On Tue 24-11-09 14:28:35, Alan Stern wrote: > On Tue, 24 Nov 2009, Jan Kara wrote: > > > My naive guess would be that those non-retried reads were actually > > readahead. That's not retried if I remember correctly. Later, when we > > really needed the data, we sent another read request... > > That would be my guess too. I don't know how to verify it, though. > > If you're interested in pursuing this farther, I can show you how to > generate equivalent errors on demand using an emulated USB drive. > At this point it's not clear how much more one could learn by doing > this, however. After digging in block layer code, it's as we suspected: In case of host error DID_ERROR (which is our case), scsi request is retried iff it is not a FAILFAST request which is set if bio is doing readahead... So this is explained and everything behaves as it should. Thanks everybody involved :). Honza -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
