btrfs: delalloc for page dirtied out-of-band in fixup worker

 We encountered an issue that was easily observable on s/390 systems but
 could really happen anywhere. The timing just seemed to hit reliably
 on s/390 with limited memory.

 The gist is that when an unexpected set_page_dirty() happened, we'd
 run into the BUG() in btrfs_writepage_fixup_worker since it wasn't
 properly set up for delalloc.

 This patch does the following:
 - Performs the missing delalloc in the fixup worker
 - Allow the start hook to return -EBUSY which informs __extent_writepage
   that it should mark the page skipped and not to redirty it. This is
   required since the fixup worker can fail with -ENOSPC and the page
   will have already been redirtied. That causes an Oops in
   drop_outstanding_extents later. Retrying the fixup worker could
   lead to an infinite loop. Deferring the page redirty also saves us
   some cycles since the page would be stuck in a resubmit-redirty loop
   until the fixup worker completes. It's not harmful, just wasteful.
 - If the fixup worker fails, we mark the page and mapping as errored,
   and end the writeback, similar to what we would do had the page
   actually been submitted to writeback.
Signed-off-by: Jeff Mahoney <jeffm@suse.com>

fs/btrfs/extent_io.c

@@ -2161,6 +2161,38 @@ static int bio_readpage_error(struct bio *failed_bio, struct page *page,
 
 /* lots and lots of room for performance fixes in the end_bio funcs */
 
+int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
+{
+	int uptodate = (err == 0);
+	struct extent_io_tree *tree;
+	int ret;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+
+	if (tree->ops && tree->ops->writepage_end_io_hook) {
+		ret = tree->ops->writepage_end_io_hook(page, start,
+					       end, NULL, uptodate);
+		if (ret)
+			uptodate = 0;
+	}
+
+	if (!uptodate && tree->ops &&
+	    tree->ops->writepage_io_failed_hook) {
+		ret = tree->ops->writepage_io_failed_hook(NULL, page,
+						 start, end, NULL);
+		/* Writeback already completed */
+		if (ret == 0)
+			return 1;
+	}
+
+	if (!uptodate) {
+		clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
+		ClearPageUptodate(page);
+		SetPageError(page);
+	}
+	return 0;
+}
+
 /*
  * after a writepage IO is done, we need to:
  * clear the uptodate bits on error
@@ -2172,13 +2204,11 @@ static int bio_readpage_error(struct bio *failed_bio, struct page *page,
  */
 static void end_bio_extent_writepage(struct bio *bio, int err)
 {
-	int uptodate = err == 0;
 	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 	struct extent_io_tree *tree;
 	u64 start;
 	u64 end;
 	int whole_page;
-	int ret;
 
 	do {
 		struct page *page = bvec->bv_page;
@@ -2195,28 +2225,9 @@ static void end_bio_extent_writepage(struct bio *bio, int err)
 
 		if (--bvec >= bio->bi_io_vec)
 			prefetchw(&bvec->bv_page->flags);
-		if (tree->ops && tree->ops->writepage_end_io_hook) {
-			ret = tree->ops->writepage_end_io_hook(page, start,
-						       end, NULL, uptodate);
-			if (ret)
-				uptodate = 0;
-		}
 
-		if (!uptodate && tree->ops &&
-		    tree->ops->writepage_io_failed_hook) {
-			ret = tree->ops->writepage_io_failed_hook(bio, page,
-							 start, end, NULL);
-			if (ret == 0) {
-				uptodate = (err == 0);
+		if (end_extent_writepage(page, err, start, end))
 			continue;
-			}
-		}
-
-		if (!uptodate) {
-			clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
-			ClearPageUptodate(page);
-			SetPageError(page);
-		}
 
 		if (whole_page)
 			end_page_writeback(page);
@@ -2818,7 +2829,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 	if (tree->ops && tree->ops->writepage_start_hook) {
 		ret = tree->ops->writepage_start_hook(page, start,
 						      page_end);
-		if (ret == -EAGAIN) {
+		if (ret) {
+			/* Fixup worker will requeue */
+			if (ret == -EBUSY)
+				wbc->pages_skipped++;
+			else
 				redirty_page_for_writepage(wbc, page);
 			update_nr_written(page, wbc, nr_written);
 			unlock_page(page);

fs/btrfs/extent_io.h

@@ -319,4 +319,5 @@ struct btrfs_mapping_tree;
 int repair_io_failure(struct btrfs_mapping_tree *map_tree, u64 start,
 			u64 length, u64 logical, struct page *page,
 			int mirror_num);
+int end_extent_writepage(struct page *page, int err, u64 start, u64 end);
 #endif

fs/btrfs/inode.c

@@ -1555,6 +1555,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
 	struct inode *inode;
 	u64 page_start;
 	u64 page_end;
+	int ret;
 
 	fixup = container_of(work, struct btrfs_writepage_fixup, work);
 	page = fixup->page;
@@ -1582,12 +1583,21 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
 				     page_end, &cached_state, GFP_NOFS);
 		unlock_page(page);
 		btrfs_start_ordered_extent(inode, ordered, 1);
+		btrfs_put_ordered_extent(ordered);
 		goto again;
 	}
 
-	BUG();
+	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
+	if (ret) {
+		mapping_set_error(page->mapping, ret);
+		end_extent_writepage(page, ret, page_start, page_end);
+		ClearPageChecked(page);
+		goto out;
+	 }
+
 	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
 	ClearPageChecked(page);
+	set_page_dirty(page);
 out:
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
 			     &cached_state, GFP_NOFS);
@@ -1630,7 +1640,7 @@ static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
 	fixup->work.func = btrfs_writepage_fixup_worker;
 	fixup->page = page;
 	btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
-	return -EAGAIN;
+	return -EBUSY;
 }
 
 static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,