Hi Roland & Co,
My apologies for the delayed response here, still catching up after the
holidays..
On Mon, 2012-01-02 at 13:38 -0800, Roland Dreier wrote:
> On Fri, Dec 23, 2011 at 1:51 PM, Nicholas A. Bellinger
> <nab@xxxxxxxxxxxxxxx> wrote:
> > So the (t_transport_complete == 0) check causing the issue above should
> > be safe to remove now.. The same is true for the !cmd->se_cmd.se_dev
> > check in tcm_qla2xxx_free_cmd() as well.
>
> Basically you're saying the following is the right fix?
>
> --- a/drivers/scsi/qla2xxx/tcm_qla2xxx.c
> +++ b/drivers/scsi/qla2xxx/tcm_qla2xxx.c
> @@ -401,21 +401,6 @@ static void tcm_qla2xxx_complete_free(struct
> work_struct *work)
> */
> void tcm_qla2xxx_free_cmd(struct qla_tgt_cmd *cmd)
> {
> - barrier();
> - /*
> - * Handle tcm_qla2xxx_init_cmd() -> transport_get_lun_for_cmd()
> - * failure case where cmd->se_cmd.se_dev was not assigned, and
> - * a call to transport_generic_free_cmd_intr() is not possible..
> - */
> - if (!cmd->se_cmd.se_dev) {
> - target_put_sess_cmd(cmd->se_cmd.se_sess, &cmd->se_cmd);
> - transport_generic_free_cmd(&cmd->se_cmd, 0);
> - return;
> - }
> -
> - if (!atomic_read(&cmd->se_cmd.t_transport_complete))
> - target_put_sess_cmd(cmd->se_cmd.se_sess, &cmd->se_cmd);
> -
> INIT_WORK(&cmd->work, tcm_qla2xxx_complete_free);
> queue_work(tcm_qla2xxx_free_wq, &cmd->work);
> }
>
> (I'm deleting the barrier() as part of this too because it's almost certainly
> wrong... an indirect call through a function pointer that gets here is just
> as much of a compiler optimization barrier anyway).
>
Yes, this should be fine. I still need to verify this internally, and
will push this into lio-core/qla_tgt-3.3 after testing.
Thanks,
--nab
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
