On 21/08/2024 00:34, Mark Wielaard wrote:
Hi Steve,
Hey Mark,
On Fri, Jun 14, 2024 at 03:47:14PM +0100, Steve Capper wrote:
This series enables Pointer Authentication (PAC) support in elfutils.
The first four patches were originally posted by German Gomez. I've
rebased to the latest elfutils and added an extra patch that was
required to debug core dumps from PAC enabled applications.
These patches were tested on Debian Testing and Fedora 40 running on an
Apple M1 MacBook Pro (the CFLAG -mbranch-protection=standard needs to be
supplied to the build).
Thanks, I found a setup to test this and it works. Nice.
Without this series applied, the following tests failed:
* run-backtrace-native.sh
* run-backtrace-dwarf.sh
* run-backtrace-native-core.sh
* run-deleted.sh
I am happy to chop/change bits as necessary.
I had some small comments on the first two patches. They look good,
just tiny nitpicks.
Thanks!
The last three introduce/depend on a new public function
dwfl_thread_state_aarch64_pauth. I rather not have such a public
architecture specific function.
Could we instead try to reuse dwfl_thread_state_registers? We could
say that negative regnums are special architecture specific settings?
There is already some precedent for that in the the
thread_state_registers_cb function given to
ebl_set_initial_registers_tid, which call with -1 to set the PC
value. We could use -2 to indicate it is an arch specific setting.
That is also slightly ugly. But given we can hide most of it in
architecture specific/private code better than special case public
architecture functions.
Yeah, I know what you mean. There's some #ifdef AARCH64 stuff too.
I have removed dwfl_thread_state_aarch64_pauth and have rolled the
functionality into dwfl_thread_state_registers. It looks a lot tidier
now thanks :-).
Just giving the revised series a few tests and will send out a V2 shortly.
Cheers,
--
Steve
