Hi Andrew, On Sat, Feb 18, 2006 at 14:14:23 -0800, Andrew Morton wrote: > > Fix below. > Thanks. > Could we please have a more complete description of this change? It's > obviosly doing more than fixing a used-uninitialised bug.. After fixing the problem in question, I discoved that the driver still didn't work with the pcmcia card I picked up. run/is_complete was sometimes not called at all or called multiple times on one interrupt. I'm not really sure why (or why that worked before). It's now ensured that run() is only added to the task queue once per interrupt (by protecting 'service' with spinlocks). Additionally interrupt sharing didn't seem to work as the interrupt handler didn't check if a interrupt condition was at all pending. In short: - change interface of the reset functions from Scsi_Cmnd to Scsi_Host. - add functions with the original interface and rename the new functions to reflect the new interface. - call these from the pcmcia driver, thereby avoiding the need to construct a (broken) Scsi_Cmnd from a Scsi_Host. - just run the bh if the interrupt is from the controller and if so ensure that it's only called once per interrupt. Jürgen - : 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
