On Wed, 7 Mar 2012, KOSAKI Motohiro wrote: > > It's a different topic, the proposal here is whether an error in > > mempolicies (either the code or flipped bit) should crash the kernel or > > not since it's a condition that can easily be recovered from and leave > > BUG() to errors that actually are fatal. Crashing the kernel offers no > > advantage. > > Should crash? The code path never reach. thus there is no ideal behavior. > In this case, BUG() is just unreachable annotation. So let's just annotate > unreachable() even though CONFIG_BUG=n. > > WARN_ON_ONCE makes code broat and no positive impact. > I think you misunderstand the difference between WARN() and BUG(). Both are intended to never be reached; the difference is that BUG() is a fatal condition and WARN() is not. All of the changes from BUG() to WARN() in this patch are not fatal and has no other side-effects other memory allocations that are not truly interleaved, for example. These should have been WARN() from the beginning. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>