]> git.karo-electronics.de Git - karo-tx-linux.git/commitdiff
Revert "x86, mm: Make spurious_fault check explicitly check the PRESENT bit"
authorAndrea Arcangeli <aarcange@redhat.com>
Wed, 20 Feb 2013 02:13:34 +0000 (13:13 +1100)
committerStephen Rothwell <sfr@canb.auug.org.au>
Wed, 20 Feb 2013 05:52:02 +0000 (16:52 +1100)
I got a report for a minor regression introduced by commit 027ef6c87853b
("mm: thp: fix pmd_present for split_huge_page and PROT_NONE with THP").

So the problem is, pageattr creates kernel pagetables (pte and pmds) that
breaks pte_present/pmd_present and the patch above exposed this invariant
breakage for pmd_present.

The same problem already existed for the pte and pte_present and it was
fixed by commit 660a293ea9be709 ("x86, mm: Make spurious_fault check
explicitly check the PRESENT bit") (if it wasn't for that commit, it
wouldn't even be a regression).  That fix avoids the pagefault to use
pte_present.  I could follow through by stopping using
pmd_present/pmd_huge too.

However I think it's more robust to fix pageattr and to clear the
PSE/GLOBAL bitflags too in addition to the present bitflag.  So the kernel
page fault can keep using the regular pte_present/pmd_present/pmd_huge.

The confusion arises because _PAGE_GLOBAL and _PAGE_PROTNONE are sharing
the same bit, and in the pmd case we pretend _PAGE_PSE to be set only in
present pmds (to facilitate split_huge_page final tlb flush).

This patch:

Revert commit 660a293ea9be709 ("x86, mm: Make spurious_fault check
explicitly check the PRESENT bit").

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: "H. Peter Anvin" <hpa@linux.intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
arch/x86/mm/fault.c

index fb674fd3fc2258f75400564acea0a835430f5988..2b97525246d43c3397f9d2149adf859b24b2842b 100644 (file)
@@ -939,14 +939,8 @@ spurious_fault(unsigned long error_code, unsigned long address)
        if (pmd_large(*pmd))
                return spurious_fault_check(error_code, (pte_t *) pmd);
 
-       /*
-        * Note: don't use pte_present() here, since it returns true
-        * if the _PAGE_PROTNONE bit is set.  However, this aliases the
-        * _PAGE_GLOBAL bit, which for kernel pages give false positives
-        * when CONFIG_DEBUG_PAGEALLOC is used.
-        */
        pte = pte_offset_kernel(pmd, address);
-       if (!(pte_flags(*pte) & _PAGE_PRESENT))
+       if (!pte_present(*pte))
                return 0;
 
        ret = spurious_fault_check(error_code, pte);