[net-next,v3,06/11] net: ethernet: ti: cpts: move tx timestamp processing to ptp worker only
diff mbox series

Message ID 20200320194244.4703-7-grygorii.strashko@ti.com
State New
Headers show
Series
  • net: ethernet: ti: cpts: add irq and HW_TS_PUSH events
Related show

Commit Message

Grygorii Strashko March 20, 2020, 7:42 p.m. UTC
Now the tx timestamp processing happens from different contexts - softirq
and thread/PTP worker. Enabling IRQ will add one more hard_irq context.
This makes over all defered TX timestamp processing and locking
overcomplicated. Move tx timestamp processing to PTP worker always instead.

napi_rx->cpts_tx_timestamp
 if ptp_packet then
    push to txq
    ptp_schedule_worker()

do_aux_work->cpts_overflow_check
 cpts_process_events()

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
---
 drivers/net/ethernet/ti/cpts.c | 165 +++++++++++++++++++--------------
 1 file changed, 94 insertions(+), 71 deletions(-)

Comments

Richard Cochran March 26, 2020, 2:29 p.m. UTC | #1
On Fri, Mar 20, 2020 at 09:42:39PM +0200, Grygorii Strashko wrote:
> Now the tx timestamp processing happens from different contexts - softirq
> and thread/PTP worker. Enabling IRQ will add one more hard_irq context.
> This makes over all defered TX timestamp processing and locking
> overcomplicated. Move tx timestamp processing to PTP worker always instead.
> 
> napi_rx->cpts_tx_timestamp
>  if ptp_packet then
>     push to txq
>     ptp_schedule_worker()
> 
> do_aux_work->cpts_overflow_check
>  cpts_process_events()

Since cpts_overflow_check() is performing new functions, consider
renaming it to match.

Thanks,
Richard
Grygorii Strashko March 26, 2020, 8:17 p.m. UTC | #2
On 26/03/2020 16:29, Richard Cochran wrote:
> On Fri, Mar 20, 2020 at 09:42:39PM +0200, Grygorii Strashko wrote:
>> Now the tx timestamp processing happens from different contexts - softirq
>> and thread/PTP worker. Enabling IRQ will add one more hard_irq context.
>> This makes over all defered TX timestamp processing and locking
>> overcomplicated. Move tx timestamp processing to PTP worker always instead.
>>
>> napi_rx->cpts_tx_timestamp
>>   if ptp_packet then
>>      push to txq
>>      ptp_schedule_worker()
>>
>> do_aux_work->cpts_overflow_check
>>   cpts_process_events()
> 
> Since cpts_overflow_check() is performing new functions, consider
> renaming it to match.

It's still performs overflow check

static long cpts_overflow_check(struct ptp_clock_info *ptp)
{
...

	mutex_lock(&cpts->ptp_clk_mutex);

	cpts_update_cur_time(cpts, -1, NULL);
^^^ here
	ns = timecounter_read(&cpts->tc);


	cpts_process_events(cpts);

What name would you suggest?

Thank you.

Patch
diff mbox series

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index f7ce482ad09c..10061e17d7b4 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -21,6 +21,8 @@ 
 #include "cpts.h"
 
 #define CPTS_SKB_TX_WORK_TIMEOUT 1 /* jiffies */
+#define CPTS_SKB_RX_TX_TMO 100 /*ms */
+#define CPTS_EVENT_RX_TX_TIMEOUT (100) /* ms */
 
 struct cpts_skb_cb_data {
 	u32 skb_mtype_seqid;
@@ -92,46 +94,6 @@  static void cpts_purge_txq(struct cpts *cpts)
 		dev_dbg(cpts->dev, "txq cleaned up %d\n", removed);
 }
 
-static bool cpts_match_tx_ts(struct cpts *cpts, struct cpts_event *event)
-{
-	struct sk_buff *skb, *tmp;
-	bool found = false;
-	u32 mtype_seqid;
-
-	mtype_seqid = event->high &
-		      ((MESSAGE_TYPE_MASK << MESSAGE_TYPE_SHIFT) |
-		       (SEQUENCE_ID_MASK << SEQUENCE_ID_SHIFT) |
-		       (EVENT_TYPE_MASK << EVENT_TYPE_SHIFT));
-
-	/* no need to grab txq.lock as access is always done under cpts->lock */
-	skb_queue_walk_safe(&cpts->txq, skb, tmp) {
-		struct skb_shared_hwtstamps ssh;
-		struct cpts_skb_cb_data *skb_cb =
-					(struct cpts_skb_cb_data *)skb->cb;
-
-		if (mtype_seqid == skb_cb->skb_mtype_seqid) {
-			memset(&ssh, 0, sizeof(ssh));
-			ssh.hwtstamp = ns_to_ktime(event->timestamp);
-			skb_tstamp_tx(skb, &ssh);
-			found = true;
-			__skb_unlink(skb, &cpts->txq);
-			dev_consume_skb_any(skb);
-			dev_dbg(cpts->dev, "match tx timestamp mtype_seqid %08x\n",
-				mtype_seqid);
-			break;
-		}
-
-		if (time_after(jiffies, skb_cb->tmo)) {
-			/* timeout any expired skbs over 1s */
-			dev_dbg(cpts->dev, "expiring tx timestamp from txq\n");
-			__skb_unlink(skb, &cpts->txq);
-			dev_consume_skb_any(skb);
-		}
-	}
-
-	return found;
-}
-
 /*
  * Returns zero if matching event type was found.
  */
@@ -151,7 +113,6 @@  static int cpts_fifo_read(struct cpts *cpts, int match)
 		}
 
 		event = list_first_entry(&cpts->pool, struct cpts_event, list);
-		event->tmo = jiffies + 2;
 		event->high = hi;
 		event->low = lo;
 		event->timestamp = timecounter_cyc2time(&cpts->tc, event->low);
@@ -169,14 +130,10 @@  static int cpts_fifo_read(struct cpts *cpts, int match)
 			}
 			break;
 		case CPTS_EV_TX:
-			if (cpts_match_tx_ts(cpts, event)) {
-				/* if the new event matches an existing skb,
-				 * then don't queue it
-				 */
-				break;
-			}
-			/* fall through */
 		case CPTS_EV_RX:
+			event->tmo = jiffies +
+				msecs_to_jiffies(CPTS_EVENT_RX_TX_TIMEOUT);
+
 			list_del_init(&event->list);
 			list_add_tail(&event->list, &cpts->events);
 			break;
@@ -297,6 +254,84 @@  static int cpts_ptp_enable(struct ptp_clock_info *ptp,
 	return -EOPNOTSUPP;
 }
 
+static bool cpts_match_tx_ts(struct cpts *cpts, struct cpts_event *event)
+{
+	struct sk_buff_head txq_list;
+	struct sk_buff *skb, *tmp;
+	unsigned long flags;
+	bool found = false;
+	u32 mtype_seqid;
+
+	mtype_seqid = event->high &
+		      ((MESSAGE_TYPE_MASK << MESSAGE_TYPE_SHIFT) |
+		       (SEQUENCE_ID_MASK << SEQUENCE_ID_SHIFT) |
+		       (EVENT_TYPE_MASK << EVENT_TYPE_SHIFT));
+
+	__skb_queue_head_init(&txq_list);
+
+	spin_lock_irqsave(&cpts->txq.lock, flags);
+	skb_queue_splice_init(&cpts->txq, &txq_list);
+	spin_unlock_irqrestore(&cpts->txq.lock, flags);
+
+	skb_queue_walk_safe(&txq_list, skb, tmp) {
+		struct skb_shared_hwtstamps ssh;
+		struct cpts_skb_cb_data *skb_cb =
+					(struct cpts_skb_cb_data *)skb->cb;
+
+		if (mtype_seqid == skb_cb->skb_mtype_seqid) {
+			memset(&ssh, 0, sizeof(ssh));
+			ssh.hwtstamp = ns_to_ktime(event->timestamp);
+			skb_tstamp_tx(skb, &ssh);
+			found = true;
+			__skb_unlink(skb, &txq_list);
+			dev_consume_skb_any(skb);
+			dev_dbg(cpts->dev, "match tx timestamp mtype_seqid %08x\n",
+				mtype_seqid);
+			break;
+		}
+
+		if (time_after(jiffies, skb_cb->tmo)) {
+			/* timeout any expired skbs over 1s */
+			dev_dbg(cpts->dev, "expiring tx timestamp from txq\n");
+			__skb_unlink(skb, &txq_list);
+			dev_consume_skb_any(skb);
+		}
+	}
+
+	spin_lock_irqsave(&cpts->txq.lock, flags);
+	skb_queue_splice(&txq_list, &cpts->txq);
+	spin_unlock_irqrestore(&cpts->txq.lock, flags);
+
+	return found;
+}
+
+static void cpts_process_events(struct cpts *cpts)
+{
+	struct list_head *this, *next;
+	struct cpts_event *event;
+	LIST_HEAD(events_free);
+	unsigned long flags;
+	LIST_HEAD(events);
+
+	spin_lock_irqsave(&cpts->lock, flags);
+	list_splice_init(&cpts->events, &events);
+	spin_unlock_irqrestore(&cpts->lock, flags);
+
+	list_for_each_safe(this, next, &events) {
+		event = list_entry(this, struct cpts_event, list);
+		if (cpts_match_tx_ts(cpts, event) ||
+		    time_after(jiffies, event->tmo)) {
+			list_del_init(&event->list);
+			list_add(&event->list, &events_free);
+		}
+	}
+
+	spin_lock_irqsave(&cpts->lock, flags);
+	list_splice_tail(&events, &cpts->events);
+	list_splice_tail(&events_free, &cpts->pool);
+	spin_unlock_irqrestore(&cpts->lock, flags);
+}
+
 static long cpts_overflow_check(struct ptp_clock_info *ptp)
 {
 	struct cpts *cpts = container_of(ptp, struct cpts, info);
@@ -305,17 +340,20 @@  static long cpts_overflow_check(struct ptp_clock_info *ptp)
 	u64 ns;
 
 	spin_lock_irqsave(&cpts->lock, flags);
-
 	cpts_update_cur_time(cpts, -1, NULL);
+	spin_unlock_irqrestore(&cpts->lock, flags);
 
 	ns = timecounter_read(&cpts->tc);
 
+	cpts_process_events(cpts);
+
+	spin_lock_irqsave(&cpts->txq.lock, flags);
 	if (!skb_queue_empty(&cpts->txq)) {
 		cpts_purge_txq(cpts);
 		if (!skb_queue_empty(&cpts->txq))
 			delay = CPTS_SKB_TX_WORK_TIMEOUT;
 	}
-	spin_unlock_irqrestore(&cpts->lock, flags);
+	spin_unlock_irqrestore(&cpts->txq.lock, flags);
 
 	dev_dbg(cpts->dev, "cpts overflow check at %lld\n", ns);
 	return (long)delay;
@@ -409,19 +447,6 @@  static u64 cpts_find_ts(struct cpts *cpts, struct sk_buff *skb,
 			break;
 		}
 	}
-
-	if (ev_type == CPTS_EV_TX && !ns) {
-		struct cpts_skb_cb_data *skb_cb =
-				(struct cpts_skb_cb_data *)skb->cb;
-		/* Not found, add frame to queue for processing later.
-		 * The periodic FIFO check will handle this.
-		 */
-		skb_get(skb);
-		/* get the timestamp for timeouts */
-		skb_cb->tmo = jiffies + msecs_to_jiffies(100);
-		__skb_queue_tail(&cpts->txq, skb);
-		ptp_schedule_worker(cpts->clock, 0);
-	}
 	spin_unlock_irqrestore(&cpts->lock, flags);
 
 	return ns;
@@ -455,9 +480,7 @@  EXPORT_SYMBOL_GPL(cpts_rx_timestamp);
 void cpts_tx_timestamp(struct cpts *cpts, struct sk_buff *skb)
 {
 	struct cpts_skb_cb_data *skb_cb = (struct cpts_skb_cb_data *)skb->cb;
-	struct skb_shared_hwtstamps ssh;
 	int ret;
-	u64 ns;
 
 	if (!(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
 		return;
@@ -471,12 +494,12 @@  void cpts_tx_timestamp(struct cpts *cpts, struct sk_buff *skb)
 	dev_dbg(cpts->dev, "%s mtype seqid %08x\n",
 		__func__, skb_cb->skb_mtype_seqid);
 
-	ns = cpts_find_ts(cpts, skb, CPTS_EV_TX, skb_cb->skb_mtype_seqid);
-	if (!ns)
-		return;
-	memset(&ssh, 0, sizeof(ssh));
-	ssh.hwtstamp = ns_to_ktime(ns);
-	skb_tstamp_tx(skb, &ssh);
+	/* Always defer TX TS processing to PTP worker */
+	skb_get(skb);
+	/* get the timestamp for timeouts */
+	skb_cb->tmo = jiffies + msecs_to_jiffies(CPTS_SKB_RX_TX_TMO);
+	skb_queue_tail(&cpts->txq, skb);
+	ptp_schedule_worker(cpts->clock, 0);
 }
 EXPORT_SYMBOL_GPL(cpts_tx_timestamp);