On Tue, Dec 13 2005, Stefan Richter wrote: > James Bottomley wrote: > >On Fri, 2005-12-09 at 20:35 +0100, Stefan Richter wrote: > >>When preparing a request in scsi_lib or in a SCSI high-level driver, > >>always set a transfer direction of DMA_NONE if data length is zero, > >>even for alleged write requests. (Extended patch derived from Jens > >>Axboe's version.) > [...] > >The problem is that I already picked up Jens' patch in rc fixes, so this > >no-longer applies. However, given that the fix needed to be in four > >separate places, which looks rather bad, I propose the following > >consolidation instead. > > Jens' patch alone does not fix the kernel panic AFAICS. "eject /dev/sdX" > goes through sd_init_command. James, could you put your patch into > scsi-rc-fixes too? And what about -stable? I trust James already included his own updated version, which is basically fixing that bug but abstracting the rq_to_scsi_dir() stuff into a helper instead. That should fix it for all cases. -- Jens Axboe - : 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
