Re: [linux-usb-devel] question on irqs and memory ordering

[Alan Stern]

On Wed, 14 Mar 2007, Oliver Neukum wrote:

> Hi,
> 
> I am looking at this code:
> 
> 	if (new_msr &
> 	    (MOS_MSR_DELTA_CTS | MOS_MSR_DELTA_DSR | MOS_MSR_DELTA_RI |
> 	     MOS_MSR_DELTA_CD)) {
> 		icount = &mos7840_port->icount;
> 
> 		/* update input line counters */
> 		if (new_msr & MOS_MSR_DELTA_CTS) {
> 			icount->cts++;
> 		}
> 		if (new_msr & MOS_MSR_DELTA_DSR) {
> 			icount->dsr++;
> 		}
> 		if (new_msr & MOS_MSR_DELTA_CD) {
> 			icount->dcd++;
> 		}
> 		if (new_msr & MOS_MSR_DELTA_RI) {
> 			icount->rng++;
> 		}
> 	}
> 
> It is called in interrupt and uses no locking. What happens if the next
> irq is processed on another cpu? Is that cpu guaranteed to see the updates
> to the incremented variables?

I _think_ it's okay.  Taking an interrupt and returning from an interrupt
ought to have explicit or implicit memory barriers.

It wouldn't hurt to ask someone else.  David Howells wrote
Documentation/memory-barriers.txt so maybe he knows.  (The document itself
doesn't say one way or the other.)

Alan Stern

-
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/