rtems gdb stub with virtex issue

[Tim]

Hi all,

I've been testing the ppc405 (virtex) patches for rtems/Till's gdb stub 
and everything seems to be going fine.

I'm going to be out of the office from Feb 19th until the end of April 
20th (woohoo!) so I'm posting the patches to the list for now:

============
RTEMS patch:
============

rtems-4-8-branch-ppc405_debug_exception.diff

The patch is against 'rtems-4-8-branch' as checked out on 2007-11-28. 
We're stuck on the other side of the gettimeofday barrier until we can 
upgrade our tools :D The files touched by the patch haven't changed 
since then, so the patch should apply fine.

1. The patch adds a support for a further ppc405 exception, vector at 
0x2000. I've called it 0x15 (rather than 0x20) so it can be included in 
the critical vectors mask.

2. The ITLB and DTLB vectors were incorrect for ppc405 - now fixed.

3. The exception prologue needed a couple of lines to remap the debug 
exception address to the appropriate exception number. Not rapt about 
adding custom code into the prologue, so any other ideas would be welcome.

I've had a look at the cvs tip - looks like significant refactoring has 
gone on in this dept. I can look at including the debug exception there 
when I get home (late april!)

===============
GDB Stub patch:
===============

RTEMS_gdb_stub_1_4-ppc405-support.diff

The patch is against Till's latest gdb stub (version 1.4) from 
http://www.slac.stanford.edu/~strauman/rtems/gdb/index.html

The file 'rtems-gdb-stub-ppc-shared.c' is now sensitive to the 
definition of 'ppc405'.

1. FP register support is swithed according to CPU_HARDWARE_FP & 
CPU_SOFTWARE_FP
2. Single-stepping uses a different mechanism (using ppc405 debug vector)

It's not pretty, but if Till's stub on ppc405/virtex is what you're 
after, then these patches will get you that far.

Comments and critism (constructive and/or hilarious) welcome. Until 
thursday. :D

Later,
Tim



Tim Cussins wrote:
> Hi everyone,
> 
> I've sussed out the issue with rtems-stub on virtex - it requires
> changes on two fronts to get it going.
> 
> AFAICT, the powerpc standard doesn't actually specify a standard for
> hardware debugging support, so the ppc405 is a slightly different beast
> to the mpc7455 (used in the mvme5500 bsp) when it comes to debug
> flags/switches/modes etc.
> 
> This means an addition to rtems (in new-exceptions) and changes in the
> gdb stub (in the arch-dependent ppc code).
> 
> I'm working with the 4.8 branch atm, and have developed a patch to add
> support for a further ppc405 exception: the 'debug' exception. Yep, I
> was as surprised as you are. ;)
> 
> The arch-dependent part of the stub needs a some modifications to
> support the new 'debug' exception.
> 
> Stepping/breakpoints are mostly all go, but there's still some oddness
> when single-stepping that I need to sort out. I'll keep you all posted.
> 
> Later,
> Tim

-------------- next part --------------
A non-text attachment was scrubbed...
Name: rtems-4-8-branch-ppc405_debug_exception.diff
Type: text/x-patch
Size: 3051 bytes
Desc: not available
URL: <http://lists.rtems.org/pipermail/users/attachments/20080219/b6a4e89a/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: RTEMS_gdb_stub_1_4-ppc405-support.diff
Type: text/x-patch
Size: 4115 bytes
Desc: not available
URL: <http://lists.rtems.org/pipermail/users/attachments/20080219/b6a4e89a/attachment-0001.bin>