[lvc-project] [PATCH 6.1] mm/memory_hotplug: fix hwpoisoned large folio handling in do_migrate_range()

[Александра Дюпина]

05/06/26 21:09, Zi Yan wrote:
> On 5 Jun 2026, at 13:26, Alexandra Diupina wrote:
>
>> From: Jinjiang Tu <tujinjiang at huawei.com>
>>
>> commit 397f6d14f9c370e4910e6885294c340f39dedbf5 upstream.
>>
>> In do_migrate_range(), the hwpoisoned folio may be large folio, which
>> can't be handled by unmap_poisoned_folio().
>>
>> I can reproduce this issue in qemu after adding delay in memory_failure()
>>
>> BUG: kernel NULL pointer dereference, address: 0000000000000000
>> Workqueue: kacpi_hotplug acpi_hotplug_work_fn
>> RIP: 0010:try_to_unmap_one+0x16a/0xfc0
>>    <TASK>
>>    rmap_walk_anon+0xda/0x1f0
>>    try_to_unmap+0x78/0x80
>>    ? __pfx_try_to_unmap_one+0x10/0x10
>>    ? __pfx_folio_not_mapped+0x10/0x10
>>    ? __pfx_folio_lock_anon_vma_read+0x10/0x10
>>    unmap_poisoned_folio+0x60/0x140
>>    do_migrate_range+0x4d1/0x600
>>    ? slab_memory_callback+0x6a/0x190
>>    ? notifier_call_chain+0x56/0xb0
>>    offline_pages+0x3e6/0x460
>>    memory_subsys_offline+0x130/0x1f0
>>    device_offline+0xba/0x110
>>    acpi_bus_offline+0xb7/0x130
>>    acpi_scan_hot_remove+0x77/0x290
>>    acpi_device_hotplug+0x1e0/0x240
>>    acpi_hotplug_work_fn+0x1a/0x30
>>    process_one_work+0x186/0x340
>>
>> Besides, do_migrate_range() may be called between memory_failure set
>> hwpoison flag and isolate the folio from lru, so remove WARN_ON(). In other
>> places, unmap_poisoned_folio() is called when the folio is isolated, obey
>> it in do_migrate_range() too.
>>
>> [david at redhat.com: don't abort offlining, fixed typo, add comment]
>> Link: https://lkml.kernel.org/r/3c214dff-9649-4015-840f-10de0e03ebe4@redhat.com
>> Fixes: b15c87263a69 ("hwpoison, memory_hotplug: allow hwpoisoned pages to be offlined")
>> Signed-off-by: Jinjiang Tu <tujinjiang at huawei.com>
>> Signed-off-by: David Hildenbrand <david at redhat.com>
>> Acked-by: Zi Yan <ziy at nvidia.com>
>> Reviewed-by: Miaohe Lin <linmiaohe at huawei.com>
>> Cc: Kefeng Wang <wangkefeng.wang at huawei.com>
>> Cc: Luis Chamberalin <mcgrof at kernel.org>
>> Cc: Matthew Wilcox (Oracle) <willy at infradead.org>
>> Cc: Michal Hocko <mhocko at kernel.org>
>> Cc: Oscar Salvador <osalvador at suse.de>
>> Cc: Pankaj Raghav <kernel at pankajraghav.com>
>> Signed-off-by: Andrew Morton <akpm at linux-foundation.org>
>> [ Alexandra: replace continue with put_folio label ]
>> Signed-off-by: Alexandra Diupina <adiupina at astralinux.ru>
>> ---
>>   mm/memory_hotplug.c | 13 ++++++++++---
>>   1 file changed, 10 insertions(+), 3 deletions(-)
>>
>> diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
>> index c8cc2f63c3ea..013db41f6ce2 100644
>> --- a/mm/memory_hotplug.c
>> +++ b/mm/memory_hotplug.c
>> @@ -1654,15 +1654,21 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
>>   		 * the unmap as the catch all safety net).
>>   		 */
>>   		if (PageHWPoison(page)) {
>> -			if (WARN_ON(folio_test_lru(folio)))
>> -				folio_isolate_lru(folio);
>> +			/*
>> +			 * unmap_poisoned_folio() cannot handle large folios
>> +			 * in all cases yet.
>> +			 */
>> +			if (folio_test_large(folio) && !folio_test_hugetlb(folio))
>> +				goto put_folio;
>> +			if (folio_test_lru(folio) && !folio_isolate_lru(folio))
>> +				goto put_folio;
>>   			if (folio_mapped(folio)) {
>>   				folio_lock(folio);
>>   				try_to_unmap(folio, TTU_IGNORE_MLOCK);
>>   				folio_unlock(folio);
>>   			}
>>
>> -			continue;
>> +			goto put_folio;
>>   		}
>>
>>   		if (!get_page_unless_zero(page))
>> @@ -1687,6 +1693,7 @@ do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
>>   				dump_page(page, "isolation failed");
>>   			}
>>   		}
>> +put_folio:
>>   		put_page(page);
>>   	}
>>   	if (!list_empty(&source)) {
>> -- 
>> 2.30.2
> I am not sure this is right.
>
> The original patch uses put_folio, because folio_try_get() is called
> before this if block. But for 6.1, get_page_unless_zero() is called
> after this if block, so the page in this if block has no elevated
> refcount compared to the original patch context. You might want to
> replace “goto put_folio” with “continue” in the original patch instead.
>
> Best Regards,
> Yan, Zi
Thanks for the comments,
I'll send the second version of the patch