Re: [PATCH v2 0/5] Add support for AArch64 MOPS instructions

[Luis Machado <luis.machado@arm.com>]

Hi Thiago,

Thanks for the series!

I just wanted to clarify the approach being used for this feature. From what I noticed,
we are implementing motion through these sequences of MOPS instructions as atomic,
right? Similar to the atomic sequences.

Is there a reason for that? From what I gathered, the Linux Kernel implements the
ability to single-step through each of the MOPS instructions. The catch is that
if the M and E instructions get interrupted, the sequence will get restarted, and
the debugger will need to be aware of that.

Was the atomic block approach for MOPS instructions used as a simplification?

I suppose displaced-stepping these sequences would be a bit tricky, but doable if
they were relocated as a block. I'm wondering what the performance impact would be
of requiring serialization of execution in a non-stop debugging scenario with
multiple threads.

One last point, doesn't record/replay require single-stepping each instruction
individually? Does that go against the atomic block approach?

I'll go stare at the code.

Regards,
Luis

On 5/7/24 03:22, Thiago Jung Bauermann wrote:
> Hello,
> 
> I'm sending v2 because Christophe made a suggestion for the
> gdb.arch/aarch64-mops-atomic-inst.exp testcase, so patch 4 incoroporates
> it.
> 
> The other patches are unchanged from v1.
> 
> Here is the original cover letter for convenience:
> 
> This patch series implements GDB support for the new instructions in
> AArch64's MOPS feature.  Patch 1 has a small overview.
> 
> What is needed from GDB is recognizing the MOPS sequences of instructions
> as atomic so that they can be stepped over during instruction single
> stepping, and also to avoid doing displaced stepping with them.  This is
> done in patch 1.
> 
> Patch 2 adds support for the new instructions to the record an replay
> target.
> 
> The other patches add testcases to test each of the aspects above, plus
> one testcase to verify the interaction of the MOPS instructions with
> watchpoints.
> 
> Tested on Ubuntu 23.10 aarch64-linux-gnu with no regressions, using the
> Arm FVP emulator as well as QEMU v8.2.
> 
> Thiago Jung Bauermann (5):
>   gdb/aarch64: Implement software single stepping for MOPS instructions
>   gdb/aarch64: Add record support for MOPS instructions.
>   gdb/testsuite: Add gdb.arch/aarch64-mops-watchpoint.exp
>   gdb/testsuite: Add gdb.arch/aarch64-mops-atomic-inst.exp
>   gdb/testsuite: Add gdb.reverse/aarch64-mops.exp
> 
>  gdb/aarch64-tdep.c                            | 191 +++++++++++++++++-
>  .../gdb.arch/aarch64-mops-atomic-inst.c       |  69 +++++++
>  .../gdb.arch/aarch64-mops-atomic-inst.exp     |  94 +++++++++
>  .../gdb.arch/aarch64-mops-watchpoint.c        |  66 ++++++
>  .../gdb.arch/aarch64-mops-watchpoint.exp      |  79 ++++++++
>  gdb/testsuite/gdb.reverse/aarch64-mops.c      |  71 +++++++
>  gdb/testsuite/gdb.reverse/aarch64-mops.exp    | 171 ++++++++++++++++
>  gdb/testsuite/lib/gdb.exp                     |  61 ++++++
>  8 files changed, 800 insertions(+), 2 deletions(-)
>  create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-atomic-inst.c
>  create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-atomic-inst.exp
>  create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-watchpoint.c
>  create mode 100644 gdb/testsuite/gdb.arch/aarch64-mops-watchpoint.exp
>  create mode 100644 gdb/testsuite/gdb.reverse/aarch64-mops.c
>  create mode 100644 gdb/testsuite/gdb.reverse/aarch64-mops.exp
> 
> 
> base-commit: 84a069db6714ddcf444095ed09dbcd7404834694