Re: [PATCH] mm: use memory barriers for non PMD-sized folios to sync

[Baolin Wang <baolin.wang@linux.alibaba.com>]

On 11/11/26 5:22 PM, David Hildenbrand (Arm) wrote:
> On 11/11/26 09:43, Baolin Wang wrote:
>>
>>
>> On 11/11/26 3:48 PM, David Hildenbrand (Arm) wrote:
>>> On 11/11/26 07:06, Baolin Wang wrote:
>>>> Commit 243cf1f67b36 ("mm: memory: fix truncation removing mapped folios")
>>>> uses PMD lock via sync_with_folio_pmd_zap() to synchronize when a cleared
>>>> PMD is found while unmapping a folio.
>>>>
>>>> However, for non PMD-sized folios, the cleared PMD can only come from
>>>> PTE table reclaim, which clears the PMD only after the rmaps for the
>>>> table have been removed. This is plain message passing, and taking
>>>> the PMD lock is more than needed, like I did in the previous patch[1].
>>>>
>>>> Without barriers, the walker can observe the cleared PMD but a stale
>>>> mapcount on weakly ordered architectures:
>>>>
>>>> CPU 0 (unmap folio)                 CPU 1 (MADV_DONTNEED, full PMD)
>>>> ================================    ================================
>>>> folio_lock()
>>>>                                       lock(ptl)
>>>>                                       clear PTEs
>>>>                                       remove rmaps   /* W mapcount */
>>>> [mapcount load satisfied early,
>>>>    folio still mapped]
>>>>                                       pmd_clear()    /* W pmd */
>>>>                                       unlock(ptl)
>>>> pmdp_get_lockless() -> none         /* R pmd */
>>>> skip the PMD
>>>> folio_mapped() -> true              /* R mapcount, stale */
>>>> unmapping assumed to have failed
>>>>
>>>> Pairing an smp_wmb() before the PMD clear with an smp_rmb() after the
>>>> PMD read forbids this outcome:
>>>>
>>>> CPU 0 (unmap folio)                 CPU 1 (MADV_DONTNEED, full PMD)
>>>> ================================    ================================
>>>>                                       remove rmaps   /* W mapcount */
>>>>                                       smp_wmb()
>>>>                                       pmd_clear()    /* W pmd */
>>>> pmdp_get_lockless() -> none         /* R pmd */
>>>> smp_rmb()
>>>> folio_mapped()                      /* R mapcount */
>>>>
>>>> If R pmd observes W pmd, R mapcount must observe W mapcount, so
>>>> folio_mapped() no longer reports the removed mappings.
>>>>
>>>> No changes for PMD-sized folios unmapping.
>>>>
>>>> [1] https://lore.kernel.org/all/542c97482eae84f86704bcee56482633d1b32321.1791616645.git.baolin.wang@linux.alibaba.com/
>>>> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
>>>> ---
>>>>    mm/internal.h        | 26 ++++++++++++++++++++++----
>>>>    mm/memory.c          | 12 ++++++++++--
>>>>    mm/page_vma_mapped.c |  2 +-
>>>>    3 files changed, 33 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/mm/internal.h b/mm/internal.h
>>>> index cd3675c044e0..a277bf4841f8 100644
>>>> --- a/mm/internal.h
>>>> +++ b/mm/internal.h
>>>> @@ -592,6 +592,7 @@ void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
>>>>     * sync_with_folio_pmd_zap - sync with concurrent zapping of folios
>>>>     * @mm: The mm_struct.
>>>>     * @pmdp: Pointer to the pmd that was found to be pmd_none().
>>>> + * @nr_pages: The number of pages in the folio.
>>>>     *
>>>>     * When we find a pmd_none() while unmapping a folio without holding the PTL,
>>>>     * zap_huge_pmd() may have cleared the PMD but not yet updated the folio
>>>> @@ -600,13 +601,30 @@ void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
>>>>     * against later mapcount reads. Skipping the PMD without synchronization
>>>>     * could make folio unmapping code assume that unmapping failed.
>>>>     *
>>>> - * Wait for concurrent zapping to complete by grabbing the PTL.
>>>> + * Wait for concurrent zapping to complete by grabbing the PTL for PMD-sized
>>>> + * folios, otherwise smp_rmb() is sufficient.
>>>>     */
>>>> -static inline void sync_with_folio_pmd_zap(struct mm_struct *mm, pmd_t *pmdp)
>>>> +static inline void sync_with_folio_pmd_zap(struct mm_struct *mm, pmd_t *pmdp,
>>>> +		unsigned int nr_pages)
>>>>    {
>>>> -	spinlock_t *ptl = pmd_lock(mm, pmdp);
>>>> +	if (nr_pages >= HPAGE_PMD_NR) {
>>>> +		spinlock_t *ptl = pmd_lock(mm, pmdp);
>>>>    
>>>> -	spin_unlock(ptl);
>>>> +		spin_unlock(ptl);
>>>> +		return;
>>>> +	}
>>>> +
>>>> +	/*
>>>> +	 * For non PMD-sized folios zapping, concurrent zap_empty_pte_table()
>>>> +	 * may clear the PMD after the rmaps for the table have been removed,
>>>> +	 * but later mapcount reads could still be observed before the PMD
>>>> +	 * read, making the folio appear to be still mapped after unmapping.
>>>> +	 * Order the PMD read before later mapcount reads, so that observing
>>>> +	 * the cleared PMD implies observing the rmap removals.
>>>> +	 *
>>>> +	 * Pairs with smp_wmb() in zap_pte_range().
>>>> +	 */
>>>> +	smp_rmb();
>>>>    }
>>>>    
>>>>    struct zap_details;
>>>> diff --git a/mm/memory.c b/mm/memory.c
>>>> index fe72e66c9c2e..1ee286cdd3af 100644
>>>> --- a/mm/memory.c
>>>> +++ b/mm/memory.c
>>>> @@ -2054,8 +2054,15 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
>>>>    	 * Clear the PMD after the rmaps for the table have been removed, in case
>>>>    	 * an incorrect mapcount is observed after unmapping the folio.
>>>>    	 */
>>>> -	if (can_reclaim_pt && direct_reclaim && addr == end)
>>>> +	if (can_reclaim_pt && direct_reclaim && addr == end) {
>>>> +		/*
>>>> +		 * rmap changes need to be observed before page tables get zapped.
>>>> +		 *
>>>> +		 * Pairs with smp_rmb() in sync_with_folio_pmd_zap().
>>>> +		 */
>>>> +		smp_wmb();
>>>>    		direct_reclaim = zap_empty_pte_table(mm, pmd, ptl, &pmdval);
>>>> +	}
>>>>    	pte_unmap_unlock(start_pte, ptl);
>>>>    
>>>>    	/*
>>>> @@ -2110,7 +2117,8 @@ static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
>>>>    			/* fall through */
>>>>    		} else if (pmd_none(*pmd)) {
>>>>    			if (details && details->single_folio)
>>>> -				sync_with_folio_pmd_zap(tlb->mm, pmd);
>>>> +				sync_with_folio_pmd_zap(tlb->mm, pmd,
>>>> +						folio_nr_pages(details->single_folio));
>>>>    			addr = next;
>>>>    			continue;
>>>>    		}
>>>> diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
>>>> index a931243ca42b..d793ad0c5364 100644
>>>> --- a/mm/page_vma_mapped.c
>>>> +++ b/mm/page_vma_mapped.c
>>>> @@ -295,7 +295,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
>>>>    			pvmw->ptl = NULL;
>>>>    		} else if (!pmd_present(pmde)) {
>>>>    			if (pvmw->flags & PVMW_SYNC)
>>>> -				sync_with_folio_pmd_zap(mm, pvmw->pmd);
>>>> +				sync_with_folio_pmd_zap(mm, pvmw->pmd, pvmw->nr_pages);
>>>>
>>>
>>> If there is no folio, why do we care about the sync? Should there be a folio
>>> here? Anyhow, I think sync_with_folio_pmd_zap() really should be consuming a folio.
>>
>> Since there are only two cases without a folio:
>> mapping_wrprotect_range() and pfn_mkclean_range(), I'm hesitant to
>> modify all page_vma_mapped_walk() callers to pass in a 'folio' parameter
>> for these 2 cases.
>>
>> Or should we add a new flag for the pfn walk?
>>
>> / Walk a PFN range which need not belong to a folio /
>> #define PVMW_PFN		(1 << 2)
>>
>> Then the changes would be:
>> } else if (!pmd_present(pmde)) {
>> 	if ((pvmw->flags & (PVMW_SYNC | PVMW_PFN)) == PVMW_SYNC)
>> 		sync_with_folio_pmd_zap(mm, pvmw->pmd,
>>                               pfn_folio(pvmw->pfn));
> If we're working on a folio, we better pass in a folio. That might require some cleanups, yes.

OK.

> The question is really whether pfn_mkclean_range + mapping_wrprotect_range really need that
> sync as well.

I don't think they need to sync. Similarly, page_mapped_in_vma() doesn't 
actually need sync either, since the subsequent operations don't check 
folio's mapcount.