Re: [PATCH v2 5/9] mm, hwpoison: make unpoison aware of raw error info in hwpoisoned hugepage

[Miaohe Lin]

On 2022/6/27 16:32, Miaohe Lin wrote:
> On 2022/6/24 7:51, Naoya Horiguchi wrote:
>> From: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
>>
>> Raw error info list needs to be removed when hwpoisoned hugetlb is
>> unpoisioned.  And unpoison handler needs to know how many errors there
>> are in the target hugepage. So add them.
>>
>> Signed-off-by: Naoya Horiguchi <naoya.horiguchi@xxxxxxx>
> 
> snip
> 
>> @@ -2255,7 +2275,7 @@ int unpoison_memory(unsigned long pfn)
>>  unlock_mutex:
>>  	mutex_unlock(&mf_mutex);
>>  	if (!ret || freeit) {
>> -		num_poisoned_pages_dec();
>> +		num_poisoned_pages_sub(count);
> 
> IIUC, num_poisoned_pages will only be incremented once for hugetlb page. If many
> subpages are hwpoisoned, they will reach the "else if (res == -EHWPOISON)" path
> in try_memory_failure_hugetlb and thus num_poisoned_pages_inc is ignored. Maybe
> that should be changed so subpages can contribute to the num_poisoned_pages
> or should we just do num_poisoned_pages_dec here? Or am I miss something?

Sorry. I re-read patch 4/9 and I found hugetlb_set_page_hwpoison will increment the
num_poisoned_pages for each subpages. Please ignore this question.

> 
> Thanks!
> 
>>  		unpoison_pr_info("Unpoison: Software-unpoisoned page %#lx\n",
>>  				 page_to_pfn(p), &unpoison_rs);
>>  	}
>>
>