| 1 |
On 30 Nov 2007 at 1:21, Alex Efros wrote: |
| 2 |
|
| 3 |
> Hi! |
| 4 |
> |
| 5 |
> On Thu, Nov 29, 2007 at 11:26:00PM +0200, pageexec@××××××××.hu wrote: |
| 6 |
> > 1. your exact kernel version + .config |
| 7 |
> |
| 8 |
> 2.6.20-hardened-r10, .config attached |
| 9 |
|
| 10 |
thanks, you could uprade to a newer kernel eventually, .20 hasn't been |
| 11 |
supported for a long time and i fixed bugs since (unrelated to the current |
| 12 |
issue though). |
| 13 |
|
| 14 |
> > 2. all PaX logs (if grsec removed address info, disable randomization |
| 15 |
> > and reproduce them that way) |
| 16 |
> |
| 17 |
> If you will not be able to reproduce this issue, notify me which options |
| 18 |
> in kernel config I should disable and I'll try it. Current logs are: |
| 19 |
|
| 20 |
thanks, i reproduced it with that site now. it seems that (probably) |
| 21 |
flash resorts to some runtime code generation, therefore chpax/paxctl |
| 22 |
-m on firefox and that opera wrapper is enough to get it to work, no |
| 23 |
need to disable non-exec pages altogether. not that it's a good solution |
| 24 |
as your browser will run without MPROTECT for all other webpages as |
| 25 |
well, but it's required unless flash is changed to not generate this |
| 26 |
code at runtime (by looking at the disasm, it seems to be a small function |
| 27 |
and the code doesn't look like that it really had to be runtime generated |
| 28 |
so chances are that this is fixable, but not by us obviously). |
| 29 |
|
| 30 |
> home ~ # paxctl -v /opt/firefox/firefox-bin |
| 31 |
> PaX control v0.5 |
| 32 |
> Copyright 2004,2005,2006,2007 PaX Team <pageexec@××××××××.hu> |
| 33 |
> |
| 34 |
> file /opt/firefox/firefox-bin does not have a PT_PAX_FLAGS program header, try conversion |
| 35 |
|
| 36 |
hm, so you didn't compile this firefox yourself nor was it provided |
| 37 |
by gentoo, correct? even then, paxctl -C should be able to handle it, |
| 38 |
ditto for the opera wrapper. |
| 39 |
|
| 40 |
> P.S. Not sure is it related to this issue, but it's usual for me to see |
| 41 |
> this in kernel log while using opera (opera continue working like nothing |
| 42 |
> is happens, but probably these messages mean some flash banners don't |
| 43 |
> working or so): |
| 44 |
|
| 45 |
these are 'normal' crashes, not due to PaX, but more likely application |
| 46 |
bugs. also hard to debug without source code/symbol info of course. |
| 47 |
|
| 48 |
> P.P.S. If I remember correctly, initial flags for both opera&ff binaries |
| 49 |
> was "PeMRxS". Now I've tried to enable S again, but looks like chpax |
| 50 |
> disable P when enable S and vice versa. So I unable to restore initial |
| 51 |
> "PeMRxS" flags. If this expected behaviour? |
| 52 |
|
| 53 |
-zex will restore the original flag state. |
| 54 |
|
| 55 |
-- |
| 56 |
gentoo-hardened@g.o mailing list |
Archives