Re: [PATCH 0/9] Add new command to print the shadow stack backtrace

[Thiago Jung Bauermann <thiago.bauermann@linaro.org>]

Hello Christina,

Sorry for the delay. I'm going over them now and will have some review
comments soon.

"Schimpe, Christina" <christina.schimpe@intel.com> writes:

> I did not test the "bt shadow" command for ARM's GCS. Does the command make sense for GCS,
> too ?

Yes, it does. I think this is useful, thanks!

> If yes, does the implementation I shared here makes sense to you in general?

I'm still in the middle of reviewing it, but so far I think so.

> I tried to address the different Aarch and Intel behavior we discussed here
> https://sourceware.org/pipermail/gdb-patches/2025-June/218947.html 
> by introducing a new gdbarch hook top_addr_empty_shadow_stack.
>
> Does that go into the right direction for your purposes?

Yes, I believe so. Thank you for addressing this difference between the
architectures.

I tested your patches and there's a bug unfortunately:

FAIL: gdb.arch/aarch64-gcs-disp-step.exp: continue to breakpoint at blr
FAIL: gdb.arch/aarch64-gcs-disp-step.exp: continue until exit
FAIL: gdb.arch/aarch64-gcs.exp: call inferior function
FAIL: gdb.arch/aarch64-gcs-return.exp: test inferior call and continue: call (int) call2()
FAIL: gdb.arch/aarch64-gcs-return.exp: test inferior call and continue: continue until exit
FAIL: gdb.arch/aarch64-gcs-return.exp: test frame level update: check gcspr of frame 1
FAIL: gdb.arch/aarch64-gcs-return.exp: test frame level update: check gcspr of frame 2
FAIL: gdb.arch/aarch64-gcs-return.exp: test return from current frame: continue until exit

In each FAIL above a GCS error happens unexpectedly in the inferior. It
appears that GDB isn't updating the shadow stack correctly. I still have
to see what exactly is the problem.

-- 
Thiago