From: Xiubo Li xiubli@redhat.com

While the mapped IOs continue if we try to flush a file's buffer we can see that the fsync() won't complete until the IOs finish.

This is analogous to Jan Kara's commit (f446daaea9d4 mm: implement writeback livelock avoidance using page tagging), we will try to avoid livelocks of writeback when some steadily creates dirty pages in a mapping we are writing out.

Cc: stable@vger.kernel.org Signed-off-by: Xiubo Li xiubli@redhat.com --- fs/ceph/addr.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 5731b82bf368..caf6ac5c1390 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -879,6 +879,7 @@ static int ceph_writepages_start(struct address_space *mapping, bool should_loop, range_whole = false; bool done = false; bool caching = ceph_is_cache_enabled(inode); + xa_mark_t tag;

if (wbc->sync_mode == WB_SYNC_NONE && fsc->write_congested) @@ -905,6 +906,11 @@ static int ceph_writepages_start(struct address_space *mapping, start_index = wbc->range_cyclic ? mapping->writeback_index : 0; index = start_index;

+ if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) { + tag = PAGECACHE_TAG_TOWRITE; + } else { + tag = PAGECACHE_TAG_DIRTY; + } retry: /* find oldest snap context with dirty data */ snapc = get_oldest_context(inode, &ceph_wbc, NULL); @@ -943,6 +949,9 @@ static int ceph_writepages_start(struct address_space *mapping, dout(" non-head snapc, range whole\n"); }

+ if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) + tag_pages_for_writeback(mapping, index, end); + ceph_put_snap_context(last_snapc); last_snapc = snapc;

@@ -959,7 +968,7 @@ static int ceph_writepages_start(struct address_space *mapping,

get_more_pages: pvec_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, - end, PAGECACHE_TAG_DIRTY); + end, tag); dout("pagevec_lookup_range_tag got %d\n", pvec_pages); if (!pvec_pages && !locked_pages) break;