[PATCH] sched: fix cond_resched_softirq() offset
authorIngo Molnar <mingo@elte.hu>
Sat, 30 Dec 2006 00:48:13 +0000 (16:48 -0800)
committerLinus Torvalds <torvalds@woody.osdl.org>
Sat, 30 Dec 2006 18:56:41 +0000 (10:56 -0800)
commit9414232fa0cc28e2f51b8c76d260f2748f7953fc
tree71717c2907f4fe2f50aba355e8cf03f3a2a8b385
parentd449db98d5d7d90f29f9f6e091b0e1d996184df1
[PATCH] sched: fix cond_resched_softirq() offset

Remove the __resched_legal() check: it is conceptually broken.  The biggest
problem it had is that it can mask buggy cond_resched() calls.  A
cond_resched() call is only legal if we are not in an atomic context, with
two narrow exceptions:

 - if the system is booting
 - a reacquire_kernel_lock() down() done while PREEMPT_ACTIVE is set

But __resched_legal() hid this and just silently returned whenever
these primitives were called from invalid contexts. (Same goes for
cond_resched_locked() and cond_resched_softirq()).

Furthermore, the __legal_resched(0) call was buggy in that it caused
unnecessarily long softirq latencies via cond_resched_softirq().  (which is
only called from softirq-off sections, hence the code did nothing.)

The fix is to resurrect the efficiency of the might_sleep checks and to
only allow the narrow exceptions.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/sched.c