkey_is_inode(k.k))
 			continue;
 
-		ret = bch2_inode_unpack(k, &u);
-		if (ret) {
-			/* Should have been caught earlier in fsck: */
-			bch_err(c, "error unpacking inode %llu: %i", k.k->p.offset, ret);
-			break;
-		}
+		BUG_ON(bch2_inode_unpack(k, &u));
 
 		if (u.bi_flags & BCH_INODE_unlinked)
 			continue;
@@ -2170,6 +2163,7 @@ int bch2_check_directory_structure(struct bch_fs *c)
 	bch2_trans_iter_exit(trans, &iter);
 	bch2_trans_put(trans);
 	darray_exit(&path);
+
 	bch_err_fn(c, ret);
 	return ret;
 }
@@ -2255,17 +2249,16 @@ static int check_nlinks_find_hardlinks(struct bch_fs *c,
 				       struct nlink_table *t,
 				       u64 start, u64 *end)
 {
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct btree_iter iter;
 	struct bkey_s_c k;
 	struct bch_inode_unpacked u;
-	int ret = 0;
 
-	for_each_btree_key(trans, iter, BTREE_ID_inodes,
+	int ret = bch2_trans_run(c,
+		for_each_btree_key2(trans, iter, BTREE_ID_inodes,
 					  POS(0, start),
 					  BTREE_ITER_INTENT|
 					  BTREE_ITER_PREFETCH|
-			   BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+					  BTREE_ITER_ALL_SNAPSHOTS, k, ({
 			if (!bkey_is_inode(k.k))
 				continue;
 
@@ -2288,14 +2281,10 @@ static int check_nlinks_find_hardlinks(struct bch_fs *c,
 				ret = 0;
 				break;
 			}
+			0;
+		})));
 
-	}
-	bch2_trans_iter_exit(trans, &iter);
-	bch2_trans_put(trans);
-
-	if (ret)
-		bch_err(c, "error in fsck: btree error %i while walking inodes", ret);
-
+	bch_err_fn(c, ret);
 	return ret;
 }
 
@@ -2303,19 +2292,18 @@ noinline_for_stack
 static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links,
 				     u64 range_start, u64 range_end)
 {
-	struct btree_trans *trans = bch2_trans_get(c);
 	struct snapshots_seen s;
 	struct btree_iter iter;
 	struct bkey_s_c k;
 	struct bkey_s_c_dirent d;
-	int ret;
 
 	snapshots_seen_init(&s);
 
-	for_each_btree_key(trans, iter, BTREE_ID_dirents, POS_MIN,
+	int ret = bch2_trans_run(c,
+		for_each_btree_key2(trans, iter, BTREE_ID_dirents, POS_MIN,
 				    BTREE_ITER_INTENT|
 				    BTREE_ITER_PREFETCH|
-			   BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+				    BTREE_ITER_ALL_SNAPSHOTS, k, ({
 			ret = snapshots_seen_update(c, &s, iter.btree_id, k.k->p);
 			if (ret)
 				break;
@@ -2331,14 +2319,12 @@ static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links
 						 bch2_snapshot_equiv(c, d.k->p.snapshot));
 				break;
 			}
-	}
-	bch2_trans_iter_exit(trans, &iter);
-
-	if (ret)
-		bch_err(c, "error in fsck: btree error %i while walking dirents", ret);
+			0;
+		})));
 
-	bch2_trans_put(trans);
 	snapshots_seen_exit(&s);
+
+	bch_err_fn(c, ret);
 	return ret;
 }
 

fs/bcachefs/inode.c

@@ -1168,29 +1168,29 @@ int bch2_delete_dead_inodes(struct bch_fs *c)
 	 * but we can't retry because the btree write buffer won't have been
 	 * flushed and we'd spin:
 	 */
-	for_each_btree_key(trans, iter, BTREE_ID_deleted_inodes, POS_MIN,
-			   BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
-		ret = commit_do(trans, NULL, NULL,
-				BCH_TRANS_COMMIT_no_enospc|
-				BCH_TRANS_COMMIT_lazy_rw,
-			may_delete_deleted_inode(trans, &iter, k.k->p, &need_another_pass));
-		if (ret < 0)
-			break;
-
-		if (ret) {
-			if (!test_bit(BCH_FS_rw, &c->flags)) {
-				bch2_trans_unlock(trans);
-				bch2_fs_lazy_rw(c);
-			}
-
+	ret = for_each_btree_key_commit(trans, iter, BTREE_ID_deleted_inodes, POS_MIN,
+					BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+					NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
+		ret = may_delete_deleted_inode(trans, &iter, k.k->p, &need_another_pass);
+		if (ret > 0) {
 			bch_verbose(c, "deleting unlinked inode %llu:%u", k.k->p.offset, k.k->p.snapshot);
 
 			ret = bch2_inode_rm_snapshot(trans, k.k->p.offset, k.k->p.snapshot);
-			if (ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart))
-				break;
+			/*
+			 * We don't want to loop here: a transaction restart
+			 * error here means we handled a transaction restart and
+			 * we're actually done, but if we loop we'll retry the
+			 * same key because the write buffer hasn't been flushed
+			 * yet
+			 */
+			if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
+				ret = 0;
+				continue;
 			}
 		}
-	bch2_trans_iter_exit(trans, &iter);
+
+		ret;
+	}));
 
 	if (!ret && need_another_pass) {
 		ret = bch2_btree_write_buffer_flush_sync(trans);

fs/bcachefs/move.c

@@ -377,8 +377,8 @@ struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
 
 		io_opts->d.nr = 0;
 
-		for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, extent_k.k->p.inode),
-				   BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+		ret = for_each_btree_key2(trans, iter, BTREE_ID_inodes, POS(0, extent_k.k->p.inode),
+					  BTREE_ITER_ALL_SNAPSHOTS, k, ({
 			if (k.k->p.offset != extent_k.k->p.inode)
 				break;
 
@@ -391,11 +391,8 @@ struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
 			struct snapshot_io_opts_entry e = { .snapshot = k.k->p.snapshot };
 			bch2_inode_opts_get(&e.io_opts, trans->c, &inode);
 
-			ret = darray_push(&io_opts->d, e);
-			if (ret)
-				break;
-		}
-		bch2_trans_iter_exit(trans, &iter);
+			darray_push(&io_opts->d, e);
+		}));
 		io_opts->cur_inum = extent_k.k->p.inode;
 	}
 

fs/bcachefs/recovery.c

@@ -534,7 +534,8 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
 	keys->gap = keys->nr;
 
 	set_bit(BCH_FS_may_go_rw, &c->flags);
-	if (keys->nr || c->opts.fsck)
+
+	if (keys->nr || c->opts.fsck || !c->sb.clean)
 		return bch2_fs_read_write_early(c);
 	return 0;
 }

fs/bcachefs/snapshot.c

@@ -1410,19 +1410,16 @@ int bch2_delete_dead_snapshots(struct bch_fs *c)
 		goto err;
 	}
 
-	for_each_btree_key(trans, iter, BTREE_ID_snapshots,
-			   POS_MIN, 0, k, ret) {
+	ret = for_each_btree_key2(trans, iter, BTREE_ID_snapshots,
+				  POS_MIN, 0, k, ({
 		if (k.k->type != KEY_TYPE_snapshot)
 			continue;
 
 		snap = bkey_s_c_to_snapshot(k);
-		if (BCH_SNAPSHOT_DELETED(snap.v)) {
-			ret = snapshot_list_add(c, &deleted, k.k->p.offset);
-			if (ret)
-				break;
-		}
-	}
-	bch2_trans_iter_exit(trans, &iter);
+		BCH_SNAPSHOT_DELETED(snap.v)
+			? snapshot_list_add(c, &deleted, k.k->p.offset)
+			: 0;
+	}));
 
 	if (ret) {
 		bch_err_msg(c, ret, "walking snapshots");
@@ -1469,18 +1466,20 @@ int bch2_delete_dead_snapshots(struct bch_fs *c)
 	bch2_trans_unlock(trans);
 	down_write(&c->snapshot_create_lock);
 
-	for_each_btree_key(trans, iter, BTREE_ID_snapshots,
-			   POS_MIN, 0, k, ret) {
+	ret = for_each_btree_key2(trans, iter, BTREE_ID_snapshots,
+				  POS_MIN, 0, k, ({
 		u32 snapshot = k.k->p.offset;
 		u32 equiv = bch2_snapshot_equiv(c, snapshot);
 
-		if (equiv != snapshot)
-			snapshot_list_add(c, &deleted_interior, snapshot);
-	}
-	bch2_trans_iter_exit(trans, &iter);
+		equiv != snapshot
+			? snapshot_list_add(c, &deleted_interior, snapshot)
+			: 0;
+	}));
 
-	if (ret)
+	if (ret) {
+		bch_err_msg(c, ret, "walking snapshots");
 		goto err_create_lock;
+	}
 
 	/*
 	 * Fixing children of deleted snapshots can't be done completely