Skip to contentSkip to conversation

scsi: target: tcmu: Add TMR notification for tcmu

PATCH Applied Bodo Stroesser 3 participants 11 messages Jul 26, 2020 – Jul 29, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v3 cover scsi: target: tcmu: Add TMR notification for tcmu

This patch series is made on top of Martin's for-next branch.

ChangeLog:

v2: in patch "scsi: target: tcmu: Implement tmr_notify callback"
changed new comment's style from "// ..." to "/* ... */" and correctly use "/** " for function doc.

V3:
- Patch 1 "scsi: target: Modify core_tmr_abort_task()":
fixed wrong spin_lock handling. Nested calls to spin_lock_irqsave and spin_lock_irqrestore used the same flags field. Inner pair replaced by spin_lock / spin_unlock

- Patches 5,7,8:
"scsi: target: tcmu: Factor out new helper ring_insert_padding" "scsi: target: tcmu: Implement tmr_notify callback" "scsi: target: tcmu: Make TMR notification optional" New definitions of struct tcmu_dev *dev renamed to *udev.

- Patch 8 "scsi: target: tcmu: Make TMR notification optional" Spacing fixed at function definition.

---

TCM/LIO core handles TMRs without involving backends. But TMR completion message is sent to initiator only after commands aborted by the TMR are completed by the backend. Especially in case of tcmu backend, if userspace executes long running commands and therefore initiator sends ABORT_TASK on timeout, the ABORT itself can time out if core still waits for userspace/tcmu to complete the command.

It would be very helpful for userspace to get a notification about received TMR and which commands were aborted by that TMR. Then userspace can decide whether to cancel command processing, and it can send command completion earlier than it would without TMR notification.
It is also helpful for userspace traces and device emulation to get notification about TMR events.

So this patch series in the first two patches implements in core the usage of a new optional backend callback for TMR notifications. The core calls it before core waits for completion of aborted commands (params: se_dev, TMR type, and list of se_cmds aborted by this TMR).
Of course other backends than tcmu can use this part of the series also to implement their own TMR notification if necessary.

The further six patches implement the TMR notify callback for tcmu. The new configFS attribute tmr_notification allows to switch on TMR messages on the cmd ring. The default of the attribute is the old behavior without TMR info on the ring, but with following changes:
- if tcmu receives an already aborted command, it immediately rejects it. So it will never appear in userspace.
- if tcmu finds, that according to TMR notification a cmd on the qfull_queue was aborted, tcmu removes it from qfull_queue and completes it immediately. So userspace will not 'see' those commands.

When attribute tmr_notification is set to 1, tcmu additionally prepares a list of cmd_ids from those commands, that are aborted by the TMR and are active in cmd ring (not timed out). This list together with the TMR type is either immediately written to cmd ring (new TMR entry type) or queued in a separate tmr queue if ring space is too small.
TMRs in the tmr queue do not time out. If ring space becomes available, tcmu moves TMRs from tmr queue to ring with higher priority than cmds from qfull queue.

This mechanism makes sure that userspace receives TMR notifications as early as possible. Userspace can use the list of cmd_ids attached to the TMR notification to identify aborted commands from its list of received and not yet completed commands. In case userspace has meanwhile completed some of the cmd_ids on the list, it can just ignore these cmd_ids. A possible new command having the same cmd_id as one of the aborted commands will always appear on the ring after the TMR notification.

Bodo Stroesser (8):
scsi: target: Modify core_tmr_abort_task()
scsi: target: Add tmr_notify backend function
scsi: target: tcmu: Use priv pointer in se_cmd
scsi: target: tcmu: Do not queue aborted commands
scsi: target: tcmu: Factor out new helper ring_insert_padding
scsi: target: tcmu: Fix and simplify timeout handling
scsi: target: tcmu: Implement tmr_notify callback
scsi: target: tcmu: Make TMR notification optional

 drivers/target/target_core_tmr.c       |  36 +++-
 drivers/target/target_core_transport.c |   1 +
 drivers/target/target_core_user.c      | 373 +++++++++++++++++++++++++++------
 include/target/target_core_backend.h   |   2 +
 include/target/target_core_base.h      |   1 +
 include/uapi/linux/target_core_user.h  |  25 +++
 6 files changed, 366 insertions(+), 72 deletions(-)
Signature · 1 line
2.12.3
#1 lore raw

PATCH v3 1/8 scsi: target: Modify core_tmr_abort_task()

This patch modifies core_tmr_abort_task() to use same looping and locking scheme as core_tmr_drain_state_list() does.

This frees the state_list element in se_cmd for later use by tmr notification handling.

Note: __target_check_io_state() now is called with param 0 instead of dev->dev_attrib.emulate_tas, because tas is not relevant since we always get ABRT on same session like the aborted command.

---
 drivers/target/target_core_tmr.c | 22 ++++++++++++++--------
 1 file changed, 14 insertions(+), 8 deletions(-)
drivers/target/target_core_tmr.c +14 -8
index 89c84d472cd7..73c4155f3c1e 100644
@@ -116,14 +116,15 @@ void core_tmr_abort_task(
 	struct se_tmr_req *tmr,
 	struct se_session *se_sess)
 {
-	struct se_cmd *se_cmd;
+	struct se_cmd *se_cmd, *next;
 	unsigned long flags;
+	bool rc;
 	u64 ref_tag;
 
-	spin_lock_irqsave(&se_sess->sess_cmd_lock, flags);
-	list_for_each_entry(se_cmd, &se_sess->sess_cmd_list, se_cmd_list) {
+	spin_lock_irqsave(&dev->execute_task_lock, flags);
+	list_for_each_entry_safe(se_cmd, next, &dev->state_list, state_list) {
 
-		if (dev != se_cmd->se_dev)
+		if (se_sess != se_cmd->se_sess)
 			continue;
 
 		/* skip task management functions, including tmr->task_cmd */
@@ -137,11 +138,16 @@ void core_tmr_abort_task(
 		printk("ABORT_TASK: Found referenced %s task_tag: %llu\n",
 			se_cmd->se_tfo->fabric_name, ref_tag);
 
-		if (!__target_check_io_state(se_cmd, se_sess,
-					     dev->dev_attrib.emulate_tas))
+		spin_lock(&se_sess->sess_cmd_lock);
+		rc = __target_check_io_state(se_cmd, se_sess, 0);
+		spin_unlock(&se_sess->sess_cmd_lock);
+		if (!rc)
 			continue;
 
-		spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+		list_del_init(&se_cmd->state_list);
+		se_cmd->state_active = false;
+
+		spin_unlock_irqrestore(&dev->execute_task_lock, flags);
 
 		/*
 		 * Ensure that this ABORT request is visible to the LU RESET
@@ -159,7 +165,7 @@ void core_tmr_abort_task(
 		atomic_long_inc(&dev->aborts_complete);
 		return;
 	}
-	spin_unlock_irqrestore(&se_sess->sess_cmd_lock, flags);
+	spin_unlock_irqrestore(&dev->execute_task_lock, flags);
 
 	printk("ABORT_TASK: Sending TMR_TASK_DOES_NOT_EXIST for ref_tag: %lld\n",
 			tmr->ref_task_tag);
Signature · 1 line
2.12.3
#2 lore raw

PATCH v3 2/8 scsi: target: Add tmr_notify backend function

In reply to Bodo Stroesser

Target core is modified to call an optional backend callback function if a TMR is received or commands are aborted implicitly after a PR command was received. The backend function takes as parameters the se_dev, the type of the TMR, and the list of aborted commands. If no commands were aborted, an empty list is supplied.

---
 drivers/target/target_core_tmr.c       | 16 +++++++++++++++-
 drivers/target/target_core_transport.c |  1 +
 include/target/target_core_backend.h   |  2 ++
 include/target/target_core_base.h      |  1 +
 4 files changed, 19 insertions(+), 1 deletion(-)
drivers/target/target_core_tmr.c +15 -1
index 73c4155f3c1e..e4513ef09159 100644
@@ -116,6 +116,7 @@ void core_tmr_abort_task(
 	struct se_tmr_req *tmr,
 	struct se_session *se_sess)
 {
+	LIST_HEAD(aborted_list);
 	struct se_cmd *se_cmd, *next;
 	unsigned long flags;
 	bool rc;
@@ -144,7 +145,7 @@ void core_tmr_abort_task(
 		if (!rc)
 			continue;
 
-		list_del_init(&se_cmd->state_list);
+		list_move_tail(&se_cmd->state_list, &aborted_list);
 		se_cmd->state_active = false;
 
 		spin_unlock_irqrestore(&dev->execute_task_lock, flags);
@@ -157,6 +158,11 @@ void core_tmr_abort_task(
 			WARN_ON_ONCE(transport_lookup_tmr_lun(tmr->task_cmd) <
 					0);
 
+		if (dev->transport->tmr_notify)
+			dev->transport->tmr_notify(dev, TMR_ABORT_TASK,
+						   &aborted_list);
+
+		list_del_init(&se_cmd->state_list);
 		target_put_cmd_and_wait(se_cmd);
 
 		printk("ABORT_TASK: Sending TMR_FUNCTION_COMPLETE for"
@@ -167,6 +173,9 @@ void core_tmr_abort_task(
 	}
 	spin_unlock_irqrestore(&dev->execute_task_lock, flags);
 
+	if (dev->transport->tmr_notify)
+		dev->transport->tmr_notify(dev, TMR_ABORT_TASK, &aborted_list);
+
 	printk("ABORT_TASK: Sending TMR_TASK_DOES_NOT_EXIST for ref_tag: %lld\n",
 			tmr->ref_task_tag);
 	tmr->response = TMR_TASK_DOES_NOT_EXIST;
@@ -318,6 +327,11 @@ static void core_tmr_drain_state_list(
 	}
 	spin_unlock_irqrestore(&dev->execute_task_lock, flags);
 
+	if (dev->transport->tmr_notify)
+		dev->transport->tmr_notify(dev, preempt_and_abort_list ?
+					   TMR_LUN_RESET_PRO : TMR_LUN_RESET,
+					   &drain_task_list);
+
 	while (!list_empty(&drain_task_list)) {
 		cmd = list_entry(drain_task_list.next, struct se_cmd, state_list);
 		list_del_init(&cmd->state_list);
drivers/target/target_core_transport.c +1 -0
index e6e1fa68de54..9fb0be0aa620 100644
@@ -2946,6 +2946,7 @@ static const char *target_tmf_name(enum tcm_tmreq_table tmf)
 	case TMR_LUN_RESET:		return "LUN_RESET";
 	case TMR_TARGET_WARM_RESET:	return "TARGET_WARM_RESET";
 	case TMR_TARGET_COLD_RESET:	return "TARGET_COLD_RESET";
+	case TMR_LUN_RESET_PRO:		return "LUN_RESET_PRO";
 	case TMR_UNKNOWN:		break;
 	}
 	return "(?)";
include/target/target_core_backend.h +2 -0
index f51452e3b984..6336780d83a7 100644
@@ -40,6 +40,8 @@ struct target_backend_ops {
 	ssize_t (*show_configfs_dev_params)(struct se_device *, char *);
 
 	sense_reason_t (*parse_cdb)(struct se_cmd *cmd);
+	void (*tmr_notify)(struct se_device *se_dev, enum tcm_tmreq_table,
+			   struct list_head *aborted_cmds);
 	u32 (*get_device_type)(struct se_device *);
 	sector_t (*get_blocks)(struct se_device *);
 	sector_t (*get_alignment_offset_lbas)(struct se_device *);
include/target/target_core_base.h +1 -0
index 18c3f277b770..549947d407cf 100644
@@ -207,6 +207,7 @@ enum tcm_tmreq_table {
 	TMR_LUN_RESET		= 5,
 	TMR_TARGET_WARM_RESET	= 6,
 	TMR_TARGET_COLD_RESET	= 7,
+	TMR_LUN_RESET_PRO	= 0x80,
 	TMR_UNKNOWN		= 0xff,
 };
 
Signature · 1 line
2.12.3
#3 lore raw

PATCH v3 3/8 scsi: target: tcmu: Use priv pointer in se_cmd

In reply to Bodo Stroesser

We initialize and clean up the se_cmd's priv pointer under cmd_ring_lock to point to the corresponding tcmu_cmd.

In the patch that implements tmr_notify callback in tcmu we will use the priv pointer.

---
 drivers/target/target_core_user.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
drivers/target/target_core_user.c +5 -2
index 366878b0b2fd..b06b18d1b135 100644
@@ -541,6 +541,8 @@ tcmu_get_block_page(struct tcmu_dev *udev, uint32_t dbi)
 
 static inline void tcmu_free_cmd(struct tcmu_cmd *tcmu_cmd)
 {
+	if (tcmu_cmd->se_cmd)
+		tcmu_cmd->se_cmd->priv = NULL;
 	kfree(tcmu_cmd->dbi);
 	kmem_cache_free(tcmu_cmd_cache, tcmu_cmd);
 }
@@ -1109,10 +1111,11 @@ tcmu_queue_cmd(struct se_cmd *se_cmd)
 		return TCM_LOGICAL_UNIT_COMMUNICATION_FAILURE;
 
 	mutex_lock(&udev->cmdr_lock);
+	se_cmd->priv = tcmu_cmd;
 	ret = queue_cmd_ring(tcmu_cmd, &scsi_ret);
-	mutex_unlock(&udev->cmdr_lock);
 	if (ret < 0)
 		tcmu_free_cmd(tcmu_cmd);
+	mutex_unlock(&udev->cmdr_lock);
 	return scsi_ret;
 }
 
@@ -1179,7 +1182,6 @@ static void tcmu_handle_completion(struct tcmu_cmd *cmd, struct tcmu_cmd_entry *
 		target_complete_cmd(cmd->se_cmd, entry->rsp.scsi_status);
 
 out:
-	cmd->se_cmd = NULL;
 	tcmu_cmd_free_data(cmd, cmd->dbi_cnt);
 	tcmu_free_cmd(cmd);
 }
@@ -1285,6 +1287,7 @@ static void tcmu_check_expired_ring_cmd(struct tcmu_cmd *cmd)
 	set_bit(TCMU_CMD_BIT_EXPIRED, &cmd->flags);
 	list_del_init(&cmd->queue_entry);
 	se_cmd = cmd->se_cmd;
+	se_cmd->priv = NULL;
 	cmd->se_cmd = NULL;
 
 	pr_debug("Timing out inflight cmd %u on dev %s.\n",
Signature · 1 line
2.12.3
#4 lore raw

PATCH v3 4/8 scsi: target: tcmu: Do not queue aborted commands

In reply to Bodo Stroesser

If tcmu receives an already aborted command, tcmu_queue_cmd() should reject it.

---
 drivers/target/target_core_user.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/target/target_core_user.c +4 -3
index b06b18d1b135..25c480fde9ee 100644
@@ -1103,8 +1103,8 @@ tcmu_queue_cmd(struct se_cmd *se_cmd)
 	struct se_device *se_dev = se_cmd->se_dev;
 	struct tcmu_dev *udev = TCMU_DEV(se_dev);
 	struct tcmu_cmd *tcmu_cmd;
-	sense_reason_t scsi_ret;
-	int ret;
+	sense_reason_t scsi_ret = TCM_CHECK_CONDITION_ABORT_CMD;
+	int ret = -1;
 
 	tcmu_cmd = tcmu_alloc_cmd(se_cmd);
 	if (!tcmu_cmd)
@@ -1112,7 +1112,8 @@ tcmu_queue_cmd(struct se_cmd *se_cmd)
 
 	mutex_lock(&udev->cmdr_lock);
 	se_cmd->priv = tcmu_cmd;
-	ret = queue_cmd_ring(tcmu_cmd, &scsi_ret);
+	if (!(se_cmd->transport_state & CMD_T_ABORTED))
+		ret = queue_cmd_ring(tcmu_cmd, &scsi_ret);
 	if (ret < 0)
 		tcmu_free_cmd(tcmu_cmd);
 	mutex_unlock(&udev->cmdr_lock);
Signature · 1 line
2.12.3
#5 lore raw

PATCH v3 5/8 scsi: target: tcmu: Factor out new helper ring_insert_padding

In reply to Bodo Stroesser

The new helper ring_insert_padding is split off from and then called by queue_cmd_ring. It inserts a padding if necessary. The new helper will in a further patch be used during writing of TMR notifications to command ring.

---
 drivers/target/target_core_user.c | 51 +++++++++++++++++++++++----------------
 1 file changed, 30 insertions(+), 21 deletions(-)
drivers/target/target_core_user.c +30 -21
index 25c480fde9ee..eb68c5fee7b7 100644
@@ -926,6 +926,34 @@ static int add_to_qfull_queue(struct tcmu_cmd *tcmu_cmd)
 	return 0;
 }
 
+static uint32_t ring_insert_padding(struct tcmu_dev *udev, size_t cmd_size)
+{
+	struct tcmu_cmd_entry_hdr *hdr;
+	struct tcmu_mailbox *mb = udev->mb_addr;
+	uint32_t cmd_head = mb->cmd_head % udev->cmdr_size; /* UAM */
+
+	/* Insert a PAD if end-of-ring space is too small */
+	if (head_to_end(cmd_head, udev->cmdr_size) < cmd_size) {
+		size_t pad_size = head_to_end(cmd_head, udev->cmdr_size);
+
+		hdr = (void *) mb + CMDR_OFF + cmd_head;
+		tcmu_hdr_set_op(&hdr->len_op, TCMU_OP_PAD);
+		tcmu_hdr_set_len(&hdr->len_op, pad_size);
+		hdr->cmd_id = 0; /* not used for PAD */
+		hdr->kflags = 0;
+		hdr->uflags = 0;
+		tcmu_flush_dcache_range(hdr, sizeof(*hdr));
+
+		UPDATE_HEAD(mb->cmd_head, pad_size, udev->cmdr_size);
+		tcmu_flush_dcache_range(mb, sizeof(*mb));
+
+		cmd_head = mb->cmd_head % udev->cmdr_size; /* UAM */
+		WARN_ON(cmd_head != 0);
+	}
+
+	return cmd_head;
+}
+
 /**
  * queue_cmd_ring - queue cmd to ring or internally
  * @tcmu_cmd: cmd to queue
@@ -941,7 +969,7 @@ static int queue_cmd_ring(struct tcmu_cmd *tcmu_cmd, sense_reason_t *scsi_err)
 	struct tcmu_dev *udev = tcmu_cmd->tcmu_dev;
 	struct se_cmd *se_cmd = tcmu_cmd->se_cmd;
 	size_t base_command_size, command_size;
-	struct tcmu_mailbox *mb;
+	struct tcmu_mailbox *mb = udev->mb_addr;
 	struct tcmu_cmd_entry *entry;
 	struct iovec *iov;
 	int iov_cnt, cmd_id;
@@ -980,8 +1008,6 @@ static int queue_cmd_ring(struct tcmu_cmd *tcmu_cmd, sense_reason_t *scsi_err)
 	if (!list_empty(&udev->qfull_queue))
 		goto queue;
 
-	mb = udev->mb_addr;
-	cmd_head = mb->cmd_head % udev->cmdr_size; /* UAM */
 	if ((command_size > (udev->cmdr_size / 2)) ||
 	    data_length > udev->data_size) {
 		pr_warn("TCMU: Request of size %zu/%zu is too big for %u/%zu "
@@ -1001,24 +1027,7 @@ static int queue_cmd_ring(struct tcmu_cmd *tcmu_cmd, sense_reason_t *scsi_err)
 		goto queue;
 	}
 
-	/* Insert a PAD if end-of-ring space is too small */
-	if (head_to_end(cmd_head, udev->cmdr_size) < command_size) {
-		size_t pad_size = head_to_end(cmd_head, udev->cmdr_size);
-
-		entry = (void *) mb + CMDR_OFF + cmd_head;
-		tcmu_hdr_set_op(&entry->hdr.len_op, TCMU_OP_PAD);
-		tcmu_hdr_set_len(&entry->hdr.len_op, pad_size);
-		entry->hdr.cmd_id = 0; /* not used for PAD */
-		entry->hdr.kflags = 0;
-		entry->hdr.uflags = 0;
-		tcmu_flush_dcache_range(entry, sizeof(entry->hdr));
-
-		UPDATE_HEAD(mb->cmd_head, pad_size, udev->cmdr_size);
-		tcmu_flush_dcache_range(mb, sizeof(*mb));
-
-		cmd_head = mb->cmd_head % udev->cmdr_size; /* UAM */
-		WARN_ON(cmd_head != 0);
-	}
+	cmd_head = ring_insert_padding(udev, command_size);
 
 	entry = (void *) mb + CMDR_OFF + cmd_head;
 	memset(entry, 0, command_size);
Signature · 1 line
2.12.3
#6 lore raw

PATCH v3 6/8 scsi: target: tcmu: Fix and simplify timeout handling

In reply to Bodo Stroesser

During cmd timeout handling in check_timedout_devices() due to a race it can happen, that tcmu_set_next_deadline() does not start a timer as expected:
1) Either tcmu_check_expired_ring_cmd() checks the inflight_queue or tcmu_check_expired_queue_cmd() checks the qfull_queue while jiffies has the value X
2) At the end of the check the queue contains one remaining command with deadline X (time_after(X, X) is false and thus the command is not handled as being timed out).
3) After tcmu_check_expired_xxxxx_cmd() a timer interrupt happens and jiffies is incremented to X+1.
4) Now tcmu_set_next_deadline() is called, but it skips the command, since time_after(X+1, X) is true. Therefore tcmu_set_next_deadline() finds no new deadline and stops the timer, which it shouldn't.

Since commands that time out are removed from inflight_queue or qfull_queue, we don't need the check with time_after() in tcmu_set_next_deadline(), but can use the deadline from the first cmd in the queue.

Additionally I replaced the remaining time_after() calls in tcmu_check_expired_xxxxx_cmd() with time_after_eq(), because it is not useful to set the timeout to deadline, but then check for jiffies being greater than deadline.

Next I simplified the end of tcmu_handle_completions() and changed the check for no more pending commands from
"mb->cmd_tail = mb->cmd_head"
to
"idr_is_empty(&udev->commands)"
because the old check doesn't work correctly if paddings or in the future TMRs are in the ring.

Finally tcmu_set_next_deadline() was shifted in the source as preparation for later implementation of tmr_notify callback.

---
 drivers/target/target_core_user.c | 59 +++++++++++++++------------------------
 1 file changed, 23 insertions(+), 36 deletions(-)
drivers/target/target_core_user.c +23 -36
index eb68c5fee7b7..bddd40f07929 100644
@@ -1129,6 +1129,18 @@ tcmu_queue_cmd(struct se_cmd *se_cmd)
 	return scsi_ret;
 }
 
+static void tcmu_set_next_deadline(struct list_head *queue,
+				   struct timer_list *timer)
+{
+	struct tcmu_cmd *cmd;
+
+	if (!list_empty(queue)) {
+		cmd = list_first_entry(queue, struct tcmu_cmd, queue_entry);
+		mod_timer(timer, cmd->deadline);
+	} else
+		del_timer(timer);
+}
+
 static void tcmu_handle_completion(struct tcmu_cmd *cmd, struct tcmu_cmd_entry *entry)
 {
 	struct se_cmd *se_cmd = cmd->se_cmd;
@@ -1196,25 +1208,6 @@ static void tcmu_handle_completion(struct tcmu_cmd *cmd, struct tcmu_cmd_entry *
 	tcmu_free_cmd(cmd);
 }
 
-static void tcmu_set_next_deadline(struct list_head *queue,
-				   struct timer_list *timer)
-{
-	struct tcmu_cmd *tcmu_cmd, *tmp_cmd;
-	unsigned long deadline = 0;
-
-	list_for_each_entry_safe(tcmu_cmd, tmp_cmd, queue, queue_entry) {
-		if (!time_after(jiffies, tcmu_cmd->deadline)) {
-			deadline = tcmu_cmd->deadline;
-			break;
-		}
-	}
-
-	if (deadline)
-		mod_timer(timer, deadline);
-	else
-		del_timer(timer);
-}
-
 static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 {
 	struct tcmu_mailbox *mb;
@@ -1267,22 +1260,16 @@ static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 		handled++;
 	}
 
-	if (mb->cmd_tail = mb->cmd_head) {
-		/* no more pending commands */
-		del_timer(&udev->cmd_timer);
-
-		if (list_empty(&udev->qfull_queue)) {
-			/*
-			 * no more pending or waiting commands so try to
-			 * reclaim blocks if needed.
-			 */
-			if (atomic_read(&global_db_count) >
-			    tcmu_global_max_blocks)
-				schedule_delayed_work(&tcmu_unmap_work, 0);
-		}
-	} else if (udev->cmd_time_out) {
-		tcmu_set_next_deadline(&udev->inflight_queue, &udev->cmd_timer);
+	if (atomic_read(&global_db_count) > tcmu_global_max_blocks &&
+	    idr_is_empty(&udev->commands) && list_empty(&udev->qfull_queue)) {
+		/*
+		 * Allocated blocks exceeded global block limit, currently no
+		 * more pending or waiting commands so try to reclaim blocks.
+		 */
+		schedule_delayed_work(&tcmu_unmap_work, 0);
 	}
+	if (udev->cmd_time_out)
+		tcmu_set_next_deadline(&udev->inflight_queue, &udev->cmd_timer);
 
 	return handled;
 }
@@ -1291,7 +1278,7 @@ static void tcmu_check_expired_ring_cmd(struct tcmu_cmd *cmd)
 {
 	struct se_cmd *se_cmd;
 
-	if (!time_after(jiffies, cmd->deadline))
+	if (!time_after_eq(jiffies, cmd->deadline))
 		return;
 
 	set_bit(TCMU_CMD_BIT_EXPIRED, &cmd->flags);
@@ -1310,7 +1297,7 @@ static void tcmu_check_expired_queue_cmd(struct tcmu_cmd *cmd)
 {
 	struct se_cmd *se_cmd;
 
-	if (!time_after(jiffies, cmd->deadline))
+	if (!time_after_eq(jiffies, cmd->deadline))
 		return;
 
 	pr_debug("Timing out queued cmd %p on dev %s.\n",
Signature · 1 line
2.12.3
#7 lore raw

PATCH v3 7/8 scsi: target: tcmu: Implement tmr_notify callback

In reply to Bodo Stroesser

This patch implements the tmr_notify callback for tcmu. When the callback is called, tcmu checks the list of aborted commands it received as parameter:
- aborted commands in the qfull_queue are removed from the queue and target_complete_command is called
- from the cmd_ids of aborted commands currently uncompleted in cmd ring it creates a list of aborted cmd_ids.
Finally a TMR notification is written to cmd ring containing TMR type and cmd_id list. If there is no space in ring, the TMR notification is queued on a TMR specific queue.

The TMR specific queue 'tmr_queue' can be seen as a extension of the cmd ring. At the end of each iexecution of tcmu_complete_commands() we check, whether tmr_queue contains TMRs and try to move them onto the ring. If tmr_queue is not empty after that, we don't call run_qfull_queue() because commands must not overtake TMRs.

Operating that way we guarantee that cmd_ids in TMR notification received by userspace either match an active, not yet completed command or are no longer valid due to userspace having complete some cmd_ids meanwhile.

New commands that were assigned to an aborted cmd_id will always appear on the cmd ring _after_ the TMR.

---
 drivers/target/target_core_user.c     | 225 ++++++++++++++++++++++++++++++++--
 include/uapi/linux/target_core_user.h |  25 ++++
 2 files changed, 241 insertions(+), 9 deletions(-)
drivers/target/target_core_user.c +216 -9
index bddd40f07929..cb5a561a46e8 100644
@@ -137,6 +137,7 @@ struct tcmu_dev {
 
 	struct mutex cmdr_lock;
 	struct list_head qfull_queue;
+	struct list_head tmr_queue;
 
 	uint32_t dbi_max;
 	uint32_t dbi_thresh;
@@ -183,6 +184,15 @@ struct tcmu_cmd {
 #define TCMU_CMD_BIT_EXPIRED 0
 	unsigned long flags;
 };
+
+struct tcmu_tmr {
+	struct list_head queue_entry;
+
+	uint8_t tmr_type;
+	uint32_t tmr_cmd_cnt;
+	int16_t tmr_cmd_ids[0];
+};
+
 /*
  * To avoid dead lock the mutex lock order should always be:
  *
@@ -844,6 +854,9 @@ static bool is_ring_space_avail(struct tcmu_dev *udev, struct tcmu_cmd *cmd,
 		return false;
 	}
 
+	if (!data_needed)
+		return true;
+
 	/* try to check and get the data blocks as needed */
 	space = spc_bitmap_free(udev->data_bitmap, udev->dbi_thresh);
 	if ((space * DATA_BLOCK_SIZE) < data_needed) {
@@ -1106,6 +1119,60 @@ static int queue_cmd_ring(struct tcmu_cmd *tcmu_cmd, sense_reason_t *scsi_err)
 	return 1;
 }
 
+/**
+ * queue_tmr_ring - queue tmr info to ring or internally
+ * @udev: related tcmu_dev
+ * @tmr: tcmu_tmr containing tmr info to queue
+ *
+ * Returns:
+ *  0 success
+ *  1 internally queued to wait for ring memory to free.
+ */
+static int
+queue_tmr_ring(struct tcmu_dev *udev, struct tcmu_tmr *tmr)
+{
+	struct tcmu_tmr_entry *entry;
+	int cmd_size;
+	int id_list_sz;
+	struct tcmu_mailbox *mb = udev->mb_addr;
+	uint32_t cmd_head;
+
+	if (test_bit(TCMU_DEV_BIT_BROKEN, &udev->flags))
+		goto out_free;
+
+	id_list_sz = sizeof(tmr->tmr_cmd_ids[0]) * tmr->tmr_cmd_cnt;
+	cmd_size = round_up(sizeof(*entry) + id_list_sz, TCMU_OP_ALIGN_SIZE);
+
+	if (!list_empty(&udev->tmr_queue) ||
+	    !is_ring_space_avail(udev, NULL, cmd_size, 0)) {
+		list_add_tail(&tmr->queue_entry, &udev->tmr_queue);
+		pr_debug("adding tmr %p on dev %s to TMR ring space wait queue\n",
+			 tmr, udev->name);
+		return 1;
+	}
+
+	cmd_head = ring_insert_padding(udev, cmd_size);
+
+	entry = (void *)mb + CMDR_OFF + cmd_head;
+	memset(entry, 0, cmd_size);
+	tcmu_hdr_set_op(&entry->hdr.len_op, TCMU_OP_TMR);
+	tcmu_hdr_set_len(&entry->hdr.len_op, cmd_size);
+	entry->tmr_type = tmr->tmr_type;
+	entry->cmd_cnt = tmr->tmr_cmd_cnt;
+	memcpy(&entry->cmd_ids[0], &tmr->tmr_cmd_ids[0], id_list_sz);
+	tcmu_flush_dcache_range(entry, cmd_size);
+
+	UPDATE_HEAD(mb->cmd_head, cmd_size, udev->cmdr_size);
+	tcmu_flush_dcache_range(mb, sizeof(*mb));
+
+	uio_event_notify(&udev->uio_info);
+
+out_free:
+	kfree(tmr);
+
+	return 0;
+}
+
 static sense_reason_t
 tcmu_queue_cmd(struct se_cmd *se_cmd)
 {
@@ -1141,6 +1208,85 @@ static void tcmu_set_next_deadline(struct list_head *queue,
 		del_timer(timer);
 }
 
+static int
+tcmu_tmr_type(enum tcm_tmreq_table tmf)
+{
+	switch (tmf) {
+	case TMR_ABORT_TASK:		return TCMU_TMR_ABORT_TASK;
+	case TMR_ABORT_TASK_SET:	return TCMU_TMR_ABORT_TASK_SET;
+	case TMR_CLEAR_ACA:		return TCMU_TMR_CLEAR_ACA;
+	case TMR_CLEAR_TASK_SET:	return TCMU_TMR_CLEAR_TASK_SET;
+	case TMR_LUN_RESET:		return TCMU_TMR_LUN_RESET;
+	case TMR_TARGET_WARM_RESET:	return TCMU_TMR_TARGET_WARM_RESET;
+	case TMR_TARGET_COLD_RESET:	return TCMU_TMR_TARGET_COLD_RESET;
+	case TMR_LUN_RESET_PRO:		return TCMU_TMR_LUN_RESET_PRO;
+	default:			return TCMU_TMR_UNKNOWN;
+	}
+}
+
+static void
+tcmu_tmr_notify(struct se_device *se_dev, enum tcm_tmreq_table tmf,
+		struct list_head *cmd_list)
+{
+	int i = 0, cmd_cnt = 0;
+	bool unqueued = false;
+	uint16_t *cmd_ids = NULL;
+	struct tcmu_cmd *cmd;
+	struct se_cmd *se_cmd;
+	struct tcmu_tmr *tmr;
+	struct tcmu_dev *udev = TCMU_DEV(se_dev);
+
+	mutex_lock(&udev->cmdr_lock);
+
+	/* First we check for aborted commands in qfull_queue */
+	list_for_each_entry(se_cmd, cmd_list, state_list) {
+		i++;
+		if (!se_cmd->priv)
+			continue;
+		cmd = se_cmd->priv;
+		/* Commands on qfull queue have no id yet */
+		if (cmd->cmd_id) {
+			cmd_cnt++;
+			continue;
+		}
+		pr_debug("Removing aborted command %p from queue on dev %s.\n",
+			 cmd, udev->name);
+
+		list_del_init(&cmd->queue_entry);
+		tcmu_free_cmd(cmd);
+		target_complete_cmd(se_cmd, SAM_STAT_TASK_ABORTED);
+		unqueued = true;
+	}
+	if (unqueued)
+		tcmu_set_next_deadline(&udev->qfull_queue, &udev->qfull_timer);
+
+	pr_debug("TMR event %d on dev %s, aborted cmds %d, afflicted cmd_ids %d\n",
+		 tcmu_tmr_type(tmf), udev->name, i, cmd_cnt);
+
+	tmr = kmalloc(sizeof(*tmr) + cmd_cnt * sizeof(*cmd_ids), GFP_KERNEL);
+	if (!tmr)
+		goto unlock;
+
+	tmr->tmr_type = tcmu_tmr_type(tmf);
+	tmr->tmr_cmd_cnt = cmd_cnt;
+
+	if (cmd_cnt != 0) {
+		cmd_cnt = 0;
+		list_for_each_entry(se_cmd, cmd_list, state_list) {
+			if (!se_cmd->priv)
+				continue;
+			cmd = se_cmd->priv;
+			if (cmd->cmd_id)
+				tmr->tmr_cmd_ids[cmd_cnt++] = cmd->cmd_id;
+		}
+	}
+
+	queue_tmr_ring(udev, tmr);
+
+unlock:
+	mutex_unlock(&udev->cmdr_lock);
+}
+
 static void tcmu_handle_completion(struct tcmu_cmd *cmd, struct tcmu_cmd_entry *entry)
 {
 	struct se_cmd *se_cmd = cmd->se_cmd;
@@ -1208,11 +1354,43 @@ static void tcmu_handle_completion(struct tcmu_cmd *cmd, struct tcmu_cmd_entry *
 	tcmu_free_cmd(cmd);
 }
 
+static int tcmu_run_tmr_queue(struct tcmu_dev *udev)
+{
+	struct tcmu_tmr *tmr, *tmp;
+	LIST_HEAD(tmrs);
+
+	if (list_empty(&udev->tmr_queue))
+		return 1;
+
+	pr_debug("running %s's tmr queue\n", udev->name);
+
+	list_splice_init(&udev->tmr_queue, &tmrs);
+
+	list_for_each_entry_safe(tmr, tmp, &tmrs, queue_entry) {
+		list_del_init(&tmr->queue_entry);
+
+		pr_debug("removing tmr %p on dev %s from queue\n",
+			 tmr, udev->name);
+
+		if (queue_tmr_ring(udev, tmr)) {
+			pr_debug("ran out of space during tmr queue run\n");
+			/*
+			 * tmr was requeued, so just put all tmrs back in
+			 * the queue
+			 */
+			list_splice_tail(&tmrs, &udev->tmr_queue);
+			return 0;
+		}
+	}
+
+	return 1;
+}
+
 static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 {
 	struct tcmu_mailbox *mb;
 	struct tcmu_cmd *cmd;
-	int handled = 0;
+	bool free_space = false;
 
 	if (test_bit(TCMU_DEV_BIT_BROKEN, &udev->flags)) {
 		pr_err("ring broken, not handling completions\n");
@@ -1235,7 +1413,10 @@ static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 		tcmu_flush_dcache_range(entry, ring_left < sizeof(*entry) ?
 					ring_left : sizeof(*entry));
 
-		if (tcmu_hdr_get_op(entry->hdr.len_op) = TCMU_OP_PAD) {
+		free_space = true;
+
+		if (tcmu_hdr_get_op(entry->hdr.len_op) = TCMU_OP_PAD ||
+		    tcmu_hdr_get_op(entry->hdr.len_op) = TCMU_OP_TMR) {
 			UPDATE_HEAD(udev->cmdr_last_cleaned,
 				    tcmu_hdr_get_len(entry->hdr.len_op),
 				    udev->cmdr_size);
@@ -1256,9 +1437,9 @@ static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 		UPDATE_HEAD(udev->cmdr_last_cleaned,
 			    tcmu_hdr_get_len(entry->hdr.len_op),
 			    udev->cmdr_size);
-
-		handled++;
 	}
+	if (free_space)
+		free_space = tcmu_run_tmr_queue(udev);
 
 	if (atomic_read(&global_db_count) > tcmu_global_max_blocks &&
 	    idr_is_empty(&udev->commands) && list_empty(&udev->qfull_queue)) {
@@ -1271,7 +1452,7 @@ static unsigned int tcmu_handle_completions(struct tcmu_dev *udev)
 	if (udev->cmd_time_out)
 		tcmu_set_next_deadline(&udev->inflight_queue, &udev->cmd_timer);
 
-	return handled;
+	return free_space;
 }
 
 static void tcmu_check_expired_ring_cmd(struct tcmu_cmd *cmd)
@@ -1381,6 +1562,7 @@ static struct se_device *tcmu_alloc_device(struct se_hba *hba, const char *name)
 	INIT_LIST_HEAD(&udev->node);
 	INIT_LIST_HEAD(&udev->timedout_entry);
 	INIT_LIST_HEAD(&udev->qfull_queue);
+	INIT_LIST_HEAD(&udev->tmr_queue);
 	INIT_LIST_HEAD(&udev->inflight_queue);
 	idr_init(&udev->commands);
 
@@ -1455,8 +1637,8 @@ static int tcmu_irqcontrol(struct uio_info *info, s32 irq_on)
 	struct tcmu_dev *udev = container_of(info, struct tcmu_dev, uio_info);
 
 	mutex_lock(&udev->cmdr_lock);
-	tcmu_handle_completions(udev);
-	run_qfull_queue(udev, false);
+	if (tcmu_handle_completions(udev))
+		run_qfull_queue(udev, false);
 	mutex_unlock(&udev->cmdr_lock);
 
 	return 0;
@@ -1609,6 +1791,16 @@ static void tcmu_blocks_release(struct radix_tree_root *blocks,
 	}
 }
 
+static void tcmu_remove_all_queued_tmr(struct tcmu_dev *udev)
+{
+	struct tcmu_tmr *tmr, *tmp;
+
+	list_for_each_entry_safe(tmr, tmp, &udev->tmr_queue, queue_entry) {
+		list_del_init(&tmr->queue_entry);
+		kfree(tmr);
+	}
+}
+
 static void tcmu_dev_kref_release(struct kref *kref)
 {
 	struct tcmu_dev *udev = container_of(kref, struct tcmu_dev, kref);
@@ -1631,6 +1823,8 @@ static void tcmu_dev_kref_release(struct kref *kref)
 		if (tcmu_check_and_free_pending_cmd(cmd) != 0)
 			all_expired = false;
 	}
+	/* There can be left over TMR cmds. Remove them. */
+	tcmu_remove_all_queued_tmr(udev);
 	if (!list_empty(&udev->qfull_queue))
 		all_expired = false;
 	idr_destroy(&udev->commands);
@@ -1885,7 +2079,9 @@ static int tcmu_configure_device(struct se_device *dev)
 	/* Initialise the mailbox of the ring buffer */
 	mb = udev->mb_addr;
 	mb->version = TCMU_MAILBOX_VERSION;
-	mb->flags = TCMU_MAILBOX_FLAG_CAP_OOOC | TCMU_MAILBOX_FLAG_CAP_READ_LEN;
+	mb->flags = TCMU_MAILBOX_FLAG_CAP_OOOC |
+		    TCMU_MAILBOX_FLAG_CAP_READ_LEN |
+		    TCMU_MAILBOX_FLAG_CAP_TMR;
 	mb->cmdr_off = CMDR_OFF;
 	mb->cmdr_size = udev->cmdr_size;
 
@@ -2055,6 +2251,15 @@ static void tcmu_reset_ring(struct tcmu_dev *udev, u8 err_level)
 
 	del_timer(&udev->cmd_timer);
 
+	/*
+	 * ring is empty and qfull queue never contains aborted commands.
+	 * So TMRs in tmr queue do not contain relevant cmd_ids.
+	 * After a ring reset userspace should do a fresh start, so
+	 * even LUN RESET message is no longer relevant.
+	 * Therefore remove all TMRs from qfull queue
+	 */
+	tcmu_remove_all_queued_tmr(udev);
+
 	run_qfull_queue(udev, false);
 
 	mutex_unlock(&udev->cmdr_lock);
@@ -2607,6 +2812,7 @@ static struct target_backend_ops tcmu_ops = {
 	.destroy_device		= tcmu_destroy_device,
 	.free_device		= tcmu_free_device,
 	.parse_cdb		= tcmu_parse_cdb,
+	.tmr_notify		= tcmu_tmr_notify,
 	.set_configfs_dev_params = tcmu_set_configfs_dev_params,
 	.show_configfs_dev_params = tcmu_show_configfs_dev_params,
 	.get_device_type	= sbc_get_device_type,
@@ -2633,7 +2839,8 @@ static void find_free_blocks(void)
 		}
 
 		/* Try to complete the finished commands first */
-		tcmu_handle_completions(udev);
+		if (tcmu_handle_completions(udev))
+			run_qfull_queue(udev, false);
 
 		/* Skip the udevs in idle */
 		if (!udev->dbi_thresh) {
include/uapi/linux/target_core_user.h +25 -0
index b7b57967d90f..95b1597f16ae 100644
@@ -45,6 +45,7 @@
 #define ALIGN_SIZE 64 /* Should be enough for most CPUs */
 #define TCMU_MAILBOX_FLAG_CAP_OOOC (1 << 0) /* Out-of-order completions */
 #define TCMU_MAILBOX_FLAG_CAP_READ_LEN (1 << 1) /* Read data length */
+#define TCMU_MAILBOX_FLAG_CAP_TMR (1 << 2) /* TMR notifications */
 
 struct tcmu_mailbox {
 	__u16 version;
@@ -62,6 +63,7 @@ struct tcmu_mailbox {
 enum tcmu_opcode {
 	TCMU_OP_PAD = 0,
 	TCMU_OP_CMD,
+	TCMU_OP_TMR,
 };
 
 /*
@@ -128,6 +130,29 @@ struct tcmu_cmd_entry {
 
 } __packed;
 
+struct tcmu_tmr_entry {
+	struct tcmu_cmd_entry_hdr hdr;
+
+#define TCMU_TMR_UNKNOWN		0
+#define TCMU_TMR_ABORT_TASK		1
+#define TCMU_TMR_ABORT_TASK_SET		2
+#define TCMU_TMR_CLEAR_ACA		3
+#define TCMU_TMR_CLEAR_TASK_SET		4
+#define TCMU_TMR_LUN_RESET		5
+#define TCMU_TMR_TARGET_WARM_RESET	6
+#define TCMU_TMR_TARGET_COLD_RESET	7
+/* Pseudo reset due to received PR OUT */
+#define TCMU_TMR_LUN_RESET_PRO		128
+	__u8 tmr_type;
+
+	__u8 __pad1;
+	__u16 __pad2;
+	__u32 cmd_cnt;
+	__u64 __pad3;
+	__u64 __pad4;
+	__u16 cmd_ids[0];
+} __packed;
+
 #define TCMU_OP_ALIGN_SIZE sizeof(__u64)
 
 enum tcmu_genl_cmd {
Signature · 1 line
2.12.3
#8 lore raw

PATCH v3 8/8 scsi: target: tcmu: Make TMR notification optional

In reply to Bodo Stroesser

Add "tmr_notification" configFS attribute to tcmu devices. If default value 0 of the attribute is used, tcmu only removes aborted commands from qfull_queue.
If user changes tmr_notification to 1, additionally TMR notifications will be written to the cmd ring.

---
 drivers/target/target_core_user.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)
drivers/target/target_core_user.c +38 -0
index cb5a561a46e8..9b7592350502 100644
@@ -118,6 +118,7 @@ struct tcmu_dev {
 #define TCMU_DEV_BIT_OPEN 0
 #define TCMU_DEV_BIT_BROKEN 1
 #define TCMU_DEV_BIT_BLOCKED 2
+#define TCMU_DEV_BIT_TMR_NOTIFY 3
 	unsigned long flags;
 
 	struct uio_info uio_info;
@@ -1260,6 +1261,9 @@ tcmu_tmr_notify(struct se_device *se_dev, enum tcm_tmreq_table tmf,
 	if (unqueued)
 		tcmu_set_next_deadline(&udev->qfull_queue, &udev->qfull_timer);
 
+	if (!test_bit(TCMU_DEV_BIT_TMR_NOTIFY, &udev->flags))
+		goto unlock;
+
 	pr_debug("TMR event %d on dev %s, aborted cmds %d, afflicted cmd_ids %d\n",
 		 tcmu_tmr_type(tmf), udev->name, i, cmd_cnt);
 
@@ -2706,6 +2710,39 @@ static ssize_t tcmu_emulate_write_cache_store(struct config_item *item,
 }
 CONFIGFS_ATTR(tcmu_, emulate_write_cache);
 
+static ssize_t tcmu_tmr_notification_show(struct config_item *item, char *page)
+{
+	struct se_dev_attrib *da = container_of(to_config_group(item),
+					struct se_dev_attrib, da_group);
+	struct tcmu_dev *udev = TCMU_DEV(da->da_dev);
+
+	return snprintf(page, PAGE_SIZE, "%i\n",
+			test_bit(TCMU_DEV_BIT_TMR_NOTIFY, &udev->flags));
+}
+
+static ssize_t tcmu_tmr_notification_store(struct config_item *item,
+					   const char *page, size_t count)
+{
+	struct se_dev_attrib *da = container_of(to_config_group(item),
+					struct se_dev_attrib, da_group);
+	struct tcmu_dev *udev = TCMU_DEV(da->da_dev);
+	u8 val;
+	int ret;
+
+	ret = kstrtou8(page, 0, &val);
+	if (ret < 0)
+		return ret;
+	if (val > 1)
+		return -EINVAL;
+
+	if (val)
+		set_bit(TCMU_DEV_BIT_TMR_NOTIFY, &udev->flags);
+	else
+		clear_bit(TCMU_DEV_BIT_TMR_NOTIFY, &udev->flags);
+	return count;
+}
+CONFIGFS_ATTR(tcmu_, tmr_notification);
+
 static ssize_t tcmu_block_dev_show(struct config_item *item, char *page)
 {
 	struct se_device *se_dev = container_of(to_config_group(item),
@@ -2787,6 +2824,7 @@ static struct configfs_attribute *tcmu_attrib_attrs[] = {
 	&tcmu_attr_dev_config,
 	&tcmu_attr_dev_size,
 	&tcmu_attr_emulate_write_cache,
+	&tcmu_attr_tmr_notification,
 	&tcmu_attr_nl_reply_supported,
 	NULL,
 };
Signature · 1 line
2.12.3
Reviewed-by on the v3 series
#9 lore raw
In reply to Bodo Stroesser
Quoted 72 lines · Bodo Stroesser

This patch series is made on top of Martin's for-next branch.

ChangeLog:

v2: in patch "scsi: target: tcmu: Implement tmr_notify callback"
changed new comment's style from "// ..." to "/* ... */" and correctly use "/** " for function doc.

V3:
- Patch 1 "scsi: target: Modify core_tmr_abort_task()":
fixed wrong spin_lock handling. Nested calls to spin_lock_irqsave and spin_lock_irqrestore used the same flags field. Inner pair replaced by spin_lock / spin_unlock

- Patches 5,7,8:
"scsi: target: tcmu: Factor out new helper ring_insert_padding" "scsi: target: tcmu: Implement tmr_notify callback" "scsi: target: tcmu: Make TMR notification optional" New definitions of struct tcmu_dev *dev renamed to *udev.

- Patch 8 "scsi: target: tcmu: Make TMR notification optional" Spacing fixed at function definition.

---

TCM/LIO core handles TMRs without involving backends. But TMR completion message is sent to initiator only after commands aborted by the TMR are completed by the backend. Especially in case of tcmu backend, if userspace executes long running commands and therefore initiator sends ABORT_TASK on timeout, the ABORT itself can time out if core still waits for userspace/tcmu to complete the command.

It would be very helpful for userspace to get a notification about received TMR and which commands were aborted by that TMR. Then userspace can decide whether to cancel command processing, and it can send command completion earlier than it would without TMR notification.
It is also helpful for userspace traces and device emulation to get notification about TMR events.

So this patch series in the first two patches implements in core the usage of a new optional backend callback for TMR notifications. The core calls it before core waits for completion of aborted commands (params: se_dev, TMR type, and list of se_cmds aborted by this TMR).
Of course other backends than tcmu can use this part of the series also to implement their own TMR notification if necessary.

The further six patches implement the TMR notify callback for tcmu. The new configFS attribute tmr_notification allows to switch on TMR messages on the cmd ring. The default of the attribute is the old behavior without TMR info on the ring, but with following changes:
- if tcmu receives an already aborted command, it immediately rejects it. So it will never appear in userspace.
- if tcmu finds, that according to TMR notification a cmd on the qfull_queue was aborted, tcmu removes it from qfull_queue and completes it immediately. So userspace will not 'see' those commands.

When attribute tmr_notification is set to 1, tcmu additionally prepares a list of cmd_ids from those commands, that are aborted by the TMR and are active in cmd ring (not timed out). This list together with the TMR type is either immediately written to cmd ring (new TMR entry type) or queued in a separate tmr queue if ring space is too small.
TMRs in the tmr queue do not time out. If ring space becomes available, tcmu moves TMRs from tmr queue to ring with higher priority than cmds from qfull queue.

Quoted 10 lines

This mechanism makes sure that userspace receives TMR notifications as early as possible. Userspace can use the list of cmd_ids attached to the TMR notification to identify aborted commands from its list of received and not yet completed commands. In case userspace has meanwhile completed some of the cmd_ids on the list, it can just ignore these cmd_ids. A possible new command having the same cmd_id as one of the aborted commands will always appear on the ring after the TMR notification.

Thanks for all the work on this.

#10 lore raw
In reply to Bodo Stroesser
Quoted 9 lines · Bodo Stroesser

This patch series is made on top of Martin's for-next branch.

ChangeLog:

v2: in patch "scsi: target: tcmu: Implement tmr_notify callback"
changed new comment's style from "// ..." to "/* ... */" and correctly use "/** " for function doc.

[...]

Applied to 5.9/scsi-queue, thanks!

[1/8] scsi: target: Modify core_tmr_abort_task()

      https://git.kernel.org/mkp/scsi/c/f5e2714ad1a6
[2/8] scsi: target: Add tmr_notify backend function
      https://git.kernel.org/mkp/scsi/c/2e45a1a9c75d
[3/8] scsi: target: tcmu: Use priv pointer in se_cmd
      https://git.kernel.org/mkp/scsi/c/a35129024e88
[4/8] scsi: target: tcmu: Do not queue aborted commands
      https://git.kernel.org/mkp/scsi/c/c96849276211
[5/8] scsi: target: tcmu: Factor out new helper ring_insert_padding
      https://git.kernel.org/mkp/scsi/c/3d3f9d56a570
[6/8] scsi: target: tcmu: Fix and simplify timeout handling
      https://git.kernel.org/mkp/scsi/c/ed212ca87897
[7/8] scsi: target: tcmu: Implement tmr_notify callback
      https://git.kernel.org/mkp/scsi/c/bc2d214af5db
[8/8] scsi: target: tcmu: Make TMR notification optional
      https://git.kernel.org/mkp/scsi/c/59526d7a187f
Signature · 1 line
Martin K. Petersen	Oracle Linux Engineering