Re: [RFC PATCH v3 3/4] mm: add do_set_pte_range()

[David Hildenbrand <david@redhat.com>]

On 03.02.23 14:30, Yin, Fengwei wrote:
> 
> 
> On 2/3/2023 9:25 PM, David Hildenbrand wrote:
>> On 03.02.23 14:16, Yin Fengwei wrote:
>>> do_set_pte_range() allows to setup page table entries for a
>>> specific range. It calls folio_add_file_rmap_range() to take
>>> advantage of batched rmap update for large folio.
>>>
>>> Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
>>> ---
>>>    include/linux/mm.h |  3 +++
>>>    mm/filemap.c       |  1 -
>>>    mm/memory.c        | 59 ++++++++++++++++++++++++++++++----------------
>>>    3 files changed, 42 insertions(+), 21 deletions(-)
>>>
>>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>>> index d6f8f41514cc..93192f04b276 100644
>>> --- a/include/linux/mm.h
>>> +++ b/include/linux/mm.h
>>> @@ -1162,6 +1162,9 @@ static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
>>>      vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
>>>    void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
>>> +void do_set_pte_range(struct vm_fault *vmf, struct folio *folio,
>>> +        unsigned long addr, pte_t *pte,
>>> +        unsigned long start, unsigned int nr);
>>>      vm_fault_t finish_fault(struct vm_fault *vmf);
>>>    vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
>>> diff --git a/mm/filemap.c b/mm/filemap.c
>>> index f444684db9f2..74046a3a0ff5 100644
>>> --- a/mm/filemap.c
>>> +++ b/mm/filemap.c
>>> @@ -3386,7 +3386,6 @@ static vm_fault_t filemap_map_folio_range(struct vm_fault *vmf,
>>>              ref_count++;
>>>            do_set_pte(vmf, page, addr);
>>> -        update_mmu_cache(vma, addr, vmf->pte);
>>>        } while (vmf->pte++, page++, addr += PAGE_SIZE, ++count < nr_pages);
>>>          /* Restore the vmf->pte */
>>> diff --git a/mm/memory.c b/mm/memory.c
>>> index 7a04a1130ec1..3754b2ef166a 100644
>>> --- a/mm/memory.c
>>> +++ b/mm/memory.c
>>> @@ -4257,36 +4257,58 @@ vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page)
>>>    }
>>>    #endif
>>>    -void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr)
>>> +void do_set_pte_range(struct vm_fault *vmf, struct folio *folio,
>>> +        unsigned long addr, pte_t *pte,
>>> +        unsigned long start, unsigned int nr)
>>>    {
>>>        struct vm_area_struct *vma = vmf->vma;
>>>        bool uffd_wp = pte_marker_uffd_wp(vmf->orig_pte);
>>>        bool write = vmf->flags & FAULT_FLAG_WRITE;
>>> +    bool cow = write && !(vma->vm_flags & VM_SHARED);
>>>        bool prefault = vmf->address != addr;
>>> +    struct page *page = folio_page(folio, start);
>>>        pte_t entry;
>>>    -    flush_icache_page(vma, page);
>>> -    entry = mk_pte(page, vma->vm_page_prot);
>>> +    if (!cow) {
>>> +        folio_add_file_rmap_range(folio, start, nr, vma, false);
>>> +        add_mm_counter(vma->vm_mm, mm_counter_file(page), nr);
>>> +    }
>>>    -    if (prefault && arch_wants_old_prefaulted_pte())
>>> -        entry = pte_mkold(entry);
>>> -    else
>>> -        entry = pte_sw_mkyoung(entry);
>>> +    do {
>>> +        flush_icache_page(vma, page);
>>> +        entry = mk_pte(page, vma->vm_page_prot);
>>>    -    if (write)
>>> -        entry = maybe_mkwrite(pte_mkdirty(entry), vma);
>>> -    if (unlikely(uffd_wp))
>>> -        entry = pte_mkuffd_wp(entry);
>>> -    /* copy-on-write page */
>>> -    if (write && !(vma->vm_flags & VM_SHARED)) {
>>> +        if (prefault && arch_wants_old_prefaulted_pte())
>>> +            entry = pte_mkold(entry);
>>> +        else
>>> +            entry = pte_sw_mkyoung(entry);
>>> +
>>> +        if (write)
>>> +            entry = maybe_mkwrite(pte_mkdirty(entry), vma);
>>> +        if (unlikely(uffd_wp))
>>> +            entry = pte_mkuffd_wp(entry);
>>> +        set_pte_at(vma->vm_mm, addr, pte, entry);
>>> +
>>> +        /* no need to invalidate: a not-present page won't be cached */
>>> +        update_mmu_cache(vma, addr, pte);
>>> +    } while (pte++, page++, addr += PAGE_SIZE, --nr > 0);
>>> +}
>>> +
>>> +void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr)
>>> +{
>>> +    struct folio *folio = page_folio(page);
>>> +    struct vm_area_struct *vma = vmf->vma;
>>> +    bool cow = (vmf->flags & FAULT_FLAG_WRITE) &&
>>> +            !(vma->vm_flags & VM_SHARED);
>>> +
>>> +    if (cow) {
>>>            inc_mm_counter(vma->vm_mm, MM_ANONPAGES);
>>>            page_add_new_anon_rmap(page, vma, addr);
>>
>> As raised, we cannot PTE-map a multi-page folio that way.
>>
>> This function only supports single-page anon folios.
>>
>> page_add_new_anon_rmap() -> folio_add_new_anon_rmap(). As that documents:
>>
>> "If the folio is large, it is accounted as a THP" -- for example, we would only increment the "entire mapcount" and set the PageAnonExclusive bit only on the head page.
>>
>> So this really doesn't work for multi-page folios and if the function would be used for that, we'd be in trouble.
>>
>> We'd want some fence here to detect that and bail out if we'd be instructed to do that. At least a WARN_ON_ONCE() I guess.
>> update_mmu_tlb(vma, vmf->address, vmf->pte);
>>
>> Right now the function looks like it might just handle that.
> You are right. I thought moving cow case out of it can make it explicit.
> But looks like it doesn't. I will add WARN_ON_ONCE(). Thanks.

I guess I would move the cow check into  do_set_pte_range() as well, and 
verify in there that we are really only dealing with a single-page 
folio, commenting that rmap code would need serious adjustment to make 
it work and that current code never passes a multi-page folio.

Thanks!

-- 
Thanks,

David / dhildenb