Re: [PATCH v5 mm-new 0/6] mm: hugetlb: allocate frozen gigantic folio

[Kefeng Wang <wangkefeng.wang@huawei.com>]

On 2026/1/8 9:05, Kefeng Wang wrote:
> 
> 
> On 2026/1/8 3:38, Zi Yan wrote:
>> On 7 Jan 2026, at 13:39, Mark Brown wrote:
>>
>>> On Wed, Jan 07, 2026 at 07:31:30PM +0200, Claudiu Beznea wrote:
>>>> On 12/30/25 09:24, Kefeng Wang wrote:
>>>
>>>>> Introduce alloc_contig_frozen_pages() and cma_alloc_frozen_compound()
>>>>> which avoid atomic operation about page refcount, and then convert to
>>>>> allocate frozen gigantic folio by the new helpers in hugetlb to 
>>>>> cleanup
>>>>> the alloc_gigantic_folio().
>>>>
>>>> I'm seeing the following issues on the Renesas RZ/G3S SoC when doing 
>>>> suspend
>>>> to idle:
>>>>
>>>> [  129.636729] Unable to handle kernel paging request at virtual 
>>>> address
>>>> dead000000000108
>>>> [  129.644674] Mem abort info:
>>>> [  129.647456]   ESR = 0x0000000096000044
>>>
>>> This is also introducing OOMs when doing at least audio tests (I don't
>>> think these are super relevant) on Raspberry Pi 3B+ running NFS root
>>> (probably more relevant):
>>>
>>> [   64.064256] Unable to handle kernel paging request at virtual 
>>> address fffffdffc1000000
>>>
>>> ...
>>>
>>> [   64.087583] Call trace:
>>> [   64.087586]  kmem_cache_free+0x88/0x434 (P)
>>> [   64.087598]  skb_free_head+0x9c/0xb8
>>> [   64.087608]  skb_release_data+0x120/0x174
>>> [   64.087615]  __kfree_skb+0x2c/0x44
>>> [   64.087622]  tcp_data_queue+0x948/0xe50
>>>
>>> Full log:
>>>
>>>    https://lava.sirena.org.uk/scheduler/job/2341856#L1721
>>>
>>> Bisection identifies:
>>>
>>> [fb9a328d30400dbc8b2ea5a57daeb28bedac398b] mm: cma: add 
>>> cma_alloc_frozen{_compound}()
>>>
>>> as being the comit that introduces the issue.  Bisect log with links to
>>> further test runs:
>>
>> Hi Mark and Claudiu,
> 
> Thanks for the reports.
> 
>>
>> Can you try the patch below to see if it fixes the issue? Basically,
>> in cma_release(), count was used to drop page ref and decreased to 0,
>> but after the loop, count becomes -1 and __cma_release_frozen()
>> is releasing unnecessary pages.
> 
> Oh，sorry for introducing the regression, my previous self-tests were
> more focused on the hugetlb part, I should be more careful about this,
> and thanks Zi Yan for the quick fix, I will do more check and test for
> the non-hugetlb part.
> 
> 

Based on the cma_debug interface, I performed quick tests to reproduce
the issue before Zi's fix. With the fix applied, the crash was resolved,
and no other issues were observed during testing.