Re: AutoNUMA15

[Andrea Arcangeli <aarcange@redhat.com>]

Hi Don,

On Thu, May 31, 2012 at 01:01:19PM -0700, Don Morris wrote:
> We're a special pte, but a non-zero pfn. Being Xorg, I'm
> assuming this is a remap of a kernel page into the user virtual
> address space, but that's just a gut instinct. Since I read

I reproduced it. The address is in /dev/mem, the other is a nonlinear
ext4 map.

> the above as "We don't expect to ever take spurious faults
> on instantiated special ptes", I would think you'd need

I would better skip VM_PFNMAP|VM_MIXEDMAP agreed, but it still
shouldn't fail like this, vma_normal_page shouldn't error out on a
pte_special.

> Of course... I'm still really ramping up on this kernel, so
> that could all be hokum, too. Hopefully it helps.

It helps a lot, thanks!

> I can dump the EFI memory map and whatnot to you if you
> need it, but I think this is more of an algorithmic issue

No need, I can reproduce.

On the bright side, it looks totally harmless and you can ignore it.

And if you run "echo 0 >/sys/kernel/mm/autonuma/knuma_scand/pmd" it
seems to go away but I suggest to keep the default and ignore it, the
pmd scan saves 1% of the overhead.

I'll push a fix in the origin/autonuma branch as soon as I figure it
out...

Andrea

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>