Re: [PATCH v1 4/6] mm/autonuma: use can_change_(pte|pmd)_writable() to replace savedwrite

[David Hildenbrand <david@redhat.com>]

On 02.11.22 22:22, Nadav Amit wrote:
> On Nov 2, 2022, at 12:12 PM, David Hildenbrand <david@redhat.com> wrote:
> 
>> !! External Email
>>
>> commit b191f9b106ea ("mm: numa: preserve PTE write permissions across a
>> NUMA hinting fault") added remembering write permissions using ordinary
>> pte_write() for PROT_NONE mapped pages to avoid write faults when
>> remapping the page !PROT_NONE on NUMA hinting faults.
>>
> 
> [ snip ]
> 
> Here’s a very shallow reviewed with some minor points...

Appreciated.

> 
>> ---
>> include/linux/mm.h |  2 ++
>> mm/huge_memory.c   | 28 +++++++++++++++++-----------
>> mm/ksm.c           |  9 ++++-----
>> mm/memory.c        | 19 ++++++++++++++++---
>> mm/mprotect.c      |  7 ++-----
>> 5 files changed, 41 insertions(+), 24 deletions(-)
>>
>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>> index 25ff9a14a777..a0deeece5e87 100644
>> --- a/include/linux/mm.h
>> +++ b/include/linux/mm.h
>> @@ -1975,6 +1975,8 @@ extern unsigned long move_page_tables(struct vm_area_struct *vma,
>> #define  MM_CP_UFFD_WP_ALL                 (MM_CP_UFFD_WP | \
>>                                             MM_CP_UFFD_WP_RESOLVE)
>>
>> +bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
>> +                            pte_t pte);
> 
> It might not be customary, but how about marking it as __pure?

Right, there is no a single use of __pure in the mm domain.

> 
>> extern unsigned long change_protection(struct mmu_gather *tlb,
>>                               struct vm_area_struct *vma, unsigned long start,
>>                               unsigned long end, pgprot_t newprot,
>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>> index 2ad68e91896a..45abd27d75a0 100644
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -1462,8 +1462,7 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
>>         unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
>>         int page_nid = NUMA_NO_NODE;
>>         int target_nid, last_cpupid = (-1 & LAST_CPUPID_MASK);
>> -       bool migrated = false;
>> -       bool was_writable = pmd_savedwrite(oldpmd);
>> +       bool try_change_writable, migrated = false;
>>         int flags = 0;
>>
>>         vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
>> @@ -1472,13 +1471,22 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
>>                 goto out;
>>         }
>>
>> +       /* See mprotect_fixup(). */
>> +       if (vma->vm_flags & VM_SHARED)
>> +               try_change_writable = vma_wants_writenotify(vma, vma->vm_page_prot);
>> +       else
>> +               try_change_writable = !!(vma->vm_flags & VM_WRITE);
> 
> Do you find it better to copy the code instead of extracting it to a
> separate function?

Yeah, you're right ;) usually the issue is coming up with a suitable name. Let me try.

vma_wants_manual_writability_change() hmm ...

> 
>> +
>>         pmd = pmd_modify(oldpmd, vma->vm_page_prot);
>>         page = vm_normal_page_pmd(vma, haddr, pmd);
>>         if (!page)
>>                 goto out_map;
>>
>>         /* See similar comment in do_numa_page for explanation */
>> -       if (!was_writable)
>> +       if (try_change_writable && !pmd_write(pmd) &&
>> +            can_change_pmd_writable(vma, vmf->address, pmd))
>> +               pmd = pmd_mkwrite(pmd);
>> +       if (!pmd_write(pmd))
>>                 flags |= TNF_NO_GROUP;
>>
>>         page_nid = page_to_nid(page);
>> @@ -1523,8 +1531,12 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
>>         /* Restore the PMD */
>>         pmd = pmd_modify(oldpmd, vma->vm_page_prot);
>>         pmd = pmd_mkyoung(pmd);
>> -       if (was_writable)
>> +
>> +       /* Similar to mprotect() protection updates, avoid write faults. */
>> +       if (try_change_writable && !pmd_write(pmd) &&
>> +            can_change_pmd_writable(vma, vmf->address, pmd))
> 
> Why do I have a deja-vu? :)
> 
> There must be a way to avoid the redundant code and specifically the call to
> can_change_pmd_writable(), no?

The issue is that as soon as we drop the page table lock, that information is stale.
Especially, after we fail migration.

So the following should work, however, if we fail migration we wouldn't map the
page writable and would have to re-calculate:

diff --git a/mm/memory.c b/mm/memory.c
index c5599a9279b1..a997625641e4 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -4674,10 +4674,10 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
         struct vm_area_struct *vma = vmf->vma;
         struct page *page = NULL;
         int page_nid = NUMA_NO_NODE;
+       bool writable = false;
         int last_cpupid;
         int target_nid;
         pte_t pte, old_pte;
-       bool was_writable = pte_savedwrite(vmf->orig_pte);
         int flags = 0;
  
         /*
@@ -4696,6 +4696,17 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
         old_pte = ptep_get(vmf->pte);
         pte = pte_modify(old_pte, vma->vm_page_prot);
  
+       /*
+        * Detect now whether the PTE is or can be writable. Note that this
+        * information is valid as long as we're holding the PT lock, so also on
+        * the remap path below.
+        */
+       writable = pte_write(pte);
+       if (!writable && vma_wants_manual_writability_change(vma) &&
+           can_change_pte_writable(vma, vmf->address, pte);
+           writable = true;
+       }
+
         page = vm_normal_page(vma, vmf->address, pte);
         if (!page || is_zone_device_page(page))
                 goto out_map;
@@ -4712,7 +4723,7 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
          * pte_dirty has unpredictable behaviour between PTE scan updates,
          * background writeback, dirty balancing and application behaviour.
          */
-       if (!was_writable)
+       if (!writable)
                 flags |= TNF_NO_GROUP;
  
         /*
@@ -4738,6 +4749,7 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
                 put_page(page);
                 goto out_map;
         }
+       writable = false;
         pte_unmap_unlock(vmf->pte, vmf->ptl);
  
         /* Migrate to the requested node */
@@ -4767,7 +4779,7 @@ static vm_fault_t do_numa_page(struct vm_fault *vmf)
         old_pte = ptep_modify_prot_start(vma, vmf->address, vmf->pte);
         pte = pte_modify(old_pte, vma->vm_page_prot);
         pte = pte_mkyoung(pte);
-       if (was_writable)
+       if (writable)
                 pte = pte_mkwrite(pte);
         ptep_modify_prot_commit(vma, vmf->address, vmf->pte, old_pte, pte);
         update_mmu_cache(vma, vmf->address, vmf->pte);


To me, the less error-prone approach is to re-calculate.


[...]
>> --- a/mm/ksm.c
>> +++ b/mm/ksm.c
>> @@ -1069,7 +1069,6 @@ static int write_protect_page(struct vm_area_struct *vma, struct page *page,
>>
>>         anon_exclusive = PageAnonExclusive(page);
>>         if (pte_write(*pvmw.pte) || pte_dirty(*pvmw.pte) ||
>> -           (pte_protnone(*pvmw.pte) && pte_savedwrite(*pvmw.pte)) ||
> 
> Not related to your code, but it does not make me comfortable that PTE’s
> status bits (which are volatile) are not accessed in this manner.
> 
> Especially since the PTE is later saved into orig_pte. It would feel safer
> to do READ_ONCE(*pvmw.pte) and work on it (probably in a separate patch).

I assume you are talking about the dirty bit. I don't immediately see how something
could go wrong here, but I agree that it might look cleaner that way.

Anyhow, independent of this series, so I'll leave that alone for now but add a
note for the future.

-- 
Thanks,

David / dhildenb