On Tue, 2017-05-30 at 10:54 -0700, Himanshu Madhani wrote: > The reason for hard coding LUN ID to 0 is that, from the FC > protocol perspective, ABTS does not have any knowledge of > LUN ID. So, there is no reason for qla2xxx driver to > manufacture the LUN ID. Sorry but I think this is completely wrong. Are you aware that target_submit_tmr() performs an exact match on the LUN specified as the fourth argument? This patch will cause ABTS requests to be ignored that apply to commands that have been submitted to another LUN than LUN 0. Please note that I had proposed a better approach three months ago on the target-devel mailing list and that I'm still waiting for someone from Cavium to review these patches: * [PATCH v6 09/33] target: Make it possible to specify I_T nexus for SCSI abort (http://www.spinics.net/lists/target-devel/msg14534.html). * [PATCH v6 10/33] tcm_qla2xxx: Let the target core look up the LUN of the aborted cmd (http://www.spinics.net/lists/target-devel/msg14563.html). Bart.
