Re: [PATCH 3/3] mm/memory-failure: send SIGBUS in the event of thp split fail

[Miaohe Lin <linmiaohe@huawei.com>]

On 2024/5/9 0:56, Jane Chu wrote:
> On 5/8/2024 2:03 AM, Miaohe Lin wrote:
> 
>> On 2024/5/2 7:24, Jane Chu wrote:
>>> When handle hwpoison in a GUP longterm pin'ed thp page,
>>> try_to_split_thp_page() will fail. And at this point, there is little else
>>> the kernel could do except sending a SIGBUS to the user process, thus
>>> give it a chance to recover.
>> It seems the user process will still receive SIGBUS via kill_accessing_process()
>> when (re-)access thp later. So they should have a chance to recover already.
>> Or am I miss something?
> 
> The concern is about real UE consumption in which case, it's desirable to kill the process ASAP without having to relying on subsequent access.  Also to honor processes' MCE-early-kill request. kill_accessing_process() is very conservative in that, it doesn't check other processes that have the poisoned page mapped.

I see. Thanks for your explanation.
Thanks.
.

> 
> thanks,
> 
> -jane
> 
>>
>> Thanks.
>> .
>>
>>
> .