Re: Linux 2.6.35.2

[Sedat Dilek]

On Sat, Aug 14, 2010 at 8:47 PM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Sat, Aug 14, 2010 at 11:27 AM, Ed Tomlinson <edt@xxxxxx> wrote:
>>
>> It seems to be helping here. ÂWithout it 35.2 would either stall in boot or give lots of tracebacks. ÂWith it things
>> start normally and my background stuff works as expected.
>
> Ok, thanks. I've committed it to the -git tree.
>
> Just to verify: you have CONFIG_HIGHPTE enabled on a x86-32 kernel,
> right? I'd have expected this to not ever show up anywhere else, and
> I'm just verifying that there isn't anything else going on. I'm
> appending the commit message, the patch hasn't changed.
>
> Â Â Â Â Â Â Â Â Â Â Â Â Â ÂLinus
> ---
>
> Author: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> Date: Â Sat Aug 14 11:44:56 2010 -0700
>
> Â Âmm: fix page table unmap for stack guard page properly
>
> Â ÂWe do in fact need to unmap the page table _before_ doing the whole
> Â Âstack guard page logic, because if it is needed (mainly 32-bit x86 with
> Â ÂPAE and CONFIG_HIGHPTE, but other architectures may use it too) then it
> Â Âwill do a kmap_atomic/kunmap_atomic.
>
> Â ÂAnd those kmaps will create an atomic region that we cannot do
> Â Âallocations in. ÂHowever, the whole stack expand code will need to do
> Â Âanon_vma_prepare() and vma_lock_anon_vma() and they cannot do that in an
> Â Âatomic region.
>
> Â ÂNow, a better model might actually be to do the anon_vma_prepare() when
> Â Â_creating_ a VM_GROWSDOWN segment, and not have to worry about any of
> Â Âthis at page fault time. ÂBut in the meantime, this is the
> Â Âstraightforward fix for the issue.
>
> Â ÂSee https://bugzilla.kernel.org/show_bug.cgi?id=16588 for details.
>
> Â ÂReported-by: Wylda <wylda@xxxxxxxx>
> Â ÂReported-by: Sedat Dilek <sedat.dilek@xxxxxxxxx>
> Â ÂReported-by: Mike Pagano <mpagano@xxxxxxxxxx>
> Â ÂReported-by: FranÃois Valenduc <francois.valenduc@xxxxxxxxxxxxx>
> Â ÂTested-by: Ed Tomlinson <edt@xxxxxx>
> Â ÂCc: Pekka Enberg <penberg@xxxxxxxxxx>
> Â ÂCc: Greg KH <gregkh@xxxxxxx>
> Â ÂCc: stable@xxxxxxxxxx
> Â ÂSigned-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
>
> Âmm/memory.c | Â 13 ++++++-------
> Â1 files changed, 6 insertions(+), 7 deletions(-)
>

Missing: Tested-by: Wylda <wylda@xxxxxxxx>

I am building now upstream (2.6.35-git15) with that patch to see if it
fixes the issues there, too. Report later.

- Sedat -
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/