On Mon, 2018-09-17 at 14:35 -0700, Bart Van Assche wrote:
> This patch does not change any functionality but makes the next patch
> easier to read.
>
> Signed-off-by: Bart Van Assche <bvanassche@xxxxxxx>
> Cc: Nicholas Bellinger <nab@xxxxxxxxxxxxxxx>
> Cc: Mike Christie <mchristi@xxxxxxxxxx>
> Cc: Christoph Hellwig <hch@xxxxxx>
> Cc: Hannes Reinecke <hare@xxxxxxx>
> ---
> drivers/target/target_core_transport.c | 69 ++++++++++++++++++++++----
> include/target/target_core_base.h | 2 +-
> 2 files changed, 60 insertions(+), 11 deletions(-)
>
<SNIP>
> @@ -2759,9 +2802,16 @@ static void target_release_cmd_kref(struct kref *kref)
> {
> struct se_cmd *se_cmd = container_of(kref, struct se_cmd, cmd_kref);
> struct se_session *se_sess = se_cmd->se_sess;
> - struct completion *compl = se_cmd->compl;
> + struct wait_queue_head release_wq;
> unsigned long flags;
>
> + init_waitqueue_head(&release_wq);
> + /*
> + * No locking is required since we are in the release function and
> + * adding to release_wq is only allowed while holding a reference.
> + */
> + list_splice_init(&se_cmd->release_wq.head, &release_wq.head);
> +
> if (se_sess) {
> spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
> list_del_init(&se_cmd->se_cmd_list);
> @@ -2770,8 +2820,7 @@ static void target_release_cmd_kref(struct kref *kref)
>
> target_free_cmd_mem(se_cmd);
> se_cmd->se_tfo->release_cmd(se_cmd);
> - if (compl)
> - complete(compl);
> + wake_up_all(&release_wq);
>
> percpu_ref_put(&se_sess->cmd_count);
> }
Your earlier commit 7b2cc7dc0 in mainline to use a local stack compl in
transport_generic_free_cmd() was a reasonable change, because fast-path
code in target_release_cmd_kref() was only doing a complete() when the
se_cmd was actually being quiesced..
However, the addition of wake_up_all() above for every se_cmd means a
__wake_up_common_lock() happens no matter what every time in fast-path
code.
Please, don't do this.
