target: Add generic active I/O shutdown logic

This patch adds the initial pieces of generic active I/O shutdown logic.
This is intended to be a 'opt-in' feature for fabric modules that
includes the following functions to provide a mechinism for fabric
modules to track se_cmd via se_session->sess_cmd_list:

*) target_get_sess_cmd() - Add se_cmd to sess->sess_cmd_list, called
   from fabric module incoming I/O path.
*) target_put_sess_cmd() - Check for completion or drop se_cmd from
   ->sess_cmd_list
*) target_splice_sess_cmd_list() - Splice active I/O list from
   ->sess_cmd_list to ->sess_wait_list, can called with HW fabric
   lock held.
*) target_wait_for_sess_cmds() - Walk ->sess_wait_list waiting on
   individual ->cmd_wait_comp.  Optional transport_wait_for_tasks()
   call.

target_splice_sess_cmd_list() is allowed to be called under HW fabric
lock, and performs the splice into se_sess->sess_wait_list and set
se_cmd->cmd_wait_set.  Then target_wait_for_sess_cmds() walks the list
waiting for individual target_put_sess_cmd() fabric callbacks to
complete.

It also adds TFO->check_release_cmd() to split the completion and memory
release calls, where a fabric module uses target_put_sess_cmd() to check
for I/O completion during session shutdown.  This is currently pushed out
into fabric modules as current fabric code may sleep here waiting for
TFO->check_stop_free() to complete in main response path, and because
target_wait_for_sess_cmds() calling TFO->release_cmd() to free fabric
descriptor memory directly.

Cc: Christoph Hellwig <hch@lst.de>
Cc: Roland Dreier <roland@purestorage.com>
Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>

4 files changed, 145 insertions, 6 deletions

diff --git a/drivers/target/target_core_transport.c b/drivers/target/target_core_transport.c
index 81bc355be31729..e84b26ffb17b2a 100644
--- a/drivers/target/target_core_transport.c
+++ b/drivers/target/target_core_transport.c
@@ -268,6 +268,9 @@ struct se_session *transport_init_session(void)
 	}
 	INIT_LIST_HEAD(&se_sess->sess_list);
 	INIT_LIST_HEAD(&se_sess->sess_acl_list);
+	INIT_LIST_HEAD(&se_sess->sess_cmd_list);
+	INIT_LIST_HEAD(&se_sess->sess_wait_list);
+	spin_lock_init(&se_sess->sess_cmd_lock);
 
 	return se_sess;
 }
@@ -1505,11 +1508,12 @@ void transport_init_se_cmd(
 	INIT_LIST_HEAD(&cmd->se_ordered_node);
 	INIT_LIST_HEAD(&cmd->se_qf_node);
 	INIT_LIST_HEAD(&cmd->se_queue_node);
-
+	INIT_LIST_HEAD(&cmd->se_cmd_list);
 	INIT_LIST_HEAD(&cmd->t_task_list);
 	init_completion(&cmd->transport_lun_fe_stop_comp);
 	init_completion(&cmd->transport_lun_stop_comp);
 	init_completion(&cmd->t_transport_stop_comp);
+	init_completion(&cmd->cmd_wait_comp);
 	spin_lock_init(&cmd->t_state_lock);
 	atomic_set(&cmd->transport_dev_active, 1);
 
@@ -3950,6 +3954,14 @@ void transport_release_cmd(struct se_cmd *cmd)
 		core_tmr_release_req(cmd->se_tmr_req);
 	if (cmd->t_task_cdb != cmd->__t_task_cdb)
 		kfree(cmd->t_task_cdb);
+	/*
+	 * Check if target_wait_for_sess_cmds() is expecting to
+	 * release se_cmd directly here..
+	 */
+	if (cmd->check_release != 0 && cmd->se_tfo->check_release_cmd)
+		if (cmd->se_tfo->check_release_cmd(cmd) != 0)
+			return;
+
 	cmd->se_tfo->release_cmd(cmd);
 }
 EXPORT_SYMBOL(transport_release_cmd);
@@ -3977,6 +3989,114 @@ void transport_generic_free_cmd(struct se_cmd *cmd, int wait_for_tasks)
 }
 EXPORT_SYMBOL(transport_generic_free_cmd);
 
+/* target_get_sess_cmd - Add command to active ->sess_cmd_list
+ * @se_sess:	session to reference
+ * @se_cmd:	command descriptor to add
+ */
+void target_get_sess_cmd(struct se_session *se_sess, struct se_cmd *se_cmd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
+	list_add_tail(&se_cmd->se_cmd_list, &se_sess->sess_cmd_list);
+	se_cmd->check_release = 1;
+	spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+}
+EXPORT_SYMBOL(target_get_sess_cmd);
+
+/* target_put_sess_cmd - Check for active I/O shutdown or list delete
+ * @se_sess: 	session to reference
+ * @se_cmd:	command descriptor to drop
+ */
+int target_put_sess_cmd(struct se_session *se_sess, struct se_cmd *se_cmd)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
+	if (list_empty(&se_cmd->se_cmd_list)) {
+		spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+		WARN_ON(1);
+		return 0;
+	}
+
+	if (se_sess->sess_tearing_down && se_cmd->cmd_wait_set) {
+		spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+		complete(&se_cmd->cmd_wait_comp);
+		return 1;
+	}
+	list_del(&se_cmd->se_cmd_list);
+	spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+
+	return 0;
+}
+EXPORT_SYMBOL(target_put_sess_cmd);
+
+/* target_splice_sess_cmd_list - Split active cmds into sess_wait_list
+ * @se_sess:	session to split
+ */
+void target_splice_sess_cmd_list(struct se_session *se_sess)
+{
+	struct se_cmd *se_cmd;
+	unsigned long flags;
+
+	WARN_ON(!list_empty(&se_sess->sess_wait_list));
+	INIT_LIST_HEAD(&se_sess->sess_wait_list);
+
+	spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
+	se_sess->sess_tearing_down = 1;
+
+	list_splice_init(&se_sess->sess_cmd_list, &se_sess->sess_wait_list);
+
+	list_for_each_entry(se_cmd, &se_sess->sess_wait_list, se_cmd_list)
+		se_cmd->cmd_wait_set = 1;
+
+	spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+}
+EXPORT_SYMBOL(target_splice_sess_cmd_list);
+
+/* target_wait_for_sess_cmds - Wait for outstanding descriptors
+ * @se_sess:    session to wait for active I/O
+ * @wait_for_tasks:	Make extra transport_wait_for_tasks call
+ */
+void target_wait_for_sess_cmds(
+	struct se_session *se_sess,
+	int wait_for_tasks)
+{
+	struct se_cmd *se_cmd, *tmp_cmd;
+	bool rc = false;
+
+	list_for_each_entry_safe(se_cmd, tmp_cmd,
+				&se_sess->sess_wait_list, se_cmd_list) {
+		list_del(&se_cmd->se_cmd_list);
+
+		pr_debug("Waiting for se_cmd: %p t_state: %d, fabric state:"
+			" %d\n", se_cmd, se_cmd->t_state,
+			se_cmd->se_tfo->get_cmd_state(se_cmd));
+
+		if (wait_for_tasks) {
+			pr_debug("Calling transport_wait_for_tasks se_cmd: %p t_state: %d,"
+				" fabric state: %d\n", se_cmd, se_cmd->t_state,
+				se_cmd->se_tfo->get_cmd_state(se_cmd));
+
+			rc = transport_wait_for_tasks(se_cmd);
+
+			pr_debug("After transport_wait_for_tasks se_cmd: %p t_state: %d,"
+				" fabric state: %d\n", se_cmd, se_cmd->t_state,
+				se_cmd->se_tfo->get_cmd_state(se_cmd));
+		}
+
+		if (!rc) {
+			wait_for_completion(&se_cmd->cmd_wait_comp);
+			pr_debug("After cmd_wait_comp: se_cmd: %p t_state: %d"
+				" fabric state: %d\n", se_cmd, se_cmd->t_state,
+				se_cmd->se_tfo->get_cmd_state(se_cmd));
+		}
+
+		se_cmd->se_tfo->release_cmd(se_cmd);
+	}
+}
+EXPORT_SYMBOL(target_wait_for_sess_cmds);
+
 /*	transport_lun_wait_for_tasks():
  *
  *	Called from ConfigFS context to stop the passed struct se_cmd to allow
@@ -4153,14 +4273,14 @@ int transport_clear_lun_from_sessions(struct se_lun *lun)
  * Called from frontend fabric context to wait for storage engine
  * to pause and/or release frontend generated struct se_cmd.
  */
-void transport_wait_for_tasks(struct se_cmd *cmd)
+bool transport_wait_for_tasks(struct se_cmd *cmd)
 {
 	unsigned long flags;
 
 	spin_lock_irqsave(&cmd->t_state_lock, flags);
 	if (!(cmd->se_cmd_flags & SCF_SE_LUN_CMD) && !(cmd->se_tmr_req)) {
 		spin_unlock_irqrestore(&cmd->t_state_lock, flags);
-		return;
+		return false;
 	}
 	/*
 	 * Only perform a possible wait_for_tasks if SCF_SUPPORTED_SAM_OPCODE
@@ -4168,7 +4288,7 @@ void transport_wait_for_tasks(struct se_cmd *cmd)
 	 */
 	if (!(cmd->se_cmd_flags & SCF_SUPPORTED_SAM_OPCODE) && !cmd->se_tmr_req) {
 		spin_unlock_irqrestore(&cmd->t_state_lock, flags);
-		return;
+		return false;
 	}
 	/*
 	 * If we are already stopped due to an external event (ie: LUN shutdown)
@@ -4211,7 +4331,7 @@ void transport_wait_for_tasks(struct se_cmd *cmd)
 	if (!atomic_read(&cmd->t_transport_active) ||
 	     atomic_read(&cmd->t_transport_aborted)) {
 		spin_unlock_irqrestore(&cmd->t_state_lock, flags);
-		return;
+		return false;
 	}
 
 	atomic_set(&cmd->t_transport_stop, 1);
@@ -4236,6 +4356,8 @@ void transport_wait_for_tasks(struct se_cmd *cmd)
 		cmd->se_tfo->get_task_tag(cmd));
 
 	spin_unlock_irqrestore(&cmd->t_state_lock, flags);
+
+	return true;
 }
 EXPORT_SYMBOL(transport_wait_for_tasks);
 
diff --git a/include/target/target_core_base.h b/include/target/target_core_base.h
index d571bcfd16ad3a..dd245b68be5a0b 100644
--- a/include/target/target_core_base.h
+++ b/include/target/target_core_base.h
@@ -425,6 +425,9 @@ struct se_cmd {
 	enum transport_state_table t_state;
 	/* Transport specific error status */
 	int			transport_error_status;
+	/* Used to signal cmd->se_tfo->check_release_cmd() usage per cmd */
+	int			check_release:1;
+	int			cmd_wait_set:1;
 	/* See se_cmd_flags_table */
 	u32			se_cmd_flags;
 	u32			se_ordered_id;
@@ -451,6 +454,8 @@ struct se_cmd {
 	struct se_session	*se_sess;
 	struct se_tmr_req	*se_tmr_req;
 	struct list_head	se_queue_node;
+	struct list_head	se_cmd_list;
+	struct completion	cmd_wait_comp;
 	struct target_core_fabric_ops *se_tfo;
 	int (*transport_emulate_cdb)(struct se_cmd *);
 	void (*transport_complete_callback)(struct se_cmd *);
@@ -558,12 +563,16 @@ struct se_node_acl {
 } ____cacheline_aligned;
 
 struct se_session {
+	int			sess_tearing_down:1;
 	u64			sess_bin_isid;
 	struct se_node_acl	*se_node_acl;
 	struct se_portal_group *se_tpg;
 	void			*fabric_sess_ptr;
 	struct list_head	sess_list;
 	struct list_head	sess_acl_list;
+	struct list_head	sess_cmd_list;
+	struct list_head	sess_wait_list;
+	spinlock_t		sess_cmd_lock;
 } ____cacheline_aligned;
 
 struct se_device;
diff --git a/include/target/target_core_fabric_ops.h b/include/target/target_core_fabric_ops.h
index 04c591da084447..0256825f923dbd 100644
--- a/include/target/target_core_fabric_ops.h
+++ b/include/target/target_core_fabric_ops.h
@@ -52,6 +52,10 @@ struct target_core_fabric_ops {
 	 * Returning 0 will signal a descriptor has not been released.
 	 */
 	int (*check_stop_free)(struct se_cmd *);
+	/*
+	 * Optional check for active I/O shutdown
+	 */
+	int (*check_release_cmd)(struct se_cmd *);
 	void (*release_cmd)(struct se_cmd *);
 	/*
 	 * Called with spin_lock_bh(struct se_portal_group->session_lock held.
diff --git a/include/target/target_core_transport.h b/include/target/target_core_transport.h
index d1b68c9ccb7592..c16e9431dd01bb 100644
--- a/include/target/target_core_transport.h
+++ b/include/target/target_core_transport.h
@@ -164,12 +164,16 @@ extern bool target_stop_task(struct se_task *task, unsigned long *flags);
 extern int transport_generic_map_mem_to_cmd(struct se_cmd *cmd, struct scatterlist *, u32,
 				struct scatterlist *, u32);
 extern int transport_clear_lun_from_sessions(struct se_lun *);
-extern void transport_wait_for_tasks(struct se_cmd *);
+extern bool transport_wait_for_tasks(struct se_cmd *);
 extern int transport_check_aborted_status(struct se_cmd *, int);
 extern int transport_send_check_condition_and_sense(struct se_cmd *, u8, int);
 extern void transport_send_task_abort(struct se_cmd *);
 extern void transport_release_cmd(struct se_cmd *);
 extern void transport_generic_free_cmd(struct se_cmd *, int);
+extern void target_get_sess_cmd(struct se_session *, struct se_cmd *);
+extern int target_put_sess_cmd(struct se_session *, struct se_cmd *);
+extern void target_splice_sess_cmd_list(struct se_session *);
+extern void target_wait_for_sess_cmds(struct se_session *, int);
 extern void transport_generic_wait_for_cmds(struct se_cmd *, int);
 extern void transport_do_task_sg_chain(struct se_cmd *);
 extern void transport_generic_process_write(struct se_cmd *);