This message is from the T13 list server.
Jens, It is real simple invalidate modes and make them exclusive. Better yet wait for host side TCQ and force the HBA to deal with drive side TCQ and then stomp on which side of the cable that fails. By waiting for host side dma-ring TCQ (similar to network cards) it is safer for the host (driver). Cheers, Andre Hedrick LAD Storage Consulting Group On Thu, 10 Jul 2003, Jens Axboe wrote: > This message is from the T13 list server. > > > On Wed, Jul 09 2003, Tony Goodfellow wrote: > > This message is from the T13 list server. > > > > > > There are several differences that I would like to see but this is the way > > it has been specified. As far as I know the current devices do this. If our > > driver receives a request that needs a non-queued command it waits until the > > current queue is empty before issuing the command, it has to comply with the > > standard. > > This is why FUA was added to the queued command list. > > Linux does the same thing, for the same reason. It worries me more than > just a little bit that the people here discussing flush cache usage is > not even aware of how that affects tcq as worded in the standard... > > -- > Jens Axboe >
