On 1/26/22 03:09, Ard Biesheuvel wrote:
On Wed, 26 Jan 2022 at 11:40, Dan Li <ashim...@linux.alibaba.com> wrote:
Thanks, Ard,
On 1/26/22 00:10, Ard Biesheuvel wrote:
On Wed, 26 Jan 2022 at 08:53, Dan Li <ashim...@linux.alibaba.com> wrote:
Hi, all,
Sorry for bothering.
I'm trying to commit aarch64 scs code to the gcc and there is an issue
that I'm not sure about, could someone give me some suggestions?
(To avoid noise, I did't cc PING^3 [1] to the kernel mail list :) )
When clang enables scs, the following instructions are usually generated:
str x30, [x18], 8
ldp x29, x30, [sp], 16
......
ldp x29, x30, [sp], 16 ## x30 pop
ldr x30, [x18, -8]! ## x30 pop again
ret
The x30 register is popped twice here, Richard suggested that we can
omit the first x30 pop here.
AFAICT, it seems fine and also safe for SCS. But I'm not sure if I'm
missing something with the kernel, could someone give some suggestions?
The previous discussion can be found here [1].
[1] https://gcc.gnu.org/pipermail/gcc-patches/2022-January/589257.html
As was pointed out in the discussion, binary patching is in fact a
concern for the Linux kernel. E.g., Android uses generic binary
builds, but we would like to be able to switch between pointer
authentication and shadow call stack at boot time, rather than always
support both, and take the SCS performance hit on systems that
implement PAC as well.
However, it seems more straight-forward to patch PACIASP and AUTIASP
instructions into SCS push/pop instructions rather than the other way
around, as we can force the use of these exact opcodes [in the NOP
space]), as well as rely on existing unwind annotations to locate any
such instruction in the binary.
Well, then I think I don't need to submit a kernel patch to
enable SCS for gcc :)
Not entirely.
BTW:
Do we have a plan to submit patches of dynamic patch PAC into
the kernel recently?
At the moment, there are just some ideas floating around. I did
implement a proof of concept that uses unwind data, but it hit some
issues with cfi_negate_ra_state being emitted imprecisely (GCC) or not
at all (Clang) in some cases. Using objtool would be another
possibility.
So in summary, getting SCS support proper into GCC is definitely worth
the effort.
Got it!
And thanks for the suggestion, Ard :)
