| 1 |
On Mon, 08 Jan 2007 13:40:56 +1030 Shawn Haggett <podge@××××××××.com> |
| 2 |
wrote: |
| 3 |
|
| 4 |
> Hmm... I've had a search of the forums. People have been having |
| 5 |
> similar problems, but not the same as mine. Seems there's a problem |
| 6 |
> with the Borland compiler (which I assume the binary was compiled |
| 7 |
> with) and it putting code into the .data section. However I have |
| 8 |
> played with chpax and paxctl (for paxctl I had to convert the binary, |
| 9 |
> since it didn't have a pax header). It seems turning protections |
| 10 |
> on/off made no difference, it still dies at the same point with the |
| 11 |
> same error. I tried diff'ing the different strace outputs, the only |
| 12 |
> thing that seemed to change were memory addresses (as I would expect |
| 13 |
> with random memory layout). |
| 14 |
> |
| 15 |
> I'm new to this level of playing with binaries, so excuse me if |
| 16 |
> anything above doesn't make sense. |
| 17 |
|
| 18 |
Are you running teamspeak on a 64-bit kernel with a processor that |
| 19 |
supports the NX bit (`grep nx /proc/cpuinfo`)? |
| 20 |
|
| 21 |
On my hardened amd64 system, ts will not run at all, even with all |
| 22 |
protections disabled via chpax. However, when using chpax to disable |
| 23 |
PAGE_EXEC on an i386 system, ts will start. This led me to believe that |
| 24 |
on architectures where the NX bit is supported natively by the |
| 25 |
kernel/Pax [1], the protection cannot be disabled on a per-binary |
| 26 |
basis. Perhaps someone more knowledgeable can clarify this? |
| 27 |
|
| 28 |
I switched to ventrilo in the end, it runs without modification and |
| 29 |
the sound quality is significantly better. |
| 30 |
|
| 31 |
-atj |
| 32 |
|
| 33 |
[1] |
| 34 |
http://marc.theaimsgroup.com/?l=gentoo-hardened&m=114987924519660&w=2 |
| 35 |
-- |
| 36 |
gentoo-hardened@g.o mailing list |
Archives