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

[Michael Eager <eager@eagercon.com>]

Hi Gopi --

I (or more accurately, git) fixed several whitespace issues.

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


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