From: Lancelot SIX <Lancelot.Six@amd.com>
To: Tankut Baris Aktemur <tankutbaris.aktemur@amd.com>,
gdb-patches@sourceware.org
Subject: Re: [PATCH v2] gdb/amd-dbgapi-target: suppress a repeated stop request
Date: Wed, 20 May 2026 21:31:30 +0100 [thread overview]
Message-ID: <d8c803a2-3f37-41a3-bbaf-8ebb747c3aaa@amd.com> (raw)
In-Reply-To: <20260520115215.2042499-1-tankutbaris.aktemur@amd.com>
Hi,
On 20/05/2026 12:52, Tankut Baris Aktemur wrote:
> This revision uses a nested gdb_test_multiple/gdb_test when defining
> the user command in the test.
>
> Regards,
> Baris
>
> ====
>
> Sending a second stop request to an AMD GPU thread before fetching the
> event caused by the first request leads to an error:
>
> wave_stop for wave_1 failed (The wave has an outstanding stop request)
>
> Prevent sending a new stop request if there already is an outstanding
> one. The fix is in amd_dbgapi_target::stop.
>
> A regression test is included. The test uses non-stop mode and
> executes the "interrupt" command twice, because in non-stop mode this
> command uses the 'stop' target op, where the fix is applied.
>
> To be able to execute two interrupt commands repeatedly, we define a
> user command.
> ---
> gdb/amd-dbgapi-target.c | 8 ++-
> gdb/testsuite/gdb.rocm/interrupt-twice.cpp | 43 +++++++++++++
> gdb/testsuite/gdb.rocm/interrupt-twice.exp | 75 ++++++++++++++++++++++
> 3 files changed, 123 insertions(+), 3 deletions(-)
> create mode 100644 gdb/testsuite/gdb.rocm/interrupt-twice.cpp
> create mode 100644 gdb/testsuite/gdb.rocm/interrupt-twice.exp
>
> diff --git a/gdb/amd-dbgapi-target.c b/gdb/amd-dbgapi-target.c
> index 421ec8599ed..d44f03d0b80 100644
> --- a/gdb/amd-dbgapi-target.c
> +++ b/gdb/amd-dbgapi-target.c
> @@ -1090,14 +1090,16 @@ amd_dbgapi_target::stop (ptid_t ptid)
> sizeof (state), &state);
> if (status == AMD_DBGAPI_STATUS_SUCCESS)
> {
> - /* If the wave is already known to be stopped then do nothing. */
> - if (state == AMD_DBGAPI_WAVE_STATE_STOP)
> + wave_info &wi = get_thread_wave_info (thread);
> +
> + /* If the wave is already known to be stopped or there is an
> + outstanding stop request, then do nothing. */
> + if (state == AMD_DBGAPI_WAVE_STATE_STOP || wi.stopping)
> return;
>
> status = amd_dbgapi_wave_stop (wave_id);
> if (status == AMD_DBGAPI_STATUS_SUCCESS)
> {
> - wave_info &wi = get_thread_wave_info (thread);
> wi.stopping = true;
> return;
> }
> diff --git a/gdb/testsuite/gdb.rocm/interrupt-twice.cpp b/gdb/testsuite/gdb.rocm/interrupt-twice.cpp
> new file mode 100644
> index 00000000000..fc8d2cca697
> --- /dev/null
> +++ b/gdb/testsuite/gdb.rocm/interrupt-twice.cpp
> @@ -0,0 +1,43 @@
> +/* Copyright 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/>. */
> +
> +#include <hip/hip_runtime.h>
> +#include "gdb_watchdog.h"
> +
> +__device__ void
> +loop ()
> +{
> + while (true)
> + __builtin_amdgcn_s_sleep (8);
> +}
> +
> +__global__ void
> +kern ()
> +{
> + loop ();
> +}
> +
> +int
> +main ()
> +{
> + /* Make sure that if anything goes wrong, the program eventually
> + gets killed. */
> + gdb_watchdog (30);
> +
> + kern<<<1, 1>>> ();
> + return hipDeviceSynchronize () != hipSuccess;
> +}
> diff --git a/gdb/testsuite/gdb.rocm/interrupt-twice.exp b/gdb/testsuite/gdb.rocm/interrupt-twice.exp
> new file mode 100644
> index 00000000000..3c653547dc2
> --- /dev/null
> +++ b/gdb/testsuite/gdb.rocm/interrupt-twice.exp
> @@ -0,0 +1,75 @@
> +# Copyright 2026 Free Software Foundation, Inc.
> +
> +# 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/>.
> +
> +# Test that sending repeated stop requests to a running GPU thread
> +# does not cause a failure. This is done in non-stop mode because
> +# "interrupt" command in this mode uses the 'stop' target op.
> +
> +load_lib rocm.exp
> +
> +require allow_hipcc_tests
> +
> +standard_testfile .cpp
> +
> +if {[build_executable "failed to prepare" $testfile $srcfile {debug hip}]} {
> + return
> +}
> +
> +with_rocm_gpu_lock {
> + save_vars { ::GDBFLAGS } {
> + append ::GDBFLAGS " -ex \"set non-stop on\""
> + clean_restart $::testfile
> + }
> +
> + gdb_breakpoint "loop" {allow-pending} {temporary}
> + gdb_run_cmd
> +
> + set gpu_thread "undefined"
> +
> + gdb_test_multiple "" "hit breakpoint" {
> + -re -wrap "Thread ($decimal) \[^\r\n\]*hit Temporary breakpoint.*" {
> + set gpu_thread $expect_out(1,string)
> + pass $gdb_test_name
> + }
> + }
> +
> + gdb_test "thread $gpu_thread" "Switching to.*" "switch to gpu thread"
> +
> + # Resume the thread in the background. It will loop. Then we
> + # interrupt twice. To be able to run the "interrupt" command back
> + # to back, we define a user command.
> + gdb_test "continue &" "Continuing." "continue async"
> +
> + gdb_test_multiple "define inttwice" "" {
> + -re "End with .*\r\n>$" {
> + gdb_test "interrupt\ninterrupt\nend" "" $gdb_test_name
> + }
> + }
> +
> + # For logging purposes.
> + gdb_test "show user inttwice"
Not sure this is needed, but if we really have it, shouldn't we check
the output?
gdb_test "show user inttwice" \
[multi_line \
"User command \"inttwice\":" \
"\[\t \]*interrupt" \
"\[\t \]*interrupt" \
""]
> +
> + gdb_test_multiple "inttwice" "interrupt twice" {
> + -re "wave_stop \[^\r\n\]+ failed \[^\r\n\]+ outstanding stop request\\)\r\n" {
> + fail $gdb_test_name
> + }
> + -re "Thread $gpu_thread \[^\r\n\]*stopped" {
> + pass $gdb_test_name
> + }
> + -re "$gdb_prompt" {
I expect this should be (with " $")
-re "$gdb_prompt $" {
> + exp_continue
> + }
> + }
> +}
With this adjusted, this LGTM, thanks. I have tested this on gfx1031,
and also applied to downstream rocgdb.
Best,
Lancelot.
Approved-by: Lancelot Six <lancelot.six@amd.com> (amdgpu)
next prev parent reply other threads:[~2026-05-20 20:32 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-05-20 8:58 [PATCH] " Tankut Baris Aktemur
2026-05-20 10:09 ` Lancelot SIX
2026-05-20 11:49 ` Aktemur, Baris
2026-05-20 11:52 ` [PATCH v2] " Tankut Baris Aktemur
2026-05-20 20:31 ` Lancelot SIX [this message]
2026-05-21 6:45 ` Aktemur, Baris
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=d8c803a2-3f37-41a3-bbaf-8ebb747c3aaa@amd.com \
--to=lancelot.six@amd.com \
--cc=gdb-patches@sourceware.org \
--cc=tankutbaris.aktemur@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox