On 9 Dec 2005 at 2:49, JM wrote:
> Linux praisenet 2.6.13-rsbac-rsbac #2 Mon Dec 5 17:06:35 CET 2005 i686 
> Pentium 
> III (Katmai) GenuineIntel GNU/Linux

as far as PaX is concerned, 2.6.14 is the last 'supported' version,
as in, i fix stuff only in there, and i'm sure i did do so since the
2.6.13 port was abandoned. so if you can give that a try and reproduce
this, it'll help confirm/eliminate PaX bug at least.

> I have Pax and some RSBAC modules compiled in, if requested I may attach my 
> configuration.
> Should I send someone else this log too?

Amon Ott probably would be interested as well.

>  Dec  9 02:35:01 hostname PREEMPT

can you try without preempt? i never really audited PaX for such use,
even if i think most of the code is not sensitive to it, it's better
to leave it off.

>  Dec  9 02:35:01 hostname Call Trace:
>  Dec  9 02:35:01 hostname Code: 65 78 00 5f 5f 64 65 76 5f 67 65 74 5f 62 79 
> 5f 6e 61 6d 65 00 5f 5f 64 65 76 5f 72 65 6d 6f 76 65 5f 70 61 63 6b 00 5f 5f 
> 73 6b <62> 5f 6c 69 6e 65 61 72 69 7a 65 00 64 65 76 5f 61 64 64 5f 70

this points to something royally hosed. the above 'code' resolves to
a plain ascii string, eip fell into the middle of '__skb_linearize',
hardly valid machine code ;-). but short of a valid stacktrace, it's
hard to tell what the kernel was doing. maybe if you disabled module
support and enabled KERNEXEC you'd get a better stacktrace, but that's
just a guess. if you can reproduce it reliably, you could also just
try PaX (on 2.6.14 as well) and RSBAC alone.

-- 
[email protected] mailing list

Reply via email to