Re: [mm-unstable PATCH v5 4/8] mm, hwpoison: make unpoison aware of raw error info in hwpoisoned hugepage

["HORIGUCHI NAOYA(堀口 直也)" <naoya.horiguchi@nec.com>]

On Mon, Jul 11, 2022 at 03:09:01PM +0800, Miaohe Lin wrote:
> On 2022/7/8 13:36, Naoya Horiguchi wrote:
> > From: Naoya Horiguchi <naoya.horiguchi@nec.com>
> > 
> > Raw error info list needs to be removed when hwpoisoned hugetlb is
> > unpoisoned.  And unpoison handler needs to know how many errors there
> > are in the target hugepage. So add them.
> > 
> > HPageVmemmapOptimized(hpage) and HPageRawHwpUnreliable(hpage)) can't be
> > unpoisoned, so let's skip them.
> > 
> > Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
> > Reported-by: kernel test robot <lkp@intel.com>
> > ---
> > v4 -> v5:
> > - fix type of return value of free_raw_hwp_pages()
> >   (found by kernel test robot),
> > - prevent unpoison for HPageVmemmapOptimized and HPageRawHwpUnreliable.
> > ---
...
> > diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> > index 6833c5e4b410..89e74ec8a95f 100644
> > --- a/mm/memory-failure.c
> > +++ b/mm/memory-failure.c
> > @@ -1720,22 +1720,41 @@ static int hugetlb_set_page_hwpoison(struct page *hpage, struct page *page)
> >  	return ret;
> >  }
> >  
> > -int hugetlb_clear_page_hwpoison(struct page *hpage)
> > +static long free_raw_hwp_pages(struct page *hpage, bool move_flag)
> 
> NO strong opinion: Maybe the return type should be "unsigned" as it always >= 0 ?

Yes, will update.

> 
> >  {
> >  	struct llist_head *head;
> >  	struct llist_node *t, *tnode;
> > +	long count = 0;
> >  
> > -	if (!HPageRawHwpUnreliable(hpage))
> > -		ClearPageHWPoison(hpage);
> > +	/*
> > +	 * HPageVmemmapOptimized hugepages can't be unpoisoned because
> > +	 * struct pages for tail pages are required to free hwpoisoned
> > +	 * hugepages.  HPageRawHwpUnreliable hugepages shouldn't be
> > +	 * unpoisoned by definition.
> > +	 */
> > +	if (HPageVmemmapOptimized(hpage) || HPageRawHwpUnreliable(hpage))
> > +		return 0;
> >  	head = raw_hwp_list_head(hpage);
> >  	llist_for_each_safe(tnode, t, head->first) {
> >  		struct raw_hwp_page *p = container_of(tnode, struct raw_hwp_page, node);
> >  
> > -		SetPageHWPoison(p->page);
> > +		if (move_flag)
> > +			SetPageHWPoison(p->page);
> >  		kfree(p);
> > +		count++;
> >  	}
> >  	llist_del_all(head);
> > -	return 0;
> > +	return count;
> > +}
> > +
> > +int hugetlb_clear_page_hwpoison(struct page *hpage)
> 
> It seems the return value is unused?

Yes, the return value is not needed now.

> 
> > +{
> > +	int ret = -EBUSY;
> > +
> > +	if (!HPageRawHwpUnreliable(hpage))
> > +		ret = !TestClearPageHWPoison(hpage);
> > +	free_raw_hwp_pages(hpage, true);
> > +	return ret;
> >  }
> >  
> >  /*
> > @@ -1879,6 +1898,10 @@ static inline int try_memory_failure_hugetlb(unsigned long pfn, int flags, int *
> >  	return 0;
> >  }
> >  
> > +static inline long free_raw_hwp_pages(struct page *hpage, bool move_flag)
> 
> If return type is changed, remember to change here too.

OK.

> > +{
> > +	return 0;
> > +}
> >  #endif	/* CONFIG_HUGETLB_PAGE */
> >  
> >  static int memory_failure_dev_pagemap(unsigned long pfn, int flags,
> > @@ -2284,6 +2307,7 @@ int unpoison_memory(unsigned long pfn)
> >  	struct page *p;
> >  	int ret = -EBUSY;
> >  	int freeit = 0;
> > +	long count = 1;
> >  	static DEFINE_RATELIMIT_STATE(unpoison_rs, DEFAULT_RATELIMIT_INTERVAL,
> >  					DEFAULT_RATELIMIT_BURST);
> >  
> > @@ -2331,6 +2355,13 @@ int unpoison_memory(unsigned long pfn)
> >  
> >  	ret = get_hwpoison_page(p, MF_UNPOISON);
> >  	if (!ret) {
> > +		if (PageHuge(p)) {
> > +			count = free_raw_hwp_pages(page, false);
> 
> It seems the current behavior is: if any subpage of a hugetlb page is unpoisoned, then all of the
> hwpoisoned subpages will be unpoisoned. I'm not sure whether this is what we want.

Basically raw_hwp_info is not available to userspace (it might be recorded
in dmesg but not available via /proc/kpageflags), so unpoisoning error
subpages one-by-one is sometimes bothering.  If someone would like to
unpoison one-by-one (I expect nobody would), I can do this.

> Thanks.

Thank you!

- Naoya Horiguchi