On Sun, 2006-02-19 at 00:31 +0100, Jürgen E. Fischer wrote:
> - just run the bh if the interrupt is from the controller and if so
> ensure that it's only called once per interrupt.
This is the bit that looks like the actual fix. It basically uses
service as a flag to debounce the interrupt.
This:
> HOSTDATA(shpnt)->in_intr++;
>
> + if( HOSTDATA(shpnt)->service==0 ) {
> + DO_UNLOCK(flags);
> + return;
> + }
Looks a bit wrong. If that ever triggered, you'd exit your task handler
with in_intr raised, which would cause a panic the next time
is_complete() was called.
The driver still seems to have an awful lot of locking confusion between
the host lock and its own internal lock (stored in the host structure).
I think there are several races and other nasties that could be cleaned
up simply by moving to using the host lock everywhere.
James
-
: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
