[PATCH v6 6/8] riscv: vector: adjust ptrace and signal behavior for INITIAL state
Paul Walmsley
pjw at kernel.org
Tue Nov 3 19:46:49 PST 2026
Hi Andy,
On Fri, 18 Sep 2026, Andy Chiu wrote:
> The last patch introduced the INITIAL vector state to avoid saving and
> restoring vector registers across syscall boundaries. However, this
> optimization did not fully account for the ptrace and signal handling
> interfaces.
>
> As a result, two issues emerged:
> 1. Ptrace reads at syscall stop could observe stale, non-nulled
> registers.
> 2. Modifications to the ucontext through signal interface during a
> syscall stop would be overwritten by the vector discaring macro.
If patch 5 introduces bugs that then need to be fixed by this patch 6,
then either the two patches should be combined, or patch 5 needs to be
rewritten so that it doesn't introduce bugs. That way, bisectability is
preserved without introducing unexpected regressions. Care to take a look
and figure out what the best approach is in this case?
thanks,
- Paul
More information about the linux-riscv
mailing list