Re: gdbsx: likely bug in xg_write_mem

[Mukesh Rathor <mukesh.rathor@oracle.com>]

On Wed, 25 Jul 2012 16:55:36 +0200
András Méhes <andras.mehes@ericsson.com> wrote:

> Hi Mukesh,
> 
> 
> Thanks for the good work on Xen debugging!
> 
> I've just stumbled upon what I believe is a bug in 'xg_write_mem';
> specifically, the assumption that 'iop->remain == 0' implies success
> seems to be false.  First, the 'memset' call sets 'iop->remain' to 0.
> Then, the hypercall fails without touching 'iop->remain' and
> 'xg_write_mem' reports success....  Here's the relevant debug output
> from gdbsx:
> 
Hi Andras,

I  am currently out of office, and will look at it next week sometimes 
after I return and catch up. Thanks for pointing it out.

> That would be a much better alternative for setting breakpoints
> *before* the kernel is actually mapped in.  The very brute-force
> workaround with only software breakpoints -after the patch- is to
> keep single stepping until setting the breakpoint finally
> succeeds.. ;-(

Yeah, it's on the list, but I've been so busy with hybrid, and other
stuff. Besides, haven't had many people ask for it, so priority remains
low. Feel free to work on it tho, my top pri afteer I return is hybrid.

thanks,
Mukesh


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel