On Wed, 2017-02-08 at 14:24 -0800, Bart Van Assche wrote:
> Whether or not a session is being torn down does not affect whether
> or not SCSI commands are in the task set. Hence remove the "tearing
> down" checks from the TMF code. The CMD_T_FABRIC_STOP test is left
> out because it overlaps with the CMD_T_COMPLETE and kref tests. The
> TRANSPORT_ISTATE_PROCESSING check is left out because it is now
> safe to wait for a command that is in that state.
>
> Signed-off-by: Bart Van Assche <bart.vanassche@xxxxxxxxxxx>
> Reviewed-by: Hannes Reinecke <hare@xxxxxxxx>
> Cc: Christoph Hellwig <hch@xxxxxx>
> Cc: Andy Grover <agrover@xxxxxxxxxx>
> Cc: David Disseldorp <ddiss@xxxxxxx>
> ---
> drivers/target/target_core_tmr.c | 30 ++++++------------------------
> 1 file changed, 6 insertions(+), 24 deletions(-)
>
> diff --git a/drivers/target/target_core_tmr.c b/drivers/target/target_core_tmr.c
> index 398526704001..e4c8ebbcb20c 100644
> --- a/drivers/target/target_core_tmr.c
> +++ b/drivers/target/target_core_tmr.c
> @@ -98,9 +98,8 @@ static bool __target_check_io_state(struct se_cmd *se_cmd,
> assert_spin_locked(&sess->sess_cmd_lock);
> WARN_ON_ONCE(!irqs_disabled());
> /*
> - * If command already reached CMD_T_COMPLETE state within
> - * target_complete_cmd() or CMD_T_FABRIC_STOP due to shutdown,
> - * this se_cmd has been passed to fabric driver and will
> + * If a command has already reached the CMD_T_COMPLETE state,
> + * it has already been passed to the fabric driver and will
> * not be aborted.
> *
> * Otherwise, obtain a local se_cmd->cmd_kref now for TMR
> @@ -108,15 +107,9 @@ static bool __target_check_io_state(struct se_cmd *se_cmd,
> * long as se_cmd->cmd_kref is still active unless zero.
> */
> spin_lock(&se_cmd->t_state_lock);
> - if (se_cmd->transport_state & (CMD_T_COMPLETE | CMD_T_FABRIC_STOP)) {
> - pr_debug("Attempted to abort io tag: %llu already complete or"
> - " fabric stop, skipping\n", se_cmd->tag);
> - spin_unlock(&se_cmd->t_state_lock);
> - return false;
> - }
> - if (sess->sess_tearing_down) {
> - pr_debug("Attempted to abort io tag: %llu already shutdown,"
> - " skipping\n", se_cmd->tag);
> + if (se_cmd->transport_state & CMD_T_COMPLETE) {
> + pr_debug("Response for command with tag %llu is being sent to initiator, skipping\n",
> + se_cmd->tag);
> spin_unlock(&se_cmd->t_state_lock);
> return false;
> }
The removal of this breaks the second order case I've mentioned at the
top of patch #18.
That is, while CMD_T_ABORTED is going on, fabric drivers must be able to
handle a session reinstatement that forces all se_cmd descriptors into a
quiesce state.
Getting rid of CMD_T_FABRIC_STOP here prevents session shutdown code
from knowing when commands are actively being CMD_T_ABORTED.
> @@ -235,18 +228,7 @@ static void core_tmr_drain_tmr_list(
>
> spin_lock(&sess->sess_cmd_lock);
> spin_lock(&cmd->t_state_lock);
> - if (!(cmd->transport_state & CMD_T_ACTIVE) ||
> - (cmd->transport_state & CMD_T_FABRIC_STOP)) {
> - spin_unlock(&cmd->t_state_lock);
> - spin_unlock(&sess->sess_cmd_lock);
> - continue;
> - }
> - if (cmd->t_state == TRANSPORT_ISTATE_PROCESSING) {
> - spin_unlock(&cmd->t_state_lock);
> - spin_unlock(&sess->sess_cmd_lock);
> - continue;
> - }
> - if (sess->sess_tearing_down) {
> + if (!(cmd->transport_state & CMD_T_ACTIVE)) {
> spin_unlock(&cmd->t_state_lock);
> spin_unlock(&sess->sess_cmd_lock);
> continue;
Likewise here. There are cases when a TMR cannot be aborted.
Like once it's been stopped during session shutdown quiesce, and once
it's already been queued into se_device->tmr_wq (eg:
TRANSPORT_ISTATE_PROCESSING) it must be allowed to proceed.
--
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
