[v5,19/22] target: Move target_remove_from_state_list() into target_release_cmd_kref()

Commit Message

Bart Van Assche Feb. 10, 2017, 1:28 a.m. UTC

The only code that iterates over the state list is the LUN reset
code. Deferring removal from the state list is fine since that
code grabs a reference to the command before it does any other
processing. Leaving out the se_cmd.state_active test is fine
because this test is already performed by
target_remove_from_state_list().

Signed-off-by: Bart Van Assche <[e-mail]>
Reviewed-by: Hannes Reinecke <[e-mail]>
Cc: Christoph Hellwig <[e-mail]>
Cc: Andy Grover <[e-mail]>
Cc: David Disseldorp <[e-mail]>
---
 drivers/target/target_core_transport.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

Patch

diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
index 9deab861529c..af72f6b40a85 100644
--- a/drivers/target/target_core_transport.c
+++ b/drivers/target/target_core_transport.c
@@ -604,8 +604,6 @@  static int transport_cmd_check_stop_to_fabric(struct se_cmd *cmd)
 {
 	unsigned long flags;
 
-	target_remove_from_state_list(cmd);
-
 	spin_lock_irqsave(&cmd->t_state_lock, flags);
 	/*
 	 * Determine if frontend context caller is requesting the stopping of
@@ -2486,13 +2484,6 @@  int transport_generic_free_cmd(struct se_cmd *cmd, int wait_for_tasks)
 	} else {
 		if (wait_for_tasks)
 			target_wait_free_cmd(cmd);
-		/*
-		 * Handle WRITE failure case where transport_generic_new_cmd()
-		 * has already added se_cmd to state_list, but fabric has
-		 * failed command before I/O submission.
-		 */
-		if (cmd->state_active)
-			target_remove_from_state_list(cmd);
 	}
 	/*
 	 * Since the iSCSI and iSER targets driver assume that a SCSI command
@@ -2565,6 +2556,8 @@  static void target_release_cmd_kref(struct kref *kref)
 
 	WARN_ON_ONCE(atomic_read(&se_cmd->tgt_ref.refcount) != 0);
 
+	target_remove_from_state_list(se_cmd);
+
 	if (se_cmd->lun_ref_active)
 		percpu_ref_put(&se_cmd->se_lun->lun_ref);