Re: [PATCH 03/13] target: add workqueue based cmd submission
From: Mike Christie <michael.christie@oracle.com>
Date: 2021-02-10 00:15:55
Also in:
target-devel
On 2/9/21 12:43 PM, Mike Christie wrote:
On 2/9/21 9:48 AM, Bodo Stroesser wrote:quoted
On 09.02.21 13:38, Mike Christie wrote:quoted
+void target_queued_submit_work(struct work_struct *work) +{ + struct se_cmd_queue *sq = container_of(work, struct se_cmd_queue, work); + struct se_cmd *se_cmd, *next_cmd; + struct llist_node *cmd_list; + unsigned char *cdb; + + cmd_list = llist_del_all(&sq->cmd_list); + if (!cmd_list) + /* Previous call took what we were queued to submit */ + return; + + cmd_list = llist_reverse_order(cmd_list); + llist_for_each_entry_safe(se_cmd, next_cmd, cmd_list, se_cmd_list) { + cdb = se_cmd->se_tfo->get_cdb(se_cmd);If I got it right, get_cdb is a new, optional callback. So, should we check, whether it is set?I think a check is not really useful. The caller has to implement it if they call target_queue_cmd_submit. It would be a BUG() and so either way you would get a crash when you are developing a patch.
I'll actually add the BUG. It looks like it's preferred in the lio code.