Re: Deferred handling of writes to /proc/irq/xx/smp_affinity
From: Zwane Mwaikambo
Date: Thu Apr 28 2005 - 01:46:52 EST
On Wed, 27 Apr 2005, Ashok Raj wrote:
> Hi Andrew and all
> It is safe to re-program rte entries in ioapic only when an intr is pending.
> Existing code does this incorrectly by reprogamming rte entries immediatly
> when a value is written to /proc/irq/xx/smp_affinity. IRQ_BALANCE code in
> kernel does this right, but /proc/irq needs to be handled the same way so that
> user mode irq_balancer wont lock up systems or loose interrupts in the race.
> This is already fixed in ia64, introduced for i386 and x86_64.
> since this touches 3 arch's managing in -mm for trial would be best to make
> sure nothing is broken, before considering for main line.
Ashok, thanks for doing this, i only have minor nitpicks, can we avoid
the wrapper for irq_desc and just index it directly? And could we use
static inline functions for MOVE_IRQ and SET_IRQ_INFO macros and use nop
functions where required.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/