From: Luis Machado <luis.machado@arm.com>
To: Thiago Jung Bauermann <thiago.bauermann@linaro.org>,
gdb-patches@sourceware.org
Cc: Christophe Lyon <christophe.lyon@linaro.org>
Subject: Re: [PATCH v3 4/5] gdb/testsuite: Add gdb.arch/aarch64-mops-single-step.exp
Date: Fri, 10 May 2024 14:07:11 +0100 [thread overview]
Message-ID: <de60db5e-7301-4a0e-9331-5b0b74d4a9b1@arm.com> (raw)
In-Reply-To: <20240510052408.2173579-5-thiago.bauermann@linaro.org>
Thanks. Just one nit.
On 5/10/24 06:24, Thiago Jung Bauermann wrote:
> The testcase verifies that MOPS sequences are correctly single-stepped.
> ---
> .../gdb.arch/aarch64-mops-single-step.c | 73 ++++++++++
> .../gdb.arch/aarch64-mops-single-step.exp | 132 ++++++++++++++++++
> 2 files changed, 205 insertions(+)
> create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-single-step.c
> create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-single-step.exp
>
> Changes in v3:
> - Renamed gdb.arch/aarch64-mops-atomic-inst.exp to
> gdb.arch/aarch64-mops-single-step.exp.
> - Adjusted test to expect the MOPS sequence to reset back to the prologue
> instruction.
> - Set size variable before the cpyf and cpy sequences, because after each
> sequence the variable is set to zero. This bug didn't affect the
> effectiveness of the test.
>
> Changes in v2:
> - Add prfm instruction after each MOPS sequence and look for it in the
> testcase to verify that the sequence was stepped through (Suggested by
> Christophe).
>
> diff --git a/gdb/testsuite/gdb.arch/aarch64-mops-single-step.c b/gdb/testsuite/gdb.arch/aarch64-mops-single-step.c
> new file mode 100644
> index 000000000000..4a27867d4b57
> --- /dev/null
> +++ b/gdb/testsuite/gdb.arch/aarch64-mops-single-step.c
> @@ -0,0 +1,73 @@
> +/* This file is part of GDB, the GNU debugger.
> +
> + Copyright 2024 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/>. */
> +
> +#define TEST_STRING "Just a test string."
> +#define BUF_SIZE sizeof(TEST_STRING)
> +
> +int
> +main (void)
> +{
> + char source[BUF_SIZE] = TEST_STRING;
> + char dest[BUF_SIZE];
> + char *p, *q;
> + long size, zero;
> +
> + /* Note: The prfm instruction in the asm statements below is there just
> + to allow the testcase to recognize when the PC is at the instruction
> + right after the MOPS sequence. */
> +
> + p = dest;
> + size = sizeof (dest);
> + zero = 0;
> + /* Break memset. */
> + /* memset implemented in MOPS instructions. */
> + __asm__ volatile ("setp [%0]!, %1!, %2\n\t"
> + "setm [%0]!, %1!, %2\n\t"
> + "sete [%0]!, %1!, %2\n\t"
> + "prfm pldl3keep, [%0, #0]\n\t"
> + : "+&r"(p), "+&r"(size)
> + : "r"(zero)
> + : "memory");
> +
> + p = dest;
> + q = source;
> + size = sizeof (dest);
> + /* Break memcpy. */
> + /* memcpy implemented in MOPS instructions. */
> + __asm__ volatile ("cpyfp [%0]!, [%1]!, %2!\n\t"
> + "cpyfm [%0]!, [%1]!, %2!\n\t"
> + "cpyfe [%0]!, [%1]!, %2!\n\t"
> + "prfm pldl3keep, [%0, #0]\n\t"
> + : "+&r" (p), "+&r" (q), "+&r" (size)
> + :
> + : "memory");
> +
> + p = dest;
> + q = source;
> + size = sizeof (dest);
> + /* Break memmove. */
> + /* memmove implemented in MOPS instructions. */
> + __asm__ volatile ("cpyp [%0]!, [%1]!, %2!\n\t"
> + "cpym [%0]!, [%1]!, %2!\n\t"
> + "cpye [%0]!, [%1]!, %2!\n\t"
> + "prfm pldl3keep, [%0, #0]\n\t"
> + : "+&r" (p), "+&r" (q), "+&r" (size)
> + :
> + : "memory");
> +
> + return 0;
> +}
> diff --git a/gdb/testsuite/gdb.arch/aarch64-mops-single-step.exp b/gdb/testsuite/gdb.arch/aarch64-mops-single-step.exp
> new file mode 100644
> index 000000000000..a6390d4bff7e
> --- /dev/null
> +++ b/gdb/testsuite/gdb.arch/aarch64-mops-single-step.exp
> @@ -0,0 +1,132 @@
> +# Copyright 2024 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/>.
> +#
> +# This file is part of the GDB testsuite.
> +
> +# Test single stepping through MOPS (memory operations) instruction sequences.
> +
> +require allow_aarch64_mops_tests
> +
> +standard_testfile
> +if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile} \
> + [list debug additional_flags=-march=armv9.3-a]] } {
> + return -1
> +}
> +
> +# INSTRUCTION should be just its mnemonic, without any arguments.
> +proc is_at_instruction { instruction } {
> + global gdb_prompt hex
> +
> + set test "pc points to $instruction"
> + gdb_test_multiple {x/i $pc} $test {
> + -re -wrap "=> $hex \[^\r\n\]+:\t$instruction\t\[^\r\n\]+" {
> + return 1
> + }
> + -re "\r\n$gdb_prompt $" {
> + return 0
> + }
> + }
> +
> + return 0
> +}
> +
> +proc arrive_at_instruction { instruction } {
> + set count 0
> +
> + while { [is_at_instruction $instruction] != 1 } {
> + gdb_test -nopass "stepi" ".*__asm__ volatile.*" \
> + "stepi #$count to reach $instruction"
> + incr count
> +
> + if { $count > 50 } {
> + fail "didn't reach $instruction"
> + return 0
> + }
> + }
> +
> + return 1
> +}
> +
> +# If the inferior is reschedule to another CPU while a main or epilogue
s/reschedule/rescheduled
> +# instruction is executed, the OS resets the inferior back to the prologue
> +# instruction, so we need to allow for that possibility.
> +proc step_through_sequence { prefix } {
> + set count 0
> +
> + while { [is_at_instruction ${prefix}p] == 1 && $count < 50 } {
> + incr count
> +
> + # The stepi output isn't useful to detect whether we stepped over
> + # the instruction.
> + gdb_test "stepi" "\[^\r\n\]+" "step over ${prefix}p"
> + if { [is_at_instruction ${prefix}m] == 1 } {
> + pass "stepped over ${prefix}p"
> + } else {
> + fail "stepped over ${prefix}e"
> + return 0
> + }
> +
> + gdb_test "stepi" "\[^\r\n\]+" "step over ${prefix}m"
> + if { [is_at_instruction ${prefix}e] == 1 } {
> + pass "stepped over ${prefix}m"
> + } elseif { [is_at_instruction ${prefix}p] == 1 } {
> + # The inferior was rescheduled to another CPU.
> + pass "${prefix}m: reset back to prologue"
> + continue
> + } else {
> + fail "stepped over ${prefix}m"
> + return 0
> + }
> +
> + gdb_test "stepi" "\[^\r\n\]+" "step over ${prefix}e"
> + if { [is_at_instruction prfm] == 1 } {
> + pass "stepped over ${prefix}e"
> + return 1
> + } elseif { [is_at_instruction ${prefix}p] == 1 } {
> + # The inferior was rescheduled to another CPU.
> + pass "${prefix}e: reset back to prologue"
> + continue
> + }
> + }
> +
> + fail "step through $prefix sequence"
> + return 0
> +}
> +
> +if ![runto_main] {
> + return -1
> +}
> +
> +gdb_breakpoint ${srcfile}:[gdb_get_line_number "Break memset"]
> +gdb_breakpoint ${srcfile}:[gdb_get_line_number "Break memcpy"]
> +gdb_breakpoint ${srcfile}:[gdb_get_line_number "Break memmove"]
> +
> +gdb_continue_to_breakpoint "memset breakpoint"
> +
> +if { [arrive_at_instruction setp] } {
> + step_through_sequence set
> +}
> +
> +gdb_continue_to_breakpoint "memcpy breakpoint"
> +
> +if { [arrive_at_instruction cpyfp] } {
> + step_through_sequence cpyf
> +}
> +
> +gdb_continue_to_breakpoint "memmove breakpoint"
> +
> +if { [arrive_at_instruction cpyp] } {
> + step_through_sequence cpy
> +}
next prev parent reply other threads:[~2024-05-10 13:07 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-10 5:24 [PATCH v3 0/5] Add support for AArch64 MOPS instructions Thiago Jung Bauermann
2024-05-10 5:24 ` [PATCH v3 1/5] gdb/aarch64: Disable displaced single-step for " Thiago Jung Bauermann
2024-05-10 5:24 ` [PATCH v3 2/5] gdb/aarch64: Add record support " Thiago Jung Bauermann
2024-05-10 12:59 ` Luis Machado
2024-05-23 2:04 ` Thiago Jung Bauermann
2024-05-10 5:24 ` [PATCH v3 3/5] gdb/testsuite: Add gdb.arch/aarch64-mops-watchpoint.exp Thiago Jung Bauermann
2024-05-10 13:04 ` Luis Machado
2024-05-23 2:06 ` Thiago Jung Bauermann
2024-05-10 5:24 ` [PATCH v3 4/5] gdb/testsuite: Add gdb.arch/aarch64-mops-single-step.exp Thiago Jung Bauermann
2024-05-10 13:07 ` Luis Machado [this message]
2024-05-23 2:08 ` Thiago Jung Bauermann
2024-05-10 5:24 ` [PATCH v3 5/5] gdb/testsuite: Add gdb.reverse/aarch64-mops.exp Thiago Jung Bauermann
2024-05-10 13:08 ` Luis Machado
2024-05-10 17:26 ` Guinevere Larsen
2024-05-23 1:55 ` Thiago Jung Bauermann
2024-05-10 14:16 ` [PATCH v3 0/5] Add support for AArch64 MOPS instructions Pedro Alves
2024-05-21 22:18 ` Thiago Jung Bauermann
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=de60db5e-7301-4a0e-9331-5b0b74d4a9b1@arm.com \
--to=luis.machado@arm.com \
--cc=christophe.lyon@linaro.org \
--cc=gdb-patches@sourceware.org \
--cc=thiago.bauermann@linaro.org \
/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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).