Remote stub can't single-step, how to tell GDB?

Remote stub can't single-step, how to tell GDB?

[Deepak Saxena]

Hello,

I am working on porting the existing ARM KGDB stub into the 
grand-unified-kgdb project (http://kgdb.sf.net) and am looking
for some info on how to cleanup our single-step handling. For
the existing stub we basically took the get_next_pc() code
from gdb and put it in the kgdb-stub so that we can execute
single-steps in the stub itself. When we receive an 's' command,
we call get_next_pc() and stuff the breakpoint at that address.
IMHO, having this in the kernel is overkill and skimming the gdb src, 
I am guessing there is away to just force the user's client to do this, 
but I am not sure how. In arm-tdep.c, I see the following comment that 
makes me wonder if right now I have no choice but handle single-step 
in the kernel stub:

  /* Single stepping.  */
  /* XXX For an RDI target we should ask the target if it can single-step.  */
  set_gdbarch_software_single_step (gdbarch, arm_software_single_step);

Any pointers appreciated.

Tnx,
~Deepak

-- 
Deepak Saxena - dsaxena at plexity dot net - http://www.plexity.net/

"Unlike me, many of you have accepted the situation of your imprisonment and
 will die here like rotten cabbages." - Number 6

Re: Remote stub can't single-step, how to tell GDB?

[Daniel Jacobowitz]

On Thu, Aug 12, 2004 at 02:49:58PM -0700, Deepak Saxena wrote:
> 
> Hello,
> 
> I am working on porting the existing ARM KGDB stub into the 
> grand-unified-kgdb project (http://kgdb.sf.net) and am looking
> for some info on how to cleanup our single-step handling. For
> the existing stub we basically took the get_next_pc() code
> from gdb and put it in the kgdb-stub so that we can execute
> single-steps in the stub itself. When we receive an 's' command,
> we call get_next_pc() and stuff the breakpoint at that address.
> IMHO, having this in the kernel is overkill and skimming the gdb src, 
> I am guessing there is away to just force the user's client to do this, 
> but I am not sure how. In arm-tdep.c, I see the following comment that 
> makes me wonder if right now I have no choice but handle single-step 
> in the kernel stub:
> 
>   /* Single stepping.  */
>   /* XXX For an RDI target we should ask the target if it can single-step.  */
>   set_gdbarch_software_single_step (gdbarch, arm_software_single_step);
> 
> Any pointers appreciated.

Hi Deepak :-)

Does GDB currently send single-step packets to the kgdb stub?  What
version of GDB are you using?  The call to set_gdbarch_software_single_step
sets up software (client-GDB-controlled) single stepping, and it should
not ask the stub to do so.

[GDB's handling of target capabilities in this area is a bit shoddy. 
But here it looks like the default matches what you want anyway.]

-- 
Daniel Jacobowitz

Re: Remote stub can't single-step, how to tell GDB?

[Deepak Saxena]

On Aug 12 2004, at 18:36, Daniel Jacobowitz was caught saying:
> On Thu, Aug 12, 2004 at 02:49:58PM -0700, Deepak Saxena wrote:
> >   /* Single stepping.  */
> >   /* XXX For an RDI target we should ask the target if it can single-step.  */
> >   set_gdbarch_software_single_step (gdbarch, arm_software_single_step);
> > 
> > Any pointers appreciated.
> 
> Hi Deepak :-)
> 
> Does GDB currently send single-step packets to the kgdb stub?  What
> version of GDB are you using?  The call to set_gdbarch_software_single_step
> sets up software (client-GDB-controlled) single stepping, and it should
> not ask the stub to do so.
> 
> [GDB's handling of target capabilities in this area is a bit shoddy. 
> But here it looks like the default matches what you want anyway.]

You're right. The original stub was written with some ancient (5.x) 
version of GDB and it must have behaved differently. Debug remote 
on 6.0 shows GDB sending memory read/write commands.  

Tnx,
~Deepak

-- 
Deepak Saxena - dsaxena at plexity dot net - http://www.plexity.net/

"Unlike me, many of you have accepted the situation of your imprisonment and
 will die here like rotten cabbages." - Number 6