On Tue, Jun 10, 2008 at 05:53:08PM +0100, Catalin Marinas wrote:

On Tue, 2008-06-10 at 08:47 -0700, Paul E. McKenney wrote:

quoted

On Tue, Jun 10, 2008 at 03:51:25PM +0100, Catalin Marinas wrote:

quoted

I was thinking whether this condition can be removed and allow the
smp_call_function*() to be called with IRQs disabled. At a quick look,
it seems to be possible if the csd_flag_wait() function calls the IPI
handlers directly when the IRQs are disabled (see the patch below).

[...]

quoted

There were objections last month:  http://lkml.org/lkml/2008/5/3/167

Thanks, I missed this discussion.

quoted

The issue was that this permits some interrupts to arrive despite
interrupts being disabled.  There seemed to be less resistance to
doing this in the wait==1 case, however.

The "(wait == 1) && irqs_disabled()" case is what I would be interested
in. In the patch you proposed, this doesn't seem to be allowed (at least
from the use of WARN_ON). However, from your post in May:

quoted

5.	If you call smp_call_function() with irqs disabled, then you
	are guaranteed that no other CPU's smp_call_function() handler
	will be invoked while smp_call_function() is executing.

this would be possible but no one need this functionality yet.

Would one use-case (ARM SMP and DMA cache maintenance) be enough to
implement this or I should add it to the ARM-specific code?

How will you implement it? You have to be able to wait *somewhere*
(either before or after the smp_call_function call) with interrupts
enabled. It is not enough just to eg. use a spinlock around
smp_call_function, because other CPUs might also be trying to call
down the same path also with interrupts disabled, and they'll wait
forever on the spinlock.