1 files changed, 68 insertions, 40 deletions

diff --git a/fs/bcachefs/io_misc.c b/fs/bcachefs/io_misc.c
index 1baf78594cca..bf72b1d2e2cb 100644
--- a/fs/bcachefs/io_misc.c
+++ b/fs/bcachefs/io_misc.c
@@ -43,7 +43,7 @@ int bch2_extent_fallocate(struct btree_trans *trans,
 	bch2_bkey_buf_init(&new);
 	closure_init_stack(&cl);
 
-	k = bch2_btree_iter_peek_slot(iter);
+	k = bch2_btree_iter_peek_slot(trans, iter);
 	ret = bkey_err(k);
 	if (ret)
 		return ret;
@@ -91,7 +91,7 @@ int bch2_extent_fallocate(struct btree_trans *trans,
 				opts.data_replicas,
 				BCH_WATERMARK_normal, 0, &cl, &wp);
 		if (bch2_err_matches(ret, BCH_ERR_operation_blocked))
-			ret = -BCH_ERR_transaction_restart_nested;
+			ret = bch_err_throw(c, transaction_restart_nested);
 		if (ret)
 			goto err;
 
@@ -113,11 +113,14 @@ int bch2_extent_fallocate(struct btree_trans *trans,
 err:
 	if (!ret && sectors_allocated)
 		bch2_increment_clock(c, sectors_allocated, WRITE);
-	if (should_print_err(ret))
-		bch_err_inum_offset_ratelimited(c,
-			inum.inum,
-			iter->pos.offset << 9,
-			"%s(): error: %s", __func__, bch2_err_str(ret));
+	if (should_print_err(ret)) {
+		struct printbuf buf = PRINTBUF;
+		lockrestart_do(trans,
+			bch2_inum_offset_err_msg_trans(trans, &buf, inum, iter->pos.offset << 9));
+		prt_printf(&buf, "fallocate error: %s", bch2_err_str(ret));
+		bch_err_ratelimited(c, "%s", buf.buf);
+		printbuf_exit(&buf);
+	}
 err_noprint:
 	bch2_open_buckets_put(c, &open_buckets);
 	bch2_disk_reservation_put(c, &disk_res);
@@ -125,8 +128,8 @@ err_noprint:
 	bch2_bkey_buf_exit(&old, c);
 
 	if (closure_nr_remaining(&cl) != 1) {
-		bch2_trans_unlock(trans);
-		closure_sync(&cl);
+		bch2_trans_unlock_long(trans);
+		bch2_wait_on_allocator(c, &cl);
 	}
 
 	return ret;
@@ -161,12 +164,12 @@ int bch2_fpunch_at(struct btree_trans *trans, struct btree_iter *iter,
 		if (ret)
 			continue;
 
-		bch2_btree_iter_set_snapshot(iter, snapshot);
+		bch2_btree_iter_set_snapshot(trans, iter, snapshot);
 
 		/*
-		 * peek_upto() doesn't have ideal semantics for extents:
+		 * peek_max() doesn't have ideal semantics for extents:
 		 */
-		k = bch2_btree_iter_peek_upto(iter, end_pos);
+		k = bch2_btree_iter_peek_max(trans, iter, end_pos);
 		if (!k.k)
 			break;
 
@@ -198,7 +201,7 @@ int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
 
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(inum.inum, start),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	ret = bch2_fpunch_at(trans, &iter, inum, end, i_sectors_delta);
 
@@ -224,13 +227,14 @@ void bch2_logged_op_truncate_to_text(struct printbuf *out, struct bch_fs *c, str
 
 static int truncate_set_isize(struct btree_trans *trans,
 			      subvol_inum inum,
-			      u64 new_i_size)
+			      u64 new_i_size,
+			      bool warn)
 {
-	struct btree_iter iter = { NULL };
+	struct btree_iter iter = {};
 	struct bch_inode_unpacked inode_u;
 	int ret;
 
-	ret   = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT) ?:
+	ret   = __bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent, warn) ?:
 		(inode_u.bi_size = new_i_size, 0) ?:
 		bch2_inode_write(trans, &iter, &inode_u);
 
@@ -247,23 +251,25 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans,
 	struct bkey_i_logged_op_truncate *op = bkey_i_to_logged_op_truncate(op_k);
 	subvol_inum inum = { le32_to_cpu(op->v.subvol), le64_to_cpu(op->v.inum) };
 	u64 new_i_size = le64_to_cpu(op->v.new_i_size);
+	bool warn_errors = i_sectors_delta != NULL;
 	int ret;
 
 	ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-			truncate_set_isize(trans, inum, new_i_size));
+			truncate_set_isize(trans, inum, new_i_size, i_sectors_delta != NULL));
 	if (ret)
 		goto err;
 
 	bch2_trans_iter_init(trans, &fpunch_iter, BTREE_ID_extents,
 			     POS(inum.inum, round_up(new_i_size, block_bytes(c)) >> 9),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 	ret = bch2_fpunch_at(trans, &fpunch_iter, inum, U64_MAX, i_sectors_delta);
 	bch2_trans_iter_exit(trans, &fpunch_iter);
 
 	if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 		ret = 0;
 err:
-	bch2_logged_op_finish(trans, op_k);
+	if (warn_errors)
+		bch_err_fn(c, ret);
 	return ret;
 }
 
@@ -287,9 +293,14 @@ int bch2_truncate(struct bch_fs *c, subvol_inum inum, u64 new_i_size, u64 *i_sec
 	 * resume only proceeding in one of the snapshots
 	 */
 	down_read(&c->snapshot_create_lock);
-	int ret = bch2_trans_run(c,
-		bch2_logged_op_start(trans, &op.k_i) ?:
-		__bch2_resume_logged_op_truncate(trans, &op.k_i, i_sectors_delta));
+	struct btree_trans *trans = bch2_trans_get(c);
+	int ret = bch2_logged_op_start(trans, &op.k_i);
+	if (ret)
+		goto out;
+	ret = __bch2_resume_logged_op_truncate(trans, &op.k_i, i_sectors_delta);
+	ret = bch2_logged_op_finish(trans, &op.k_i) ?: ret;
+out:
+	bch2_trans_put(trans);
 	up_read(&c->snapshot_create_lock);
 
 	return ret;
@@ -307,7 +318,8 @@ void bch2_logged_op_finsert_to_text(struct printbuf *out, struct bch_fs *c, stru
 	prt_printf(out, " src_offset=%llu",	le64_to_cpu(op.v->src_offset));
 }
 
-static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset, s64 len)
+static int adjust_i_size(struct btree_trans *trans, subvol_inum inum,
+			 u64 offset, s64 len, bool warn)
 {
 	struct btree_iter iter;
 	struct bch_inode_unpacked inode_u;
@@ -316,7 +328,7 @@ static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset
 	offset	<<= 9;
 	len	<<= 9;
 
-	ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT);
+	ret = __bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent, warn);
 	if (ret)
 		return ret;
 
@@ -356,15 +368,25 @@ static int __bch2_resume_logged_op_finsert(struct btree_trans *trans,
 	u64 len = abs(shift);
 	u64 pos = le64_to_cpu(op->v.pos);
 	bool insert = shift > 0;
+	u32 snapshot;
+	bool warn_errors = i_sectors_delta != NULL;
 	int ret = 0;
 
 	ret = bch2_inum_opts_get(trans, inum, &opts);
 	if (ret)
 		return ret;
 
+	/*
+	 * check for missing subvolume before fpunch, as in resume we don't want
+	 * it to be a fatal error
+	 */
+	ret = lockrestart_do(trans, __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot, warn_errors));
+	if (ret)
+		return ret;
+
 	bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
 			     POS(inum.inum, 0),
-			     BTREE_ITER_INTENT);
+			     BTREE_ITER_intent);
 
 	switch (op->v.state) {
 case LOGGED_OP_FINSERT_start:
@@ -372,12 +394,12 @@ case LOGGED_OP_FINSERT_start:
 
 	if (insert) {
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, src_offset, len) ?:
+				adjust_i_size(trans, inum, src_offset, len, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 		if (ret)
 			goto err;
 	} else {
-		bch2_btree_iter_set_pos(&iter, POS(inum.inum, src_offset));
+		bch2_btree_iter_set_pos(trans, &iter, POS(inum.inum, src_offset));
 
 		ret = bch2_fpunch_at(trans, &iter, inum, src_offset + len, i_sectors_delta);
 		if (ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart))
@@ -395,20 +417,20 @@ case LOGGED_OP_FINSERT_shift_extents:
 		struct bkey_i delete, *copy;
 		struct bkey_s_c k;
 		struct bpos src_pos = POS(inum.inum, src_offset);
-		u32 snapshot;
 
 		bch2_trans_begin(trans);
 
-		ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+		ret = __bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot,
+						    warn_errors);
 		if (ret)
 			goto btree_err;
 
-		bch2_btree_iter_set_snapshot(&iter, snapshot);
-		bch2_btree_iter_set_pos(&iter, SPOS(inum.inum, pos, snapshot));
+		bch2_btree_iter_set_snapshot(trans, &iter, snapshot);
+		bch2_btree_iter_set_pos(trans, &iter, SPOS(inum.inum, pos, snapshot));
 
 		k = insert
-			? bch2_btree_iter_peek_prev(&iter)
-			: bch2_btree_iter_peek_upto(&iter, POS(inum.inum, U64_MAX));
+			? bch2_btree_iter_peek_prev_min(trans, &iter, POS(inum.inum, 0))
+			: bch2_btree_iter_peek_max(trans, &iter, POS(inum.inum, U64_MAX));
 		if ((ret = bkey_err(k)))
 			goto btree_err;
 
@@ -442,7 +464,7 @@ case LOGGED_OP_FINSERT_shift_extents:
 
 		op->v.pos = cpu_to_le64(insert ? bkey_start_offset(&delete.k) : delete.k.p.offset);
 
-		ret =   bch2_bkey_set_needs_rebalance(c, copy, &opts) ?:
+		ret =   bch2_bkey_set_needs_rebalance(c, &opts, copy) ?:
 			bch2_btree_insert_trans(trans, BTREE_ID_extents, &delete, 0) ?:
 			bch2_btree_insert_trans(trans, BTREE_ID_extents, copy, 0) ?:
 			bch2_logged_op_update(trans, &op->k_i) ?:
@@ -462,12 +484,12 @@ btree_err:
 
 	if (!insert) {
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, src_offset, shift) ?:
+				adjust_i_size(trans, inum, src_offset, shift, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 	} else {
 		/* We need an inode update to update bi_journal_seq for fsync: */
 		ret = commit_do(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
-				adjust_i_size(trans, inum, 0, 0) ?:
+				adjust_i_size(trans, inum, 0, 0, warn_errors) ?:
 				bch2_logged_op_update(trans, &op->k_i));
 	}
 
@@ -476,8 +498,9 @@ case LOGGED_OP_FINSERT_finish:
 	break;
 	}
 err:
-	bch2_logged_op_finish(trans, op_k);
 	bch2_trans_iter_exit(trans, &iter);
+	if (warn_errors)
+		bch_err_fn(c, ret);
 	return ret;
 }
 
@@ -506,9 +529,14 @@ int bch2_fcollapse_finsert(struct bch_fs *c, subvol_inum inum,
 	 * resume only proceeding in one of the snapshots
 	 */
 	down_read(&c->snapshot_create_lock);
-	int ret = bch2_trans_run(c,
-		bch2_logged_op_start(trans, &op.k_i) ?:
-		__bch2_resume_logged_op_finsert(trans, &op.k_i, i_sectors_delta));
+	struct btree_trans *trans = bch2_trans_get(c);
+	int ret = bch2_logged_op_start(trans, &op.k_i);
+	if (ret)
+		goto out;
+	ret = __bch2_resume_logged_op_finsert(trans, &op.k_i, i_sectors_delta);
+	ret = bch2_logged_op_finish(trans, &op.k_i) ?: ret;
+out:
+	bch2_trans_put(trans);
 	up_read(&c->snapshot_create_lock);
 
 	return ret;