On Thu, Jun 10, 2021 at 1:58 PM Eric W. Biederman <ebiederm@xxxxxxxxxxxx> wrote:
The problem is sometimes we read all of the registers from
a context where they are not all saved.
Ouch. Yes. And this is really painful because none of the *normal*
architectures do this, so it gets absolutely no coverage.
I think at this point we need to say that the architectures that have a
do this need to be fixed to at least call do_exit and the kernel
function in create_io_thread with the deeper stack.
Yeah. We traditionally have that requirement for fork() and friends
too (vfork/clone), so adding exit and io_uring to do so seems like the
most straightforward thing.
But I really wish we had some way to test and trigger this so that we
wouldn't get caught on this before. Something in task_pt_regs() that
catches "this doesn't actually work" and does a WARN_ON_ONCE() on the
affected architectures?
Linus
