Sergei Shtylyov wrote, on 03/02/2011 11:09 PM:

I wrote:

quoted

quoted

Request the handler irq such that there is no nesting for calls.
the notifiers are not expected to be nested, further the interrupt
events for status change should be handled prior to the next event
else there is a risk of loosing events.

quoted

quoted

Signed-off-by: Nishanth Menon<nm@ti.com>
---
arch/arm/mach-omap2/smartreflex.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)

quoted

quoted

diff --git a/arch/arm/mach-omap2/smartreflex.c

b/arch/arm/mach-omap2/smartreflex.c
index 99e4c4f..a4e9f2d 100644

--- a/arch/arm/mach-omap2/smartreflex.c
+++ b/arch/arm/mach-omap2/smartreflex.c

@@ -278,7 +278,7 @@ static int sr_late_init(struct omap_sr *sr_info)

goto error;
}
ret = request_irq(sr_info->irq, sr_interrupt,
- 0, name, (void *)sr_info);
+ IRQF_DISABLED, name, (void *)sr_info);

quoted

Isn't this flag a nop now?

Indeed, it is, according to the comment in <linux/interrupt.h>. So what
does this patch really achieve?

Carry over since I had implemented this driver originally on an ancient 
kernel and other than the showing off the obvious fact that I am not 
reading lwn.net enough, I dont think it achieved anything much :(

The original intent was as following: since bootloaders may have used 
the same h/w logic which could generate pending interrupt status, I 
wanted to explicitly flag that requesting the irq should also ensure the 
IRQ is in disabled status before I get a chance to clean things up.


-- 
Regards,
Nishanth Menon