Excerpts from Florian Weimer's message of April 22, 2020 5:15 pm: > * Nicholas Piggin: > >> Another option would be to use a different signal. I don't see that any >> are more suitable. > > SIGSYS comes to my mind. But I don't know how exclusively it is > associated with seccomp these days.
SIGSYS is entirely seccomp now. There looks like a single obscure MIPS user of it in Linux that's not seccomp, but it would be entirely new for powerpc (or any of the common platforms, arm, x86 etc). So I would be disinclined to use SIGSYS unless there are no other better signal types, and we don't want to use SIGILL for some good reason -- is there a good reason to add complexity for userspace by differentiating these two situations? Thanks, Nick