Re: [Bochs-developers] bochs memory trace, found a bug?

[Stanislav S. <stl...@gm...>]

I don't understood which error you refer to (sorry can't guess it from the
screenshot).

Might be you mean the bug

 


2864794 

BX_INSTR_OPCODE
<https://sourceforge.net/tracker/?func=detail&aid=2864794&group_id=12580&ati
d=112580>  in "cpu_loop" causes crash in x86_64 host 

 

If Bochs crash for you - this is definitely Bochs issue and should be fixed.

 

My issue that I didn't hit the failure yet and therefore cannot track it
down.

If you can run Bochs with gdb and supply some info about segmentation fault
(where it happen, which function/line of code, why and etc) - it will help
me to fix a problem as soon as possible.

 

Thanks,

Stanislav

 

From: Eduardo Cruz [mailto:edu...@gm...] 
Sent: Wednesday, August 11, 2010 8:01 PM
To: boc...@li...
Subject: [Bochs-developers] bochs memory trace, found a bug?

 

Ok, I now have a recent debian amd64 system running in bochs.

I have hacked the linux kernel to save in a special register the task id of
the task running on the cpu.

By adding this support into bochs (which I am almost finishing) and using
the memory instrumentation macros, I can generate memory traces of specific
process.

I have already done this mod into qemu, but the results were weird, since
qemu executes hundreds of thousands instructions of a cpu before switching
to another cpu in smp environments.

I hope in bochs, by setting the cpu quantum to a small value, I get correct
results.

 

But, sometimes, I get an annoying error.

I took a screenshot, you can see it here:

http://img713.imageshack.us/img713/3848/capturadeteladn.png

Bochs logs the error (left window), and the kernel detects it too (right
window).

Sometimes, after this error appears, I get a segmentation fault.

 

It's not my hack's fault, since it was already happening before my
modifications.


-- 
Eduardo Henrique Molina da Cruz
MSc student
Parallel and Distributed Processing Group
Federal University of Rio Grande do Sul (UFRGS)