Messages in this thread |  | | Date | Fri, 30 Aug 2002 10:03:15 -0700 (PDT) | From | Linus Torvalds <> | Subject | Re: [patch] scheduler fixes, 2.5.32-BK |
| |
On Fri, 30 Aug 2002, Ingo Molnar wrote: > > actually, i think the race does not exist. up() is perfectly safely done > on the on-stack semaphore, because both the wake_up() done by __up() and > the __down() path takes the waitqueue spinlock, so i cannot see where > the up() touches the semaphore after the down()-ed task has been woken > up.
It touches the _spinlock_.
Which may no longer be a spinlock, since the waiter may never have blocked on it at all, and may have just succeeded directly with the atomic decrement of the counter.
Trust me, semaphores on the stack do not work unless there is some other synchronization above the semaphore level (or unless we make semaphores much slower, ie we take the spinlock even in the fast path).
Linus
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
|  |