Skip to contentSkip to conversation

net: dsa: loop: Preparatory changes for

PATCH In review Florian Fainelli 3 participants 15 messages Aug 3, 2020 – Aug 4, 2020

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

cover net: dsa: loop: Preparatory changes for

Hi David,

These patches are all meant to help pave the way for a 802.1Q data path added to the mockup driver, making it more useful than just testing for configuration. Sending those out now since there is no real need to wait.

Thanks

Florian Fainelli (5):
net: dsa: loop: PVID should be per-port
net: dsa: loop: Support 4K VLANs
net: dsa: loop: Move data structures to header
net: dsa: loop: Wire-up MTU callbacks
net: dsa: loop: Set correct number of ports

 drivers/net/dsa/dsa_loop.c | 61 ++++++++++++++------------------------
 include/linux/dsa/loop.h   | 41 +++++++++++++++++++++++++
 2 files changed, 64 insertions(+), 38 deletions(-)
 create mode 100644 include/linux/dsa/loop.h
Signature · 1 line
2.25.1
#1 lore raw

PATCH 1/5 net: dsa: loop: PVID should be per-port

The PVID should be per-port, this is a preliminary change to support a 802.1Q data path in the driver.

---
 drivers/net/dsa/dsa_loop.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/net/dsa/dsa_loop.c +4 -4
index f8bc85a6e670..4a57238cdfd8 100644
@@ -45,6 +45,7 @@ static struct dsa_loop_mib_entry dsa_loop_mibs[] = {
 
 struct dsa_loop_port {
 	struct dsa_loop_mib_entry mib[__DSA_LOOP_CNT_MAX];
+	u16 pvid;
 };
 
 #define DSA_LOOP_VLANS	5
@@ -55,7 +56,6 @@ struct dsa_loop_priv {
 	struct dsa_loop_vlan vlans[DSA_LOOP_VLANS];
 	struct net_device *netdev;
 	struct dsa_loop_port ports[DSA_MAX_PORTS];
-	u16 pvid;
 };
 
 static struct phy_device *phydevs[PHY_MAX_ADDR];
@@ -224,7 +224,7 @@ static void dsa_loop_port_vlan_add(struct dsa_switch *ds, int port,
 	}
 
 	if (pvid)
-		ps->pvid = vid;
+		ps->ports[port].pvid = vid;
 }
 
 static int dsa_loop_port_vlan_del(struct dsa_switch *ds, int port,
@@ -234,7 +234,7 @@ static int dsa_loop_port_vlan_del(struct dsa_switch *ds, int port,
 	struct dsa_loop_priv *ps = ds->priv;
 	struct mii_bus *bus = ps->bus;
 	struct dsa_loop_vlan *vl;
-	u16 vid, pvid = ps->pvid;
+	u16 vid, pvid = ps->ports[port].pvid;
 
 	/* Just do a sleeping operation to make lockdep checks effective */
 	mdiobus_read(bus, ps->port_base + port, MII_BMSR);
@@ -252,7 +252,7 @@ static int dsa_loop_port_vlan_del(struct dsa_switch *ds, int port,
 		dev_dbg(ds->dev, "%s: port: %d vlan: %d, %stagged, pvid: %d\n",
 			__func__, port, vid, untagged ? "un" : "", pvid);
 	}
-	ps->pvid = pvid;
+	ps->ports[port].pvid = pvid;
 
 	return 0;
 }
Signature · 1 line
2.25.1
#3 lore raw

PATCH 2/5 net: dsa: loop: Support 4K VLANs

In reply to Florian Fainelli

Allocate a 4K array of VLANs instead of limiting ourselves to just 5 which is arbitrary.

---
 drivers/net/dsa/dsa_loop.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/net/dsa/dsa_loop.c +2 -4
index 4a57238cdfd8..6e97b44c6f3f 100644
@@ -48,12 +48,10 @@ struct dsa_loop_port {
 	u16 pvid;
 };
 
-#define DSA_LOOP_VLANS	5
-
 struct dsa_loop_priv {
 	struct mii_bus	*bus;
 	unsigned int	port_base;
-	struct dsa_loop_vlan vlans[DSA_LOOP_VLANS];
+	struct dsa_loop_vlan vlans[VLAN_N_VID];
 	struct net_device *netdev;
 	struct dsa_loop_port ports[DSA_MAX_PORTS];
 };
@@ -191,7 +189,7 @@ dsa_loop_port_vlan_prepare(struct dsa_switch *ds, int port,
 	/* Just do a sleeping operation to make lockdep checks effective */
 	mdiobus_read(bus, ps->port_base + port, MII_BMSR);
 
-	if (vlan->vid_end > DSA_LOOP_VLANS)
+	if (vlan->vid_end > ARRAY_SIZE(ps->vlans))
 		return -ERANGE;
 
 	return 0;
Signature · 1 line
2.25.1
maintainer Reviewed-by on v1 2/5
#4 lore raw

Re: [PATCH net-next 2/5] net: dsa: loop: Support 4K VLANs

Quoted 15 lines · Florian Fainelli

Allocate a 4K array of VLANs instead of limiting ourselves to just 5 which is arbitrary.

---
 drivers/net/dsa/dsa_loop.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/net/dsa/dsa_loop.c +0 -0
index 4a57238cdfd8..6e97b44c6f3f 100644
@@ -48,12 +48,10 @@ struct dsa_loop_port {
 	u16 pvid;
 };

drivers/net/dsa/dsa_loop.c struct dsa_loop_port { · line 56

Quoted 10 lines
drivers/net/dsa/dsa_loop.c +1 -3
 
-#define DSA_LOOP_VLANS	5
-
 struct dsa_loop_priv {
 	struct mii_bus	*bus;
 	unsigned int	port_base;
-	struct dsa_loop_vlan vlans[DSA_LOOP_VLANS];
+	struct dsa_loop_vlan vlans[VLAN_N_VID];
 	struct net_device *netdev;
 	struct dsa_loop_port ports[DSA_MAX_PORTS];

That is 4K x (2 x u16) = 16K RAM. I suppose for a test driver which is never expected to be used in production, that is O.K.

Andrew

#5 lore raw

Re: [PATCH net-next 2/5] net: dsa: loop: Support 4K VLANs

Quoted 17 lines · Andrew Lunn
Quoted 16 lines · Florian Fainelli

Allocate a 4K array of VLANs instead of limiting ourselves to just 5 which is arbitrary.

---
 drivers/net/dsa/dsa_loop.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
drivers/net/dsa/dsa_loop.c +0 -0
index 4a57238cdfd8..6e97b44c6f3f 100644
@@ -48,12 +48,10 @@ struct dsa_loop_port {
 	u16 pvid;
 };
 
Quoted 12 lines

drivers/net/dsa/dsa_loop.c struct dsa_loop_port { · line 56

Quoted 9 lines
drivers/net/dsa/dsa_loop.c +1 -3
-#define DSA_LOOP_VLANS	5
-
 struct dsa_loop_priv {
 	struct mii_bus	*bus;
 	unsigned int	port_base;
-	struct dsa_loop_vlan vlans[DSA_LOOP_VLANS];
+	struct dsa_loop_vlan vlans[VLAN_N_VID];
 	struct net_device *netdev;
 	struct dsa_loop_port ports[DSA_MAX_PORTS];

That is 4K x (2 x u16) = 16K RAM. I suppose for a test driver which is never expected to be used in production, that is O.K.

I think so too, if we are worried we could switch to vmalloc() down the road.

Quoted 5 lines

Andrew

Signature · 1 line
Florian
#6 lore raw

PATCH 3/5 net: dsa: loop: Move data structures to header

In reply to Florian Fainelli

In preparation for adding support for a mockup data path, move the driver data structures to include/linux/dsa/loop.h such that we can share them between net/dsa/ and drivers/net/dsa/ later on.

---
 drivers/net/dsa/dsa_loop.c | 32 +-----------------------------
 include/linux/dsa/loop.h   | 40 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 41 insertions(+), 31 deletions(-)
 create mode 100644 include/linux/dsa/loop.h
drivers/net/dsa/dsa_loop.c +1 -31
index 6e97b44c6f3f..ed0b580c9944 100644
@@ -14,28 +14,11 @@
 #include <linux/workqueue.h>
 #include <linux/module.h>
 #include <linux/if_bridge.h>
+#include <linux/dsa/loop.h>
 #include <net/dsa.h>
 
 #include "dsa_loop.h"
 
-struct dsa_loop_vlan {
-	u16 members;
-	u16 untagged;
-};
-
-struct dsa_loop_mib_entry {
-	char name[ETH_GSTRING_LEN];
-	unsigned long val;
-};
-
-enum dsa_loop_mib_counters {
-	DSA_LOOP_PHY_READ_OK,
-	DSA_LOOP_PHY_READ_ERR,
-	DSA_LOOP_PHY_WRITE_OK,
-	DSA_LOOP_PHY_WRITE_ERR,
-	__DSA_LOOP_CNT_MAX,
-};
-
 static struct dsa_loop_mib_entry dsa_loop_mibs[] = {
 	[DSA_LOOP_PHY_READ_OK]	= { "phy_read_ok", },
 	[DSA_LOOP_PHY_READ_ERR]	= { "phy_read_err", },
@@ -43,19 +26,6 @@ static struct dsa_loop_mib_entry dsa_loop_mibs[] = {
 	[DSA_LOOP_PHY_WRITE_ERR] = { "phy_write_err", },
 };
 
-struct dsa_loop_port {
-	struct dsa_loop_mib_entry mib[__DSA_LOOP_CNT_MAX];
-	u16 pvid;
-};
-
-struct dsa_loop_priv {
-	struct mii_bus	*bus;
-	unsigned int	port_base;
-	struct dsa_loop_vlan vlans[VLAN_N_VID];
-	struct net_device *netdev;
-	struct dsa_loop_port ports[DSA_MAX_PORTS];
-};
-
 static struct phy_device *phydevs[PHY_MAX_ADDR];
 
 static enum dsa_tag_protocol dsa_loop_get_protocol(struct dsa_switch *ds,
include/linux/dsa/loop.h new file +40 -0
new file mode 100644
index 000000000000..bb39401a8056
@@ -0,0 +1,40 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef DSA_LOOP_H
+#define DSA_LOOP_H
+
+#include <linux/types.h>
+#include <linux/ethtool.h>
+#include <net/dsa.h>
+
+struct dsa_loop_vlan {
+	u16 members;
+	u16 untagged;
+};
+
+struct dsa_loop_mib_entry {
+	char name[ETH_GSTRING_LEN];
+	unsigned long val;
+};
+
+enum dsa_loop_mib_counters {
+	DSA_LOOP_PHY_READ_OK,
+	DSA_LOOP_PHY_READ_ERR,
+	DSA_LOOP_PHY_WRITE_OK,
+	DSA_LOOP_PHY_WRITE_ERR,
+	__DSA_LOOP_CNT_MAX,
+};
+
+struct dsa_loop_port {
+	struct dsa_loop_mib_entry mib[__DSA_LOOP_CNT_MAX];
+	u16 pvid;
+};
+
+struct dsa_loop_priv {
+	struct mii_bus	*bus;
+	unsigned int	port_base;
+	struct dsa_loop_vlan vlans[VLAN_N_VID];
+	struct net_device *netdev;
+	struct dsa_loop_port ports[DSA_MAX_PORTS];
+};
+
+#endif /* DSA_LOOP_H */
Signature · 1 line
2.25.1
#7 lore raw

PATCH 4/5 net: dsa: loop: Wire-up MTU callbacks

In reply to Florian Fainelli

For now we simply store the port MTU into a per-port member.

---
 drivers/net/dsa/dsa_loop.c | 17 +++++++++++++++++
 include/linux/dsa/loop.h   |  1 +
 2 files changed, 18 insertions(+)
drivers/net/dsa/dsa_loop.c +17 -0
index ed0b580c9944..6a7d661b5a59 100644
@@ -225,6 +225,21 @@ static int dsa_loop_port_vlan_del(struct dsa_switch *ds, int port,
 	return 0;
 }
 
+static int dsa_loop_port_change_mtu(struct dsa_switch *ds, int port,
+				    int new_mtu)
+{
+	struct dsa_loop_priv *priv = ds->priv;
+
+	priv->ports[port].mtu = new_mtu;
+
+	return 0;
+}
+
+static int dsa_loop_port_max_mtu(struct dsa_switch *ds, int port)
+{
+	return ETH_MAX_MTU;
+}
+
 static const struct dsa_switch_ops dsa_loop_driver = {
 	.get_tag_protocol	= dsa_loop_get_protocol,
 	.setup			= dsa_loop_setup,
@@ -241,6 +256,8 @@ static const struct dsa_switch_ops dsa_loop_driver = {
 	.port_vlan_prepare	= dsa_loop_port_vlan_prepare,
 	.port_vlan_add		= dsa_loop_port_vlan_add,
 	.port_vlan_del		= dsa_loop_port_vlan_del,
+	.port_change_mtu	= dsa_loop_port_change_mtu,
+	.port_max_mtu		= dsa_loop_port_max_mtu,
 };
 
 static int dsa_loop_drv_probe(struct mdio_device *mdiodev)
include/linux/dsa/loop.h +1 -0
index bb39401a8056..5a3470bcc8a7 100644
@@ -27,6 +27,7 @@ enum dsa_loop_mib_counters {
 struct dsa_loop_port {
 	struct dsa_loop_mib_entry mib[__DSA_LOOP_CNT_MAX];
 	u16 pvid;
+	int mtu;
 };
 
 struct dsa_loop_priv {
Signature · 1 line
2.25.1
#9 lore raw

PATCH 5/5 net: dsa: loop: Set correct number of ports

In reply to Florian Fainelli

We only support DSA_LOOP_NUM_PORTS in the switch, do not tell the DSA core to allocate up to DSA_MAX_PORTS which is nearly the double (6 vs. 11).

---
 drivers/net/dsa/dsa_loop.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/dsa/dsa_loop.c +1 -1
index 6a7d661b5a59..eb600b3dbf26 100644
@@ -275,7 +275,7 @@ static int dsa_loop_drv_probe(struct mdio_device *mdiodev)
 		return -ENOMEM;
 
 	ds->dev = &mdiodev->dev;
-	ds->num_ports = DSA_MAX_PORTS;
+	ds->num_ports = DSA_LOOP_NUM_PORTS;
 
 	ps = devm_kzalloc(&mdiodev->dev, sizeof(*ps), GFP_KERNEL);
 	if (!ps)
Signature · 1 line
2.25.1
maintainer
#11 lore raw
In reply to Florian Fainelli
Quoted 5 lines · Florian Fainelli

Hi David,

These patches are all meant to help pave the way for a 802.1Q data path added to the mockup driver, making it more useful than just testing for configuration.

Could you give some more details. I assume the tag driver is going be to more active. At least, that would make sense with the data structure moves.

Andrew

#12 lore raw
Quoted 10 lines · Andrew Lunn
Quoted 5 lines · Florian Fainelli

Hi David,

These patches are all meant to help pave the way for a 802.1Q data path added to the mockup driver, making it more useful than just testing for configuration.

Could you give some more details. I assume the tag driver is going be to more active. At least, that would make sense with the data structure moves.

Still working on a 802.1Q data path that allows to carry VLAN tags from the DSA CPU interface (guest side) to a tap (host side) and from there you can ping, bridge etc. devices. I am having some problems with the responses sent back to the guest which are incorrectly tagged on ingress.

You can see the WIP here:
https://github.com/ffainelli/linux/commits/dsa-loop-8021q

Signature · 1 line
Florian
maintainer
#13 lore raw
In reply to Florian Fainelli

From: Florian Fainelli <[email protected]>
Date: Mon, 3 Aug 2020 13:03:49 -0700

Quoted 4 lines

These patches are all meant to help pave the way for a 802.1Q data path added to the mockup driver, making it more useful than just testing for configuration. Sending those out now since there is no real need to wait.

Series applied, I added "a 802.1Q data path" to the subject line I integrated into the merge commit for the series as it seems your Subject line here was chopped off.

#14 lore raw
Quoted 11 lines · David Miller

From: Florian Fainelli <[email protected]>
Date: Mon, 3 Aug 2020 13:03:49 -0700

Quoted 4 lines

These patches are all meant to help pave the way for a 802.1Q data path added to the mockup driver, making it more useful than just testing for configuration. Sending those out now since there is no real need to wait.

Series applied, I added "a 802.1Q data path" to the subject line I integrated into the merge commit for the series as it seems your Subject line here was chopped off.

Yes it was, the merge commit looks good to me, thank you.

Signature · 1 line
Florian