mm/filemap: support readpage splitting a page
authorMatthew Wilcox (Oracle) <willy@infradead.org>
Wed, 24 Feb 2021 20:02:05 +0000 (12:02 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 24 Feb 2021 21:38:28 +0000 (13:38 -0800)
commitbd8a1f3655a704b9a1924fb3feffa3ecd6e5f8ae
tree530ff4ebcab5f12b72b4066071489a42383730cc
parent4805462598113f350838d612d0895db2dbb3992b
mm/filemap: support readpage splitting a page

For page splitting to succeed, the thread asking to split the page has to
be the only one with a reference to the page.  Calling
wait_on_page_locked() while holding a reference to the page will
effectively prevent this from happening with sufficient threads waiting on
the same page.  Use put_and_wait_on_page_locked() to sleep without holding
a reference to the page, then retry the page lookup after the page is
unlocked.

Since we now get the page lock a little earlier in filemap_update_page(),
we can eliminate a number of duplicate checks.  The original intent
(commit ebded02788b5 ("avoid unnecessary calls to lock_page when waiting
for IO to complete during a read")) behind getting the page lock later was
to avoid re-locking the page after it has been brought uptodate by another
thread.  We still avoid that because we go through the normal lookup path
again after the winning thread has brought the page uptodate.

Link: https://lkml.kernel.org/r/20210122160140.223228-7-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Kent Overstreet <kent.overstreet@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/filemap.c