Re: [PATCH 9/9] gdb, mi: Add -shadow-stack-list-frames command

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

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

>> -----Original Message-----
>> From: Thiago Jung Bauermann <thiago.bauermann@linaro.org>
>> Sent: Mittwoch, 26. November 2025 05:27
>> To: Schimpe, Christina <christina.schimpe@intel.com>
>> Cc: gdb-patches@sourceware.org
>> Subject: Re: [PATCH 9/9] gdb, mi: Add -shadow-stack-list-frames command
>> 
>> Christina Schimpe <christina.schimpe@intel.com> writes:
>> 
>> > diff --git a/gdb/testsuite/gdb.mi/mi-shadow-stack-signal.exp
>> > b/gdb/testsuite/gdb.mi/mi-shadow-stack-signal.exp
>> > new file mode 100644
>> > index 00000000000..219f1b8ad78
>> > --- /dev/null
>> > +++ b/gdb/testsuite/gdb.mi/mi-shadow-stack-signal.exp
>> 
>> My only suggestion for this testcase is to add "amd64" to the filename, since
>> it can only run on x86. Perhaps mi-amd64-shadow-stack-signal.exp?
>> 
>> > diff --git a/gdb/testsuite/gdb.mi/mi-shadow-stack.exp
>> > b/gdb/testsuite/gdb.mi/mi-shadow-stack.exp
>> > new file mode 100644
>> > index 00000000000..bbbc7ec5460
>> > --- /dev/null
>> > +++ b/gdb/testsuite/gdb.mi/mi-shadow-stack.exp
>> 
>> Samme suggestion here, about adding "amd64" to the filename.
>
> I applied all of it except the filename, since for mi I don't see any file
> starting with amd64-* in the testsuite folder gdb.mi/*.
>
> What happens if you run this for GCS ? Could we create a source file
> which is architecture independent?

I tried that but didn't get very far. There is the tunable
"glibc.cpu.aarch64_gcs" which can be used to force GCS to be enabled,
but then GDB gets a SIGSEGV at startup if it's not itself built to use
GCS. I need to make the inferior use the prctl syscall to enable shadow
stack explicitly.

I assume this doesn't happen in your case because for Intel you need
both the tunable and the compiler flag "-fcf-protection=return" to
activate shadow stacks?

Speaking of tunables, now that I looked closer into the testcase I was
surprised that you need to explicitly enable the SHSTK bit in
glibc.cpu.hwcaps. 

On AArch64, if the CPU and the kernel both support GCS, then the GCS
hwcap will be automatically enabled. This doesn't mean that GCS will be
used by applications though.

Anyway, nothing to worry about. I'm just noting the difference.

My suggestion is to leave the testcase as you currently have it, then
I can adapt it later for AArch64 with a separate patch.

-- 
Thiago