Messages in this thread Patch in this message | | | Date | Thu, 28 May 2009 13:41:00 +0200 | From | Oleg Nesterov <> | Subject | Re: [RFC PATCH 11/12 v2] ptrace: mv task_struct->ptrace_message ptrace_ctx->message |
| |
On 05/28, Oleg Nesterov wrote: > > ptrace_event() can use current->ptrace_ctx safely. If mask == 0 and we do > not check task_ptrace() we are called from tracehook_report_clone_complete(), > in this case trace == T means current was traced when it called do_fork().
Perhaps it is better to optimize out "mask != 0" check in ptrace_event(). This special !mask case is only needed for _report_clone(), we can use mask == PT_PTRACED instead.
Oleg.
--- PTRACE/include/linux/ptrace.h~10_HOOK_COMPLETE 2009-05-28 11:19:33.000000000 +0200 +++ PTRACE/include/linux/ptrace.h 2009-05-28 11:49:29.000000000 +0200 @@ -157,7 +157,7 @@ int generic_ptrace_pokedata(struct task_ */ static inline int ptrace_event(int mask, int event, unsigned long message) { - if (mask && likely(!(task_ptrace(current) & mask))) + if (likely(!(task_ptrace(current) & mask))) return 0; current->ptrace_message = message; ptrace_notify((event << 8) | SIGTRAP); --- PTRACE/include/linux/tracehook.h~10_HOOK_COMPLETE 2009-05-28 08:41:11.000000000 +0200 +++ PTRACE/include/linux/tracehook.h 2009-05-28 11:50:14.000000000 +0200 @@ -344,7 +344,7 @@ static inline void tracehook_report_clon struct task_struct *child) { if (unlikely(trace)) - ptrace_event(0, trace, pid); + ptrace_event(PT_PTRACED, trace, pid); } /**
| |