This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 00/11] Aarch64 linux GDB native multi-arch debugging (part 1)
- From: Pedro Alves <palves at redhat dot com>
- To: Yao Qi <qiyaoltc at gmail dot com>, gdb-patches at sourceware dot org
- Date: Wed, 01 Jul 2015 18:00:18 +0100
- Subject: Re: [PATCH 00/11] Aarch64 linux GDB native multi-arch debugging (part 1)
- Authentication-results: sourceware.org; auth=none
- References: <1435759111-22856-1-git-send-email-yao dot qi at linaro dot org>
On 07/01/2015 02:58 PM, Yao Qi wrote:
> This patch series is to enable Aarch64 GDB (64-bit) debug arm program
> (32-bit) rudimentarily, which means with these patches applied, Aarch64
> GDB is able to do some basic debugging to arm program . However, some
> features are still missing,
>
> - Debug multi-threaded program,
> - Use HW watchpoint and breakpoint,
>
> because current linux kernel ptrace doesn't have right requests to get
> the TLS area and to set HW watchpoint registers. I am still working
> with kernel hackers to get ptrace requests ready. Our plan is to be
> clear on aarch64 multi-arch debugging requirements to kernel first,
> get them upstream to kernel, and then, post the rest of aarch64 gdb
> multi-arch patches. Afterwards, I'll add aarch64 multi-arch support
> in gdbserver too.
I'd be fantastic if you managed to merge the gdb and gdbserver
arm/aarch64 files, like we did for x86.
>
> Patch #1 is to align arm-linux-nat.c to aarch64-linux-nat.c, it is a
> refactor patch. Patch #2 and #3 starts the basic multi-arch debugging
> support, in which some code is shared between arm and aarch64. Patch
> #4 is to adjust all arm-*.xml target descriptions to fix some fails
> I've seen. Patch #5 - #8 are various tweaks to gdb testsuite for
> aarch64 multi-arch debugging.
I skimmed these and they looked fine to me.
>
> Patch #9 - #11 can be regarded as an RFC, which teaches GDB to use
> hardware single step in software single step architecture if target
> supports hardware single step. They are useful to aarch64 multi-arch
> debugging, because hardware single step is available for arm32 process
> on aarch64.
These I think are problematic as is.
>
> The whole series is tested on aarch64-linux with both aarch64 gcc and
> arm gcc. It is also tested on arm-linux and x86_64-linux. Since we'll
> branch 7.10 soon, I don't like to have such big change at last moment
> of branching and releasing. I plan to push them (or part of them) in
> after 7.10 branch is created, if there are no objections.
>
Thanks,
Pedro Alves