loop: set PF_MEMALLOC_NOIO for the worker thread

A deadlock with this stacktrace was observed.

The loop thread does a GFP_KERNEL allocation, it calls into dm-bufio
shrinker and the shrinker depends on I/O completion in the dm-bufio
subsystem.

In order to fix the deadlock (and other similar ones), we set the flag
PF_MEMALLOC_NOIO at loop thread entry.

PID: 474    TASK: ffff8813e11f4600  CPU: 10  COMMAND: "kswapd0"
   #0 [ffff8813dedfb938] __schedule at ffffffff8173f405
   #1 [ffff8813dedfb990] schedule at ffffffff8173fa27
   #2 [ffff8813dedfb9b0] schedule_timeout at ffffffff81742fec
   hardkernel#3 [ffff8813dedfba60] io_schedule_timeout at ffffffff8173f186
   hardkernel#4 [ffff8813dedfbaa0] bit_wait_io at ffffffff8174034f
   hardkernel#5 [ffff8813dedfbac0] __wait_on_bit at ffffffff8173fec8
   hardkernel#6 [ffff8813dedfbb10] out_of_line_wait_on_bit at ffffffff8173ff81
   hardkernel#7 [ffff8813dedfbb90] __make_buffer_clean at ffffffffa038736f [dm_bufio]
   hardkernel#8 [ffff8813dedfbbb0] __try_evict_buffer at ffffffffa0387bb8 [dm_bufio]
   hardkernel#9 [ffff8813dedfbbd0] dm_bufio_shrink_scan at ffffffffa0387cc3 [dm_bufio]
  hardkernel#10 [ffff8813dedfbc40] shrink_slab at ffffffff811a87ce
  hardkernel#11 [ffff8813dedfbd30] shrink_zone at ffffffff811ad778
  hardkernel#12 [ffff8813dedfbdc0] kswapd at ffffffff811ae92f
  hardkernel#13 [ffff8813dedfbec0] kthread at ffffffff810a8428
  hardkernel#14 [ffff8813dedfbf50] ret_from_fork at ffffffff81745242

  PID: 14127  TASK: ffff881455749c00  CPU: 11  COMMAND: "loop1"
   #0 [ffff88272f5af228] __schedule at ffffffff8173f405
   #1 [ffff88272f5af280] schedule at ffffffff8173fa27
   #2 [ffff88272f5af2a0] schedule_preempt_disabled at ffffffff8173fd5e
   hardkernel#3 [ffff88272f5af2b0] __mutex_lock_slowpath at ffffffff81741fb5
   hardkernel#4 [ffff88272f5af330] mutex_lock at ffffffff81742133
   hardkernel#5 [ffff88272f5af350] dm_bufio_shrink_count at ffffffffa03865f9 [dm_bufio]
   hardkernel#6 [ffff88272f5af380] shrink_slab at ffffffff811a86bd
   hardkernel#7 [ffff88272f5af470] shrink_zone at ffffffff811ad778
   hardkernel#8 [ffff88272f5af500] do_try_to_free_pages at ffffffff811adb34
   hardkernel#9 [ffff88272f5af590] try_to_free_pages at ffffffff811adef8
  hardkernel#10 [ffff88272f5af610] __alloc_pages_nodemask at ffffffff811a09c3
  hardkernel#11 [ffff88272f5af710] alloc_pages_current at ffffffff811e8b71
  hardkernel#12 [ffff88272f5af760] new_slab at ffffffff811f4523
  hardkernel#13 [ffff88272f5af7b0] __slab_alloc at ffffffff8173a1b5
  hardkernel#14 [ffff88272f5af880] kmem_cache_alloc at ffffffff811f484b
  hardkernel#15 [ffff88272f5af8d0] do_blockdev_direct_IO at ffffffff812535b3
  hardkernel#16 [ffff88272f5afb00] __blockdev_direct_IO at ffffffff81255dc3
  hardkernel#17 [ffff88272f5afb30] xfs_vm_direct_IO at ffffffffa01fe3fc [xfs]
  hardkernel#18 [ffff88272f5afb90] generic_file_read_iter at ffffffff81198994
  hardkernel#19 [ffff88272f5afc50] __dta_xfs_file_read_iter_2398 at ffffffffa020c970 [xfs]
  hardkernel#20 [ffff88272f5afcc0] lo_rw_aio at ffffffffa0377042 [loop]
  hardkernel#21 [ffff88272f5afd70] loop_queue_work at ffffffffa0377c3b [loop]
  hardkernel#22 [ffff88272f5afe60] kthread_worker_fn at ffffffff810a8a0c
  hardkernel#23 [ffff88272f5afec0] kthread at ffffffff810a8428
  hardkernel#24 [ffff88272f5aff50] ret_from_fork at ffffffff81745242

Signed-off-by: Mikulas Patocka <[email protected]>
Cc: [email protected]
Signed-off-by: Jens Axboe <[email protected]>

drivers/block/loop.c

@@ -885,7 +885,7 @@ static void loop_unprepare_queue(struct loop_device *lo)
 
 static int loop_kthread_worker_fn(void *worker_ptr)
 {
-	current->flags |= PF_LESS_THROTTLE;
+	current->flags |= PF_LESS_THROTTLE | PF_MEMALLOC_NOIO;
 	return kthread_worker_fn(worker_ptr);
 }