On Wed, Oct 13, 2021 at 01:05:26PM -0500, Mike Christie wrote: > On 10/13/21 12:51 PM, Mike Christie wrote: > Another alternative would be to have iscsi check if it was waiting on > data before sending a TMR to LIO core. If it is, then it would just cleanup > internally and complete the TMR with success/failed depending on the TMR or > drop it and let the initiator escalate. I was thinking about that, but there is some TAS code that requires the proper response from the aborted command. Also the ABORT TASK execution path kinda allows ABORT TASK to be aborted itself. I'm not sure if that's code has any meaning now, but if it does, then clearing WRITE_10 early may not be a valid action from the SCSI protocol point of view.
