From 9a9794a81a8a1be8670aa673ec0f0fbfbddeccae Mon Sep 17 00:00:00 2001 From: Kairui Song Date: Tue, 29 Apr 2025 17:48:03 +0800 Subject: [PATCH] mm, swap: fix false warning for large allocation with !THP_SWAP The !CONFIG_THP_SWAP check existed before just fine because slot cache would reject high order allocation and let the caller split all folios and try again. But slot cache is gone, so large allocation will directly go to the allocator, and the allocator should just fail silently to inform caller to do the folio split, this is totally fine and expected. Remove this meaningless warning. Link: https://lkml.kernel.org/r/20250429094803.85518-1-ryncsn@gmail.com Fixes: 0ff67f990bd4 ("mm, swap: remove swap slot cache") Signed-off-by: Kairui Song Reported-by: Heiko Carstens Closes: https://lore.kernel.org/linux-mm/20250428135252.25453B17-hca@linux.ibm.com/ Tested-by: Heiko Carstens Cc: Baoquan He Signed-off-by: Andrew Morton --- mm/swapfile.c | 23 ++++++++++++++++------- 1 file changed, 16 insertions(+), 7 deletions(-) diff --git a/mm/swapfile.c b/mm/swapfile.c index 2eff8b51a945..f214843612dc 100644 --- a/mm/swapfile.c +++ b/mm/swapfile.c @@ -1272,13 +1272,22 @@ int folio_alloc_swap(struct folio *folio, gfp_t gfp) VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio); VM_BUG_ON_FOLIO(!folio_test_uptodate(folio), folio); - /* - * Should not even be attempting large allocations when huge - * page swap is disabled. Warn and fail the allocation. - */ - if (order && (!IS_ENABLED(CONFIG_THP_SWAP) || size > SWAPFILE_CLUSTER)) { - VM_WARN_ON_ONCE(1); - return -EINVAL; + if (order) { + /* + * Reject large allocation when THP_SWAP is disabled, + * the caller should split the folio and try again. + */ + if (!IS_ENABLED(CONFIG_THP_SWAP)) + return -EAGAIN; + + /* + * Allocation size should never exceed cluster size + * (HPAGE_PMD_SIZE). + */ + if (size > SWAPFILE_CLUSTER) { + VM_WARN_ON_ONCE(1); + return -EINVAL; + } } local_lock(&percpu_swap_cluster.lock);