[PATCH v2] gdb/MicroBlaze: Add support for native linux gdb
Gopi Kumar Bulusu
gopi@sankhya.com
Tue Jun 2 05:05:00 GMT 2026
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
> >
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: v2-0001-gdb-MicroBlaze-Add-support-for-native-linux-gdb.patch
Type: text/x-patch
Size: 22426 bytes
Desc: not available
URL: <https://sourceware.org/pipermail/binutils/attachments/20260602/df7435e8/attachment-0001.bin>
More information about the Binutils
mailing list