On Tue 25-03-14 10:56:34, Michal Hocko wrote: > On Tue 25-03-14 12:06:36, Christoph Lameter wrote: > > On Tue, 25 Mar 2014, Michal Hocko wrote: > > > > > You are right. The function even does VM_BUG_ON(!irqs_disabled())... > > > Unfortunatelly we do not seem to have an _irq alternative of the bit > > > spinlock. > > > Not sure what to do about it. Christoph? > > > > > > Btw. it seems to go way back to 3.1 (1d07171c5e58e). > > > > Well there is a preempt_enable() (bit_spin_lock) and a preempt_disable() > > bit_spin_unlock() within a piece of code where irqs are disabled. > > > > Is that a problem? Has been there for a long time. > > It is because preempt_enable calls __preempt_schedule when the preempt > count drops down to 0. You would need to call preempt_disable before you > disable interrupts or use an irq safe bit spin unlock which doesn't > enabled preemption unconditionally. Hmm, now that I am looking into the code more closely it seems that preempt_schedule bails out when interrupts are disabled. -- Michal Hocko SUSE Labs -- 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/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>