On 04/04/2012 03:20 PM, Nicholas A. Bellinger wrote: > Mmmm, due to the requirements of Immediate Data and Unsolicited Data Out > for WRITE payloads, I'm not sure if a conversion of iscsi-target to > target_submit_cmd() in it's current form with a single call is going to > work as you expect here. > > The reason we have historically needed the logical split of what is now > in target_submit_cmd() logic for iscsi-target is to decode the CDB and > perform memory allocation ahead of actually dispatching for backend > device execution of command in iSCSI per session CmdSN order.. (eg: the > command submission may not always happen on the same RX context it was > received upon) > > The majority of these cases where the RX context execution may be > delayed are related to multiple connection (MC/S) operation, but > Immediate Data and Unsolicited Data Out does also have special > considerations (see iscsi_target_erl1.c:iscsit_execute_cmd) that I can > see having issues with a target_submit_cmd() conversion. > > I've considered an iscsi-target conversion to target_submit_cmd() in the > past months, but AFAICT this would seem to cause more problems that it's > worth for all of the iscsi-target specific cases. Of course, I'm happy > to be proved wrong on this, but I'd likely not merge a conversion patch > until these immediate data + MC/S special cases can addressed. Naturally -- those are part of LIO's feature set and refactoring needs to take them into account. In fact I think much of my desire to hack iscsit is to help the code flow as gracefully as possible even though so many features are present. (Also btw, thanks for reviewing all these patches) -- Andy -- 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
