James Bottomley <mailto:James.Bottomley@xxxxxxxxxxxx> sez: >On Tue, 2007-06-19 at 11:41 -0400, Salyzyn, Mark wrote: >>It was discovered that if we had a busy status return >>from the Adapter for the SCSI srb command to a physical >>component, that we returned DID_NO_CONNECT rather than >>what one would expect DID_BUS_BUSY. >Are you sure you want DID_BUS_BUSY? I'm just asking because >I'm not sure of the firmware ramifications. DID_BUS_BUSY >will turn the command around for an immediate retry. >If there's a firmware resource issue, you should return >something like DID_REQUEUE which will throttle the >queue and reissue this command when another one returns. Thanks for noting this. I believe that this is the behavior we want. This is related to a SCSI pass-through to the physical targets. I see no dummied-up returns of SRB_STATUS_BUSY from the top all the way down to the CHIM. This is a report from the physical bus or end device and thus does not represent an Adapter resource limit. Immediate re-queuing is indicated. I noticed this issue when we were talking internally about mitigating the sequential queuing of commands to SATA ATAPI devices at the CHIM level. A long command (erase CD for example) issued by an application was followed by a test unit ready with a short timeout issued by the Linux device class layer for media checking and we ended up timing out the test unit ready. I had suggested, in violation of the sat specification, to return the test unit ready with SRB_STATUS_BUSY when timed out prior to even making it to the physical transport while sitting in the sequential queue. Thus I noticed the shortcoming in the driver regarding the recent (> 2.6.10) addition of this return value. The CHIM folks balked at a spoofed SRB_STATUS_BUSY response. Even though the workaround was not accepted, this scenario would still be acceptable for immediate re-queuing. Sincerely -- Mark Salyzyn - To unsubscribe from this list: 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
