Re: PAE problems was [RFC] Reproducible OOM with just a few sleeps
From: H. Peter Anvin
Date: Fri Feb 01 2013 - 11:58:09 EST
On 02/01/2013 02:25 AM, Pavel Machek wrote:
On Fri 2013-02-01 11:20:44, Pavel Machek wrote:
On Thu 2013-01-31 23:38:27, Phil Turmel wrote:
On 01/31/2013 10:13 PM, paul.szabo@xxxxxxxxxxxxx wrote:
[trim /] Does not that prove that PAE is broken?
Please, Paul, take *yes* for an answer. It is broken. You've received
multiple dissertations on why it is going to stay that way. Unless you
fix it yourself, and everyone seems to be politely wishing you the best
of luck with that.
It is not Paul's job to fix PAE. It is job of whoever broke it to do
so.
If it is broken with 2GB of RAM, it is clearly not the known "lowmem
starvation" issue, it is something else... and probably worth
debugging.
So, Paul, if you have time and interest... Try to find some old kernel
version where sleep test works with PAE. Hopefully there is one. Then
do bisection... author of the patch should then fix it. (And if not,
at least you have patch you can revert.)
rjw is worth cc-ing at that point.
Ouch, and... IIRC (hpa should know for sure), PAE is neccessary for
R^X support on x86, thus getting more common, not less. If it does not
work, that's bad news.
Actually, if PAE is known broken, it should probably get marked as
such in Kconfig. That's sure to get some discussion started...
Pavel
OK, so by the time this thread gets to me there is of course no
information in it.
The vast majority of all 32-bit kernels compiled these days are PAE, so
it would seem rather odd if PAE was totally broken.
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
--
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/