Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1539817 > unrolled thread

[PATCH 1/5] net: ethernet: ti: cpsw: improve re-split policy

Back to article view | Back to linux.kernel

Contents

  [PATCH 1/5] net: ethernet: ti: cpsw: improve re-split policy Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> - 2016-12-10 13:30 +0100
    [PATCH 1/5] net: ethernet: ti: cpsw: use same macros to get active slave Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> - 2016-12-10 13:30 +0100
    [PATCH 4/5] net: ethernet: ti: cpsw: re-split res only when speed is changed Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> - 2016-12-10 13:30 +0100
      Re: [PATCH 4/5] net: ethernet: ti: cpsw: re-split res only when  speed is changed David Miller <davem@davemloft.net> - 2016-12-10 23:40 +0100
    [PATCH 5/5] net: ethernet: ti: cpsw: sync rates for channels in dual emac mode Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> - 2016-12-10 13:30 +0100
      Re: [PATCH 5/5] net: ethernet: ti: cpsw: sync rates for channels  in dual emac mode David Miller <davem@davemloft.net> - 2016-12-10 23:40 +0100
    [PATCH 3/5] net: ethernet: ti: cpsw: combine budget and weight split and check Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org> - 2016-12-10 13:30 +0100
      Re: [PATCH 3/5] net: ethernet: ti: cpsw: combine budget and weight  split and check David Miller <davem@davemloft.net> - 2016-12-10 23:40 +0100
    Re: [PATCH 1/5] net: ethernet: ti: cpsw: improve re-split policy David Miller <davem@davemloft.net> - 2016-12-10 23:40 +0100

#1539817 — [PATCH 1/5] net: ethernet: ti: cpsw: improve re-split policy

This patches add several simplifications and improvements to set
maximum rate for channels taking in account switch and dual emac mode.

Don't re-split res in the following cases:
- speed of phys is not changed
- speed of phys is changed and no rate limited channels
- speed of phys is changed and all channels are rate limited
- phy is unlinked while dev is open
- phy is linked back but speed is not changed

The maximum speed is sum of "linked" phys, thus res are split taken
into account two interfaces, both for dual emac mode and for
switch mode.

Tested on am572x

Based on net-next/master

Ivan Khoronzhuk (5):
  net: ethernet: ti: cpsw: use same macros to get active slave
  net: ethernet: ti: cpsw: don't start queue twice
  net: ethernet: ti: cpsw: combine budget and weight split and check
  net: ethernet: ti: cpsw: re-split res only when speed is changed
  net: ethernet: ti: cpsw: sync rates for channels in dual emac mode

 drivers/net/ethernet/ti/cpsw.c | 178 +++++++++++++++++++++++------------------
 1 file changed, 99 insertions(+), 79 deletions(-)

-- 
2.7.4

[toc] | [next] | [standalone]

#1539818 — [PATCH 1/5] net: ethernet: ti: cpsw: use same macros to get active slave

Use the same, more convenient macros, to get active slave.

Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
---
 drivers/net/ethernet/ti/cpsw.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index b62d958..c45f7d2 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -1624,10 +1624,7 @@ static void cpsw_hwtstamp_v2(struct cpsw_priv *priv)
 	struct cpsw_common *cpsw = priv->cpsw;
 	u32 ctrl, mtype;
 
-	if (cpsw->data.dual_emac)
-		slave = &cpsw->slaves[priv->emac_port];
-	else
-		slave = &cpsw->slaves[cpsw->data.active_slave];
+	slave = &cpsw->slaves[cpsw_slave_index(cpsw, priv)];
 
 	ctrl = slave_read(slave, CPSW2_CONTROL);
 	switch (cpsw->version) {
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1539819 — [PATCH 4/5] net: ethernet: ti: cpsw: re-split res only when speed is changed

Don't re-split res in the following cases:
- speed of phys is not changed
- speed of phys is changed and no rate limited channels
- speed of phys is changed and all channels are rate limited
- phy is unlinked while dev is open
- phy is linked back but speed is not changed

The maximum speed is sum of "linked" phys, thus res are split taken
in account two interfaces, both for dual emac mode and for
switch mode.

Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
---
 drivers/net/ethernet/ti/cpsw.c | 64 ++++++++++++++++++++++++++++++++++++++----
 1 file changed, 59 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index a2c2c06..7ccfa63 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -394,6 +394,7 @@ struct cpsw_common {
 	u32 irqs_table[IRQ_NUM];
 	struct cpts			*cpts;
 	int				rx_ch_num, tx_ch_num;
+	int				speed;
 };
 
 struct cpsw_priv {
@@ -761,7 +762,6 @@ static void cpsw_split_res(struct net_device *ndev)
 	struct cpsw_vector *txv = cpsw->txv;
 	int i, ch_weight, rlim_ch_num = 0;
 	int budget, bigest_rate_ch = 0;
-	struct cpsw_slave *slave;
 	u32 ch_rate, max_rate;
 	int ch_budget = 0;
 
@@ -781,8 +781,16 @@ static void cpsw_split_res(struct net_device *ndev)
 		bigest_rate = 0;
 		max_rate = consumed_rate;
 	} else {
-		slave = &cpsw->slaves[cpsw_slave_index(cpsw, priv)];
-		max_rate = slave->phy->speed * 1000;
+		max_rate = cpsw->speed * 1000;
+
+		/* if max_rate is less then expected due to reduced link speed,
+		 * split proportionally according next potential max speed
+		 */
+		if (max_rate < consumed_rate)
+			max_rate *= 10;
+
+		if (max_rate < consumed_rate)
+			max_rate *= 10;
 
 		ch_budget = (consumed_rate * CPSW_POLL_WEIGHT) / max_rate;
 		ch_budget = (CPSW_POLL_WEIGHT - ch_budget) /
@@ -1013,15 +1021,56 @@ static void _cpsw_adjust_link(struct cpsw_slave *slave,
 	slave->mac_control = mac_control;
 }
 
+static int cpsw_get_common_speed(struct cpsw_common *cpsw)
+{
+	int i, speed;
+
+	for (i = 0, speed = 0; i < cpsw->data.slaves; i++)
+		if (cpsw->slaves[i].phy && cpsw->slaves[i].phy->link)
+			speed += cpsw->slaves[i].phy->speed;
+
+	return speed;
+}
+
+static int cpsw_need_resplit(struct cpsw_common *cpsw)
+{
+	int i, rlim_ch_num;
+	int speed, ch_rate;
+
+	/* re-split resources only in case speed was changed */
+	speed = cpsw_get_common_speed(cpsw);
+	if (speed == cpsw->speed || !speed)
+		return 0;
+
+	cpsw->speed = speed;
+
+	for (i = 0, rlim_ch_num = 0; i < cpsw->tx_ch_num; i++) {
+		ch_rate = cpdma_chan_get_rate(cpsw->txv[i].ch);
+		if (!ch_rate)
+			break;
+
+		rlim_ch_num++;
+	}
+
+	/* cases not dependent on speed */
+	if (!rlim_ch_num || rlim_ch_num == cpsw->tx_ch_num)
+		return 0;
+
+	return 1;
+}
+
 static void cpsw_adjust_link(struct net_device *ndev)
 {
 	struct cpsw_priv	*priv = netdev_priv(ndev);
+	struct cpsw_common	*cpsw = priv->cpsw;
 	bool			link = false;
 
 	for_each_slave(priv, _cpsw_adjust_link, priv, &link);
 
 	if (link) {
-		cpsw_split_res(priv->ndev);
+		if (cpsw_need_resplit(cpsw))
+			cpsw_split_res(ndev);
+
 		netif_carrier_on(ndev);
 		if (netif_running(ndev))
 			netif_tx_wake_all_queues(ndev);
@@ -1538,6 +1587,10 @@ static int cpsw_ndo_stop(struct net_device *ndev)
 		cpsw_ale_stop(cpsw->ale);
 	}
 	for_each_slave(priv, cpsw_slave_stop, cpsw);
+
+	if (cpsw_need_resplit(cpsw))
+		cpsw_split_res(ndev);
+
 	pm_runtime_put_sync(cpsw->dev);
 	if (cpsw->data.dual_emac)
 		cpsw->slaves[priv->emac_port].open_stat = false;
@@ -1983,7 +2036,7 @@ static int cpsw_ndo_set_tx_maxrate(struct net_device *ndev, int queue, u32 rate)
 		return -EINVAL;
 	}
 
-	if (rate > 2000) {
+	if (rate > cpsw->speed) {
 		dev_err(priv->dev, "The channel rate cannot be more than 2Gbps");
 		return -EINVAL;
 	}
@@ -2998,6 +3051,7 @@ static int cpsw_probe(struct platform_device *pdev)
 	ndev->ethtool_ops = &cpsw_ethtool_ops;
 	netif_napi_add(ndev, &cpsw->napi_rx, cpsw_rx_poll, CPSW_POLL_WEIGHT);
 	netif_tx_napi_add(ndev, &cpsw->napi_tx, cpsw_tx_poll, CPSW_POLL_WEIGHT);
+	cpsw_split_res(ndev);
 
 	/* register the network device */
 	SET_NETDEV_DEV(ndev, &pdev->dev);
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1539924 — Re: [PATCH 4/5] net: ethernet: ti: cpsw: re-split res only when speed is changed

From: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Date: Sat, 10 Dec 2016 14:23:49 +0200

> Don't re-split res in the following cases:
> - speed of phys is not changed
> - speed of phys is changed and no rate limited channels
> - speed of phys is changed and all channels are rate limited
> - phy is unlinked while dev is open
> - phy is linked back but speed is not changed
> 
> The maximum speed is sum of "linked" phys, thus res are split taken
> in account two interfaces, both for dual emac mode and for
> switch mode.
> 
> Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>

Applied.

[toc] | [prev] | [next] | [standalone]

#1539820 — [PATCH 5/5] net: ethernet: ti: cpsw: sync rates for channels in dual emac mode

The channels are common for both ndevs in dual emac mode. Hence, keep
in sync their rates.

Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
---
 drivers/net/ethernet/ti/cpsw.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index 7ccfa63..b203143 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -2020,9 +2020,10 @@ static int cpsw_ndo_set_tx_maxrate(struct net_device *ndev, int queue, u32 rate)
 {
 	struct cpsw_priv *priv = netdev_priv(ndev);
 	struct cpsw_common *cpsw = priv->cpsw;
+	struct cpsw_slave *slave;
 	u32 min_rate;
 	u32 ch_rate;
-	int ret;
+	int i, ret;
 
 	ch_rate = netdev_get_tx_queue(ndev, queue)->tx_maxrate;
 	if (ch_rate == rate)
@@ -2053,6 +2054,15 @@ static int cpsw_ndo_set_tx_maxrate(struct net_device *ndev, int queue, u32 rate)
 	if (ret)
 		return ret;
 
+	/* update rates for slaves tx queues */
+	for (i = 0; i < cpsw->data.slaves; i++) {
+		slave = &cpsw->slaves[i];
+		if (!slave->ndev)
+			continue;
+
+		netdev_get_tx_queue(slave->ndev, queue)->tx_maxrate = rate;
+	}
+
 	cpsw_split_res(ndev);
 	return ret;
 }
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1539926 — Re: [PATCH 5/5] net: ethernet: ti: cpsw: sync rates for channels in dual emac mode

From: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Date: Sat, 10 Dec 2016 14:23:50 +0200

> The channels are common for both ndevs in dual emac mode. Hence, keep
> in sync their rates.
> 
> Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>

Applied.

[toc] | [prev] | [next] | [standalone]

#1539823 — [PATCH 3/5] net: ethernet: ti: cpsw: combine budget and weight split and check

Re-split weight along with budget. It simplify code a little
and update state after every rate change. Also it's necessarily
to move arguments checks to this combined function. Replace
maximum rate check for an interface on maximum possible rate.

Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
---
 drivers/net/ethernet/ti/cpsw.c | 107 +++++++++++++----------------------------
 1 file changed, 34 insertions(+), 73 deletions(-)

diff --git a/drivers/net/ethernet/ti/cpsw.c b/drivers/net/ethernet/ti/cpsw.c
index 23213a3..a2c2c06 100644
--- a/drivers/net/ethernet/ti/cpsw.c
+++ b/drivers/net/ethernet/ti/cpsw.c
@@ -753,27 +753,18 @@ static void cpsw_rx_handler(void *token, int len, int status)
 		dev_kfree_skb_any(new_skb);
 }
 
-/* split budget depending on channel rates */
-static void cpsw_split_budget(struct net_device *ndev)
+static void cpsw_split_res(struct net_device *ndev)
 {
 	struct cpsw_priv *priv = netdev_priv(ndev);
+	u32 consumed_rate = 0, bigest_rate = 0;
 	struct cpsw_common *cpsw = priv->cpsw;
 	struct cpsw_vector *txv = cpsw->txv;
-	u32 consumed_rate, bigest_rate = 0;
+	int i, ch_weight, rlim_ch_num = 0;
 	int budget, bigest_rate_ch = 0;
 	struct cpsw_slave *slave;
-	int i, rlim_ch_num = 0;
 	u32 ch_rate, max_rate;
 	int ch_budget = 0;
 
-	if (cpsw->data.dual_emac)
-		slave = &cpsw->slaves[priv->emac_port];
-	else
-		slave = &cpsw->slaves[cpsw->data.active_slave];
-
-	max_rate = slave->phy->speed * 1000;
-
-	consumed_rate = 0;
 	for (i = 0; i < cpsw->tx_ch_num; i++) {
 		ch_rate = cpdma_chan_get_rate(txv[i].ch);
 		if (!ch_rate)
@@ -785,7 +776,14 @@ static void cpsw_split_budget(struct net_device *ndev)
 
 	if (cpsw->tx_ch_num == rlim_ch_num) {
 		max_rate = consumed_rate;
+	} else if (!rlim_ch_num) {
+		ch_budget = CPSW_POLL_WEIGHT / cpsw->tx_ch_num;
+		bigest_rate = 0;
+		max_rate = consumed_rate;
 	} else {
+		slave = &cpsw->slaves[cpsw_slave_index(cpsw, priv)];
+		max_rate = slave->phy->speed * 1000;
+
 		ch_budget = (consumed_rate * CPSW_POLL_WEIGHT) / max_rate;
 		ch_budget = (CPSW_POLL_WEIGHT - ch_budget) /
 			    (cpsw->tx_ch_num - rlim_ch_num);
@@ -793,22 +791,28 @@ static void cpsw_split_budget(struct net_device *ndev)
 			      (cpsw->tx_ch_num - rlim_ch_num);
 	}
 
-	/* split tx budget */
+	/* split tx weight/budget */
 	budget = CPSW_POLL_WEIGHT;
 	for (i = 0; i < cpsw->tx_ch_num; i++) {
 		ch_rate = cpdma_chan_get_rate(txv[i].ch);
 		if (ch_rate) {
 			txv[i].budget = (ch_rate * CPSW_POLL_WEIGHT) / max_rate;
 			if (!txv[i].budget)
-				txv[i].budget = 1;
+				txv[i].budget++;
 			if (ch_rate > bigest_rate) {
 				bigest_rate_ch = i;
 				bigest_rate = ch_rate;
 			}
+
+			ch_weight = (ch_rate * 100) / max_rate;
+			if (!ch_weight)
+				ch_weight++;
+			cpdma_chan_set_weight(cpsw->txv[i].ch, ch_weight);
 		} else {
 			txv[i].budget = ch_budget;
 			if (!bigest_rate_ch)
 				bigest_rate_ch = i;
+			cpdma_chan_set_weight(cpsw->txv[i].ch, 0);
 		}
 
 		budget -= txv[i].budget;
@@ -1017,7 +1021,7 @@ static void cpsw_adjust_link(struct net_device *ndev)
 	for_each_slave(priv, _cpsw_adjust_link, priv, &link);
 
 	if (link) {
-		cpsw_split_budget(priv->ndev);
+		cpsw_split_res(priv->ndev);
 		netif_carrier_on(ndev);
 		if (netif_running(ndev))
 			netif_tx_wake_all_queues(ndev);
@@ -1962,64 +1966,25 @@ static int cpsw_ndo_vlan_rx_kill_vid(struct net_device *ndev,
 static int cpsw_ndo_set_tx_maxrate(struct net_device *ndev, int queue, u32 rate)
 {
 	struct cpsw_priv *priv = netdev_priv(ndev);
-	int tx_ch_num = ndev->real_num_tx_queues;
-	u32 consumed_rate, min_rate, max_rate;
 	struct cpsw_common *cpsw = priv->cpsw;
-	struct cpsw_slave *slave;
-	int ret, i, weight;
-	int rlim_num = 0;
+	u32 min_rate;
 	u32 ch_rate;
+	int ret;
 
 	ch_rate = netdev_get_tx_queue(ndev, queue)->tx_maxrate;
 	if (ch_rate == rate)
 		return 0;
 
-	if (cpsw->data.dual_emac)
-		slave = &cpsw->slaves[priv->emac_port];
-	else
-		slave = &cpsw->slaves[cpsw->data.active_slave];
-	max_rate = slave->phy->speed;
-
-	consumed_rate = 0;
-	for (i = 0; i < tx_ch_num; i++) {
-		if (i == queue)
-			ch_rate = rate;
-		else
-			ch_rate = netdev_get_tx_queue(ndev, i)->tx_maxrate;
-		if (!ch_rate)
-			continue;
-
-		rlim_num++;
-		consumed_rate += ch_rate;
-	}
-
-	if (consumed_rate > max_rate)
-		dev_info(priv->dev, "The common rate shouldn't be more than %dMbps",
-			 max_rate);
-
-	if (consumed_rate > max_rate) {
-		if (max_rate == 10 && consumed_rate <= 100) {
-			max_rate = 100;
-		} else if (max_rate <= 100 && consumed_rate <= 1000) {
-			max_rate = 1000;
-		} else {
-			dev_err(priv->dev, "The common rate cannot be more than %dMbps",
-				max_rate);
-			return -EINVAL;
-		}
-	}
-
-	if (consumed_rate > max_rate) {
-		dev_err(priv->dev, "The common rate cannot be more than %dMbps",
-			max_rate);
+	ch_rate = rate * 1000;
+	min_rate = cpdma_chan_get_min_rate(cpsw->dma);
+	if ((ch_rate < min_rate && ch_rate)) {
+		dev_err(priv->dev, "The channel rate cannot be less than %dMbps",
+			min_rate);
 		return -EINVAL;
 	}
 
-	rate *= 1000;
-	min_rate = cpdma_chan_get_min_rate(cpsw->dma);
-	if ((rate < min_rate && rate)) {
-		dev_err(priv->dev, "The common rate cannot be less than %dMbps",
-			min_rate);
+	if (rate > 2000) {
+		dev_err(priv->dev, "The channel rate cannot be more than 2Gbps");
 		return -EINVAL;
 	}
 
@@ -2029,17 +1994,13 @@ static int cpsw_ndo_set_tx_maxrate(struct net_device *ndev, int queue, u32 rate)
 		return ret;
 	}
 
-	if (rlim_num == tx_ch_num)
-		max_rate = consumed_rate;
+	ret = cpdma_chan_set_rate(cpsw->txv[queue].ch, ch_rate);
+	pm_runtime_put(cpsw->dev);
 
-	weight = (rate * 100) / (max_rate * 1000);
-	cpdma_chan_set_weight(cpsw->txv[queue].ch, weight);
-	ret = cpdma_chan_set_rate(cpsw->txv[queue].ch, rate);
+	if (ret)
+		return ret;
 
-	/* re-split budget between channels */
-	if (!rate)
-		cpsw_split_budget(ndev);
-	pm_runtime_put(cpsw->dev);
+	cpsw_split_res(ndev);
 	return ret;
 }
 
@@ -2399,7 +2360,7 @@ static int cpsw_set_channels(struct net_device *ndev,
 		if (ret)
 			goto err;
 
-		cpsw_split_budget(ndev);
+		cpsw_split_res(ndev);
 
 		/* After this receive is started */
 		cpdma_ctlr_start(cpsw->dma);
-- 
2.7.4

[toc] | [prev] | [next] | [standalone]

#1539929 — Re: [PATCH 3/5] net: ethernet: ti: cpsw: combine budget and weight split and check

From: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Date: Sat, 10 Dec 2016 14:23:48 +0200

> Re-split weight along with budget. It simplify code a little
> and update state after every rate change. Also it's necessarily
> to move arguments checks to this combined function. Replace
> maximum rate check for an interface on maximum possible rate.
> 
> Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>

Applied.

[toc] | [prev] | [next] | [standalone]

#1539927

From: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Date: Sat, 10 Dec 2016 14:23:45 +0200

> This patches add several simplifications and improvements to set
> maximum rate for channels taking in account switch and dual emac mode.
> 
> Don't re-split res in the following cases:
> - speed of phys is not changed
> - speed of phys is changed and no rate limited channels
> - speed of phys is changed and all channels are rate limited
> - phy is unlinked while dev is open
> - phy is linked back but speed is not changed
> 
> The maximum speed is sum of "linked" phys, thus res are split taken
> into account two interfaces, both for dual emac mode and for
> switch mode.
> 
> Tested on am572x
> 
> Based on net-next/master

Applied.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web