On Thu, 2017-02-09 at 02:07 -0800, Nicholas A. Bellinger wrote: > I'm strongly opposed to adding another reference count and struct > completion to every se_cmd, and using a special version of > __target_put_sess_cmd() sprinkled all over the place. > > We've already got a se_cmd->cmd_kref, se_cmd->t_transport_stop_comp used > for se_cmd quiesce (CMD_T_STOP) during CMD_T_ABORT and > transport_generic_free_cmd() shutdown, and se_cmd->cmd_wait_set for > drivers doing active I/O shutdown via target_sess_cmd_list_set_waiting. > > Adding a new reference count, a third struct completion, and fast path > complete_all() for everyone (not just CMD_T_ABORT) just to get rid of > se_cmd->cmd_wait_set doesn't simplify anything. > > So that said, dropping this patch until you can figure out how to use > the single existing se_cmd->cmd_kref. It is very ugly that there are multiple direct calls of .release_cmd() in the target code. That is just not acceptable. The only way to address this is by introducing a second reference count in struct se_cmd. We need one reference count to keep track of the lifetime of struct se_cmd and a second reference count to keep track of the number of references held by the target driver. Regarding t_transport_stop_comp: my plan is to remove that completion later on. Bart.-- 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
