1 files changed, 57 insertions, 8 deletions

diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
index 46d847f..e08b6bc 100644
--- a/fs/btrfs/scrub.c
+++ b/fs/btrfs/scrub.c
@@ -1504,8 +1504,9 @@ static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
 				sblock->no_io_error_seen = 0;
 		} else {
 			bio->bi_iter.bi_sector = page->physical >> 9;
+			bio_set_op_attrs(bio, REQ_OP_READ, 0);
 
-			if (btrfsic_submit_bio_wait(READ, bio))
+			if (btrfsic_submit_bio_wait(bio))
 				sblock->no_io_error_seen = 0;
 		}
 
@@ -1583,6 +1584,7 @@ static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
 			return -EIO;
 		bio->bi_bdev = page_bad->dev->bdev;
 		bio->bi_iter.bi_sector = page_bad->physical >> 9;
+		bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
 
 		ret = bio_add_page(bio, page_good->page, PAGE_SIZE, 0);
 		if (PAGE_SIZE != ret) {
@@ -1590,7 +1592,7 @@ static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
 			return -EIO;
 		}
 
-		if (btrfsic_submit_bio_wait(WRITE, bio)) {
+		if (btrfsic_submit_bio_wait(bio)) {
 			btrfs_dev_stat_inc_and_print(page_bad->dev,
 				BTRFS_DEV_STAT_WRITE_ERRS);
 			btrfs_dev_replace_stats_inc(
@@ -1684,6 +1686,7 @@ again:
 		bio->bi_end_io = scrub_wr_bio_end_io;
 		bio->bi_bdev = sbio->dev->bdev;
 		bio->bi_iter.bi_sector = sbio->physical >> 9;
+		bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
 		sbio->err = 0;
 	} else if (sbio->physical + sbio->page_count * PAGE_SIZE !=
 		   spage->physical_for_dev_replace ||
@@ -1731,7 +1734,7 @@ static void scrub_wr_submit(struct scrub_ctx *sctx)
 	 * orders the requests before sending them to the driver which
 	 * doubled the write performance on spinning disks when measured
 	 * with Linux 3.5 */
-	btrfsic_submit_bio(WRITE, sbio->bio);
+	btrfsic_submit_bio(sbio->bio);
 }
 
 static void scrub_wr_bio_end_io(struct bio *bio)
@@ -2041,7 +2044,7 @@ static void scrub_submit(struct scrub_ctx *sctx)
 	sbio = sctx->bios[sctx->curr];
 	sctx->curr = -1;
 	scrub_pending_bio_inc(sctx);
-	btrfsic_submit_bio(READ, sbio->bio);
+	btrfsic_submit_bio(sbio->bio);
 }
 
 static int scrub_add_page_to_rd_bio(struct scrub_ctx *sctx,
@@ -2088,6 +2091,7 @@ again:
 		bio->bi_end_io = scrub_bio_end_io;
 		bio->bi_bdev = sbio->dev->bdev;
 		bio->bi_iter.bi_sector = sbio->physical >> 9;
+		bio_set_op_attrs(bio, REQ_OP_READ, 0);
 		sbio->err = 0;
 	} else if (sbio->physical + sbio->page_count * PAGE_SIZE !=
 		   spage->physical ||
@@ -3582,6 +3586,46 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 		 */
 		scrub_pause_on(fs_info);
 		ret = btrfs_inc_block_group_ro(root, cache);
+		if (!ret && is_dev_replace) {
+			/*
+			 * If we are doing a device replace wait for any tasks
+			 * that started dellaloc right before we set the block
+			 * group to RO mode, as they might have just allocated
+			 * an extent from it or decided they could do a nocow
+			 * write. And if any such tasks did that, wait for their
+			 * ordered extents to complete and then commit the
+			 * current transaction, so that we can later see the new
+			 * extent items in the extent tree - the ordered extents
+			 * create delayed data references (for cow writes) when
+			 * they complete, which will be run and insert the
+			 * corresponding extent items into the extent tree when
+			 * we commit the transaction they used when running
+			 * inode.c:btrfs_finish_ordered_io(). We later use
+			 * the commit root of the extent tree to find extents
+			 * to copy from the srcdev into the tgtdev, and we don't
+			 * want to miss any new extents.
+			 */
+			btrfs_wait_block_group_reservations(cache);
+			btrfs_wait_nocow_writers(cache);
+			ret = btrfs_wait_ordered_roots(fs_info, -1,
+						       cache->key.objectid,
+						       cache->key.offset);
+			if (ret > 0) {
+				struct btrfs_trans_handle *trans;
+
+				trans = btrfs_join_transaction(root);
+				if (IS_ERR(trans))
+					ret = PTR_ERR(trans);
+				else
+					ret = btrfs_commit_transaction(trans,
+								       root);
+				if (ret) {
+					scrub_pause_off(fs_info);
+					btrfs_put_block_group(cache);
+					break;
+				}
+			}
+		}
 		scrub_pause_off(fs_info);
 
 		if (ret == 0) {
@@ -3602,9 +3646,11 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 			break;
 		}
 
+		btrfs_dev_replace_lock(&fs_info->dev_replace, 1);
 		dev_replace->cursor_right = found_key.offset + length;
 		dev_replace->cursor_left = found_key.offset;
 		dev_replace->item_needs_writeback = 1;
+		btrfs_dev_replace_unlock(&fs_info->dev_replace, 1);
 		ret = scrub_chunk(sctx, scrub_dev, chunk_offset, length,
 				  found_key.offset, cache, is_dev_replace);
 
@@ -3640,6 +3686,11 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 
 		scrub_pause_off(fs_info);
 
+		btrfs_dev_replace_lock(&fs_info->dev_replace, 1);
+		dev_replace->cursor_left = dev_replace->cursor_right;
+		dev_replace->item_needs_writeback = 1;
+		btrfs_dev_replace_unlock(&fs_info->dev_replace, 1);
+
 		if (ro_set)
 			btrfs_dec_block_group_ro(root, cache);
 
@@ -3677,9 +3728,6 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 			ret = -ENOMEM;
 			break;
 		}
-
-		dev_replace->cursor_left = dev_replace->cursor_right;
-		dev_replace->item_needs_writeback = 1;
 skip:
 		key.offset = found_key.offset + length;
 		btrfs_release_path(path);
@@ -4392,6 +4440,7 @@ static int write_page_nocow(struct scrub_ctx *sctx,
 	bio->bi_iter.bi_size = 0;
 	bio->bi_iter.bi_sector = physical_for_dev_replace >> 9;
 	bio->bi_bdev = dev->bdev;
+	bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_SYNC);
 	ret = bio_add_page(bio, page, PAGE_SIZE, 0);
 	if (ret != PAGE_SIZE) {
 leave_with_eio:
@@ -4400,7 +4449,7 @@ leave_with_eio:
 		return -EIO;
 	}
 
-	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio))
+	if (btrfsic_submit_bio_wait(bio))
 		goto leave_with_eio;
 
 	bio_put(bio);