mm/memory-failure: reject unsupported non-folio compound page

When !CONFIG_TRANSPARENT_HUGEPAGE, a non-folio compound page can appear in
a userspace mapping via either vm_insert_*() functions or
vm_operatios_struct->fault(). They are not folios, thus should not be
considered for folio operations like split. To reject these pages, make
sure get_hwpoison_page() is always called as HWPoisonHandlable() will do
the right work.

[Some commit log borrowed from Zi Yan. Thanks.]

Link: https://lkml.kernel.org/r/20260205075328.523211-1-linmiaohe@huawei.com
Fixes: 689b8986776c ("mm/memory-failure: improve large block size folio handling")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reported-by: 是参差 <shicenci@gmail.com>
Closes: https://lore.kernel.org/all/PS1PPF7E1D7501F1E4F4441E7ECD056DEADAB98A@PS1PPF7E1D7501F.apcprd02.prod.outlook.com/
Reviewed-by: Zi Yan <ziy@nvidia.com>
Tested-by: Zi Yan <ziy@nvidia.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

authored by Miaohe Lin and committed by Andrew Morton ae9fd76c b5cbacd7

+20 -22
+20 -22
mm/memory-failure.c
··· 2411 * In fact it's dangerous to directly bump up page count from 0, 2412 * that may make page_ref_freeze()/page_ref_unfreeze() mismatch. 2413 */ 2414 - if (!(flags & MF_COUNT_INCREASED)) { 2415 - res = get_hwpoison_page(p, flags); 2416 - if (!res) { 2417 - if (is_free_buddy_page(p)) { 2418 - if (take_page_off_buddy(p)) { 2419 - page_ref_inc(p); 2420 - res = MF_RECOVERED; 2421 - } else { 2422 - /* We lost the race, try again */ 2423 - if (retry) { 2424 - ClearPageHWPoison(p); 2425 - retry = false; 2426 - goto try_again; 2427 - } 2428 - res = MF_FAILED; 2429 - } 2430 - res = action_result(pfn, MF_MSG_BUDDY, res); 2431 } else { 2432 - res = action_result(pfn, MF_MSG_KERNEL_HIGH_ORDER, MF_IGNORED); 2433 } 2434 - goto unlock_mutex; 2435 - } else if (res < 0) { 2436 - res = action_result(pfn, MF_MSG_GET_HWPOISON, MF_IGNORED); 2437 - goto unlock_mutex; 2438 } 2439 } 2440 2441 folio = page_folio(p);
··· 2411 * In fact it's dangerous to directly bump up page count from 0, 2412 * that may make page_ref_freeze()/page_ref_unfreeze() mismatch. 2413 */ 2414 + res = get_hwpoison_page(p, flags); 2415 + if (!res) { 2416 + if (is_free_buddy_page(p)) { 2417 + if (take_page_off_buddy(p)) { 2418 + page_ref_inc(p); 2419 + res = MF_RECOVERED; 2420 } else { 2421 + /* We lost the race, try again */ 2422 + if (retry) { 2423 + ClearPageHWPoison(p); 2424 + retry = false; 2425 + goto try_again; 2426 + } 2427 + res = MF_FAILED; 2428 } 2429 + res = action_result(pfn, MF_MSG_BUDDY, res); 2430 + } else { 2431 + res = action_result(pfn, MF_MSG_KERNEL_HIGH_ORDER, MF_IGNORED); 2432 } 2433 + goto unlock_mutex; 2434 + } else if (res < 0) { 2435 + res = action_result(pfn, MF_MSG_GET_HWPOISON, MF_IGNORED); 2436 + goto unlock_mutex; 2437 } 2438 2439 folio = page_folio(p);