Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1516664 > unrolled thread

[PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6 Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> - 2016-11-08 00:40 +0100
    Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to  bit 6 Dave Hansen <dave.hansen@intel.com> - 2016-11-11 00:40 +0100
      Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7  to bit 6 Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> - 2016-11-11 02:20 +0100
      Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7  to bit 6 "Kirill A. Shutemov" <kirill@shutemov.name> - 2016-11-11 12:20 +0100

#1516664 — [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6

pmd_present() checks _PAGE_PSE along with _PAGE_PRESENT to avoid false negative
return when it races with thp spilt (during which _PAGE_PRESENT is temporary
cleared.) I don't think that dropping _PAGE_PSE check in pmd_present() works
well because it can hurt optimization of tlb handling in thp split.
In the current kernel, bit 6 is not used in non-present format because nonlinear
file mapping is obsolete, so let's move _PAGE_SWP_SOFT_DIRTY to that bit.
Bit 7 is used as reserved (always clear), so please don't use it for other
purpose.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
---
 arch/x86/include/asm/pgtable_types.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git v4.9-rc2-mmotm-2016-10-27-18-27/arch/x86/include/asm/pgtable_types.h v4.9-rc2-mmotm-2016-10-27-18-27_patched/arch/x86/include/asm/pgtable_types.h
index 8b4de22..89e88f1 100644
--- v4.9-rc2-mmotm-2016-10-27-18-27/arch/x86/include/asm/pgtable_types.h
+++ v4.9-rc2-mmotm-2016-10-27-18-27_patched/arch/x86/include/asm/pgtable_types.h
@@ -98,14 +98,14 @@
  * Tracking soft dirty bit when a page goes to a swap is tricky.
  * We need a bit which can be stored in pte _and_ not conflict
  * with swap entry format. On x86 bits 6 and 7 are *not* involved
- * into swap entry computation, but bit 6 is used for nonlinear
- * file mapping, so we borrow bit 7 for soft dirty tracking.
+ * into swap entry computation, but bit 7 is used for thp migration,
+ * so we borrow bit 6 for soft dirty tracking.
  *
  * Please note that this bit must be treated as swap dirty page
- * mark if and only if the PTE has present bit clear!
+ * mark if and only if the PTE/PMD has present bit clear!
  */
 #ifdef CONFIG_MEM_SOFT_DIRTY
-#define _PAGE_SWP_SOFT_DIRTY	_PAGE_PSE
+#define _PAGE_SWP_SOFT_DIRTY	_PAGE_DIRTY
 #else
 #define _PAGE_SWP_SOFT_DIRTY	(_AT(pteval_t, 0))
 #endif
-- 
2.7.0

[toc] | [next] | [standalone]

#1519442 — Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6

On 11/07/2016 03:31 PM, Naoya Horiguchi wrote:
> pmd_present() checks _PAGE_PSE along with _PAGE_PRESENT to avoid false negative
> return when it races with thp spilt (during which _PAGE_PRESENT is temporary
> cleared.) I don't think that dropping _PAGE_PSE check in pmd_present() works
> well because it can hurt optimization of tlb handling in thp split.
> In the current kernel, bit 6 is not used in non-present format because nonlinear
> file mapping is obsolete, so let's move _PAGE_SWP_SOFT_DIRTY to that bit.
> Bit 7 is used as reserved (always clear), so please don't use it for other
> purpose.
...
>  #ifdef CONFIG_MEM_SOFT_DIRTY
> -#define _PAGE_SWP_SOFT_DIRTY	_PAGE_PSE
> +#define _PAGE_SWP_SOFT_DIRTY	_PAGE_DIRTY
>  #else
>  #define _PAGE_SWP_SOFT_DIRTY	(_AT(pteval_t, 0))
>  #endif

I'm not sure this works.  Take a look at commit 00839ee3b29 and the
erratum it works around.  I _think_ this means that a system affected by
the erratum might see an erroneous _PAGE_SWP_SOFT_DIRTY/_PAGE_DIRTY get
set in swap ptes.

There are much worse things that can happen, but I don't think bits 5
(Accessed) and 6 (Dirty) are good choices since they're affected by the
erratum.

[toc] | [prev] | [next] | [standalone]

#1519498 — Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6

On Thu, Nov 10, 2016 at 03:29:51PM -0800, Dave Hansen wrote:
> On 11/07/2016 03:31 PM, Naoya Horiguchi wrote:
> > pmd_present() checks _PAGE_PSE along with _PAGE_PRESENT to avoid false negative
> > return when it races with thp spilt (during which _PAGE_PRESENT is temporary
> > cleared.) I don't think that dropping _PAGE_PSE check in pmd_present() works
> > well because it can hurt optimization of tlb handling in thp split.
> > In the current kernel, bit 6 is not used in non-present format because nonlinear
> > file mapping is obsolete, so let's move _PAGE_SWP_SOFT_DIRTY to that bit.
> > Bit 7 is used as reserved (always clear), so please don't use it for other
> > purpose.
> ...
> >  #ifdef CONFIG_MEM_SOFT_DIRTY
> > -#define _PAGE_SWP_SOFT_DIRTY	_PAGE_PSE
> > +#define _PAGE_SWP_SOFT_DIRTY	_PAGE_DIRTY
> >  #else
> >  #define _PAGE_SWP_SOFT_DIRTY	(_AT(pteval_t, 0))
> >  #endif
> 
> I'm not sure this works.  Take a look at commit 00839ee3b29 and the
> erratum it works around.  I _think_ this means that a system affected by
> the erratum might see an erroneous _PAGE_SWP_SOFT_DIRTY/_PAGE_DIRTY get
> set in swap ptes.
> 
> There are much worse things that can happen, but I don't think bits 5
> (Accessed) and 6 (Dirty) are good choices since they're affected by the
> erratum.

Thank you for the information. According to 00839ee3b29, some bits which
are safe from the errata are reclaimed, so assigning one of such bits for
_PAGE_SWP_SOFT_DIRTY seems to work. And I'll update the description.

Thanks,
Naoya Horiguchi

[toc] | [prev] | [next] | [standalone]

#1519699 — Re: [PATCH v2 01/12] mm: x86: move _PAGE_SWP_SOFT_DIRTY from bit 7 to bit 6

On Thu, Nov 10, 2016 at 03:29:51PM -0800, Dave Hansen wrote:
> On 11/07/2016 03:31 PM, Naoya Horiguchi wrote:
> > pmd_present() checks _PAGE_PSE along with _PAGE_PRESENT to avoid false negative
> > return when it races with thp spilt (during which _PAGE_PRESENT is temporary
> > cleared.) I don't think that dropping _PAGE_PSE check in pmd_present() works
> > well because it can hurt optimization of tlb handling in thp split.
> > In the current kernel, bit 6 is not used in non-present format because nonlinear
> > file mapping is obsolete, so let's move _PAGE_SWP_SOFT_DIRTY to that bit.
> > Bit 7 is used as reserved (always clear), so please don't use it for other
> > purpose.
> ...
> >  #ifdef CONFIG_MEM_SOFT_DIRTY
> > -#define _PAGE_SWP_SOFT_DIRTY	_PAGE_PSE
> > +#define _PAGE_SWP_SOFT_DIRTY	_PAGE_DIRTY
> >  #else
> >  #define _PAGE_SWP_SOFT_DIRTY	(_AT(pteval_t, 0))
> >  #endif
> 
> I'm not sure this works.  Take a look at commit 00839ee3b29 and the
> erratum it works around.  I _think_ this means that a system affected by
> the erratum might see an erroneous _PAGE_SWP_SOFT_DIRTY/_PAGE_DIRTY get
> set in swap ptes.

But, is it destructive in any way? What is the harm if we mark swap entry
dirty by mistake?

Pavel?

-- 
 Kirill A. Shutemov

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web