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>
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index fcf77e1..89ba79f 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2161,6 +2161,38 @@
/* 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 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 @@
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);
- continue;
- }
- }
-
- if (!uptodate) {
- clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
- ClearPageUptodate(page);
- SetPageError(page);
- }
+ if (end_extent_writepage(page, err, start, end))
+ continue;
if (whole_page)
end_page_writeback(page);
@@ -2818,8 +2829,12 @@
if (tree->ops && tree->ops->writepage_start_hook) {
ret = tree->ops->writepage_start_hook(page, start,
page_end);
- if (ret == -EAGAIN) {
- redirty_page_for_writepage(wbc, page);
+ 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);
ret = 0;