On Thu, 2011-06-02 at 11:42 +0200, Christoph Hellwig wrote: > On Mon, May 30, 2011 at 11:07:33PM -0700, Nicholas A. Bellinger wrote: > > Also, I think we will still need to keep the CDB splitting logic for at > > least PSCSI and IBLOCK backends. > > only PSCSI really needs the splitting of the CDB, as no one elses uses > the CDB after it's been split. > > I guess what you mean is splitting commands into multiple se_tasks, > which in some form will still be needed, Yep, this is what I meant > but I suspect it's more efficient > to not create multiple data structures for the iblock case, but rather > create bios on the fly while parsing the S/G list, both for IBLOCK and PSCSI. > The main issue with something like this is that HW target mode still needs a sg_next() walkable chained SGL list built from N possible se_task->task_sg[] payloads from a single se_cmd. That said, I am not exactly sure how this type of change would look.. > If we can't completely get rid of the se_tasks at least only using them > for data I/O, and avoiding their creation for the internally handled metadata > CDBs seems like another nice optimization. Also sounds like a reasonable cleanup for v4.1. Thanks, --nab -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
