Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Thiago Jung Bauermann <thiago.bauermann@linaro.org>
To: <Ezra.Sitorus@arm.com>
Cc: <gdb-patches@sourceware.org>,  <luis.machado.foss@gmail.com>
Subject: Re: [PATCH 1/4] gdb/aarch64: record/replay support for LRCPC3
Date: Sat, 25 Apr 2026 02:08:36 -0300	[thread overview]
Message-ID: <87o6j7bp9n.fsf@linaro.org> (raw)
In-Reply-To: <20260420215232.68675-2-Ezra.Sitorus@arm.com> (Ezra Sitorus's message of "Mon, 20 Apr 2026 22:52:29 +0100")

Hello Ezra,

<Ezra.Sitorus@arm.com> writes:

> From: Ezra Sitorus <ezra.sitorus@arm.com>
>
> FEAT_LRCPC3 introduces various load/store instructions with release
> consistency for cases where ordering is required. This patch teaches GDB
> to decode these instructions for recording and reversing.
>
> The gdb.reverse/aarch64-lrcpc3.exp testcase verifies that the
> instructions are recorded and correctly reversed. In particular, there
> are some interesting cases to note:
> * ldapur/stlur are SIMD instructions, but are not decoded in the simd
>   function.
> * There are writeback cases to cover too. These were taken from the
>   binutils testcases: gas/testsuite/gas/aarch64/rcpc3.s.
>
> The full testsuite was done on aarch64-none-linux-gnu without LRCPC3.
> The gdb.arch and gdb.reverse tests were run on Shrinkwrap with LRCPC3
> support.
>
> Please note:
> 1) There is no support for LRCPC and LRCPC2 instructions
> 2) LRCPC3 is gated with +rcpc3 in GCC/binutils.
> ---
>  gdb/aarch64-tdep.c                           |  87 +++++
>  gdb/testsuite/gdb.reverse/aarch64-lrcpc3.c   | 319 +++++++++++++++++++
>  gdb/testsuite/gdb.reverse/aarch64-lrcpc3.exp | 203 ++++++++++++
>  gdb/testsuite/lib/gdb.exp                    |  63 ++++
>  4 files changed, 672 insertions(+)
>  create mode 100644 gdb/testsuite/gdb.reverse/aarch64-lrcpc3.c
>  create mode 100644 gdb/testsuite/gdb.reverse/aarch64-lrcpc3.exp

Thanks! I have just a few nits.

I appreciate the well-commented changes in aarch64-tdep.c. They made the
code easier to follow.

> diff --git a/gdb/aarch64-tdep.c b/gdb/aarch64-tdep.c
> index 4befaa2720d..81d4f160f8f 100644
> --- a/gdb/aarch64-tdep.c
> +++ b/gdb/aarch64-tdep.c
> @@ -5812,6 +5812,93 @@ aarch64_record_load_store (aarch64_insn_decode_record *aarch64_insn_r)
>  	  aarch64_insn_r->reg_rec_count = 1;
>  	}
>      }
> +  /* LRCPC3 instructions.  This covers ldiapp/stilp, ldapur/stlur (FP/SIMD),
> +     ldapr/stlr.  */
> +  else if ((insn_bits24_27 & 0x0b) == 0x09 && insn_bits28_29 == 0x01
> +	   && insn_bits10_11 == 0x02 && !insn_bit21)
> +    {
> +      /* ldapur/stlur (FP/SIMD), ldapr/stlr.  We can differentiate between the
> +	 2 types by checking the vector flag.  */
> +      if (insn_bit23 || vector_flag)
> +	{
> +	  /* For the vector instruction, the offset comes from the imm9
> +	     bitfield, whereas the other can only take possible values from the
> +	     size bitfield.  */
> +	  int16_t imm9_off = sbits (aarch64_insn_r->aarch64_insn, 12, 20);
> +	  offset = vector_flag ? imm9_off : -(1 << size_bits);
> +	  uint32_t regnum_offset = vector_flag ? AARCH64_V0_REGNUM : 0;
> +	  if (ld_flag)
> +	    {
> +	      record_buf[0] = reg_rt + regnum_offset;
> +	      aarch64_insn_r->reg_rec_count = 1;
> +	      if (!vector_flag)
> +		{
> +		  /* The Rn register always has writeback in LRCPC3. This is
> +		     not the case in LRCPC.  */
> +		  record_buf[1] = reg_rn;
> +		  aarch64_insn_r->reg_rec_count = 2;
> +		}
> +	    }
> +	  else
> +	    {
> +	      regcache_raw_read_unsigned (aarch64_insn_r->regcache, reg_rn,
> +					  &address);
> +	      /* (vector_flag && insn_bit23) is the STLUR instruction with Q
> +		 register.  */
> +	      datasize = (vector_flag && insn_bit23) ? 128 : (8 << size_bits);
> +	      /* LRCPC3 adds STLR with a pre-indexed offset. There is another
> +		 STLR variant without offset but this has a different encoding.  */
> +	      if (!vector_flag)
> +		{
> +		  record_buf[0] = reg_rn;
> +		  aarch64_insn_r->reg_rec_count = 1;
> +		}
> +	      record_buf_mem[0] = datasize >> 3;
> +	      record_buf_mem[1] = address + offset;
> +	      aarch64_insn_r->mem_rec_count = 1;
> +	    }
> +	}
> +      else
> +	{
> +	  /* ldiapp/stilp.  */
> +	  uint8_t opc2 = bits (aarch64_insn_r->aarch64_insn, 12, 15);
> +	  reg_rt2 = bits (aarch64_insn_r->aarch64_insn, 16, 20);
> +	  if (ld_flag)
> +	    {
> +	      record_buf[0] = reg_rt;
> +	      record_buf[1] = reg_rt2;
> +	      aarch64_insn_r->reg_rec_count = 2;
> +
> +	      /* If the registers don't match and there's no offset then
> +	         there's WB.  */

The line above is indented with 8 spaces which can be substituted
by a \t. git am complained about it.

> +	      if (reg_rn != reg_rt && reg_rn != reg_rt2 && opc2 == 0)
> +		{
> +		  record_buf[2] = reg_rn;
> +		  aarch64_insn_r->reg_rec_count = 3;
> +		}
> +	    }
> +	  else
> +	    {
> +	      datasize = 8 << size_bits;
> +	      regcache_raw_read_unsigned (aarch64_insn_r->regcache, reg_rn,
> +					  &address);
> +	      offset = (opc2 == 0) ? (2 << size_bits) : 0;
> +	      address -= offset;
> +
> +	      record_buf_mem[0] = datasize >> 3;
> +	      record_buf_mem[1] = address;
> +	      record_buf_mem[2] = datasize >> 3;
> +	      record_buf_mem[3] = address + (datasize >> 3);
> +	      aarch64_insn_r->mem_rec_count = 2;
> +
> +	      if (offset != 0)
> +		{
> +		  record_buf[0] = reg_rn;
> +		  aarch64_insn_r->reg_rec_count = 1;
> +		}
> +	    }
> +	}
> +    }
>    /* Load/store register (register offset) instructions.  */
>    else if ((insn_bits24_27 & 0x0b) == 0x08 && insn_bits28_29 == 0x03
>  	   && insn_bits10_11 == 0x02 && insn_bit21)
> diff --git a/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.c b/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.c
> new file mode 100644
> index 00000000000..9265489ff70
> --- /dev/null
> +++ b/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.c
> @@ -0,0 +1,319 @@
> +/* This test program is part of GDB, the GNU debugger.
> +
> +   Copyright 2024-2026 Free Software Foundation, Inc.

The copyright should start with 2026.

> diff --git a/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.exp b/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.exp
> new file mode 100644
> index 00000000000..7d7a8c45445
> --- /dev/null
> +++ b/gdb/testsuite/gdb.reverse/aarch64-lrcpc3.exp
> @@ -0,0 +1,203 @@
> +# Copyright 2024-2026 Free Software Foundation, Inc.

Here too, the copyright should start with 2026.

> +
> +# 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 instruction record for AArch64 FEAT_LRCPC3 instructions.
> +# Based on gdb.reverse/aarch64-mops.exp
> +#
> +# The basic flow of the record tests are:
> +#    1) Stop before executing the instructions of interest.  Record
> +#       the initial value of the registers that the instruction will
> +#       change, i.e. the destination register.
> +#    2) Execute the instructions.  Record the new value of the
> +#       registers that changed.
> +#    3) Reverse the direction of the execution and execute back to
> +#       just before the instructions of interest.  Record the final
> +#       value of the registers of interest.
> +#    4) Check that the initial and new values of the registers are
> +#       different, i.e. the instruction changed the registers as expected.
> +#    5) Check that the initial and final values of the registers are
> +#       the same, i.e. GDB record restored the registers to their
> +#       original values.
> +
> +require allow_aarch64_lrcpc3_tests
> +
> +standard_testfile
> +
> +if {
> +    [prepare_for_testing "failed to prepare" ${testfile} ${srcfile} \
> +	[list debug additional_flags=-march=armv8-a+rcpc3]]
> +} {

The style we use in Tcl puts the { and } of the if condition in the same
line as the condition expression.

> +    return -1

Top-level returns from .exp files don't need to return anything anymore.

The current style is to have a bare "return".

Applies to this return ...

> +}
> +
> +if {![runto_main]} {
> +    return -1

... and this return.

gdb.reverse/aarch64-mops.exp is wrong in that regard. Sorry for the bad influence.

> +set ldiapp_cases {
> +	{ ldiapp-0 { x19 x20 } {     } { } { } { ptr } { src } }
> +	{ ldiapp-1 { w19 w20 } {     } { } { } { ptr } { src } }
> +	{ ldiapp-2 { x19 x20 } { ptr } { } { } {     } { src } }
> +	{ ldiapp-3 { w19 w20 } { ptr } { } { } {     } { src } }
> +	{ ldiapp-4 { x21 x20 } {     } { } { } {     } { src } }
> +	{ ldiapp-5 { w21 w20 } {     } { } { } {     } { src } }
> +}
> +
> +set stilp_cases {
> +	{ stilp-0  { } {     } { src } { x19 x20 } { ptr } { } }
> +	{ stilp-1  { } {     } { src } { w19 w20 } { ptr } { } }
> +	{ stilp-2  { } { ptr } { src } { x19 x20 } {     } { } }
> +	{ stilp-3  { } { ptr } { src } { x19 x20 } {     } { } }
> +	{ stilp-4  { } {     } { src } { x20     } { ptr } { } }
> +	{ stilp-5  { } {     } { src } { w20     } { ptr } { } }
> +}
> +
> +set ldapr_stlr_cases {
> +	{ ldapr-0  { x19 } { ptr } {     } {     } { } { src } }
> +	{ ldapr-1  { w19 } { ptr } {     } {     } { } { src } }
> +	{ stlr-0   {     } { ptr } { src } { x19 } { } {     } }
> +	{ stlr-1   {     } { ptr } { src } { w19 } { } {     } }
> +}
> +
> +set ldap1_stl1_cases {
> +	{ ldap1-0  { v22 } { } {     } { ptr } { } { src } }
> +	{ stl1-0   {     } { } { src } { v22 } { } {     } }
> +}
> +
> +set ldapur_stlur_cases {
> +	{ ldapur-0 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-0  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-1 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-1  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-2 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-2  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-3 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-3  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-4 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-4  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-5 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-5  {     } { } { src } { v22 } { ptr } {     } }
> +	{ ldapur-6 { v22 } { } {     } { ptr } {     } { src } }
> +	{ stlur-6  {     } { } { src } { v22 } { ptr } {     } }
> +}
> +
> +set all_cases [concat \
> +    $ldiapp_cases $stilp_cases $ldapr_stlr_cases \
> +    $ldap1_stl1_cases $ldapur_stlur_cases]
> +
> +foreach c $all_cases {
> +    lassign $c name diff_reg diff_var diff_mem same_reg same_var same_mem
> +    test_single_asm $name $diff_reg $diff_var $diff_mem $same_reg $same_var $same_mem
> +}

Nice way of structuring all the cases!

-- 
Thiago

  parent reply	other threads:[~2026-04-25  5:09 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-04-20 21:52 [PATCH 0/4] gdb/aarch64: record/replay support for AArch64 Ezra.Sitorus
2026-04-20 21:52 ` [PATCH 1/4] gdb/aarch64: record/replay support for LRCPC3 Ezra.Sitorus
2026-04-24  7:00   ` Luis
2026-04-24 15:09   ` Guinevere Larsen
2026-04-25  5:08   ` Thiago Jung Bauermann [this message]
2026-04-25  7:21   ` Thiago Jung Bauermann
2026-04-20 21:52 ` [PATCH 2/4] gdb/aarch64: Test record/replay support for CSSC Ezra.Sitorus
2026-04-24  7:07   ` Luis
2026-04-24 16:53   ` Guinevere Larsen
2026-04-25  6:10   ` Thiago Jung Bauermann
2026-04-25  7:28   ` Thiago Jung Bauermann
2026-04-20 21:52 ` [PATCH 3/4] gdb/aarch64: record/replay support for RPRFM and PRFM (reg) Ezra.Sitorus
2026-04-24  7:08   ` Luis
2026-04-24 16:56   ` Guinevere Larsen
2026-04-25  8:12   ` Thiago Jung Bauermann
2026-04-20 21:52 ` [PATCH 4/4] gdb/aarch64: record/replay support for LSE128 Ezra.Sitorus
2026-04-24  7:11   ` Luis
2026-04-24 17:12   ` Guinevere Larsen
2026-04-25  8:38   ` Thiago Jung Bauermann
2026-04-23 17:31 ` [PATCH 0/4] gdb/aarch64: record/replay support for AArch64 Guinevere Larsen
2026-04-23 22:20   ` Ezra Sitorus
2026-04-24  5:33   ` Thiago Jung Bauermann
2026-04-24  7:16 ` Luis

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=87o6j7bp9n.fsf@linaro.org \
    --to=thiago.bauermann@linaro.org \
    --cc=Ezra.Sitorus@arm.com \
    --cc=gdb-patches@sourceware.org \
    --cc=luis.machado.foss@gmail.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