On Tue, Apr 26 2022 at 18:42, Alexander Potapenko wrote:
Can you please use 'entry:' as prefix. Slapping kmsan in front of
everything does not really make sense.
> Replace instrumentation_begin() with instrumentation_begin_with_regs()
> to let KMSAN handle the non-instrumented code and unpoison pt_regs
> passed from the instrumented part.
That should be:
from the non-instrumented part
or
passed to the instrumented part
right?
> --- a/kernel/entry/common.c
> +++ b/kernel/entry/common.c
> @@ -23,7 +23,7 @@ static __always_inline void __enter_from_user_mode(struct pt_regs *regs)
> CT_WARN_ON(ct_state() != CONTEXT_USER);
> user_exit_irqoff();
>
> - instrumentation_begin();
> + instrumentation_begin_with_regs(regs);
I can see what you are trying to do, but this will end up doing the same
thing over and over. Let's just look at a syscall.
__visible noinstr void do_syscall_64(struct pt_regs *regs, int nr)
{
...
nr = syscall_enter_from_user_mode(regs, nr)
__enter_from_user_mode(regs)
.....
instrumentation_begin_with_regs(regs);
....
instrumentation_begin_with_regs(regs);
....
instrumentation_begin_with_regs(regs);
if (!do_syscall_x64(regs, nr) && !do_syscall_x32(regs, nr) && nr != -1) {
/* Invalid system call, but still a system call. */
regs->ax = __x64_sys_ni_syscall(regs);
}
instrumentation_end();
syscall_exit_to_user_mode(regs);
instrumentation_begin_with_regs(regs);
__syscall_exit_to_user_mode_work(regs);
instrumentation_end();
__exit_to_user_mode();
That means you memset state four times and unpoison regs four times. I'm
not sure whether that's desired.
instrumentation_begin()/end() are not really suitable IMO. They were
added to allow objtool to validate that nothing escapes into
instrumentable code unless annotated accordingly.
Thanks,
tglx
