Re: [PATCH v4 1/3] mm/hwpoison: mf_mutex for soft offline and unpoison

[Mike Kravetz <mike.kravetz@oracle.com>]

On 11/15/21 00:40, Naoya Horiguchi wrote:
> From: Naoya Horiguchi <naoya.horiguchi@nec.com>
> 
> Originally mf_mutex is introduced to serialize multiple MCE events, but
> it is not that useful to allow unpoison to run in parallel with memory_failure()
> and soft offline.  So apply mf_mutex to soft offline and unpoison.
> The memory failure handler and soft offline handler get simpler with this.
> 
> Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
> Reviewed-by: Yang Shi <shy828301@gmail.com>
> ---
> ChangeLog v4:
> - fix type in commit description.
> 
> ChangeLog v3:
> - merge with "mm/hwpoison: remove race consideration"
> - update description
> 
> ChangeLog v2:
> - add mutex_unlock() in "page already poisoned" path in soft_offline_page().
>   (Thanks to Ding Hui)
> ---
>  mm/memory-failure.c | 62 +++++++++++++--------------------------------
>  1 file changed, 18 insertions(+), 44 deletions(-)
> 
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index e8c38e27b753..d29c79de6034 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c

Thanks for working on this.  I tried to exercise memory error handling for
hugetlb pages and ran into issues addressed by these patches.

> @@ -1507,14 +1507,6 @@ static int memory_failure_hugetlb(unsigned long pfn, int flags)
>  	lock_page(head);
>  	page_flags = head->flags;
>  
> -	if (!PageHWPoison(head)) {
> -		pr_err("Memory failure: %#lx: just unpoisoned\n", pfn);
> -		num_poisoned_pages_dec();
> -		unlock_page(head);
> -		put_page(head);
> -		return 0;
> -	}
> -
>  	/*
>  	 * TODO: hwpoison for pud-sized hugetlb doesn't work right now, so
>  	 * simply disable it. In order to make it work properly, we need
> @@ -1628,6 +1620,8 @@ static int memory_failure_dev_pagemap(unsigned long pfn, int flags,
>  	return rc;
>  }
>  
> +static DEFINE_MUTEX(mf_mutex);

There are only two places other places where PageHWPoison is modified without
the mutex.  They are:
- test_and_clear_pmem_poison
  I 'think' pmem error handling is done separately so this does not apply.
- clear_hwpoisoned_pages
  Called before removing memory (and deleting memmap) to reconcile count
  of poisoned pages.  Should not be an issue and technically I do not think
  the ClearPageHWPoison() is actually needed in this routine.

Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
-- 
Mike Kravetz