On Tue, Feb 20, 2018 at 3:12 PM, Peter Lieven <[email protected]> wrote: > Am 15.02.2018 um 18:27 schrieb Stefan Hajnoczi: >> >> On Thu, Feb 15, 2018 at 03:24:54PM +0100, Peter Lieven wrote: >>> >>> Am 15.02.2018 um 12:15 schrieb Stefan Hajnoczi: >>>> >>>> The libiscsi iscsi_task_mgmt_async() API documentation says: >>>> >>>> abort_task will also cancel the scsi task. The callback for the scsi >>>> task will be invoked with SCSI_STATUS_CANCELLED >>>> >>>> The libiscsi implementation does not fulfil this promise. The task's >>>> callback is not invoked and its struct iscsi_pdu remains in the internal >>>> list (effectively leaked). >>> >>> If that contract is fixed in libiscsi, will the Qemu iSCSI driver still >>> work? >> >> In >> >> + /* If the command callback hasn't been called yet, drop the task */ >> + if (!acb->bh) { >> >> and >> >> + if (status == SCSI_STATUS_CANCELLED) { >> + if (!acb->bh) { >> >> we're mindful of the fact that the callback may have been invoked by >> libiscsi already. There is no risk of double-completion. > > > Hi Stefan, > > thanks for the clarification. I am fine with this change. I will check with > Ronnie for the > libiscsi fix.
Great, then this patch can go via Paolo's SCSI tree. Stefan
