mm: shmem: use proper gfp flags for shmem_writepage()
authorYang Shi <yang.shi@linux.alibaba.com>
Sun, 1 Dec 2019 01:58:07 +0000 (17:58 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 1 Dec 2019 20:59:10 +0000 (12:59 -0800)
The shmem_writepage() uses GFP_ATOMIC to allocate swap cache.  GFP_ATOMIC
used to mean __GFP_HIGH, but now it means __GFP_HIGH | __GFP_ATOMIC |
__GFP_KSWAPD_RECLAIM.  However, shmem_writepage() should write out to swap
only in response to memory pressure, so __GFP_KSWAPD_RECLAIM looks useless
since the caller may be kswapd itself or in direct reclaim already.

In addition, XArray node allocations from PF_MEMALLOC contexts could
completely exhaust the page allocator, __GFP_NOMEMALLOC stops emergency
reserves from being allocated.

Here just copy the gfp flags used by add_to_swap().

Hugh:
 "a cleanup to make the two calls look the same when they don't need to
  be different (whereas the call from __read_swap_cache_async() rightly
  uses a lower priority gfp)".

Link: http://lkml.kernel.org/r/1572991351-86061-1-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/shmem.c

index 6e4e742db5c2a050865aa47ea1c916812322c4b5..3c336b02cf08735faf9ccdc122ae44eff2d97738 100644 (file)
@@ -1369,7 +1369,8 @@ static int shmem_writepage(struct page *page, struct writeback_control *wbc)
        if (list_empty(&info->swaplist))
                list_add(&info->swaplist, &shmem_swaplist);
 
-       if (add_to_swap_cache(page, swap, GFP_ATOMIC) == 0) {
+       if (add_to_swap_cache(page, swap,
+                       __GFP_HIGH | __GFP_NOMEMALLOC | __GFP_NOWARN) == 0) {
                spin_lock_irq(&info->lock);
                shmem_recalc_inode(inode);
                info->swapped++;