Messages in this thread | | | Date | Mon, 9 Nov 2009 17:24:08 +0100 | From | Oleg Nesterov <> | Subject | Re: [PATCH] ptrace: copy_process() should disable stepping |
| |
On 11/08, Roland McGrath wrote: > > That is inappropriate use of arch details in generic code. It might > happen to be harmless fritter in practice on the arch's we have but it > is certainly not the correct way to go about things. You should just > call user_disable_single_step() unconditionally. Even on an arch with > no such machinery at all that should be defined safely as a no-op (see > linux/ptrace.h). If there is some reason not to do that, please > explain it.
Yes, we have arch_has_single_step.
I added test_tsk_thread_flag(TIF_SINGLESTEP) check for 2 reasons: to optimize out user_disable_single_step() in the likely case, and because I wasn't sure it is safe to call user_disable_single_step() unconditionally.
OK, will resend, thanks.
Oleg.
| |