diff mbox series

[v5,1/2] mm/khugepaged: recover from poisoned anonymous memory

Message ID 20221010160142.1087120-2-jiaqiyan@google.com (mailing list archive)
State New
Headers show
Series Memory poison recovery in khugepaged collapsing | expand

Commit Message

Jiaqi Yan Oct. 10, 2022, 4:01 p.m. UTC
Make __collapse_huge_page_copy return whether
collapsing/copying anonymous pages succeeded,
and make collapse_huge_page handle the return status.

Break existing PTE scan loop into two for-loops.
The first loop copies source pages into target huge page,
and can fail gracefully when running into memory errors in
source pages. If copying all pages succeeds, the second loop
releases and clears up these normal pages.
Otherwise, the second loop does the following to
roll back the page table and page states:
1) re-establish the original PTEs-to-PMD connection.
2) release source pages back to their LRU list.

Signed-off-by: Jiaqi Yan <jiaqiyan@google.com>
---
 include/linux/highmem.h            |  19 +++++
 include/trace/events/huge_memory.h |   3 +-
 mm/khugepaged.c                    | 130 ++++++++++++++++++++++-------
 3 files changed, 121 insertions(+), 31 deletions(-)

Comments

Kirill A. Shutemov Oct. 11, 2022, 11:59 p.m. UTC | #1
On Mon, Oct 10, 2022 at 09:01:41AM -0700, Jiaqi Yan wrote:
> Make __collapse_huge_page_copy return whether
> collapsing/copying anonymous pages succeeded,
> and make collapse_huge_page handle the return status.
> 
> Break existing PTE scan loop into two for-loops.
> The first loop copies source pages into target huge page,
> and can fail gracefully when running into memory errors in
> source pages. If copying all pages succeeds, the second loop
> releases and clears up these normal pages.
> Otherwise, the second loop does the following to
> roll back the page table and page states:
> 1) re-establish the original PTEs-to-PMD connection.
> 2) release source pages back to their LRU list.
> 
> Signed-off-by: Jiaqi Yan <jiaqiyan@google.com>
> ---
>  include/linux/highmem.h            |  19 +++++
>  include/trace/events/huge_memory.h |   3 +-
>  mm/khugepaged.c                    | 130 ++++++++++++++++++++++-------
>  3 files changed, 121 insertions(+), 31 deletions(-)
> 
> diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> index 25679035ca283..91a65bdabcb33 100644
> --- a/include/linux/highmem.h
> +++ b/include/linux/highmem.h
> @@ -332,6 +332,25 @@ static inline void copy_highpage(struct page *to, struct page *from)
>  
>  #endif
>  
> +/*
> + * Machine check exception handled version of copy_highpage.
> + * Return true if copying page content failed; otherwise false.
> + * Note handling #MC requires arch opt-in.
> + */
> +static inline bool copy_highpage_mc(struct page *to, struct page *from)
> +{
> +	char *vfrom, *vto;
> +	unsigned long ret;
> +
> +	vfrom = kmap_local_page(from);
> +	vto = kmap_local_page(to);
> +	ret = copy_mc_to_kernel(vto, vfrom, PAGE_SIZE);
> +	kunmap_local(vto);
> +	kunmap_local(vfrom);
> +
> +	return ret > 0;
> +}
> +
>  static inline void memcpy_page(struct page *dst_page, size_t dst_off,
>  			       struct page *src_page, size_t src_off,
>  			       size_t len)
> diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
> index d651f3437367d..756e991366639 100644
> --- a/include/trace/events/huge_memory.h
> +++ b/include/trace/events/huge_memory.h
> @@ -33,7 +33,8 @@
>  	EM( SCAN_ALLOC_HUGE_PAGE_FAIL,	"alloc_huge_page_failed")	\
>  	EM( SCAN_CGROUP_CHARGE_FAIL,	"ccgroup_charge_failed")	\
>  	EM( SCAN_TRUNCATED,		"truncated")			\
> -	EMe(SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
> +	EM( SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
> +	EMe(SCAN_COPY_MC,		"copy_poisoned_page")		\
>  
>  #undef EM
>  #undef EMe
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 70b7ac66411c0..552e7cb4c8b42 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -51,6 +51,7 @@ enum scan_result {
>  	SCAN_CGROUP_CHARGE_FAIL,
>  	SCAN_TRUNCATED,
>  	SCAN_PAGE_HAS_PRIVATE,
> +	SCAN_COPY_MC,
>  };
>  
>  #define CREATE_TRACE_POINTS
> @@ -673,44 +674,99 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>  	return 0;
>  }
>  
> -static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> -				      struct vm_area_struct *vma,
> -				      unsigned long address,
> -				      spinlock_t *ptl,
> -				      struct list_head *compound_pagelist)
> +/*
> + * __collapse_huge_page_copy - attempts to copy memory contents from normal
> + * pages to a hugepage. Cleanup the normal pages if copying succeeds;
> + * otherwise restore the original page table and release isolated normal pages.
> + * Returns true if copying succeeds, otherwise returns false.
> + *
> + * @pte: starting of the PTEs to copy from
> + * @page: the new hugepage to copy contents to
> + * @pmd: pointer to the new hugepage's PMD
> + * @rollback: the original normal pages' PMD
> + * @vma: the original normal pages' virtual memory area
> + * @address: starting address to copy
> + * @pte_ptl: lock on normal pages' PTEs
> + * @compound_pagelist: list that stores compound pages
> + */
> +static bool __collapse_huge_page_copy(pte_t *pte,
> +				struct page *page,
> +				pmd_t *pmd,
> +				pmd_t rollback,
> +				struct vm_area_struct *vma,
> +				unsigned long address,
> +				spinlock_t *pte_ptl,
> +				struct list_head *compound_pagelist)
>  {
>  	struct page *src_page, *tmp;
>  	pte_t *_pte;
> -	for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
> -				_pte++, page++, address += PAGE_SIZE) {
> -		pte_t pteval = *_pte;
> +	pte_t pteval;
> +	unsigned long _address;
> +	spinlock_t *pmd_ptl;
> +	bool copy_succeeded = true;
>  
> +	/*
> +	 * Copying pages' contents is subject to memory poison at any iteration.
> +	 */
> +	for (_pte = pte, _address = address;
> +			_pte < pte + HPAGE_PMD_NR;
> +			_pte++, page++, _address += PAGE_SIZE) {
> +		pteval = *_pte;
> +
> +		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval)))
> +			clear_user_highpage(page, _address);
> +		else {
> +			src_page = pte_page(pteval);
> +			if (copy_highpage_mc(page, src_page)) {
> +				copy_succeeded = false;
> +				break;
> +			}
> +		}
> +	}
> +
> +	if (!copy_succeeded) {
> +		/*
> +		 * Copying failed, re-establish the regular PMD that points to
> +		 * the regular page table. Restoring PMD needs to be done prior
> +		 * to releasing pages. Since pages are still isolated and locked
> +		 * here, acquiring anon_vma_lock_write is unnecessary.
> +		 */
> +		pmd_ptl = pmd_lock(vma->vm_mm, pmd);
> +		pmd_populate(vma->vm_mm, pmd, pmd_pgtable(rollback));
> +		spin_unlock(pmd_ptl);
> +	}

Initially I expected return here, but you handle copy_succeeded below. Hm.

> +
> +	for (_pte = pte, _address = address; _pte < pte + HPAGE_PMD_NR;
> +			_pte++, _address += PAGE_SIZE) {
> +		pteval = *_pte;
>  		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
> -			clear_user_highpage(page, address);
> -			add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> -			if (is_zero_pfn(pte_pfn(pteval))) {
> -				/*
> -				 * ptl mostly unnecessary.
> -				 */
> -				spin_lock(ptl);
> -				ptep_clear(vma->vm_mm, address, _pte);
> -				spin_unlock(ptl);
> +			if (copy_succeeded) {
> +				add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> +				if (is_zero_pfn(pte_pfn(pteval))) {
> +					/*
> +					 * pte_ptl mostly unnecessary.
> +					 */
> +					spin_lock(pte_ptl);
> +					pte_clear(vma->vm_mm, _address, _pte);
> +					spin_unlock(pte_ptl);
> +				}
>  			}

So this branch is NOP if !copy_succeeded, right?

>  		} else {
>  			src_page = pte_page(pteval);
> -			copy_user_highpage(page, src_page, address, vma);
>  			if (!PageCompound(src_page))
>  				release_pte_page(src_page);

And this branch only calls release_pte_page() (Which I believe would screw
up statistic).

Looks very broken. Or just hard to follow. Or both.

Please consider rework code more to streamline handling !copy_succeeded
situation.

> -			/*
> -			 * ptl mostly unnecessary, but preempt has to
> -			 * be disabled to update the per-cpu stats
> -			 * inside page_remove_rmap().
> -			 */
> -			spin_lock(ptl);
> -			ptep_clear(vma->vm_mm, address, _pte);
> -			page_remove_rmap(src_page, vma, false);
> -			spin_unlock(ptl);
> -			free_page_and_swap_cache(src_page);
> +			if (copy_succeeded) {
> +				/*
> +				 * pte_ptl mostly unnecessary, but preempt
> +				 * has to be disabled to update the per-cpu
> +				 * stats inside page_remove_rmap().
> +				 */
> +				spin_lock(pte_ptl);
> +				ptep_clear(vma->vm_mm, _address, _pte);
> +				page_remove_rmap(src_page, vma, false);
> +				spin_unlock(pte_ptl);
> +				free_page_and_swap_cache(src_page);
> +			}
>  		}
>  	}
>  
> @@ -723,6 +779,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>  		free_swap_cache(src_page);
>  		putback_lru_page(src_page);
>  	}
> +
> +	return copy_succeeded;
>  }
>  
>  static void khugepaged_alloc_sleep(void)
> @@ -1009,6 +1067,7 @@ static void collapse_huge_page(struct mm_struct *mm,
>  	struct vm_area_struct *vma;
>  	struct mmu_notifier_range range;
>  	gfp_t gfp;
> +	bool copied = false;
>  
>  	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
>  
> @@ -1121,9 +1180,13 @@ static void collapse_huge_page(struct mm_struct *mm,
>  	 */
>  	anon_vma_unlock_write(vma->anon_vma);
>  
> -	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
> -			&compound_pagelist);
> +	copied = __collapse_huge_page_copy(pte, new_page, pmd, _pmd,
> +			vma, address, pte_ptl, &compound_pagelist);
>  	pte_unmap(pte);
> +	if (!copied) {
> +		result = SCAN_COPY_MC;
> +		goto out_up_write;
> +	}
>  	/*
>  	 * spin_lock() below is not the equivalent of smp_wmb(), but
>  	 * the smp_wmb() inside __SetPageUptodate() can be reused to
> @@ -2129,6 +2192,13 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
>  				khugepaged_scan_file(mm, file, pgoff, hpage);
>  				fput(file);
>  			} else {
> +				/*
> +				 * mmap_read_lock is
> +				 * 1) still held if scan failed;
> +				 * 2) released if scan succeeded.
> +				 * It is not affected by collapsing or copying
> +				 * operations.
> +				 */
>  				ret = khugepaged_scan_pmd(mm, vma,
>  						khugepaged_scan.address,
>  						hpage);
> -- 
> 2.38.0.rc1.362.ged0d419d3c-goog
> 
>
Jiaqi Yan Oct. 14, 2022, 6:28 p.m. UTC | #2
On Tue, Oct 11, 2022 at 4:59 PM Kirill A. Shutemov <kirill@shutemov.name> wrote:
>
> On Mon, Oct 10, 2022 at 09:01:41AM -0700, Jiaqi Yan wrote:
> > Make __collapse_huge_page_copy return whether
> > collapsing/copying anonymous pages succeeded,
> > and make collapse_huge_page handle the return status.
> >
> > Break existing PTE scan loop into two for-loops.
> > The first loop copies source pages into target huge page,
> > and can fail gracefully when running into memory errors in
> > source pages. If copying all pages succeeds, the second loop
> > releases and clears up these normal pages.
> > Otherwise, the second loop does the following to
> > roll back the page table and page states:
> > 1) re-establish the original PTEs-to-PMD connection.
> > 2) release source pages back to their LRU list.
> >
> > Signed-off-by: Jiaqi Yan <jiaqiyan@google.com>
> > ---
> >  include/linux/highmem.h            |  19 +++++
> >  include/trace/events/huge_memory.h |   3 +-
> >  mm/khugepaged.c                    | 130 ++++++++++++++++++++++-------
> >  3 files changed, 121 insertions(+), 31 deletions(-)
> >
> > diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> > index 25679035ca283..91a65bdabcb33 100644
> > --- a/include/linux/highmem.h
> > +++ b/include/linux/highmem.h
> > @@ -332,6 +332,25 @@ static inline void copy_highpage(struct page *to, struct page *from)
> >
> >  #endif
> >
> > +/*
> > + * Machine check exception handled version of copy_highpage.
> > + * Return true if copying page content failed; otherwise false.
> > + * Note handling #MC requires arch opt-in.
> > + */
> > +static inline bool copy_highpage_mc(struct page *to, struct page *from)
> > +{
> > +     char *vfrom, *vto;
> > +     unsigned long ret;
> > +
> > +     vfrom = kmap_local_page(from);
> > +     vto = kmap_local_page(to);
> > +     ret = copy_mc_to_kernel(vto, vfrom, PAGE_SIZE);
> > +     kunmap_local(vto);
> > +     kunmap_local(vfrom);
> > +
> > +     return ret > 0;
> > +}
> > +
> >  static inline void memcpy_page(struct page *dst_page, size_t dst_off,
> >                              struct page *src_page, size_t src_off,
> >                              size_t len)
> > diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
> > index d651f3437367d..756e991366639 100644
> > --- a/include/trace/events/huge_memory.h
> > +++ b/include/trace/events/huge_memory.h
> > @@ -33,7 +33,8 @@
> >       EM( SCAN_ALLOC_HUGE_PAGE_FAIL,  "alloc_huge_page_failed")       \
> >       EM( SCAN_CGROUP_CHARGE_FAIL,    "ccgroup_charge_failed")        \
> >       EM( SCAN_TRUNCATED,             "truncated")                    \
> > -     EMe(SCAN_PAGE_HAS_PRIVATE,      "page_has_private")             \
> > +     EM( SCAN_PAGE_HAS_PRIVATE,      "page_has_private")             \
> > +     EMe(SCAN_COPY_MC,               "copy_poisoned_page")           \
> >
> >  #undef EM
> >  #undef EMe
> > diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> > index 70b7ac66411c0..552e7cb4c8b42 100644
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -51,6 +51,7 @@ enum scan_result {
> >       SCAN_CGROUP_CHARGE_FAIL,
> >       SCAN_TRUNCATED,
> >       SCAN_PAGE_HAS_PRIVATE,
> > +     SCAN_COPY_MC,
> >  };
> >
> >  #define CREATE_TRACE_POINTS
> > @@ -673,44 +674,99 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
> >       return 0;
> >  }
> >
> > -static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> > -                                   struct vm_area_struct *vma,
> > -                                   unsigned long address,
> > -                                   spinlock_t *ptl,
> > -                                   struct list_head *compound_pagelist)
> > +/*
> > + * __collapse_huge_page_copy - attempts to copy memory contents from normal
> > + * pages to a hugepage. Cleanup the normal pages if copying succeeds;
> > + * otherwise restore the original page table and release isolated normal pages.
> > + * Returns true if copying succeeds, otherwise returns false.
> > + *
> > + * @pte: starting of the PTEs to copy from
> > + * @page: the new hugepage to copy contents to
> > + * @pmd: pointer to the new hugepage's PMD
> > + * @rollback: the original normal pages' PMD
> > + * @vma: the original normal pages' virtual memory area
> > + * @address: starting address to copy
> > + * @pte_ptl: lock on normal pages' PTEs
> > + * @compound_pagelist: list that stores compound pages
> > + */
> > +static bool __collapse_huge_page_copy(pte_t *pte,
> > +                             struct page *page,
> > +                             pmd_t *pmd,
> > +                             pmd_t rollback,
> > +                             struct vm_area_struct *vma,
> > +                             unsigned long address,
> > +                             spinlock_t *pte_ptl,
> > +                             struct list_head *compound_pagelist)
> >  {
> >       struct page *src_page, *tmp;
> >       pte_t *_pte;
> > -     for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
> > -                             _pte++, page++, address += PAGE_SIZE) {
> > -             pte_t pteval = *_pte;
> > +     pte_t pteval;
> > +     unsigned long _address;
> > +     spinlock_t *pmd_ptl;
> > +     bool copy_succeeded = true;
> >
> > +     /*
> > +      * Copying pages' contents is subject to memory poison at any iteration.
> > +      */
> > +     for (_pte = pte, _address = address;
> > +                     _pte < pte + HPAGE_PMD_NR;
> > +                     _pte++, page++, _address += PAGE_SIZE) {
> > +             pteval = *_pte;
> > +
> > +             if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval)))
> > +                     clear_user_highpage(page, _address);
> > +             else {
> > +                     src_page = pte_page(pteval);
> > +                     if (copy_highpage_mc(page, src_page)) {
> > +                             copy_succeeded = false;
> > +                             break;
> > +                     }
> > +             }
> > +     }
> > +
> > +     if (!copy_succeeded) {
> > +             /*
> > +              * Copying failed, re-establish the regular PMD that points to
> > +              * the regular page table. Restoring PMD needs to be done prior
> > +              * to releasing pages. Since pages are still isolated and locked
> > +              * here, acquiring anon_vma_lock_write is unnecessary.
> > +              */
> > +             pmd_ptl = pmd_lock(vma->vm_mm, pmd);
> > +             pmd_populate(vma->vm_mm, pmd, pmd_pgtable(rollback));
> > +             spin_unlock(pmd_ptl);
> > +     }
>
> Initially I expected return here, but you handle copy_succeeded below. Hm.

If !copy_succeeded, after re-populate PMD, I believe the remaining
rollback operation is to undo 3 things in
__collapse_huge_page_isolate(). Please see my next 2 replies for
details.

>
> > +
> > +     for (_pte = pte, _address = address; _pte < pte + HPAGE_PMD_NR;
> > +                     _pte++, _address += PAGE_SIZE) {
> > +             pteval = *_pte;
> >               if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
> > -                     clear_user_highpage(page, address);
> > -                     add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> > -                     if (is_zero_pfn(pte_pfn(pteval))) {
> > -                             /*
> > -                              * ptl mostly unnecessary.
> > -                              */
> > -                             spin_lock(ptl);
> > -                             ptep_clear(vma->vm_mm, address, _pte);
> > -                             spin_unlock(ptl);
> > +                     if (copy_succeeded) {
> > +                             add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
> > +                             if (is_zero_pfn(pte_pfn(pteval))) {
> > +                                     /*
> > +                                      * pte_ptl mostly unnecessary.
> > +                                      */
> > +                                     spin_lock(pte_ptl);
> > +                                     pte_clear(vma->vm_mm, _address, _pte);
> > +                                     spin_unlock(pte_ptl);
> > +                             }
> >                       }
>
> So this branch is NOP if !copy_succeeded, right?

Yes, if (pte_none || zero_pfn) we don't need any rollback operations,
because the 1st loop doesn't add_mm_counter() and doesn't
ptep_clear().

>
> >               } else {
> >                       src_page = pte_page(pteval);
> > -                     copy_user_highpage(page, src_page, address, vma);
> >                       if (!PageCompound(src_page))
> >                               release_pte_page(src_page);
>
> And this branch only calls release_pte_page() (Which I believe would screw
> up statistic).

Elsewise (!pte_none && !zero_pfn), rollback needs to do these things
if !copy_succeeded:
1. unlock_page(src_page), which is locked in __collapse_huge_page_isolate
2. putback_lru_page(src_page), which is isolated in __collapse_huge_page_isolate
3. decrease compound_nr(src_page), which is increased in
__collapse_huge_page_isolate

For compound pages, these rollback operations are done in the
list_for_each_entry_safe loop before return.
For non-compound pages, release_pte_page() does 1-3.

>
> Looks very broken. Or just hard to follow. Or both.

Do you think the logic looks correct (and just hard to follow) after
these explanation? Currently the rollback operations are soaked with
the copy_succeed code to minimize lines of code. But I can definitely
rewrite to something like this if you think the rollback logic is
correct:

if (copy_succeeded) {
  ...
} else {
  re-populate PMD;
  if (!pte_none && !zero_pfn && !PageCompound(src_page))
    release_pte_page(src_page);
  for(page in compound_list) {
    release_pte_page(page);
  }
}
return copy_succeeded;

>
> Please consider rework code more to streamline handling !copy_succeeded
> situation.
>
> > -                     /*
> > -                      * ptl mostly unnecessary, but preempt has to
> > -                      * be disabled to update the per-cpu stats
> > -                      * inside page_remove_rmap().
> > -                      */
> > -                     spin_lock(ptl);
> > -                     ptep_clear(vma->vm_mm, address, _pte);
> > -                     page_remove_rmap(src_page, vma, false);
> > -                     spin_unlock(ptl);
> > -                     free_page_and_swap_cache(src_page);
> > +                     if (copy_succeeded) {
> > +                             /*
> > +                              * pte_ptl mostly unnecessary, but preempt
> > +                              * has to be disabled to update the per-cpu
> > +                              * stats inside page_remove_rmap().
> > +                              */
> > +                             spin_lock(pte_ptl);
> > +                             ptep_clear(vma->vm_mm, _address, _pte);
> > +                             page_remove_rmap(src_page, vma, false);
> > +                             spin_unlock(pte_ptl);
> > +                             free_page_and_swap_cache(src_page);
> > +                     }
> >               }
> >       }
> >
> > @@ -723,6 +779,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
> >               free_swap_cache(src_page);

I think in the case of !copy_succeed, doing free_swap_cache is not correct.
Changing to the new if-flow and just release_pte_page() on compound
pages should fix this.

> >               putback_lru_page(src_page);
> >       }
> > +
> > +     return copy_succeeded;
> >  }
> >
> >  static void khugepaged_alloc_sleep(void)
> > @@ -1009,6 +1067,7 @@ static void collapse_huge_page(struct mm_struct *mm,
> >       struct vm_area_struct *vma;
> >       struct mmu_notifier_range range;
> >       gfp_t gfp;
> > +     bool copied = false;
> >
> >       VM_BUG_ON(address & ~HPAGE_PMD_MASK);
> >
> > @@ -1121,9 +1180,13 @@ static void collapse_huge_page(struct mm_struct *mm,
> >        */
> >       anon_vma_unlock_write(vma->anon_vma);
> >
> > -     __collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
> > -                     &compound_pagelist);
> > +     copied = __collapse_huge_page_copy(pte, new_page, pmd, _pmd,
> > +                     vma, address, pte_ptl, &compound_pagelist);
> >       pte_unmap(pte);
> > +     if (!copied) {
> > +             result = SCAN_COPY_MC;
> > +             goto out_up_write;
> > +     }
> >       /*
> >        * spin_lock() below is not the equivalent of smp_wmb(), but
> >        * the smp_wmb() inside __SetPageUptodate() can be reused to
> > @@ -2129,6 +2192,13 @@ static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
> >                               khugepaged_scan_file(mm, file, pgoff, hpage);
> >                               fput(file);
> >                       } else {
> > +                             /*
> > +                              * mmap_read_lock is
> > +                              * 1) still held if scan failed;
> > +                              * 2) released if scan succeeded.
> > +                              * It is not affected by collapsing or copying
> > +                              * operations.
> > +                              */
> >                               ret = khugepaged_scan_pmd(mm, vma,
> >                                               khugepaged_scan.address,
> >                                               hpage);
> > --
> > 2.38.0.rc1.362.ged0d419d3c-goog
> >
> >
>
> --
>   Kiryl Shutsemau / Kirill A. Shutemov
Kirill A. Shutemov Oct. 31, 2022, 3:31 p.m. UTC | #3
On Fri, Oct 14, 2022 at 11:28:53AM -0700, Jiaqi Yan wrote:
> >
> > Looks very broken. Or just hard to follow. Or both.
> 
> Do you think the logic looks correct (and just hard to follow) after
> these explanation?

Yes.

> Currently the rollback operations are soaked with
> the copy_succeed code to minimize lines of code. But I can definitely
> rewrite to something like this if you think the rollback logic is
> correct:
> 
> if (copy_succeeded) {
>   ...
> } else {
>   re-populate PMD;
>   if (!pte_none && !zero_pfn && !PageCompound(src_page))
>     release_pte_page(src_page);
>   for(page in compound_list) {
>     release_pte_page(page);
>   }
> }
> return copy_succeeded;

Looks sane to me.
diff mbox series

Patch

diff --git a/include/linux/highmem.h b/include/linux/highmem.h
index 25679035ca283..91a65bdabcb33 100644
--- a/include/linux/highmem.h
+++ b/include/linux/highmem.h
@@ -332,6 +332,25 @@  static inline void copy_highpage(struct page *to, struct page *from)
 
 #endif
 
+/*
+ * Machine check exception handled version of copy_highpage.
+ * Return true if copying page content failed; otherwise false.
+ * Note handling #MC requires arch opt-in.
+ */
+static inline bool copy_highpage_mc(struct page *to, struct page *from)
+{
+	char *vfrom, *vto;
+	unsigned long ret;
+
+	vfrom = kmap_local_page(from);
+	vto = kmap_local_page(to);
+	ret = copy_mc_to_kernel(vto, vfrom, PAGE_SIZE);
+	kunmap_local(vto);
+	kunmap_local(vfrom);
+
+	return ret > 0;
+}
+
 static inline void memcpy_page(struct page *dst_page, size_t dst_off,
 			       struct page *src_page, size_t src_off,
 			       size_t len)
diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
index d651f3437367d..756e991366639 100644
--- a/include/trace/events/huge_memory.h
+++ b/include/trace/events/huge_memory.h
@@ -33,7 +33,8 @@ 
 	EM( SCAN_ALLOC_HUGE_PAGE_FAIL,	"alloc_huge_page_failed")	\
 	EM( SCAN_CGROUP_CHARGE_FAIL,	"ccgroup_charge_failed")	\
 	EM( SCAN_TRUNCATED,		"truncated")			\
-	EMe(SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
+	EM( SCAN_PAGE_HAS_PRIVATE,	"page_has_private")		\
+	EMe(SCAN_COPY_MC,		"copy_poisoned_page")		\
 
 #undef EM
 #undef EMe
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 70b7ac66411c0..552e7cb4c8b42 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -51,6 +51,7 @@  enum scan_result {
 	SCAN_CGROUP_CHARGE_FAIL,
 	SCAN_TRUNCATED,
 	SCAN_PAGE_HAS_PRIVATE,
+	SCAN_COPY_MC,
 };
 
 #define CREATE_TRACE_POINTS
@@ -673,44 +674,99 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 	return 0;
 }
 
-static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
-				      struct vm_area_struct *vma,
-				      unsigned long address,
-				      spinlock_t *ptl,
-				      struct list_head *compound_pagelist)
+/*
+ * __collapse_huge_page_copy - attempts to copy memory contents from normal
+ * pages to a hugepage. Cleanup the normal pages if copying succeeds;
+ * otherwise restore the original page table and release isolated normal pages.
+ * Returns true if copying succeeds, otherwise returns false.
+ *
+ * @pte: starting of the PTEs to copy from
+ * @page: the new hugepage to copy contents to
+ * @pmd: pointer to the new hugepage's PMD
+ * @rollback: the original normal pages' PMD
+ * @vma: the original normal pages' virtual memory area
+ * @address: starting address to copy
+ * @pte_ptl: lock on normal pages' PTEs
+ * @compound_pagelist: list that stores compound pages
+ */
+static bool __collapse_huge_page_copy(pte_t *pte,
+				struct page *page,
+				pmd_t *pmd,
+				pmd_t rollback,
+				struct vm_area_struct *vma,
+				unsigned long address,
+				spinlock_t *pte_ptl,
+				struct list_head *compound_pagelist)
 {
 	struct page *src_page, *tmp;
 	pte_t *_pte;
-	for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
-				_pte++, page++, address += PAGE_SIZE) {
-		pte_t pteval = *_pte;
+	pte_t pteval;
+	unsigned long _address;
+	spinlock_t *pmd_ptl;
+	bool copy_succeeded = true;
 
+	/*
+	 * Copying pages' contents is subject to memory poison at any iteration.
+	 */
+	for (_pte = pte, _address = address;
+			_pte < pte + HPAGE_PMD_NR;
+			_pte++, page++, _address += PAGE_SIZE) {
+		pteval = *_pte;
+
+		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval)))
+			clear_user_highpage(page, _address);
+		else {
+			src_page = pte_page(pteval);
+			if (copy_highpage_mc(page, src_page)) {
+				copy_succeeded = false;
+				break;
+			}
+		}
+	}
+
+	if (!copy_succeeded) {
+		/*
+		 * Copying failed, re-establish the regular PMD that points to
+		 * the regular page table. Restoring PMD needs to be done prior
+		 * to releasing pages. Since pages are still isolated and locked
+		 * here, acquiring anon_vma_lock_write is unnecessary.
+		 */
+		pmd_ptl = pmd_lock(vma->vm_mm, pmd);
+		pmd_populate(vma->vm_mm, pmd, pmd_pgtable(rollback));
+		spin_unlock(pmd_ptl);
+	}
+
+	for (_pte = pte, _address = address; _pte < pte + HPAGE_PMD_NR;
+			_pte++, _address += PAGE_SIZE) {
+		pteval = *_pte;
 		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
-			clear_user_highpage(page, address);
-			add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
-			if (is_zero_pfn(pte_pfn(pteval))) {
-				/*
-				 * ptl mostly unnecessary.
-				 */
-				spin_lock(ptl);
-				ptep_clear(vma->vm_mm, address, _pte);
-				spin_unlock(ptl);
+			if (copy_succeeded) {
+				add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
+				if (is_zero_pfn(pte_pfn(pteval))) {
+					/*
+					 * pte_ptl mostly unnecessary.
+					 */
+					spin_lock(pte_ptl);
+					pte_clear(vma->vm_mm, _address, _pte);
+					spin_unlock(pte_ptl);
+				}
 			}
 		} else {
 			src_page = pte_page(pteval);
-			copy_user_highpage(page, src_page, address, vma);
 			if (!PageCompound(src_page))
 				release_pte_page(src_page);
-			/*
-			 * ptl mostly unnecessary, but preempt has to
-			 * be disabled to update the per-cpu stats
-			 * inside page_remove_rmap().
-			 */
-			spin_lock(ptl);
-			ptep_clear(vma->vm_mm, address, _pte);
-			page_remove_rmap(src_page, vma, false);
-			spin_unlock(ptl);
-			free_page_and_swap_cache(src_page);
+			if (copy_succeeded) {
+				/*
+				 * pte_ptl mostly unnecessary, but preempt
+				 * has to be disabled to update the per-cpu
+				 * stats inside page_remove_rmap().
+				 */
+				spin_lock(pte_ptl);
+				ptep_clear(vma->vm_mm, _address, _pte);
+				page_remove_rmap(src_page, vma, false);
+				spin_unlock(pte_ptl);
+				free_page_and_swap_cache(src_page);
+			}
 		}
 	}
 
@@ -723,6 +779,8 @@  static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
 		free_swap_cache(src_page);
 		putback_lru_page(src_page);
 	}
+
+	return copy_succeeded;
 }
 
 static void khugepaged_alloc_sleep(void)
@@ -1009,6 +1067,7 @@  static void collapse_huge_page(struct mm_struct *mm,
 	struct vm_area_struct *vma;
 	struct mmu_notifier_range range;
 	gfp_t gfp;
+	bool copied = false;
 
 	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
 
@@ -1121,9 +1180,13 @@  static void collapse_huge_page(struct mm_struct *mm,
 	 */
 	anon_vma_unlock_write(vma->anon_vma);
 
-	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
-			&compound_pagelist);
+	copied = __collapse_huge_page_copy(pte, new_page, pmd, _pmd,
+			vma, address, pte_ptl, &compound_pagelist);
 	pte_unmap(pte);
+	if (!copied) {
+		result = SCAN_COPY_MC;
+		goto out_up_write;
+	}
 	/*
 	 * spin_lock() below is not the equivalent of smp_wmb(), but
 	 * the smp_wmb() inside __SetPageUptodate() can be reused to
@@ -2129,6 +2192,13 @@  static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
 				khugepaged_scan_file(mm, file, pgoff, hpage);
 				fput(file);
 			} else {
+				/*
+				 * mmap_read_lock is
+				 * 1) still held if scan failed;
+				 * 2) released if scan succeeded.
+				 * It is not affected by collapsing or copying
+				 * operations.
+				 */
 				ret = khugepaged_scan_pmd(mm, vma,
 						khugepaged_scan.address,
 						hpage);