Hi! > Ah, freeze(p) need not be under the lock. > > Besides, I'm now thinking the previous version is also correct, because even > if SIGCONT comes after we have forced SIGSTOP, it will remove the SIGSTOP > from the queue and the task's state will change to TASK_RUNNING, so the > next signal_wake_up() will do what it should. > > I prefer that one, because it's shorter and doesn't affect sched.h. ;-) Okay... consider your latest patches ACKed. (But it would be nice to make them wait in -mm to 2.6.21, I believe this needs testing). Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
