Messages in this thread | | | Date | Tue, 6 Dec 2011 12:18:39 +0800 | Subject | Re: [PATCH][mm] adjust the logic of checking THP | From | GuanJun He <> |
| |
On Wed, Nov 23, 2011 at 8:04 AM, Andrea Arcangeli <aarcange@redhat.com> wrote: > On Wed, Nov 02, 2011 at 02:34:30PM +0800, Guanjun He wrote: >> >> Acturally, pmd_trans_huge(orig_pmd) only checks the _PAGE_PSE bits, >> it's a pmd entry bits, only mark a size, not a flag;As one can easily >> create the same pmd entry bits for some special use,then the check >> will get confused.And this patch is to adjust the logic to use the flag, >> it can perfectly avoid this potential issuse,and basically no impact >> to the current code. > > You can't use _PAGE_PSE for special use for the pmd. Besides this is > common code, archs without such bit can define pmd_trans_huge to > return 0 like it happens with TRANSPARENT_HUGEPAGE=n.
ok, even _PAGE_PSE can not be used for special use for the pmd, This check is more efficient. the original code always check 2 items " pmd_none(*pmd) && transparent_hugepage_enabled(vma) ", even TRANSPARENT_HUGEPAGE not enabled.
> >> diff --git a/mm/memory.c b/mm/memory.c >> index a56e3ba..a76b17f 100644 >> --- a/mm/memory.c >> +++ b/mm/memory.c >> @@ -3465,20 +3465,22 @@ int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, >> pmd = pmd_alloc(mm, pud, address); >> if (!pmd) >> return VM_FAULT_OOM; >> - if (pmd_none(*pmd) && transparent_hugepage_enabled(vma)) { >> - if (!vma->vm_ops) >> - return do_huge_pmd_anonymous_page(mm, vma, address, >> - pmd, flags); >> - } else { >> - pmd_t orig_pmd = *pmd; >> - barrier(); >> - if (pmd_trans_huge(orig_pmd)) { >> - if (flags & FAULT_FLAG_WRITE && >> - !pmd_write(orig_pmd) && >> - !pmd_trans_splitting(orig_pmd)) >> - return do_huge_pmd_wp_page(mm, vma, address, >> - pmd, orig_pmd); >> + if (transparent_hugepage_enabled(vma)) { >> + if (pmd_none(*pmd)) { >> + if (!vma->vm_ops) >> + return do_huge_pmd_anonymous_page(mm, vma, address, >> + pmd, flags); >> + } else { >> + pmd_t orig_pmd = *pmd; >> + barrier(); >> + if (pmd_trans_huge(orig_pmd)) { >> + if (flags & FAULT_FLAG_WRITE && >> + !pmd_write(orig_pmd) && >> + !pmd_trans_splitting(orig_pmd)) >> + return do_huge_pmd_wp_page(mm, vma, address, >> + pmd, orig_pmd); >> return 0; >> + } > > This will infinite loop if you disable THP at runtime while some > mapping needing cow is established. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |