It looks to me like there's an extra se_cmd kref "put"
(target_put_sess_cmd()) after transport_lookup_cmd_lun() fails and
sends check condition via the fabric driver. With the extra call
to target_put_sess_cmd(), the given se_cmd is "released" prior
to my fabric driver completing the check condition initiated from
the prev line. When I make the following change, the se_cmd is
not released until the fabric driver completes the check
condition and calls transport_generic_free_cmd() -- which is
what I expect.
--- a/drivers/target/target_core_transport.c 2014-02-13 14:00:14.000000000 -0800
+++ b/drivers/target/target_core_transport.c 2014-03-17 15:50:50.235119894 -0700
@@ -1336,7 +1336,6 @@ int target_submit_cmd_map_sgls(struct se
rc = transport_lookup_cmd_lun(se_cmd, unpacked_lun);
if (rc) {
transport_send_check_condition_and_sense(se_cmd, rc, 0);
- target_put_sess_cmd(se_sess, se_cmd);
return 0;
}
Having said that, it looks like there was a commit that, among other
changes, specifically added the call to target_put_sess_cmd(), so I'm
wondering if my understanding is correct. Here is the commit that adds
the call to target_put_sess_cmd(), with the relevant comment snippet:
commit 735703cac08f34a197e1c6331ca9adc8a52d6046
Author: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
Date: Fri Jan 20 19:02:56 2012 -0800
target: Fix target_submit_cmd() exception handling
<SNIP>
It also makes transport_lookup_cmd_lun() failures in
target_submit_cmd() use transport_send_check_condition_and_sense() and
target_put_sess_cmd() directly to avoid se_cmd->se_dev reference in
transport_generic_request_failure() handling.
<SNIP>
Thanks,
Alex
--
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
