1 |
Hi! |
2 |
|
3 |
On Fri, Feb 24, 2012 at 03:41:27PM +0200, PaX Team wrote: |
4 |
> well, as i suggested it in bugzilla, i'd need to capture information about |
5 |
> the crash (probably triple fault), and the best approach would be some nested |
6 |
> virtualization setup. i have no idea how to do it easily (one way would be to |
7 |
> use bochs to run vmware if its vmx emulation is good enough but i guess the |
8 |
> resulting speed would be unbearable). |
9 |
|
10 |
That sounds too complex and slow. |
11 |
|
12 |
> another (but still time consuming) approach would be to do a binary search on |
13 |
> vmmon by stopping it at various points as it is about to launch a virtual machine, |
14 |
> that would eventually narrow down the failing code too. |
15 |
|
16 |
I probably can do this if you give me an example of how this should be |
17 |
done - i.e. example of code snippet to insert into vmmon source to stop at |
18 |
various point and start/end lines in code or involved function names where |
19 |
to place that snippet. |
20 |
|
21 |
BTW, I think it probably makes more sense to do a binary search from other |
22 |
side - remove parts of hardened changes in kernel. We already know it's |
23 |
not related to code disabled when PAX and GRSEC completely disabled in |
24 |
kernel config, so probably there is not so many hardened changes left |
25 |
which still active and may affect virtualization. |
26 |
|
27 |
-- |
28 |
WBR, Alex. |