Re: gdbsx bug : fail on a breakpoint

Re: gdbsx bug : fail on a breakpoint

[Mukesh Rathor]

Hmm... I don't know whats the real cause. Is it during boot of the HVM
guest, or is the guest completely up and you are trying to set
breakpoing? 

For immediate use, can you just rebuilt the kernel with module built
in? I don't have a setup right now to debug/fix it, and am very busy
with PVH and preparing for xen summit, so need some time to debug and
fix it.

thanks,
mukesh

PS: please cc xen-devel on all such issues.



On Mon, 20 Aug 2012 07:46:45 +0000
"Gao, Ping" <Ping.Gao@amd.com> wrote:

> Hi Mukesh,
> 
> Is any progress on this gdbsx issue reported by Andras? I'm interest
> to take use of the gdbsx but suffer the problem seems the same.
> 
> Description:
> 
>         Set gdbsx to debug a HVM kernel module, fail to set a
> breakpoint.
> 
>         Kernel :
> konrad/xen.git<http://git.kernel.org/?p=linux/kernel/git/konrad/xen.git;a=summary>
> testing<http://git.kernel.org/?p=linux/kernel/git/konrad/xen.git;a=shortlog;h=refs/heads/testing>
> Xen    :  stable 4.1.2 (ubuntu 12.04) Result :  breakpoint failed,
> gdbsx show ERROR:xg_write_mem:ERROR: failed to write 0 bytes.
> errno:14 rc:-1
> 
> I'm not the sure the issue in the below link is due to the same root
> cause or not https://bugzilla.redhat.com/show_bug.cgi?id=678571
> 
> Wait for your good news.
> 
> Br,
> Ping
> 
> -----Original Message-----
> On Wed, 25 Jul 2012 16:55:36 +0200
> AndrÃs MÃhes <andras.mehes@xxxxxxxxxxxx> 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