Re: [PATCH v2] gdb/MicroBlaze: Add support for native linux gdb

[Gopi Kumar Bulusu <gopi@sankhya.com>]

[-- Attachment #1: Type: text/plain, Size: 8401 bytes --]

namaskaaram

On Thu, Jun 11, 2026, 6:09 AM Michael Eager <eager@eagercon.com> wrote:

> Hi Gopi --
>
> I (or more accurately, git) fixed several whitespace issues.
>

Please share. I did check all the whitespace issues flagged by git-diff.


> Since this is a revised patch, has it been retested?
>


Yes, it has been retested.

dhanyavaadaaha
gopi


> On 6/1/26 10:05 PM, Gopi Kumar Bulusu wrote:
> > namaskaaram
> >
> > Attached is the updated patch. All the changes suggested on PATCH v1
> > have been made and tested.
> >
> > I did not remove the (sal.line != 0) check as it seemed trivial.
> > However, I can make that a separate patch if necessary.
> >
> > Our sysadm In the process of setting up a git compatible email - the
> > process may take more time; until then appreciate your patience with
> > the attachments.
> >
> > Important Note:
> >
> > 1) gdb appears to have an elaborate scheme to handle C++ exceptions
> > and "tunnel" them through exception unaware C code. This code is with
> > readline. Anytime
> > an error is discovered (even a syntax error) by a module,
> > gdb_exception_error() thrown is tunneled through readline using the
> > TRY_SJLJ/CATCH_SJLJ macros
> > implemented using setjmp/longjmp.
> >
> > 2) There is an issue that I came across while testing; native gdb goes
> > into an infinite loop when handling errors. I analyzed it and found it
> > to be an obvious glibc
> > sysdeps/unix/sysv/linux/microblaze/____longjmp_chk.S issue and
> > reported it to AMD. It is possible I missed out a glibc patch; At this
> > time - this is not a gdb issue.
> >
> > 3) Plan to follow through [2] and ensure a gcc/glibc defect report is
> > created if needed.
> >
> > dhanyavaadaaha
> > gopi
> >
> > On Tue, Jun 2, 2026 at 5:58 AM Michael Eager <eager@eagerm.com> wrote:
> >>
> >> Please post an updated patch for review.
> >>
> >> On 5/31/26 11:38 PM, Gopi Kumar Bulusu wrote:
> >>> On Fri, May 29, 2026 at 8:26 PM Tom Tromey <tom@tromey.com> wrote:
> >>>>
> >>>>>>>>> Gopi Kumar Bulusu <gopi@sankhya.com> writes:
> >>>>
> >>>>> Attached is a patch to add native linux support for gdb including
> >>>>> cache target support..
> >>>>
> >>>> Thanks for the patch.
> >>>>
> >>>> I read through it and have some nits.  There's nothing very serious
> >>>> except this:
> >>>>
> >>>>> The original sources for the patch come from Xilinx/AMD Yocto git
> >>>>> repository (2025.2)
> >>>>
> >>>> Who wrote them and are they covered by the copyright assignment?
> >>>>
> >>>> This is the most important consideration, the patch can't land without
> >>>> this being clear.
> >>>
> >>> This work is being performed under a contract with Xilinx/AMD as
> >>> confirmed in a different response to
> >>> Simon and therefore covered by the copyright assignment.
> >>>
> >>> I will make the changes as described below and retest.
> >>>
> >>> Approved for commit ?
> >>>
> >>> dhanyavaadaaha
> >>> gopi
> >>>
> >>>>
> >>>>>   From a3328b4cdfc58783b5f290f22255b36e26a6fb14 Mon Sep 17 00:00:00
> 2001
> >>>>> From: Gopi Kumar Bulusu <gopi@sankhya.com>
> >>>>> Date: Thu, 7 May 2026 17:47:40 +0530
> >>>>> Subject: [PATCH] gdb/MicroBlaze: Add support for native linux gdb
> >>>>
> >>>> It's best to just git send-email rather than using an attachment.
> >>>>
> >>>>>       bfd/ChangeLog:
> >>>>
> >>>>>       * elf32-microblaze.c (microblaze_elf_grok_prstatus): New
> function.
> >>>>>         (microblaze_elf_grok_psinfo): Likewise.
> >>>>
> >>>>>       gdb/ChangeLog:
> >>>>
> >>>>>       * Makefile.in (HFILES_NO_SRCDIR): Add microblaze-linux-tdep.h
> >>>>>         ALLDEPFILES: Add microblaze-linux-nat.c
> >>>>
> >>>> gdb doesn't use ChangeLogs any more and this text should all be
> removed.
> >>>>
> >>>>> +#define MICROBLAZE_TARGET_HAS_GETREGS 0
> >>>>> +#define MICROBLAZE_TARGET_HAS_SETREGS 0
> >>>>
> >>>> This seems weird:
> >>>>
> >>>>> +/* Wrapper function around ptrace.  */
> >>>>> +
> >>>>> +static void
> >>>>> +fetch_target_gp_regs (int tid, elf_gregset_t *gregs)
> >>>>> +{
> >>>>> +  elf_greg_t *gregp = *gregs;
> >>>>> +
> >>>>> +#if MICROBLAZE_TARGET_HAS_GETREGS
> >>>>> +  if (ptrace (PTRACE_GETREGS, tid, 0, (long) gregp) < 0)
> >>>>> +    {
> >>>>> +      perror_with_name (_("Couldn't get registers"));
> >>>>> +      return;
> >>>>> +    }
> >>>>> +#error "this configuration did not work during testing"
> >>>>
> >>>> Normally we don't put new dead code into gdb.
> >>>>
> >>>> I think this and the defines could just be replaced with a comment
> >>>> explaining that PTRACE_GETREGS doesn't work on this platform.
> >>>>
> >>>>> +++ b/gdb/microblaze-linux-tdep.h
> >>>>> @@ -0,0 +1,24 @@
> >>>>> +/* Target-dependent code for GNU/Linux on MicroBlaze.
> >>>>> +
> >>>>> +   Copyright (C) 2021-2026 Free Software Foundation, Inc.
> >>>>> +
> >>>>> +   This file is part of GDB.
> >>>>> +
> >>>>> +   This program is free software; you can redistribute it and/or
> modify
> >>>>> +   it under the terms of the GNU General Public License as
> published by
> >>>>> +   the Free Software Foundation; either version 3 of the License, or
> >>>>> +   (at your option) any later version.
> >>>>> +
> >>>>> +   This program is distributed in the hope that it will be useful,
> >>>>> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> >>>>> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> >>>>> +   GNU General Public License for more details.
> >>>>> +
> >>>>> +   You should have received a copy of the GNU General Public License
> >>>>> +   along with this program.  If not, see <
> http://www.gnu.org/licenses/>.  */
> >>>>> +#ifndef MICROBLAZE_LINUX_TDEP_H
> >>>>> +#define MICROBLAZE_LINUX_TDEP_H
> >>>>> + /* Target descriptions.  */
> >>>>> + extern struct target_desc *tdesc_microblaze_linux;
> >>>>
> >>>> Blank line between the #define and the comment.
> >>>>
> >>>> Also this define should have a slightly different name, see
> >>>> check-include-guards.py.
> >>>>
> >>>>>           non_stack_instruction_found = 0;
> >>>>> +          cache->register_offsets[rd] = -imm;
> >>>>>           continue;
> >>>>
> >>>> Indentation looks wrong, probably not using tabs.
> >>>>
> >>>>> @@ -399,8 +400,7 @@ microblaze_skip_prologue (struct gdbarch
> *gdbarch, CORE_ADDR start_pc)
> >>>>>        {
> >>>>>          sal = find_sal_for_pc (func_start, 0);
> >>>>
> >>>>> -      if (sal.end < func_end
> >>>>> -       && start_pc <= sal.end)
> >>>>> +      if (sal.line != 0 && sal.end <= func_end  && start_pc <=
> sal.end)
> >>>>>         start_pc = sal.end;
> >>>>
> >>>> No objection from me but this seems somewhat unrelated.
> >>>>
> >>>>
> >>>>>      /* Call for side effects.  */
> >>>>> -  get_frame_func (next_frame);
> >>>>> +  cache->pc = get_frame_func (next_frame);
> >>>>
> >>>> Please remove that comment as it is now wrong.
> >>>>
> >>>>> +  if (regnum == MICROBLAZE_SP_REGNUM)
> >>>>> +        regnum = 1;
> >>>>
> >>>> Wrong indentation.
> >>>>
> >>>>> +  if (regnum == -1) {
> >>>>> +    int i;
> >>>>> +
> >>>>> +    for (i = 0; i < MICROBLAZE_REDR_REGNUM; i++) {
> >>>>> +      regcache->raw_supply (i, regs + i * MICROBLAZE_REGISTER_SIZE);
> >>>>> +    }
> >>>>
> >>>> Wrong brace placement in this hunk.
> >>>>
> >>>>> +  cb (".reg", tdep->sizeof_gregset, tdep->sizeof_gregset,
> tdep->gregset, NULL,
> >>>>> +      cb_data);
> >>>>
> >>>> Normally in new code we're using 'nullptr' now.  Best would be to go
> >>>> through the whole patch and check for this.  You don't have to fix
> >>>> pre-existing code that is using NULL though.
> >>>>
> >>>>>      gdbarch *gdbarch
> >>>>>        = gdbarch_alloc (&info, gdbarch_tdep_up (new
> microblaze_gdbarch_tdep));
> >>>>
> >>>>> +  microblaze_gdbarch_tdep *tdep
> >>>>> +    = gdbarch_tdep<microblaze_gdbarch_tdep> (gdbarch);
> >>>>> +
> >>>>> +  tdep->gregset = NULL;
> >>>>> +  tdep->sizeof_gregset = 0;
> >>>>> +  tdep->fpregset = NULL;
> >>>>> +  tdep->sizeof_fpregset = 0;
> >>>>
> >>>> Here it would be better to just add inline initializers to the new
> >>>> fields in microblaze_gdbarch_tdep.  Then they'll automatically be
> >>>> initialized by the 'new'.
> >>>>
> >>>> Tom
> >>>
> >>
>
> --
> Michael Eager
>
>

[-- Attachment #2: Type: text/html, Size: 12934 bytes --]