On Mon, 2012-10-01 at 10:49 -0400, Christoph Hellwig wrote: > There still are various fabric drivers using old command submission > sequences. > > target_setup_cmd_from_cdb and transport_handle_cdb_direct are still used by: > > - srpt - IIRC Nick mentioned he had patches for this one, which need to > be resurrected Yes, this still needs to be extracted from lio-core-2.6.git, so it will most likely be an v3.8 item at this point. > - tcm_loop - seems like the S/G list reuse here is the problem, we'd > just need an additional argument to target_submit_cmd to make > this happen > - tcm_vhost - same as tcm_loop, but it's really sad that all the old > copy & paste submission path still gets added to a new driver > OK, just sent out patches to convert tcm_loop + tcm_vhost to the generic I/O submission path. Please have a look. > The iscsi target not only uses those two above but also has a really > odd call straight though to transport_generic_new_cmd. In fact the > whole command submission in the iscsi target is very non-standard > copared to the other drivers, and I can't find an obvious excuse > why. For iscsi-target this ends up being overly complicated with the multiple combinations of immediate data + unsolicited data-out in the existing code paths. Andy mentioned doing this once before and I thought it would end up adding too much complexity to target_submit_cmd() for the benefit, but am still happy to be proved wrong here. ;) --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
