Skip to contentSkip to conversation

2.6.12-rc2 in_atomic() picks up preempt_disable()

DISCUSSION Keith Owens 5 participants 5 messages Apr 7, 2005

lore
Share & tools
lore
#0 lore raw

2.6.12-rc2, with CONFIG_PREEMPT and CONFIG_PREEMPT_DEBUG. The in_atomic() macro thinks that preempt_disable() indicates an atomic region so calls to __might_sleep() result in a stack trace. preempt_count() returns 1, no soft or hard irqs are running and no spinlocks are held. It looks like there is no way to distinguish between the use of preempt_disable() in the lock functions (atomic) and preempt_disable() outside the lock functions (do nothing that might migrate me).

#1 lore raw
Quoted 3 lines · Keith Owens

2.6.12-rc2, with CONFIG_PREEMPT and CONFIG_PREEMPT_DEBUG. The in_atomic() macro thinks that preempt_disable() indicates an atomic region so calls to __might_sleep() result in a stack trace.

but you're not allowed to schedule when preempt is disabled!

Quoted 5 lines

preempt_count() returns 1, no soft or hard irqs are running and no spinlocks are held. It looks like there is no way to distinguish between the use of preempt_disable() in the lock functions (atomic) and preempt_disable() outside the lock functions (do nothing that might migrate me).

in what code are you seeing this?

#2 lore raw
Quoted 7 lines · Arjan van de Ven
Quoted 3 lines · Keith Owens

2.6.12-rc2, with CONFIG_PREEMPT and CONFIG_PREEMPT_DEBUG. The in_atomic() macro thinks that preempt_disable() indicates an atomic region so calls to __might_sleep() result in a stack trace.

but you're not allowed to schedule when preempt is disabled!

Could it be related to this:

http://marc.theaimsgroup.com/?l=linux-kernel&m=111277325629959&w=2

Signature · 2 lines
Romano Giannetti             -  Univ. Pontificia Comillas (Madrid, Spain)
Electronic Engineer - phone +34 915 422 800 ext 2416  fax +34 915 596 569
#3 lore raw
In reply to Keith Owens
Quoted 9 lines · Keith Owens

2.6.12-rc2, with CONFIG_PREEMPT and CONFIG_PREEMPT_DEBUG. The in_atomic() macro thinks that preempt_disable() indicates an atomic region so calls to __might_sleep() result in a stack trace. preempt_count() returns 1, no soft or hard irqs are running and no spinlocks are held. It looks like there is no way to distinguish between the use of preempt_disable() in the lock functions (atomic) and preempt_disable() outside the lock functions (do nothing that might migrate me).

Is this new behaviour?

It sounds correct to me:

	preempt_disable();
	do_something_which_might_sleep();
	preempt_enable();

Is buggy?

#4 lore raw
In reply to Keith Owens
Quoted 8 lines · Keith Owens

2.6.12-rc2, with CONFIG_PREEMPT and CONFIG_PREEMPT_DEBUG. The in_atomic() macro thinks that preempt_disable() indicates an atomic region so calls to __might_sleep() result in a stack trace. preempt_count() returns 1, no soft or hard irqs are running and no spinlocks are held. It looks like there is no way to distinguish between the use of preempt_disable() in the lock functions (atomic) and preempt_disable() outside the lock functions (do nothing that might migrate me).

preempt_disable() sections are just as much atomic as spinlocked regions. Like the name suggests it.

Ingo