]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/commitdiff
mm: fs: fix lru_cache_disabled race in bh_lru
authorMinchan Kim <minchan@kernel.org>
Tue, 22 Mar 2022 21:39:34 +0000 (14:39 -0700)
committerStefan Bader <stefan.bader@canonical.com>
Fri, 20 May 2022 12:37:31 +0000 (14:37 +0200)
BugLink: https://bugs.launchpad.net/bugs/1969110
commit c0226eb8bde854e016a594a16f5c0d98aca426fa upstream.

Check lru_cache_disabled under bh_lru_lock.  Otherwise, it could introduce
race below and it fails to migrate pages containing buffer_head.

   CPU 0 CPU 1

bh_lru_install
                                       lru_cache_disable
  lru_cache_disabled = false
                                       atomic_inc(&lru_disable_count);
       invalidate_bh_lrus_cpu of CPU 0
       bh_lru_lock
       __invalidate_bh_lrus
       bh_lru_unlock
  bh_lru_lock
  install the bh
  bh_lru_unlock

WHen this race happens a CMA allocation fails, which is critical for
the workload which depends on CMA.

Link: https://lkml.kernel.org/r/20220308180709.2017638-1-minchan@kernel.org
Fixes: 8cc621d2f45d ("mm: fs: invalidate BH LRU during page migration")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Chris Goldsworthy <cgoldswo@codeaurora.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: John Dias <joaodias@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ab657a29c3e9f02a7c7ca9ddaa7b176dec04f86e)
Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
fs/buffer.c

index c615387aedcae8bd9e94fbc3faef44a2bbb2496b..f6d2835794918a9868f487b6225c19465dcbfca9 100644 (file)
@@ -1235,16 +1235,18 @@ static void bh_lru_install(struct buffer_head *bh)
        int i;
 
        check_irqs_on();
+       bh_lru_lock();
+
        /*
         * the refcount of buffer_head in bh_lru prevents dropping the
         * attached page(i.e., try_to_free_buffers) so it could cause
         * failing page migration.
         * Skip putting upcoming bh into bh_lru until migration is done.
         */
-       if (lru_cache_disabled())
+       if (lru_cache_disabled()) {
+               bh_lru_unlock();
                return;
-
-       bh_lru_lock();
+       }
 
        b = this_cpu_ptr(&bh_lrus);
        for (i = 0; i < BH_LRU_SIZE; i++) {