On Mon, Feb 27, 2023 at 02:29:56PM -0800, Rick Edgecombe wrote: > From: Mike Rapoport <rppt@xxxxxxxxxxxxx> > > Userspace loaders may lock features before a CRIU restore operation has > the chance to set them to whatever state is required by the process > being restored. Allow a way for CRIU to unlock features. Add it as an > arch_prctl() like the other shadow stack operations, but restrict it being > called by the ptrace arch_pctl() interface. > > Tested-by: Pengfei Xu <pengfei.xu@xxxxxxxxx> > Tested-by: John Allen <john.allen@xxxxxxx> > Tested-by: Kees Cook <keescook@xxxxxxxxxxxx> > Acked-by: Mike Rapoport (IBM) <rppt@xxxxxxxxxx> That tag is kinda implicit here. Unless he doesn't ACK his own patch. :-P > Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx> > Signed-off-by: Mike Rapoport <rppt@xxxxxxxxxxxxx> > [Merged into recent API changes, added commit log and docs] > Signed-off-by: Rick Edgecombe <rick.p.edgecombe@xxxxxxxxx> ... > diff --git a/arch/x86/kernel/shstk.c b/arch/x86/kernel/shstk.c > index 2faf9b45ac72..3197ff824809 100644 > --- a/arch/x86/kernel/shstk.c > +++ b/arch/x86/kernel/shstk.c > @@ -451,9 +451,14 @@ long shstk_prctl(struct task_struct *task, int option, unsigned long features) > return 0; > } > > - /* Don't allow via ptrace */ > - if (task != current) > + /* Only allow via ptrace */ > + if (task != current) { Is that the only case? task != current means ptrace and there's no other way to do this from userspace? Isn't there some flag which says that task is ptraced? I think we should check that one too... -- Regards/Gruss, Boris. https://people.kernel.org/tglx/notes-about-netiquette