[PATCH 0/2] net: ethernet: ti: cpts: fix tx timestamping timeout

[PATCH 0/2] net: ethernet: ti: cpts: fix tx timestamping timeout

[Grygorii Strashko]

With the low Ethernet connection speed cpdma notification about packet
processing can be received before CPTS TX timestamp event, which is set
when packet actually left CPSW while cpdma notification is sent when packet
pushed in CPSW fifo. As result, when connection is slow and CPU is fast
enough TX timestamping is not working properly.
Issue was discovered using timestamping tool on am57x boards with Ethernet link
speed forced to 100M and on am335x-evm with Ethernet link speed forced to 10M.

This series fixes it by introducing TX SKB queue to store PTP SKBs for which
Ethernet Transmit Event hasn't been received yet and then re-check this queue
with new Ethernet Transmit Events by scheduling CPTS overflow
work more often until TX SKB queue is not empty.
The internal CPTS workqueues are also converted to use kthread_worker.

Grygorii Strashko (2):
  net: ethernet: ti: cpts: convert to use kthread_worker
  net: ethernet: ti: cpts: fix tx timestamping timeout

 drivers/net/ethernet/ti/cpts.c | 112 +++++++++++++++++++++++++++++++++++++----
 drivers/net/ethernet/ti/cpts.h |   5 +-
 2 files changed, 106 insertions(+), 11 deletions(-)

-- 
2.10.1

[PATCH 1/2] net: ethernet: ti: cpts: convert to use kthread_worker

[Grygorii Strashko]

There could be significant delay in CPTS work schedule under high system
load and on -RT which could cause CPTS misbehavior due to internal counter
overflow. Usage of own kthread_worker allows to avoid such kind of issues
and makes it possible to tune priority of CPTS kthread_worker thread on -RT
(thread name "cpts").

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
---
 drivers/net/ethernet/ti/cpts.c | 23 +++++++++++++++++------
 drivers/net/ethernet/ti/cpts.h |  4 +++-
 2 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index 32279d2..6a520ae 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -20,12 +20,12 @@
 #include <linux/err.h>
 #include <linux/if.h>
 #include <linux/hrtimer.h>
+#include <linux/kthread.h>
 #include <linux/module.h>
 #include <linux/net_tstamp.h>
 #include <linux/ptp_classify.h>
 #include <linux/time.h>
 #include <linux/uaccess.h>
-#include <linux/workqueue.h>
 #include <linux/if_ether.h>
 #include <linux/if_vlan.h>
 
@@ -238,14 +238,15 @@ static struct ptp_clock_info cpts_info = {
 	.enable		= cpts_ptp_enable,
 };
 
-static void cpts_overflow_check(struct work_struct *work)
+static void cpts_overflow_check(struct kthread_work *work)
 {
 	struct timespec64 ts;
 	struct cpts *cpts = container_of(work, struct cpts, overflow_work.work);
 
 	cpts_ptp_gettime(&cpts->info, &ts);
 	pr_debug("cpts overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
-	schedule_delayed_work(&cpts->overflow_work, cpts->ov_check_period);
+	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work,
+				   cpts->ov_check_period);
 }
 
 static int cpts_match(struct sk_buff *skb, unsigned int ptp_class,
@@ -378,7 +379,8 @@ int cpts_register(struct cpts *cpts)
 	}
 	cpts->phc_index = ptp_clock_index(cpts->clock);
 
-	schedule_delayed_work(&cpts->overflow_work, cpts->ov_check_period);
+	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work,
+				   cpts->ov_check_period);
 	return 0;
 
 err_ptp:
@@ -392,7 +394,7 @@ void cpts_unregister(struct cpts *cpts)
 	if (WARN_ON(!cpts->clock))
 		return;
 
-	cancel_delayed_work_sync(&cpts->overflow_work);
+	kthread_cancel_delayed_work_sync(&cpts->overflow_work);
 
 	ptp_clock_unregister(cpts->clock);
 	cpts->clock = NULL;
@@ -476,7 +478,6 @@ struct cpts *cpts_create(struct device *dev, void __iomem *regs,
 	cpts->dev = dev;
 	cpts->reg = (struct cpsw_cpts __iomem *)regs;
 	spin_lock_init(&cpts->lock);
-	INIT_DELAYED_WORK(&cpts->overflow_work, cpts_overflow_check);
 
 	ret = cpts_of_parse(cpts, node);
 	if (ret)
@@ -488,6 +489,14 @@ struct cpts *cpts_create(struct device *dev, void __iomem *regs,
 		return ERR_PTR(PTR_ERR(cpts->refclk));
 	}
 
+	kthread_init_delayed_work(&cpts->overflow_work, cpts_overflow_check);
+	cpts->kworker = kthread_create_worker(0, "cpts");
+	if (IS_ERR(cpts->kworker)) {
+		dev_err(dev, "failed to create cpts overflow_work task %ld\n",
+			PTR_ERR(cpts->kworker));
+		return ERR_CAST(cpts->kworker);
+	}
+
 	clk_prepare(cpts->refclk);
 
 	cpts->cc.read = cpts_systim_read;
@@ -513,6 +522,8 @@ void cpts_release(struct cpts *cpts)
 		return;
 
 	clk_unprepare(cpts->refclk);
+
+	kthread_destroy_worker(cpts->kworker);
 }
 EXPORT_SYMBOL_GPL(cpts_release);
 
diff --git a/drivers/net/ethernet/ti/cpts.h b/drivers/net/ethernet/ti/cpts.h
index 01ea82b..e8128e8 100644
--- a/drivers/net/ethernet/ti/cpts.h
+++ b/drivers/net/ethernet/ti/cpts.h
@@ -27,6 +27,7 @@
 #include <linux/clocksource.h>
 #include <linux/device.h>
 #include <linux/list.h>
+#include <linux/kthread.h>
 #include <linux/of.h>
 #include <linux/ptp_clock_kernel.h>
 #include <linux/skbuff.h>
@@ -119,13 +120,14 @@ struct cpts {
 	u32 cc_mult; /* for the nominal frequency */
 	struct cyclecounter cc;
 	struct timecounter tc;
-	struct delayed_work overflow_work;
 	int phc_index;
 	struct clk *refclk;
 	struct list_head events;
 	struct list_head pool;
 	struct cpts_event pool_data[CPTS_MAX_EVENTS];
 	unsigned long ov_check_period;
+	struct kthread_worker *kworker;
+	struct kthread_delayed_work overflow_work;
 };
 
 void cpts_rx_timestamp(struct cpts *cpts, struct sk_buff *skb);
-- 
2.10.1

Re: [PATCH 1/2] net: ethernet: ti: cpts: convert to use kthread_worker

[Richard Cochran]

On Fri, Jul 21, 2017 at 06:49:42PM -0500, Grygorii Strashko wrote:
> There could be significant delay in CPTS work schedule under high system
> load and on -RT which could cause CPTS misbehavior due to internal counter
> overflow. Usage of own kthread_worker allows to avoid such kind of issues
> and makes it possible to tune priority of CPTS kthread_worker thread on -RT
> (thread name "cpts").

I have also seen excessive delays in the time stamp work from the
dp83640 under certain loads.  Can we please implement this time stamp
kthread_worker in the PHC subsystem?  That way, the facility can be
used by other drivers without code duplication.

Thanks,
Richard

Re: [PATCH 1/2] net: ethernet: ti: cpts: convert to use kthread_worker

[Grygorii Strashko]

Hi Richard,

On 07/22/2017 12:29 AM, Richard Cochran wrote:
> On Fri, Jul 21, 2017 at 06:49:42PM -0500, Grygorii Strashko wrote:
>> There could be significant delay in CPTS work schedule under high system
>> load and on -RT which could cause CPTS misbehavior due to internal counter
>> overflow. Usage of own kthread_worker allows to avoid such kind of issues
>> and makes it possible to tune priority of CPTS kthread_worker thread on -RT
>> (thread name "cpts").
> 
> I have also seen excessive delays in the time stamp work from the
> dp83640 under certain loads.  Can we please implement this time stamp
> kthread_worker in the PHC subsystem?  That way, the facility can be
> used by other drivers without code duplication.

Below if pure TBD/RFC version of patch which add kthread worker to PTP core.
I'm sending it to get you opinion about implementation in general, before 
continue with more changes. Pls, take a look if you have time?
- are you ok with names (API, callbacks, ptp structs members)?

I can prepare, update and resend proper patches tom if feedback is positive.
I also can convert dp83640 driver to use new feature, but I can't test it.

>From 48212c4564ae38d633d95723b1f4616dee195b47 Mon Sep 17 00:00:00 2001
From: Grygorii Strashko <grygorii.strashko@ti.com>
Date: Mon, 24 Jul 2017 19:19:50 -0500
Subject: [PATCH] [rfc] ptp: add auxiliary worker

Add auxiliary kthread worker to PTP core so drivers can
re-use it and benefit from common implementation which is RT
friendly... TBD

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
---
 drivers/net/ethernet/ti/cpts.c   | 53 +++++++++++++++-------------------------
 drivers/net/ethernet/ti/cpts.h   |  2 --
 drivers/ptp/ptp_clock.c          | 42 +++++++++++++++++++++++++++++++
 drivers/ptp/ptp_private.h        |  3 +++
 include/linux/ptp_clock_kernel.h | 15 ++++++++++++
 5 files changed, 80 insertions(+), 35 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index 746dd1d..e05a1b4 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -280,23 +280,9 @@ static int cpts_ptp_enable(struct ptp_clock_info *ptp,
 	return -EOPNOTSUPP;
 }
 
-static struct ptp_clock_info cpts_info = {
-	.owner		= THIS_MODULE,
-	.name		= "CTPS timer",
-	.max_adj	= 1000000,
-	.n_ext_ts	= 0,
-	.n_pins		= 0,
-	.pps		= 0,
-	.adjfreq	= cpts_ptp_adjfreq,
-	.adjtime	= cpts_ptp_adjtime,
-	.gettime64	= cpts_ptp_gettime,
-	.settime64	= cpts_ptp_settime,
-	.enable		= cpts_ptp_enable,
-};
-
-static void cpts_overflow_check(struct kthread_work *work)
+static long cpts_overflow_check(struct ptp_clock_info *ptp)
 {
-	struct cpts *cpts = container_of(work, struct cpts, overflow_work.work);
+	struct cpts *cpts = container_of(ptp, struct cpts, info);
 	unsigned long delay = cpts->ov_check_period;
 	struct timespec64 ts;
 	unsigned long flags;
@@ -309,9 +295,24 @@ static void cpts_overflow_check(struct kthread_work *work)
 	spin_unlock_irqrestore(&cpts->lock, flags);
 
 	pr_debug("cpts overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
-	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work, delay);
+	return (long)delay;
 }
 
+static struct ptp_clock_info cpts_info = {
+	.owner		= THIS_MODULE,
+	.name		= "CTPS timer",
+	.max_adj	= 1000000,
+	.n_ext_ts	= 0,
+	.n_pins		= 0,
+	.pps		= 0,
+	.adjfreq	= cpts_ptp_adjfreq,
+	.adjtime	= cpts_ptp_adjtime,
+	.gettime64	= cpts_ptp_gettime,
+	.settime64	= cpts_ptp_settime,
+	.enable		= cpts_ptp_enable,
+	.do_aux_work	= cpts_overflow_check,
+};
+
 static int cpts_match(struct sk_buff *skb, unsigned int ptp_class,
 		      u16 ts_seqid, u8 ts_msgtype)
 {
@@ -392,8 +393,7 @@ static u64 cpts_find_ts(struct cpts *cpts, struct sk_buff *skb, int ev_type)
 		/* get the timestamp for timeouts */
 		skb_cb->tmo = jiffies + msecs_to_jiffies(100);
 		__skb_queue_tail(&cpts->txq, skb);
-		kthread_mod_delayed_work(cpts->kworker,
-					 &cpts->overflow_work, 0);
+		ptp_schedule_worker(cpts->clock, 0);
 	}
 	spin_unlock_irqrestore(&cpts->lock, flags);
 
@@ -457,8 +457,7 @@ int cpts_register(struct cpts *cpts)
 	}
 	cpts->phc_index = ptp_clock_index(cpts->clock);
 
-	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work,
-				   cpts->ov_check_period);
+	ptp_schedule_worker(cpts->clock, cpts->ov_check_period);
 	return 0;
 
 err_ptp:
@@ -472,8 +471,6 @@ void cpts_unregister(struct cpts *cpts)
 	if (WARN_ON(!cpts->clock))
 		return;
 
-	kthread_cancel_delayed_work_sync(&cpts->overflow_work);
-
 	ptp_clock_unregister(cpts->clock);
 	cpts->clock = NULL;
 
@@ -570,14 +567,6 @@ struct cpts *cpts_create(struct device *dev, void __iomem *regs,
 		return ERR_PTR(PTR_ERR(cpts->refclk));
 	}
 
-	kthread_init_delayed_work(&cpts->overflow_work, cpts_overflow_check);
-	cpts->kworker = kthread_create_worker(0, "cpts");
-	if (IS_ERR(cpts->kworker)) {
-		dev_err(dev, "failed to create cpts overflow_work task %ld\n",
-			PTR_ERR(cpts->kworker));
-		return ERR_CAST(cpts->kworker);
-	}
-
 	clk_prepare(cpts->refclk);
 
 	cpts->cc.read = cpts_systim_read;
@@ -603,8 +592,6 @@ void cpts_release(struct cpts *cpts)
 		return;
 
 	clk_unprepare(cpts->refclk);
-
-	kthread_destroy_worker(cpts->kworker);
 }
 EXPORT_SYMBOL_GPL(cpts_release);
 
diff --git a/drivers/net/ethernet/ti/cpts.h b/drivers/net/ethernet/ti/cpts.h
index e94b829..a9f4eec 100644
--- a/drivers/net/ethernet/ti/cpts.h
+++ b/drivers/net/ethernet/ti/cpts.h
@@ -126,8 +126,6 @@ struct cpts {
 	struct list_head pool;
 	struct cpts_event pool_data[CPTS_MAX_EVENTS];
 	unsigned long ov_check_period;
-	struct kthread_worker *kworker;
-	struct kthread_delayed_work overflow_work;
 	struct sk_buff_head txq;
 };
 
diff --git a/drivers/ptp/ptp_clock.c b/drivers/ptp/ptp_clock.c
index b774357..9bea56a 100644
--- a/drivers/ptp/ptp_clock.c
+++ b/drivers/ptp/ptp_clock.c
@@ -28,6 +28,7 @@
 #include <linux/slab.h>
 #include <linux/syscalls.h>
 #include <linux/uaccess.h>
+#include <uapi/linux/sched/types.h>
 
 #include "ptp_private.h"
 
@@ -184,6 +185,19 @@ static void delete_ptp_clock(struct posix_clock *pc)
 	kfree(ptp);
 }
 
+static void ptp_aux_kworker(struct kthread_work *work)
+{
+	struct ptp_clock *ptp = container_of(work, struct ptp_clock,
+					     aux_work.work);
+	struct ptp_clock_info *info = ptp->info;
+	long delay;
+
+	delay = info->do_aux_work(info);
+
+	if (delay >= 0)
+		kthread_queue_delayed_work(ptp->kworker, &ptp->aux_work, delay);
+}
+
 /* public interface */
 
 struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info,
@@ -217,6 +231,23 @@ struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info,
 	mutex_init(&ptp->pincfg_mux);
 	init_waitqueue_head(&ptp->tsev_wq);
 
+	if (ptp->info->do_aux_work) {
+		struct sched_param param = {
+			.sched_priority = MAX_RT_PRIO - 1 };
+
+		kthread_init_delayed_work(&ptp->aux_work, ptp_aux_kworker);
+		ptp->kworker = kthread_create_worker(0, info->name);
+		if (IS_ERR(ptp->kworker)) {
+			pr_err("failed to create ptp aux_worker task %ld\n",
+			       PTR_ERR(ptp->kworker));
+			return ERR_CAST(ptp->kworker);
+		}
+		err = sched_setscheduler_nocheck(ptp->kworker->task,
+						 SCHED_FIFO, &param);
+		if (err)
+			pr_err("sched_setscheduler_nocheck err %d\n", err);
+	}
+
 	err = ptp_populate_pin_groups(ptp);
 	if (err)
 		goto no_pin_groups;
@@ -274,6 +305,9 @@ int ptp_clock_unregister(struct ptp_clock *ptp)
 	ptp->defunct = 1;
 	wake_up_interruptible(&ptp->tsev_wq);
 
+	kthread_cancel_delayed_work_sync(&ptp->aux_work);
+	kthread_destroy_worker(ptp->kworker);
+
 	/* Release the clock's resources. */
 	if (ptp->pps_source)
 		pps_unregister_source(ptp->pps_source);
@@ -339,6 +373,14 @@ int ptp_find_pin(struct ptp_clock *ptp,
 }
 EXPORT_SYMBOL(ptp_find_pin);
 
+int ptp_schedule_worker(struct ptp_clock *ptp, unsigned long delay)
+{
+	if (!ptp->kworker)
+		return -EOPNOTSUPP;
+	return kthread_mod_delayed_work(ptp->kworker, &ptp->aux_work, delay);
+}
+EXPORT_SYMBOL(ptp_schedule_worker);
+
 /* module operations */
 
 static void __exit ptp_exit(void)
diff --git a/drivers/ptp/ptp_private.h b/drivers/ptp/ptp_private.h
index d958889..b86f1bf 100644
--- a/drivers/ptp/ptp_private.h
+++ b/drivers/ptp/ptp_private.h
@@ -22,6 +22,7 @@
 
 #include <linux/cdev.h>
 #include <linux/device.h>
+#include <linux/kthread.h>
 #include <linux/mutex.h>
 #include <linux/posix-clock.h>
 #include <linux/ptp_clock.h>
@@ -56,6 +57,8 @@ struct ptp_clock {
 	struct attribute_group pin_attr_group;
 	/* 1st entry is a pointer to the real group, 2nd is NULL terminator */
 	const struct attribute_group *pin_attr_groups[2];
+	struct kthread_worker *kworker;
+	struct kthread_delayed_work aux_work;
 };
 
 /*
diff --git a/include/linux/ptp_clock_kernel.h b/include/linux/ptp_clock_kernel.h
index a026bfd..1b8832a 100644
--- a/include/linux/ptp_clock_kernel.h
+++ b/include/linux/ptp_clock_kernel.h
@@ -98,6 +98,10 @@ struct system_device_crosststamp;
  *            parameter pin: index of the pin in question.
  *            parameter func: the desired function to use.
  *            parameter chan: the function channel index to use.
+ * @do_work:  Request driver to perform auxiliary (periodic) operations
+ *	      Driver should return delay of the next auxiliary work scheduling
+ *	      time (>=0) or negative value in case further scheduling
+ *	      is not required.
  *
  * Drivers should embed their ptp_clock_info within a private
  * structure, obtaining a reference to it using container_of().
@@ -126,6 +130,7 @@ struct ptp_clock_info {
 		      struct ptp_clock_request *request, int on);
 	int (*verify)(struct ptp_clock_info *ptp, unsigned int pin,
 		      enum ptp_pin_function func, unsigned int chan);
+	long (*do_aux_work)(struct ptp_clock_info *ptp);
 };
 
 struct ptp_clock;
@@ -211,6 +216,16 @@ extern int ptp_clock_index(struct ptp_clock *ptp);
 int ptp_find_pin(struct ptp_clock *ptp,
 		 enum ptp_pin_function func, unsigned int chan);
 
+/**
+ * ptp_schedule_worker() - schedule ptp auxiliary work
+ *
+ * @ptp:    The clock obtained from ptp_clock_register().
+ * @delay:  number of jiffies to wait before queuing
+ *          See kthread_queue_delayed_work() for more info.
+ */
+
+int ptp_schedule_worker(struct ptp_clock *ptp, unsigned long delay);
+
 #else
 static inline struct ptp_clock *ptp_clock_register(struct ptp_clock_info *info,
 						   struct device *parent)
-- 
2.10.1

-- 
regards,
-grygorii

Re: [PATCH 1/2] net: ethernet: ti: cpts: convert to use kthread_worker

[Richard Cochran]

On Mon, Jul 24, 2017 at 07:34:38PM -0500, Grygorii Strashko wrote:
> Below if pure TBD/RFC version of patch which add kthread worker to PTP core.
> I'm sending it to get you opinion about implementation in general, before 
> continue with more changes. Pls, take a look if you have time?
> - are you ok with names (API, callbacks, ptp structs members)?

The API and naming looks good to me.
 
> I can prepare, update and resend proper patches tom if feedback is positive.

Please do.

> I also can convert dp83640 driver to use new feature, but I can't test it.

No need for that.  It would be enough to have cpts as the first user
and example.
 
> +	if (ptp->info->do_aux_work) {
> +		struct sched_param param = {
> +			.sched_priority = MAX_RT_PRIO - 1 };
> +
> +		kthread_init_delayed_work(&ptp->aux_work, ptp_aux_kworker);
> +		ptp->kworker = kthread_create_worker(0, info->name);
> +		if (IS_ERR(ptp->kworker)) {
> +			pr_err("failed to create ptp aux_worker task %ld\n",
> +			       PTR_ERR(ptp->kworker));
> +			return ERR_CAST(ptp->kworker);
> +		}
> +		err = sched_setscheduler_nocheck(ptp->kworker->task,
> +						 SCHED_FIFO, &param);

I think we should not hard code the scheduler and priority here but
rather leave it to the sysadmin to configure these using chrt(1).
After all, a normal work item is has served just in many situations.

> +		if (err)
> +			pr_err("sched_setscheduler_nocheck err %d\n", err);
> +	}
> +
>  	err = ptp_populate_pin_groups(ptp);
>  	if (err)
>  		goto no_pin_groups;
> @@ -274,6 +305,9 @@ int ptp_clock_unregister(struct ptp_clock *ptp)
>  	ptp->defunct = 1;
>  	wake_up_interruptible(&ptp->tsev_wq);
>  
> +	kthread_cancel_delayed_work_sync(&ptp->aux_work);
> +	kthread_destroy_worker(ptp->kworker);

These can't be called unconditionally.

>  	/* Release the clock's resources. */
>  	if (ptp->pps_source)
>  		pps_unregister_source(ptp->pps_source);

Thanks,
Richard

[PATCH 2/2] net: ethernet: ti: cpts: fix tx timestamping timeout

[Grygorii Strashko]

With the low speed Ethernet connection CPDMA notification about packet
processing can be received before CPTS TX timestamp event, which is set
when packet actually left CPSW while cpdma notification is sent when packet
pushed in CPSW fifo.  As result, when connection is slow and CPU is fast
enough TX timestamping is not working properly.

Fix it, by introducing TX SKB queue to store PTP SKBs for which Ethernet
Transmit Event hasn't been received yet and then re-check this queue
with new Ethernet Transmit Events by scheduling CPTS overflow
work more often (every 1 jiffies) until TX SKB queue is not empty.

Side effect of this change is:
 - User space tools require to take into account possible delay in TX
timestamp processing (for example ptp4l works with tx_timestamp_timeout=400
under net traffic and tx_timestamp_timeout=25 in idle).

Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
---
 drivers/net/ethernet/ti/cpts.c | 93 +++++++++++++++++++++++++++++++++++++++---
 drivers/net/ethernet/ti/cpts.h |  1 +
 2 files changed, 88 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpts.c b/drivers/net/ethernet/ti/cpts.c
index 6a520ae..746dd1d 100644
--- a/drivers/net/ethernet/ti/cpts.c
+++ b/drivers/net/ethernet/ti/cpts.c
@@ -31,9 +31,18 @@
 
 #include "cpts.h"
 
+#define CPTS_SKB_TX_WORK_TIMEOUT 1 /* jiffies */
+
+struct cpts_skb_cb_data {
+	unsigned long tmo;
+};
+
 #define cpts_read32(c, r)	readl_relaxed(&c->reg->r)
 #define cpts_write32(c, v, r)	writel_relaxed(v, &c->reg->r)
 
+static int cpts_match(struct sk_buff *skb, unsigned int ptp_class,
+		      u16 ts_seqid, u8 ts_msgtype);
+
 static int event_expired(struct cpts_event *event)
 {
 	return time_after(jiffies, event->tmo);
@@ -77,6 +86,47 @@ static int cpts_purge_events(struct cpts *cpts)
 	return removed ? 0 : -1;
 }
 
+static bool cpts_match_tx_ts(struct cpts *cpts, struct cpts_event *event)
+{
+	struct sk_buff *skb, *tmp;
+	u16 seqid;
+	u8 mtype;
+	bool found = false;
+
+	mtype = (event->high >> MESSAGE_TYPE_SHIFT) & MESSAGE_TYPE_MASK;
+	seqid = (event->high >> SEQUENCE_ID_SHIFT) & SEQUENCE_ID_MASK;
+
+	/* 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;
+		unsigned int class = ptp_classify_raw(skb);
+		struct cpts_skb_cb_data *skb_cb =
+					(struct cpts_skb_cb_data *)skb->cb;
+
+		if (cpts_match(skb, class, seqid, mtype)) {
+			u64 ns = timecounter_cyc2time(&cpts->tc, event->low);
+
+			memset(&ssh, 0, sizeof(ssh));
+			ssh.hwtstamp = ns_to_ktime(ns);
+			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 %u seqid %04x\n",
+				mtype, seqid);
+		} else if (time_after(jiffies, skb_cb->tmo)) {
+			/* timeout any expired skbs over 1s */
+			dev_dbg(cpts->dev,
+				"expiring tx timestamp mtype %u seqid %04x\n",
+				mtype, seqid);
+			__skb_unlink(skb, &cpts->txq);
+			dev_consume_skb_any(skb);
+		}
+	}
+
+	return found;
+}
+
 /*
  * Returns zero if matching event type was found.
  */
@@ -101,9 +151,15 @@ static int cpts_fifo_read(struct cpts *cpts, int match)
 		event->low = lo;
 		type = event_type(event);
 		switch (type) {
+		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;
+			}
 		case CPTS_EV_PUSH:
 		case CPTS_EV_RX:
-		case CPTS_EV_TX:
 			list_del_init(&event->list);
 			list_add_tail(&event->list, &cpts->events);
 			break;
@@ -240,13 +296,20 @@ static struct ptp_clock_info cpts_info = {
 
 static void cpts_overflow_check(struct kthread_work *work)
 {
-	struct timespec64 ts;
 	struct cpts *cpts = container_of(work, struct cpts, overflow_work.work);
+	unsigned long delay = cpts->ov_check_period;
+	struct timespec64 ts;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cpts->lock, flags);
+	ts = ns_to_timespec64(timecounter_read(&cpts->tc));
+
+	if (!skb_queue_empty(&cpts->txq))
+		delay = CPTS_SKB_TX_WORK_TIMEOUT;
+	spin_unlock_irqrestore(&cpts->lock, flags);
 
-	cpts_ptp_gettime(&cpts->info, &ts);
 	pr_debug("cpts overflow check at %lld.%09lu\n", ts.tv_sec, ts.tv_nsec);
-	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work,
-				   cpts->ov_check_period);
+	kthread_queue_delayed_work(cpts->kworker, &cpts->overflow_work, delay);
 }
 
 static int cpts_match(struct sk_buff *skb, unsigned int ptp_class,
@@ -300,7 +363,7 @@ static u64 cpts_find_ts(struct cpts *cpts, struct sk_buff *skb, int ev_type)
 		return 0;
 
 	spin_lock_irqsave(&cpts->lock, flags);
-	cpts_fifo_read(cpts, CPTS_EV_PUSH);
+	cpts_fifo_read(cpts, -1);
 	list_for_each_safe(this, next, &cpts->events) {
 		event = list_entry(this, struct cpts_event, list);
 		if (event_expired(event)) {
@@ -318,6 +381,20 @@ static u64 cpts_find_ts(struct cpts *cpts, struct sk_buff *skb, int ev_type)
 			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);
+		kthread_mod_delayed_work(cpts->kworker,
+					 &cpts->overflow_work, 0);
+	}
 	spin_unlock_irqrestore(&cpts->lock, flags);
 
 	return ns;
@@ -359,6 +436,7 @@ int cpts_register(struct cpts *cpts)
 {
 	int err, i;
 
+	skb_queue_head_init(&cpts->txq);
 	INIT_LIST_HEAD(&cpts->events);
 	INIT_LIST_HEAD(&cpts->pool);
 	for (i = 0; i < CPTS_MAX_EVENTS; i++)
@@ -402,6 +480,9 @@ void cpts_unregister(struct cpts *cpts)
 	cpts_write32(cpts, 0, int_enable);
 	cpts_write32(cpts, 0, control);
 
+	/* Drop all packet */
+	skb_queue_purge(&cpts->txq);
+
 	clk_disable(cpts->refclk);
 }
 EXPORT_SYMBOL_GPL(cpts_unregister);
diff --git a/drivers/net/ethernet/ti/cpts.h b/drivers/net/ethernet/ti/cpts.h
index e8128e8..e94b829 100644
--- a/drivers/net/ethernet/ti/cpts.h
+++ b/drivers/net/ethernet/ti/cpts.h
@@ -128,6 +128,7 @@ struct cpts {
 	unsigned long ov_check_period;
 	struct kthread_worker *kworker;
 	struct kthread_delayed_work overflow_work;
+	struct sk_buff_head txq;
 };
 
 void cpts_rx_timestamp(struct cpts *cpts, struct sk_buff *skb);
-- 
2.10.1