Skip to contentSkip to conversation

PTP support for the SJA1105 DSA driver

PATCH Superseded Vladimir Oltean 5 participants 40 messages Jun 4, 2019 – Jun 7, 2019

Subsystems (4)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Vladimir Oltean

author maintainer
#0 lore raw

v3 cover PTP support for the SJA1105 DSA driver

This patchset adds the following:

- A timecounter/cyclecounter based PHC for the free-running timestamping clock of this switch.

- A state machine implemented in the DSA tagger for SJA1105, which keeps track of metadata follow-up Ethernet frames (the switch's way of transmitting RX timestamps).

Clock manipulations on the actual hardware PTP clock will have to be implemented anyway, for the TTEthernet block and the time-based ingress policer.

This depends upon the "FDB updates for SJA1105 DSA driver" series at:
https://patchwork.ozlabs.org/project/netdev/list/?series=111354&state=*

v2 patchset can be found at:
https://lkml.org/lkml/2019/6/2/146

Changes from v2:

- Broke previous 09/10 patch (timestamping) into multiple smaller patches.

- Every patch in the series compiles.

v1 patchset can be found at:
https://lkml.org/lkml/2019/5/28/1093

Changes from v1:

- Removed the addition of the DSA .can_timestamp callback.

- Waiting for meta frames is done completely inside the tagger, and all frames emitted on RX are already partially timestamped.

- Added a global data structure for the tagger common to all ports.

- Made PTP work with ports in standalone mode, by limiting use of the DMAC-mangling "incl_srcpt" mode only when ports are bridged, aka when the DSA master is already promiscuous and can receive anything. Also changed meta frames to be sent at the 01-80-C2-00-00-0E DMAC.

- Made some progress w.r.t. observed negative path delay. Apparently it only appears when the delay mechanism is the delay request-response (end-to-end) one. If peer delay is used (-P), the path delay is positive and appears reasonable for an 1000Base-T link (485 ns in steady state).

SJA1105 as PTP slave (OC) with E2E path delay:

ptp4l[55.600]: master offset          8 s2 freq  +83677 path delay     -2390
ptp4l[56.600]: master offset         17 s2 freq  +83688 path delay     -2391
ptp4l[57.601]: master offset          6 s2 freq  +83682 path delay     -2391
ptp4l[58.601]: master offset         -1 s2 freq  +83677 path delay     -2391

SJA1105 as PTP slave (OC) with P2P path delay:

ptp4l[48.343]: master offset          5 s2 freq  +83715 path delay       484
ptp4l[48.468]: master offset         -3 s2 freq  +83705 path delay       485
ptp4l[48.593]: master offset          0 s2 freq  +83708 path delay       485
ptp4l[48.718]: master offset          1 s2 freq  +83710 path delay       485
ptp4l[48.844]: master offset          1 s2 freq  +83710 path delay       485
ptp4l[48.969]: master offset         -5 s2 freq  +83702 path delay       485
ptp4l[49.094]: master offset          3 s2 freq  +83712 path delay       485
ptp4l[49.219]: master offset          4 s2 freq  +83714 path delay       485
ptp4l[49.344]: master offset         -5 s2 freq  +83702 path delay       485
ptp4l[49.469]: master offset          3 s2 freq  +83713 path delay       487

Vladimir Oltean (17):
net: dsa: Keep a pointer to the skb clone for TX timestamping
net: dsa: Add teardown callback for drivers
net: dsa: tag_8021q: Create helper function for removing VLAN header
net: dsa: sja1105: Move sja1105_change_tpid into

    sja1105_vlan_filtering

net: dsa: sja1105: Reverse TPID and TPID2
net: dsa: sja1105: Limit use of incl_srcpt to bridge+vlan mode
net: dsa: sja1105: Add support for the PTP clock
net: dsa: sja1105: Move sja1105_is_link_local to include/linux
net: dsa: sja1105: Add logic for TX timestamping
net: dsa: sja1105: Build a minimal understanding of meta frames
net: dsa: sja1105: Add support for the AVB Parameters Table
net: dsa: sja1105: Make sja1105_is_link_local not match meta frames
net: dsa: sja1105: Receive and decode meta frames
net: dsa: sja1105: Add a global sja1105_tagger_data structure
net: dsa: sja1105: Increase priority of CPU-trapped frames
net: dsa: sja1105: Add a state machine for RX timestamping
net: dsa: sja1105: Expose PTP timestamping ioctls to userspace

 drivers/net/dsa/sja1105/Kconfig               |   7 +
 drivers/net/dsa/sja1105/Makefile              |   1 +
 drivers/net/dsa/sja1105/sja1105.h             |  29 ++
 .../net/dsa/sja1105/sja1105_dynamic_config.c  |   2 +
 drivers/net/dsa/sja1105/sja1105_main.c        | 319 ++++++++++++--
 drivers/net/dsa/sja1105/sja1105_ptp.c         | 392 ++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_ptp.h         |  64 +++
 drivers/net/dsa/sja1105/sja1105_spi.c         |  33 ++
 .../net/dsa/sja1105/sja1105_static_config.c   |  59 +++
 .../net/dsa/sja1105/sja1105_static_config.h   |  10 +
 include/linux/dsa/8021q.h                     |   7 +
 include/linux/dsa/sja1105.h                   |  51 +++
 include/net/dsa.h                             |   1 +
 net/dsa/dsa2.c                                |   3 +
 net/dsa/slave.c                               |   3 +
 net/dsa/tag_8021q.c                           |  15 +
 net/dsa/tag_sja1105.c                         | 203 ++++++++-
 17 files changed, 1152 insertions(+), 47 deletions(-)
 create mode 100644 drivers/net/dsa/sja1105/sja1105_ptp.c
 create mode 100644 drivers/net/dsa/sja1105/sja1105_ptp.h
Signature · 1 line
2.17.1
author maintainer
#1 lore raw

PATCH v3 1/17 net: dsa: Keep a pointer to the skb clone for TX timestamping

For drivers that use deferred_xmit for PTP frames (such as sja1105), there is no need to perform matching between PTP frames and their egress timestamps, since the sending process can be serialized.

In that case, it makes sense to have the pointer to the skb clone that DSA made directly in the skb->cb. It will be used for pushing the egress timestamp back in the application socket's error queue.

---
Changes in v3:

None.

Changes in v2:

Patch is new. Forgot to send in v1.

 net/dsa/slave.c | 3 +++
 1 file changed, 3 insertions(+)
net/dsa/slave.c +3 -0
index 1e2ae9d59b88..59d7c9e0270f 100644
@@ -427,6 +427,8 @@ static void dsa_skb_tx_timestamp(struct dsa_slave_priv *p,
 	if (!clone)
 		return;
 
+	DSA_SKB_CB(skb)->clone = clone;
+
 	if (ds->ops->port_txtstamp(ds, p->dp->index, clone, type))
 		return;
 
@@ -464,6 +466,7 @@ static netdev_tx_t dsa_slave_xmit(struct sk_buff *skb, struct net_device *dev)
 	u64_stats_update_end(&s->syncp);
 
 	DSA_SKB_CB(skb)->deferred_xmit = false;
+	DSA_SKB_CB(skb)->clone = NULL;
 
 	/* Identify PTP protocol packets, clone them, and pass them to the
 	 * switch driver
Signature · 1 line
2.17.1
Reviewed-by on v3 1/17
#2 lore raw

Re: [PATCH v3 net-next 01/17] net: dsa: Keep a pointer to the skb clone for TX timestamping

Quoted 9 lines · Vladimir Oltean

For drivers that use deferred_xmit for PTP frames (such as sja1105), there is no need to perform matching between PTP frames and their egress timestamps, since the sending process can be serialized.

In that case, it makes sense to have the pointer to the skb clone that DSA made directly in the skb->cb. It will be used for pushing the egress timestamp back in the application socket's error queue.

Signature · 1 line
Florian
author maintainer
#3 lore raw

PATCH v3 2/17 net: dsa: Add teardown callback for drivers

In reply to Vladimir Oltean

This is helpful for e.g. draining per-driver (not per-port) tagger queues.

---
Changes in v3:

Moved after dsa_switch_unregister_notifier, which is symmetrical to where the setup callback is.

Changes in v2:

Patch is new.

 include/net/dsa.h | 1 +
 net/dsa/dsa2.c    | 3 +++
 2 files changed, 4 insertions(+)
include/net/dsa.h +1 -0
index a7f36219904f..4033e0677be4 100644
@@ -361,6 +361,7 @@ struct dsa_switch_ops {
 						  int port);
 
 	int	(*setup)(struct dsa_switch *ds);
+	void	(*teardown)(struct dsa_switch *ds);
 	u32	(*get_phy_flags)(struct dsa_switch *ds, int port);
 
 	/*
net/dsa/dsa2.c +3 -0
index b70befe8a3c8..d98e0e8ee8aa 100644
@@ -412,6 +412,9 @@ static void dsa_switch_teardown(struct dsa_switch *ds)
 
 	dsa_switch_unregister_notifier(ds);
 
+	if (ds->ops->teardown)
+		ds->ops->teardown(ds);
+
 	if (ds->devlink) {
 		devlink_unregister(ds->devlink);
 		devlink_free(ds->devlink);
Signature · 1 line
2.17.1
author maintainer
#6 lore raw

PATCH v3 3/17 net: dsa: tag_8021q: Create helper function for removing VLAN header

In reply to Vladimir Oltean

This removes the existing implementation from tag_sja1105, which was partially incorrect (it was not changing the MAC header offset, thereby leaving it to point 4 bytes earlier than it should have).

This overwrites the VLAN tag by moving the Ethernet source and destination MACs 4 bytes to the right. Then skb->data (assumed to be pointing immediately after the EtherType) is temporarily pushed to the beginning of the new Ethernet header, the new Ethernet header offset and length are recorded, then skb->data is moved back to where it was.

---
Changes in v3:

None.

Changes in v2:

Patch is new.

 include/linux/dsa/8021q.h |  7 +++++++
 net/dsa/tag_8021q.c       | 15 +++++++++++++++
 net/dsa/tag_sja1105.c     |  3 +--
 3 files changed, 23 insertions(+), 2 deletions(-)
include/linux/dsa/8021q.h +7 -0
index 3911e0586478..463378812f18 100644
@@ -31,6 +31,8 @@ int dsa_8021q_rx_switch_id(u16 vid);
 
 int dsa_8021q_rx_source_port(u16 vid);
 
+struct sk_buff *dsa_8021q_remove_header(struct sk_buff *skb);
+
 #else
 
 int dsa_port_setup_8021q_tagging(struct dsa_switch *ds, int index,
@@ -71,6 +73,11 @@ int dsa_8021q_rx_source_port(u16 vid)
 	return 0;
 }
 
+struct sk_buff *dsa_8021q_remove_header(struct sk_buff *skb)
+{
+	return NULL;
+}
+
 #endif /* IS_ENABLED(CONFIG_NET_DSA_TAG_8021Q) */
 
 #endif /* _NET_DSA_8021Q_H */
net/dsa/tag_8021q.c +15 -0
index 65a35e976d7b..0ce680ef8e83 100644
@@ -261,6 +261,21 @@ struct sk_buff *dsa_8021q_rcv(struct sk_buff *skb, struct net_device *netdev,
 }
 EXPORT_SYMBOL_GPL(dsa_8021q_rcv);
 
+struct sk_buff *dsa_8021q_remove_header(struct sk_buff *skb)
+{
+	u8 *from = skb_mac_header(skb);
+	u8 *dest = from + VLAN_HLEN;
+
+	memmove(dest, from, ETH_HLEN - VLAN_HLEN);
+	skb_push(skb, ETH_HLEN);
+	skb_reset_mac_header(skb);
+	skb_reset_mac_len(skb);
+	skb_pull(skb, ETH_HLEN);
+
+	return skb;
+}
+EXPORT_SYMBOL_GPL(dsa_8021q_remove_header);
+
 static const struct dsa_device_ops dsa_8021q_netdev_ops = {
 	.name		= "8021q",
 	.proto		= DSA_TAG_PROTO_8021Q,
net/dsa/tag_sja1105.c +1 -2
index d43737e6c3fb..535d8a1aabe1 100644
@@ -106,8 +106,7 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 	 * it there, see dsa_switch_rcv: skb_push(skb, ETH_HLEN).
 	 */
 	if (is_tagged)
-		memmove(skb->data - ETH_HLEN, skb->data - ETH_HLEN - VLAN_HLEN,
-			ETH_HLEN - VLAN_HLEN);
+		skb = dsa_8021q_remove_header(skb);
 
 	return skb;
 }
Signature · 1 line
2.17.1
author maintainer
#7 lore raw

PATCH v3 4/17 net: dsa: sja1105: Move sja1105_change_tpid into sja1105_vlan_filtering

In reply to Vladimir Oltean

This is a cosmetic patch, pre-cursor to making another change to the General Parameters Table (incl_srcpt) which does not logically pertain to the sja1105_change_tpid function name, but not putting it there would otherwise create a need of resetting the switch twice.

So simply move the existing code into the .port_vlan_filtering callback, where the incl_srcpt change will be added as well.

---
Changes in v3:

None.

Changes in v2:

Patch is new.

 drivers/net/dsa/sja1105/sja1105_main.c | 42 +++++++++++++-------------
 1 file changed, 21 insertions(+), 21 deletions(-)
drivers/net/dsa/sja1105/sja1105_main.c +21 -21
index b151a8fafb9e..909497aa4b6f 100644
@@ -1289,23 +1289,6 @@ static int sja1105_static_config_reload(struct sja1105_private *priv)
 	return rc;
 }
 
-/* The TPID setting belongs to the General Parameters table,
- * which can only be partially reconfigured at runtime (and not the TPID).
- * So a switch reset is required.
- */
-static int sja1105_change_tpid(struct sja1105_private *priv,
-			       u16 tpid, u16 tpid2)
-{
-	struct sja1105_general_params_entry *general_params;
-	struct sja1105_table *table;
-
-	table = &priv->static_config.tables[BLK_IDX_GENERAL_PARAMS];
-	general_params = table->entries;
-	general_params->tpid = tpid;
-	general_params->tpid2 = tpid2;
-	return sja1105_static_config_reload(priv);
-}
-
 static int sja1105_pvid_apply(struct sja1105_private *priv, int port, u16 pvid)
 {
 	struct sja1105_mac_config_entry *mac;
@@ -1424,17 +1407,34 @@ static int sja1105_vlan_prepare(struct dsa_switch *ds, int port,
 	return 0;
 }
 
+/* The TPID setting belongs to the General Parameters table,
+ * which can only be partially reconfigured at runtime (and not the TPID).
+ * So a switch reset is required.
+ */
 static int sja1105_vlan_filtering(struct dsa_switch *ds, int port, bool enabled)
 {
+	struct sja1105_general_params_entry *general_params;
 	struct sja1105_private *priv = ds->priv;
+	struct sja1105_table *table;
+	u16 tpid, tpid2;
 	int rc;
 
-	if (enabled)
+	if (enabled) {
 		/* Enable VLAN filtering. */
-		rc = sja1105_change_tpid(priv, ETH_P_8021Q, ETH_P_8021AD);
-	else
+		tpid  = ETH_P_8021Q;
+		tpid2 = ETH_P_8021AD;
+	} else {
 		/* Disable VLAN filtering. */
-		rc = sja1105_change_tpid(priv, ETH_P_SJA1105, ETH_P_SJA1105);
+		tpid  = ETH_P_SJA1105;
+		tpid2 = ETH_P_SJA1105;
+	}
+
+	table = &priv->static_config.tables[BLK_IDX_GENERAL_PARAMS];
+	general_params = table->entries;
+	general_params->tpid = tpid;
+	general_params->tpid2 = tpid2;
+
+	rc = sja1105_static_config_reload(priv);
 	if (rc)
 		dev_err(ds->dev, "Failed to change VLAN Ethertype\n");
 
Signature · 1 line
2.17.1
Reviewed-by on v3 4/17
#8 lore raw

Re: [PATCH v3 net-next 04/17] net: dsa: sja1105: Move sja1105_change_tpid into sja1105_vlan_filtering

Quoted 9 lines · Vladimir Oltean

This is a cosmetic patch, pre-cursor to making another change to the General Parameters Table (incl_srcpt) which does not logically pertain to the sja1105_change_tpid function name, but not putting it there would otherwise create a need of resetting the switch twice.

So simply move the existing code into the .port_vlan_filtering callback, where the incl_srcpt change will be added as well.

Signature · 1 line
Florian
author maintainer
#9 lore raw

PATCH v3 5/17 net: dsa: sja1105: Reverse TPID and TPID2

In reply to Vladimir Oltean

From reading the P/Q/R/S user manual, it appears that TPID is used by the switch for detecting S-tags and TPID2 for C-tags. Their meaning is not clear from the E/T manual.

---
Changes in v3:

Patch is new.

 drivers/net/dsa/sja1105/sja1105_main.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/dsa/sja1105/sja1105_main.c +4 -2
index 909497aa4b6f..0f34e713c408 100644
@@ -1421,8 +1421,8 @@ static int sja1105_vlan_filtering(struct dsa_switch *ds, int port, bool enabled)
 
 	if (enabled) {
 		/* Enable VLAN filtering. */
-		tpid  = ETH_P_8021Q;
-		tpid2 = ETH_P_8021AD;
+		tpid  = ETH_P_8021AD;
+		tpid2 = ETH_P_8021Q;
 	} else {
 		/* Disable VLAN filtering. */
 		tpid  = ETH_P_SJA1105;
@@ -1431,7 +1431,9 @@ static int sja1105_vlan_filtering(struct dsa_switch *ds, int port, bool enabled)
 
 	table = &priv->static_config.tables[BLK_IDX_GENERAL_PARAMS];
 	general_params = table->entries;
+	/* EtherType used to identify outer tagged (S-tag) VLAN traffic */
 	general_params->tpid = tpid;
+	/* EtherType used to identify inner tagged (C-tag) VLAN traffic */
 	general_params->tpid2 = tpid2;
 
 	rc = sja1105_static_config_reload(priv);
Signature · 1 line
2.17.1
author maintainer
#11 lore raw

PATCH v3 6/17 net: dsa: sja1105: Limit use of incl_srcpt to bridge+vlan mode

In reply to Vladimir Oltean

The incl_srcpt setting makes the switch mangle the destination MACs of multicast frames trapped to the CPU - a primitive tagging mechanism that works even when we cannot use the 802.1Q software features.

The downside is that the two multicast MAC addresses that the switch traps for L2 PTP (01-80-C2-00-00-0E and 01-1B-19-00-00-00) quickly turn into a lot more, as the switch encodes the source port and switch id into bytes 3 and 4 of the MAC. The resulting range of MAC addresses would need to be installed manually into the DSA master port's multicast MAC filter, and even then, most devices might not have a large enough MAC filtering table.

As a result, only limit use of incl_srcpt to when it's strictly
necessary: when under a VLAN filtering bridge. This fixes PTP in non-bridged mode (standalone ports). Otherwise, PTP frames, as well as metadata follow-up frames holding RX timestamps won't be received because they will be blocked by the master port's MAC filter. Linuxptp doesn't help, because it only requests the addition of the unmodified PTP MACs to the multicast filter.
This issue is not seen in bridged mode because the master port is put in promiscuous mode when the slave ports are enslaved to a bridge. Therefore, there is no downside to having the incl_srcpt mechanism active there.

---
Changes in v3:

None.

Changes in v2:

Patch is new.

 drivers/net/dsa/sja1105/sja1105_main.c |  9 +++++++--
 net/dsa/tag_sja1105.c                  | 20 +++++++++++---------
 2 files changed, 18 insertions(+), 11 deletions(-)
drivers/net/dsa/sja1105/sja1105_main.c +7 -2
index 0f34e713c408..a30b89455421 100644
@@ -392,11 +392,11 @@ static int sja1105_init_general_params(struct sja1105_private *priv)
 		.hostprio = 0,
 		.mac_fltres1 = SJA1105_LINKLOCAL_FILTER_A,
 		.mac_flt1    = SJA1105_LINKLOCAL_FILTER_A_MASK,
-		.incl_srcpt1 = true,
+		.incl_srcpt1 = false,
 		.send_meta1  = false,
 		.mac_fltres0 = SJA1105_LINKLOCAL_FILTER_B,
 		.mac_flt0    = SJA1105_LINKLOCAL_FILTER_B_MASK,
-		.incl_srcpt0 = true,
+		.incl_srcpt0 = false,
 		.send_meta0  = false,
 		/* The destination for traffic matching mac_fltres1 and
 		 * mac_fltres0 on all ports except host_port. Such traffic
@@ -1435,6 +1435,11 @@ static int sja1105_vlan_filtering(struct dsa_switch *ds, int port, bool enabled)
 	general_params->tpid = tpid;
 	/* EtherType used to identify inner tagged (C-tag) VLAN traffic */
 	general_params->tpid2 = tpid2;
+	/* When VLAN filtering is on, we need to at least be able to
+	 * decode management traffic through the "backup plan".
+	 */
+	general_params->incl_srcpt1 = enabled;
+	general_params->incl_srcpt0 = enabled;
 
 	rc = sja1105_static_config_reload(priv);
 	if (rc)
net/dsa/tag_sja1105.c +11 -9
index 535d8a1aabe1..b35cf5c2d01c 100644
@@ -66,8 +66,7 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 				   struct net_device *netdev,
 				   struct packet_type *pt)
 {
-	struct ethhdr *hdr = eth_hdr(skb);
-	u64 source_port, switch_id;
+	int source_port, switch_id;
 	struct sk_buff *nskb;
 	u16 tpid, vid, tci;
 	bool is_tagged;
@@ -75,12 +74,17 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 	nskb = dsa_8021q_rcv(skb, netdev, pt, &tpid, &tci);
 	is_tagged = (nskb && tpid == ETH_P_SJA1105);
 
-	skb->priority = (tci & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
-	vid = tci & VLAN_VID_MASK;
-
 	skb->offload_fwd_mark = 1;
 
-	if (sja1105_is_link_local(skb)) {
+	if (is_tagged) {
+		/* Normal traffic path. */
+		vid = tci & VLAN_VID_MASK;
+		source_port = dsa_8021q_rx_source_port(vid);
+		switch_id = dsa_8021q_rx_switch_id(vid);
+		skb->priority = (tci & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
+	} else if (sja1105_is_link_local(skb)) {
+		struct ethhdr *hdr = eth_hdr(skb);
+
 		/* Management traffic path. Switch embeds the switch ID and
 		 * port ID into bytes of the destination MAC, courtesy of
 		 * the incl_srcpt options.
@@ -91,9 +95,7 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 		hdr->h_dest[3] = 0;
 		hdr->h_dest[4] = 0;
 	} else {
-		/* Normal traffic path. */
-		source_port = dsa_8021q_rx_source_port(vid);
-		switch_id = dsa_8021q_rx_switch_id(vid);
+		return NULL;
 	}
 
 	skb->dev = dsa_master_find_slave(netdev, switch_id, source_port);
Signature · 1 line
2.17.1
Reviewed-by on v3 6/17
#12 lore raw

Re: [PATCH v3 net-next 06/17] net: dsa: sja1105: Limit use of incl_srcpt to bridge+vlan mode

Quoted 24 lines · Vladimir Oltean

The incl_srcpt setting makes the switch mangle the destination MACs of multicast frames trapped to the CPU - a primitive tagging mechanism that works even when we cannot use the 802.1Q software features.

The downside is that the two multicast MAC addresses that the switch traps for L2 PTP (01-80-C2-00-00-0E and 01-1B-19-00-00-00) quickly turn into a lot more, as the switch encodes the source port and switch id into bytes 3 and 4 of the MAC. The resulting range of MAC addresses would need to be installed manually into the DSA master port's multicast MAC filter, and even then, most devices might not have a large enough MAC filtering table.

As a result, only limit use of incl_srcpt to when it's strictly
necessary: when under a VLAN filtering bridge. This fixes PTP in non-bridged mode (standalone ports). Otherwise, PTP frames, as well as metadata follow-up frames holding RX timestamps won't be received because they will be blocked by the master port's MAC filter. Linuxptp doesn't help, because it only requests the addition of the unmodified PTP MACs to the multicast filter.
This issue is not seen in bridged mode because the master port is put in promiscuous mode when the slave ports are enslaved to a bridge. Therefore, there is no downside to having the incl_srcpt mechanism active there.

Quoted 1 line
Signature · 1 line
Florian
author maintainer
#13 lore raw

PATCH v3 7/17 net: dsa: sja1105: Add support for the PTP clock

In reply to Vladimir Oltean

The design of this PHC driver is influenced by the switch's behavior w.r.t. timestamping. It exposes two PTP counters, one free-running (PTPTSCLK) and the other offset- and frequency-corrected in hardware through PTPCLKVAL, PTPCLKADD and PTPCLKRATE. The MACs can sample either of these for frame timestamps.

However, the user manual warns that taking timestamps based on the corrected clock is less than useful, as the switch can deliver corrupted timestamps in a variety of circumstances.

Therefore, this PHC uses the free-running PTPTSCLK together with a timecounter/cyclecounter structure that translates it into a software time domain. Thus, the settime/adjtime and adjfine callbacks are hardware no-ops.

The timestamps (introduced in a further patch) will also be translated to the correct time domain before being handed over to the userspace PTP stack.

The introduction of a second set of PHC operations that operate on the hardware PTPCLKVAL/PTPCLKADD/PTPCLKRATE in the future is somewhat unavoidable, as the TTEthernet core uses the corrected PTP time domain. However, the free-running counter + timecounter structure combination will suffice for now, as the resulting timestamps yield a sub-50 ns synchronization offset in steady state using linuxptp.

For this patch, in absence of frame timestamping, the operations of the switch PHC were tested by syncing it to the system time as a local slave clock with:

phc2sys -s CLOCK_REALTIME -c swp2 -O 0 -m -S 0.01

---
Changes in v3:

None.

Changes in v2:

Now resetting the PTP clock each time the switch core itself gets reset.

 drivers/net/dsa/sja1105/Kconfig        |   7 +
 drivers/net/dsa/sja1105/Makefile       |   1 +
 drivers/net/dsa/sja1105/sja1105.h      |  18 ++
 drivers/net/dsa/sja1105/sja1105_main.c |   7 +
 drivers/net/dsa/sja1105/sja1105_ptp.c  | 286 +++++++++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_ptp.h  |  47 ++++
 drivers/net/dsa/sja1105/sja1105_spi.c  |  19 ++
 7 files changed, 385 insertions(+)
 create mode 100644 drivers/net/dsa/sja1105/sja1105_ptp.c
 create mode 100644 drivers/net/dsa/sja1105/sja1105_ptp.h
drivers/net/dsa/sja1105/Kconfig +7 -0
index 1144fc5f61a8..049cea8240e4 100644
@@ -16,3 +16,10 @@ tristate "NXP SJA1105 Ethernet switch family support"
 	    - SJA1105Q (Gen. 2, No SGMII, TT-Ethernet)
 	    - SJA1105R (Gen. 2, SGMII, No TT-Ethernet)
 	    - SJA1105S (Gen. 2, SGMII, TT-Ethernet)
+
+config NET_DSA_SJA1105_PTP
+tristate "Support for the PTP clock on the NXP SJA1105 Ethernet switch"
+	depends on NET_DSA_SJA1105
+	help
+	  This enables support for timestamping and PTP clock manipulations in
+	  the SJA1105 DSA driver.
drivers/net/dsa/sja1105/Makefile +1 -0
index 941848de8b46..946eea7d8480 100644
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
 obj-$(CONFIG_NET_DSA_SJA1105) += sja1105.o
+obj-$(CONFIG_NET_DSA_SJA1105_PTP) += sja1105_ptp.o
 
 sja1105-objs := \
     sja1105_spi.o \
drivers/net/dsa/sja1105/sja1105.h +18 -0
index 61d00682de60..3c6296203c21 100644
@@ -5,6 +5,8 @@
 #ifndef _SJA1105_H
 #define _SJA1105_H
 
+#include <linux/ptp_clock_kernel.h>
+#include <linux/timecounter.h>
 #include <linux/dsa/sja1105.h>
 #include <net/dsa.h>
 #include <linux/mutex.h>
@@ -27,6 +29,10 @@ struct sja1105_regs {
 	u64 rgu;
 	u64 config;
 	u64 rmii_pll1;
+	u64 ptp_control;
+	u64 ptpclk;
+	u64 ptpclkrate;
+	u64 ptptsclk;
 	u64 pad_mii_tx[SJA1105_NUM_PORTS];
 	u64 cgu_idiv[SJA1105_NUM_PORTS];
 	u64 rgmii_pad_mii_tx[SJA1105_NUM_PORTS];
@@ -53,6 +59,7 @@ struct sja1105_info {
 	const struct sja1105_dynamic_table_ops *dyn_ops;
 	const struct sja1105_table_ops *static_ops;
 	const struct sja1105_regs *regs;
+	int (*ptp_cmd)(const void *ctx, const void *data);
 	int (*reset_cmd)(const void *ctx, const void *data);
 	int (*setup_rgmii_delay)(const void *ctx, int port);
 	/* Prototypes from include/net/dsa.h */
@@ -72,6 +79,16 @@ struct sja1105_private {
 	struct spi_device *spidev;
 	struct dsa_switch *ds;
 	struct sja1105_port ports[SJA1105_NUM_PORTS];
+	struct ptp_clock_info ptp_caps;
+	struct ptp_clock *clock;
+	/* The cycle counter translates the PTP timestamps (based on
+	 * a free-running counter) into a software time domain.
+	 */
+	struct cyclecounter tstamp_cc;
+	struct timecounter tstamp_tc;
+	struct delayed_work refresh_work;
+	/* Serializes all operations on the cycle counter */
+	struct mutex ptp_lock;
 	/* Serializes transmission of management frames so that
 	 * the switch doesn't confuse them with one another.
 	 */
@@ -79,6 +96,7 @@ struct sja1105_private {
 };
 
 #include "sja1105_dynamic_config.h"
+#include "sja1105_ptp.h"
 
 struct sja1105_spi_message {
 	u64 access;
drivers/net/dsa/sja1105/sja1105_main.c +7 -0
index a30b89455421..5ed2adbf3505 100644
@@ -1530,6 +1530,11 @@ static int sja1105_setup(struct dsa_switch *ds)
 		return rc;
 	}
 
+	rc = sja1105_ptp_clock_register(priv);
+	if (rc < 0) {
+		dev_err(ds->dev, "Failed to register PTP clock: %d\n", rc);
+		return rc;
+	}
 	/* Create and send configuration down to device */
 	rc = sja1105_static_config_load(priv, ports);
 	if (rc < 0) {
@@ -1681,6 +1686,7 @@ static const struct dsa_switch_ops sja1105_switch_ops = {
 	.get_strings		= sja1105_get_strings,
 	.get_ethtool_stats	= sja1105_get_ethtool_stats,
 	.get_sset_count		= sja1105_get_sset_count,
+	.get_ts_info		= sja1105_get_ts_info,
 	.port_fdb_dump		= sja1105_fdb_dump,
 	.port_fdb_add		= sja1105_fdb_add,
 	.port_fdb_del		= sja1105_fdb_del,
@@ -1805,6 +1811,7 @@ static int sja1105_remove(struct spi_device *spi)
 {
 	struct sja1105_private *priv = spi_get_drvdata(spi);
 
+	sja1105_ptp_clock_unregister(priv);
 	dsa_unregister_switch(priv->ds);
 	sja1105_static_config_free(&priv->static_config);
 	return 0;
drivers/net/dsa/sja1105/sja1105_ptp.c new file +286 -0

286 lines not on this page · raw message

drivers/net/dsa/sja1105/sja1105_ptp.h new file +47 -0

47 lines not on this page · raw message

drivers/net/dsa/sja1105/sja1105_spi.c +19 -0
index b1344ed1697f..d226996881b8 100644
@@ -478,7 +478,12 @@ int sja1105_static_config_upload(struct sja1105_private *priv)
 		dev_info(dev, "Succeeded after %d tried\n", RETRIES - retries);
 	}
 
+	rc = sja1105_ptp_reset(priv);
+	if (rc < 0)
+		dev_err(dev, "Failed to reset PTP clock: %d\n", rc);
+
 	dev_info(dev, "Reset switch and programmed static config\n");
+
 out:
 	kfree(config_buf);
 	return rc;
@@ -507,6 +512,10 @@ static struct sja1105_regs sja1105et_regs = {
 	.rgmii_tx_clk = {0x100016, 0x10001D, 0x100024, 0x10002B, 0x100032},
 	.rmii_ref_clk = {0x100015, 0x10001C, 0x100023, 0x10002A, 0x100031},
 	.rmii_ext_tx_clk = {0x100018, 0x10001F, 0x100026, 0x10002D, 0x100034},
+	.ptp_control = 0x17,
+	.ptpclk = 0x18, /* Spans 0x18 to 0x19 */
+	.ptpclkrate = 0x1A,
+	.ptptsclk = 0x1B, /* Spans 0x1B to 0x1C */
 };
 
 static struct sja1105_regs sja1105pqrs_regs = {
@@ -533,6 +542,10 @@ static struct sja1105_regs sja1105pqrs_regs = {
 	.rmii_ref_clk = {0x100015, 0x10001B, 0x100021, 0x100027, 0x10002D},
 	.rmii_ext_tx_clk = {0x100017, 0x10001D, 0x100023, 0x100029, 0x10002F},
 	.qlevel = {0x604, 0x614, 0x624, 0x634, 0x644},
+	.ptp_control = 0x18,
+	.ptpclk = 0x19,
+	.ptpclkrate = 0x1B,
+	.ptptsclk = 0x1C,
 };
 
 struct sja1105_info sja1105e_info = {
@@ -543,6 +556,7 @@ struct sja1105_info sja1105e_info = {
 	.reset_cmd		= sja1105et_reset_cmd,
 	.fdb_add_cmd		= sja1105et_fdb_add,
 	.fdb_del_cmd		= sja1105et_fdb_del,
+	.ptp_cmd		= sja1105et_ptp_cmd,
 	.regs			= &sja1105et_regs,
 	.name			= "SJA1105E",
 };
@@ -554,6 +568,7 @@ struct sja1105_info sja1105t_info = {
 	.reset_cmd		= sja1105et_reset_cmd,
 	.fdb_add_cmd		= sja1105et_fdb_add,
 	.fdb_del_cmd		= sja1105et_fdb_del,
+	.ptp_cmd		= sja1105et_ptp_cmd,
 	.regs			= &sja1105et_regs,
 	.name			= "SJA1105T",
 };
@@ -565,6 +580,7 @@ struct sja1105_info sja1105p_info = {
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
+	.ptp_cmd		= sja1105pqrs_ptp_cmd,
 	.regs			= &sja1105pqrs_regs,
 	.name			= "SJA1105P",
 };
@@ -576,6 +592,7 @@ struct sja1105_info sja1105q_info = {
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
+	.ptp_cmd		= sja1105pqrs_ptp_cmd,
 	.regs			= &sja1105pqrs_regs,
 	.name			= "SJA1105Q",
 };
@@ -587,6 +604,7 @@ struct sja1105_info sja1105r_info = {
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
+	.ptp_cmd		= sja1105pqrs_ptp_cmd,
 	.regs			= &sja1105pqrs_regs,
 	.name			= "SJA1105R",
 };
@@ -599,5 +617,6 @@ struct sja1105_info sja1105s_info = {
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
+	.ptp_cmd		= sja1105pqrs_ptp_cmd,
 	.name			= "SJA1105S",
 };
Signature · 1 line
2.17.1
author maintainer
#14 lore raw

PATCH v3 8/17 net: dsa: sja1105: Move sja1105_is_link_local to include/linux

In reply to Vladimir Oltean

This function will be reused from the .port_rxtstamp callback to see if the received SKB can be timestamped by the switch.

---
Changes in v3:

None.

Changes in v2:

Patch is new.

 include/linux/dsa/sja1105.h | 15 +++++++++++++++
 net/dsa/tag_sja1105.c       | 15 ---------------
 2 files changed, 15 insertions(+), 15 deletions(-)
include/linux/dsa/sja1105.h +15 -0
index e46e18c47d41..f3237afed35a 100644
@@ -25,4 +25,19 @@ struct sja1105_port {
 	int mgmt_slot;
 };
 
+/* Similar to is_link_local_ether_addr(hdr->h_dest) but also covers PTP */
+static inline bool sja1105_is_link_local(const struct sk_buff *skb)
+{
+	const struct ethhdr *hdr = eth_hdr(skb);
+	u64 dmac = ether_addr_to_u64(hdr->h_dest);
+
+	if ((dmac & SJA1105_LINKLOCAL_FILTER_A_MASK) ==
+		    SJA1105_LINKLOCAL_FILTER_A)
+		return true;
+	if ((dmac & SJA1105_LINKLOCAL_FILTER_B_MASK) ==
+		    SJA1105_LINKLOCAL_FILTER_B)
+		return true;
+	return false;
+}
+
 #endif /* _NET_DSA_SJA1105_H */
net/dsa/tag_sja1105.c +0 -15
index b35cf5c2d01c..7a56a27a1fd2 100644
@@ -7,21 +7,6 @@
 #include <linux/packing.h>
 #include "dsa_priv.h"
 
-/* Similar to is_link_local_ether_addr(hdr->h_dest) but also covers PTP */
-static inline bool sja1105_is_link_local(const struct sk_buff *skb)
-{
-	const struct ethhdr *hdr = eth_hdr(skb);
-	u64 dmac = ether_addr_to_u64(hdr->h_dest);
-
-	if ((dmac & SJA1105_LINKLOCAL_FILTER_A_MASK) ==
-		    SJA1105_LINKLOCAL_FILTER_A)
-		return true;
-	if ((dmac & SJA1105_LINKLOCAL_FILTER_B_MASK) ==
-		    SJA1105_LINKLOCAL_FILTER_B)
-		return true;
-	return false;
-}
-
 /* This is the first time the tagger sees the frame on RX.
  * Figure out if we can decode it, and if we can, annotate skb->cb with how we
  * plan to do that, so we don't need to check again in the rcv function.
Signature · 1 line
2.17.1
author maintainer
#16 lore raw

PATCH v3 9/17 net: dsa: sja1105: Add logic for TX timestamping

In reply to Vladimir Oltean

On TX, timestamping is performed synchronously from the port_deferred_xmit worker thread.
In management routes, the switch is requested to take egress timestamps (again partial), which are reconstructed and appended to a clone of the skb that was just sent. The cloning is done by DSA and we retrieve the pointer from the structure that DSA keeps in skb->cb.
Then these clones are enqueued to the socket's error queue for application-level processing.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 drivers/net/dsa/sja1105/sja1105.h      |  10 +++
 drivers/net/dsa/sja1105/sja1105_main.c |  55 ++++++++++++-
 drivers/net/dsa/sja1105/sja1105_ptp.c  | 104 +++++++++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_ptp.h  |  17 ++++
 drivers/net/dsa/sja1105/sja1105_spi.c  |  14 ++++
 include/linux/dsa/sja1105.h            |   1 +
 6 files changed, 199 insertions(+), 2 deletions(-)
drivers/net/dsa/sja1105/sja1105.h +10 -0
index 3c6296203c21..5a4f83a3417b 100644
@@ -33,6 +33,7 @@ struct sja1105_regs {
 	u64 ptpclk;
 	u64 ptpclkrate;
 	u64 ptptsclk;
+	u64 ptpegr_ts[SJA1105_NUM_PORTS];
 	u64 pad_mii_tx[SJA1105_NUM_PORTS];
 	u64 cgu_idiv[SJA1105_NUM_PORTS];
 	u64 rgmii_pad_mii_tx[SJA1105_NUM_PORTS];
@@ -56,6 +57,15 @@ struct sja1105_info {
 	 * switch core and device_id)
 	 */
 	u64 part_no;
+	/* E/T and P/Q/R/S have partial timestamps of different sizes.
+	 * They must be reconstructed on both families anyway to get the full
+	 * 64-bit values back.
+	 */
+	int ptp_ts_bits;
+	/* Also SPI commands are of different sizes to retrieve
+	 * the egress timestamps.
+	 */
+	int ptpegr_ts_bytes;
 	const struct sja1105_dynamic_table_ops *dyn_ops;
 	const struct sja1105_table_ops *static_ops;
 	const struct sja1105_regs *regs;
drivers/net/dsa/sja1105/sja1105_main.c +53 -2

93 lines not on this page · raw message

drivers/net/dsa/sja1105/sja1105_ptp.c +104 -0

110 lines not on this page · raw message

drivers/net/dsa/sja1105/sja1105_ptp.h +17 -0
index 137ffbb0a233..af456b0a4d27 100644
@@ -10,6 +10,8 @@ int sja1105_ptp_clock_register(struct sja1105_private *priv);
 
 void sja1105_ptp_clock_unregister(struct sja1105_private *priv);
 
+int sja1105_ptpegr_ts_poll(struct sja1105_private *priv, int port, u64 *ts);
+
 int sja1105et_ptp_cmd(const void *ctx, const void *data);
 
 int sja1105pqrs_ptp_cmd(const void *ctx, const void *data);
@@ -17,6 +19,9 @@ int sja1105pqrs_ptp_cmd(const void *ctx, const void *data);
 int sja1105_get_ts_info(struct dsa_switch *ds, int port,
 			struct ethtool_ts_info *ts);
 
+u64 sja1105_tstamp_reconstruct(struct sja1105_private *priv, u64 now,
+			       u64 ts_partial);
+
 int sja1105_ptp_reset(struct sja1105_private *priv);
 
 #else
@@ -31,6 +36,18 @@ static inline void sja1105_ptp_clock_unregister(struct sja1105_private *priv)
 	return;
 }
 
+static inline int
+sja1105_ptpegr_ts_poll(struct sja1105_private *priv, int port, u64 *ts)
+{
+	return 0;
+}
+
+static inline u64 sja1105_tstamp_reconstruct(struct sja1105_private *priv,
+					     u64 now, u64 ts_partial)
+{
+	return 0;
+}
+
 static inline int sja1105_ptp_reset(struct sja1105_private *priv)
 {
 	return 0;
drivers/net/dsa/sja1105/sja1105_spi.c +14 -0
index d226996881b8..a71b16c5ae5d 100644
@@ -512,6 +512,7 @@ static struct sja1105_regs sja1105et_regs = {
 	.rgmii_tx_clk = {0x100016, 0x10001D, 0x100024, 0x10002B, 0x100032},
 	.rmii_ref_clk = {0x100015, 0x10001C, 0x100023, 0x10002A, 0x100031},
 	.rmii_ext_tx_clk = {0x100018, 0x10001F, 0x100026, 0x10002D, 0x100034},
+	.ptpegr_ts = {0xC0, 0xC2, 0xC4, 0xC6, 0xC8},
 	.ptp_control = 0x17,
 	.ptpclk = 0x18, /* Spans 0x18 to 0x19 */
 	.ptpclkrate = 0x1A,
@@ -542,6 +543,7 @@ static struct sja1105_regs sja1105pqrs_regs = {
 	.rmii_ref_clk = {0x100015, 0x10001B, 0x100021, 0x100027, 0x10002D},
 	.rmii_ext_tx_clk = {0x100017, 0x10001D, 0x100023, 0x100029, 0x10002F},
 	.qlevel = {0x604, 0x614, 0x624, 0x634, 0x644},
+	.ptpegr_ts = {0xC0, 0xC4, 0xC8, 0xCC, 0xD0},
 	.ptp_control = 0x18,
 	.ptpclk = 0x19,
 	.ptpclkrate = 0x1B,
@@ -553,6 +555,8 @@ struct sja1105_info sja1105e_info = {
 	.part_no		= SJA1105ET_PART_NO,
 	.static_ops		= sja1105e_table_ops,
 	.dyn_ops		= sja1105et_dyn_ops,
+	.ptp_ts_bits		= 24,
+	.ptpegr_ts_bytes	= 4,
 	.reset_cmd		= sja1105et_reset_cmd,
 	.fdb_add_cmd		= sja1105et_fdb_add,
 	.fdb_del_cmd		= sja1105et_fdb_del,
@@ -565,6 +569,8 @@ struct sja1105_info sja1105t_info = {
 	.part_no		= SJA1105ET_PART_NO,
 	.static_ops		= sja1105t_table_ops,
 	.dyn_ops		= sja1105et_dyn_ops,
+	.ptp_ts_bits		= 24,
+	.ptpegr_ts_bytes	= 4,
 	.reset_cmd		= sja1105et_reset_cmd,
 	.fdb_add_cmd		= sja1105et_fdb_add,
 	.fdb_del_cmd		= sja1105et_fdb_del,
@@ -577,6 +583,8 @@ struct sja1105_info sja1105p_info = {
 	.part_no		= SJA1105P_PART_NO,
 	.static_ops		= sja1105p_table_ops,
 	.dyn_ops		= sja1105pqrs_dyn_ops,
+	.ptp_ts_bits		= 32,
+	.ptpegr_ts_bytes	= 8,
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
@@ -589,6 +597,8 @@ struct sja1105_info sja1105q_info = {
 	.part_no		= SJA1105Q_PART_NO,
 	.static_ops		= sja1105q_table_ops,
 	.dyn_ops		= sja1105pqrs_dyn_ops,
+	.ptp_ts_bits		= 32,
+	.ptpegr_ts_bytes	= 8,
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
@@ -601,6 +611,8 @@ struct sja1105_info sja1105r_info = {
 	.part_no		= SJA1105R_PART_NO,
 	.static_ops		= sja1105r_table_ops,
 	.dyn_ops		= sja1105pqrs_dyn_ops,
+	.ptp_ts_bits		= 32,
+	.ptpegr_ts_bytes	= 8,
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
@@ -614,6 +626,8 @@ struct sja1105_info sja1105s_info = {
 	.static_ops		= sja1105s_table_ops,
 	.dyn_ops		= sja1105pqrs_dyn_ops,
 	.regs			= &sja1105pqrs_regs,
+	.ptp_ts_bits		= 32,
+	.ptpegr_ts_bytes	= 8,
 	.reset_cmd		= sja1105pqrs_reset_cmd,
 	.fdb_add_cmd		= sja1105pqrs_fdb_add,
 	.fdb_del_cmd		= sja1105pqrs_fdb_del,
include/linux/dsa/sja1105.h +1 -0
index f3237afed35a..f5d5f7823da6 100644
@@ -22,6 +22,7 @@
 
 struct sja1105_port {
 	struct dsa_port *dp;
+	bool hwts_tx_en;
 	int mgmt_slot;
 };
 
Signature · 1 line
2.17.1
author maintainer
#17 lore raw

PATCH v3 10/17 net: dsa: sja1105: Build a minimal understanding of meta frames

In reply to Vladimir Oltean

Meta frames are sent on the CPU port by the switch if RX timestamping is enabled. They contain a partial timestamp of the previous frame.

They are Ethernet frames with the Ethernet header constructed out of:

- SJA1105_META_DMAC
- SJA1105_META_SMAC
- ETH_P_SJA1105_META

The Ethernet payload will be decoded in a follow-up patch.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 include/linux/dsa/sja1105.h | 11 +++++++++++
 net/dsa/tag_sja1105.c       | 15 +++++++++++++++
 2 files changed, 26 insertions(+)
include/linux/dsa/sja1105.h +11 -0
index f5d5f7823da6..4f09743b4713 100644
@@ -12,6 +12,7 @@
 #include <net/dsa.h>
 
 #define ETH_P_SJA1105				ETH_P_DSA_8021Q
+#define ETH_P_SJA1105_META			0x0008
 
 /* IEEE 802.3 Annex 57A: Slow Protocols PDUs (01:80:C2:xx:xx:xx) */
 #define SJA1105_LINKLOCAL_FILTER_A		0x0180C2000000ull
@@ -20,6 +21,16 @@
 #define SJA1105_LINKLOCAL_FILTER_B		0x011B19000000ull
 #define SJA1105_LINKLOCAL_FILTER_B_MASK		0xFFFFFF000000ull
 
+/* Source and Destination MAC of follow-up meta frames.
+ * Whereas the choice of SMAC only affects the unique identification of the
+ * switch as sender of meta frames, the DMAC must be an address that is present
+ * in the DSA master port's multicast MAC filter.
+ * 01-80-C2-00-00-0E is a good choice for this, as all profiles of IEEE 1588
+ * over L2 use this address for some purpose already.
+ */
+#define SJA1105_META_SMAC			0x222222222222ull
+#define SJA1105_META_DMAC			0x0180C200000Eull
+
 struct sja1105_port {
 	struct dsa_port *dp;
 	bool hwts_tx_en;
net/dsa/tag_sja1105.c +15 -0
index 7a56a27a1fd2..9ff7cfa6ea20 100644
@@ -7,6 +7,21 @@
 #include <linux/packing.h>
 #include "dsa_priv.h"
 
+static inline bool sja1105_is_meta_frame(const struct sk_buff *skb)
+{
+	const struct ethhdr *hdr = eth_hdr(skb);
+	u64 smac = ether_addr_to_u64(hdr->h_source);
+	u64 dmac = ether_addr_to_u64(hdr->h_dest);
+
+	if (smac != SJA1105_META_SMAC)
+		return false;
+	if (dmac != SJA1105_META_DMAC)
+		return false;
+	if (ntohs(hdr->h_proto) != ETH_P_SJA1105_META)
+		return false;
+	return true;
+}
+
 /* This is the first time the tagger sees the frame on RX.
  * Figure out if we can decode it, and if we can, annotate skb->cb with how we
  * plan to do that, so we don't need to check again in the rcv function.
Signature · 1 line
2.17.1
Reviewed-by on v3 10/17
#18 lore raw

Re: [PATCH v3 net-next 10/17] net: dsa: sja1105: Build a minimal understanding of meta frames

Quoted 5 lines · Vladimir Oltean

Meta frames are sent on the CPU port by the switch if RX timestamping is enabled. They contain a partial timestamp of the previous frame.

They are Ethernet frames with the Ethernet header constructed out of:

Quoted 7 lines

- SJA1105_META_DMAC
- SJA1105_META_SMAC
- ETH_P_SJA1105_META

The Ethernet payload will be decoded in a follow-up patch.

Signature · 1 line
Florian
author maintainer
#19 lore raw

PATCH v3 11/17 net: dsa: sja1105: Add support for the AVB Parameters Table

In reply to Vladimir Oltean

This table is used to program the switch to emit "meta" follow-up Ethernet frames (which contain partial RX timestamps) after each link-local frame that was trapped to the CPU port through MAC filtering. This includes PTP frames.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 .../net/dsa/sja1105/sja1105_dynamic_config.c  |  2 +
 drivers/net/dsa/sja1105/sja1105_main.c        | 36 +++++++++++
 .../net/dsa/sja1105/sja1105_static_config.c   | 59 +++++++++++++++++++
 .../net/dsa/sja1105/sja1105_static_config.h   | 10 ++++
 4 files changed, 107 insertions(+)
drivers/net/dsa/sja1105/sja1105_dynamic_config.c +2 -0
index 352bb6e89297..56c83b9d52e4 100644
@@ -378,6 +378,7 @@ struct sja1105_dynamic_table_ops sja1105et_dyn_ops[BLK_IDX_MAX_DYN] = {
 		.addr = 0x38,
 	},
 	[BLK_IDX_L2_FORWARDING_PARAMS] = {0},
+	[BLK_IDX_AVB_PARAMS] = {0},
 	[BLK_IDX_GENERAL_PARAMS] = {
 		.entry_packing = sja1105et_general_params_entry_packing,
 		.cmd_packing = sja1105et_general_params_cmd_packing,
@@ -441,6 +442,7 @@ struct sja1105_dynamic_table_ops sja1105pqrs_dyn_ops[BLK_IDX_MAX_DYN] = {
 		.addr = 0x38,
 	},
 	[BLK_IDX_L2_FORWARDING_PARAMS] = {0},
+	[BLK_IDX_AVB_PARAMS] = {0},
 	[BLK_IDX_GENERAL_PARAMS] = {
 		.entry_packing = sja1105et_general_params_entry_packing,
 		.cmd_packing = sja1105et_general_params_cmd_packing,
drivers/net/dsa/sja1105/sja1105_main.c +36 -0
index 45c65e77b61d..f83f924c0833 100644
@@ -508,6 +508,39 @@ static int sja1105_init_l2_policing(struct sja1105_private *priv)
 	return 0;
 }
 
+static int sja1105_init_avb_params(struct sja1105_private *priv,
+				   bool on)
+{
+	struct sja1105_avb_params_entry *avb;
+	struct sja1105_table *table;
+
+	table = &priv->static_config.tables[BLK_IDX_AVB_PARAMS];
+
+	/* Discard previous AVB Parameters Table */
+	if (table->entry_count) {
+		kfree(table->entries);
+		table->entry_count = 0;
+	}
+
+	/* Configure the reception of meta frames only if requested */
+	if (!on)
+		return 0;
+
+	table->entries = kcalloc(SJA1105_MAX_AVB_PARAMS_COUNT,
+				 table->ops->unpacked_entry_size, GFP_KERNEL);
+	if (!table->entries)
+		return -ENOMEM;
+
+	table->entry_count = SJA1105_MAX_AVB_PARAMS_COUNT;
+
+	avb = table->entries;
+
+	avb->destmeta = SJA1105_META_DMAC;
+	avb->srcmeta  = SJA1105_META_SMAC;
+
+	return 0;
+}
+
 static int sja1105_static_config_load(struct sja1105_private *priv,
 				      struct sja1105_dt_port *ports)
 {
@@ -546,6 +579,9 @@ static int sja1105_static_config_load(struct sja1105_private *priv,
 	if (rc < 0)
 		return rc;
 	rc = sja1105_init_general_params(priv);
+	if (rc < 0)
+		return rc;
+	rc = sja1105_init_avb_params(priv, false);
 	if (rc < 0)
 		return rc;
 
drivers/net/dsa/sja1105/sja1105_static_config.c +59 -0

107 lines not on this page · raw message

drivers/net/dsa/sja1105/sja1105_static_config.h +10 -0
index d513b1c91b98..a9586d0b4b3b 100644
@@ -20,10 +20,12 @@
 #define SJA1105ET_SIZE_MAC_CONFIG_ENTRY			28
 #define SJA1105ET_SIZE_L2_LOOKUP_PARAMS_ENTRY		4
 #define SJA1105ET_SIZE_GENERAL_PARAMS_ENTRY		40
+#define SJA1105ET_SIZE_AVB_PARAMS_ENTRY			12
 #define SJA1105PQRS_SIZE_L2_LOOKUP_ENTRY		20
 #define SJA1105PQRS_SIZE_MAC_CONFIG_ENTRY		32
 #define SJA1105PQRS_SIZE_L2_LOOKUP_PARAMS_ENTRY		16
 #define SJA1105PQRS_SIZE_GENERAL_PARAMS_ENTRY		44
+#define SJA1105PQRS_SIZE_AVB_PARAMS_ENTRY		16
 
 /* UM10944.pdf Page 11, Table 2. Configuration Blocks */
 enum {
@@ -34,6 +36,7 @@ enum {
 	BLKID_MAC_CONFIG				= 0x09,
 	BLKID_L2_LOOKUP_PARAMS				= 0x0D,
 	BLKID_L2_FORWARDING_PARAMS			= 0x0E,
+	BLKID_AVB_PARAMS				= 0x10,
 	BLKID_GENERAL_PARAMS				= 0x11,
 	BLKID_XMII_PARAMS				= 0x4E,
 };
@@ -46,6 +49,7 @@ enum sja1105_blk_idx {
 	BLK_IDX_MAC_CONFIG,
 	BLK_IDX_L2_LOOKUP_PARAMS,
 	BLK_IDX_L2_FORWARDING_PARAMS,
+	BLK_IDX_AVB_PARAMS,
 	BLK_IDX_GENERAL_PARAMS,
 	BLK_IDX_XMII_PARAMS,
 	BLK_IDX_MAX,
@@ -64,6 +68,7 @@ enum sja1105_blk_idx {
 #define SJA1105_MAX_L2_FORWARDING_PARAMS_COUNT		1
 #define SJA1105_MAX_GENERAL_PARAMS_COUNT		1
 #define SJA1105_MAX_XMII_PARAMS_COUNT			1
+#define SJA1105_MAX_AVB_PARAMS_COUNT			1
 
 #define SJA1105_MAX_FRAME_MEMORY			929
 
@@ -179,6 +184,11 @@ struct sja1105_l2_policing_entry {
 	u64 partition;
 };
 
+struct sja1105_avb_params_entry {
+	u64 destmeta;
+	u64 srcmeta;
+};
+
 struct sja1105_mac_config_entry {
 	u64 top[8];
 	u64 base[8];
Signature · 1 line
2.17.1
author maintainer
#20 lore raw

PATCH v3 12/17 net: dsa: sja1105: Make sja1105_is_link_local not match meta frames

In reply to Vladimir Oltean

Although meta frames are configured to be sent at SJA1105_META_DMAC (01-80-C2-00-00-0E) which is a multicast MAC address that would also be trapped by the switch to the CPU, were it to receive it on a front-panel port, meta frames are conceptually not link-local frames, they only carry their RX timestamps.

The choice of sending meta frames at a multicast DMAC is a pragmatic one, to avoid installing an extra entry to the DSA master port's multicast MAC filter.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 include/linux/dsa/sja1105.h | 2 ++
 1 file changed, 2 insertions(+)
include/linux/dsa/sja1105.h +2 -0
index 4f09743b4713..b9619d9fca4c 100644
@@ -43,6 +43,8 @@ static inline bool sja1105_is_link_local(const struct sk_buff *skb)
 	const struct ethhdr *hdr = eth_hdr(skb);
 	u64 dmac = ether_addr_to_u64(hdr->h_dest);
 
+	if (ntohs(hdr->h_proto) == ETH_P_SJA1105_META)
+		return false;
 	if ((dmac & SJA1105_LINKLOCAL_FILTER_A_MASK) ==
 		    SJA1105_LINKLOCAL_FILTER_A)
 		return true;
Signature · 1 line
2.17.1
author maintainer
#21 lore raw

PATCH v3 13/17 net: dsa: sja1105: Receive and decode meta frames

In reply to Vladimir Oltean

This adds support in the tagger for understanding the source port and switch id of meta frames. Their timestamp is also extracted but not used yet - this needs to be done in a state machine that modifies the previously received timestampable frame - will be added in a follow-up patch.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 net/dsa/tag_sja1105.c | 43 ++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 42 insertions(+), 1 deletion(-)
net/dsa/tag_sja1105.c +42 -1
index 9ff7cfa6ea20..6ec9a32dda62 100644
@@ -7,6 +7,36 @@
 #include <linux/packing.h>
 #include "dsa_priv.h"
 
+struct sja1105_meta {
+	u64 tstamp;
+	u64 dmac_byte_4;
+	u64 dmac_byte_3;
+	u64 source_port;
+	u64 switch_id;
+};
+
+static void sja1105_meta_unpack(const struct sk_buff *skb,
+				struct sja1105_meta *meta)
+{
+	u8 *buf = skb_mac_header(skb) + ETH_HLEN;
+
+	/* UM10944.pdf section 4.2.17 AVB Parameters:
+	 * Structure of the meta-data follow-up frame.
+	 * It is in network byte order, so there are no quirks
+	 * while unpacking the meta frame.
+	 *
+	 * Also SJA1105 E/T only populates bits 23:0 of the timestamp
+	 * whereas P/Q/R/S does 32 bits. Since the structure is the
+	 * same and the E/T puts zeroes in the high-order byte, use
+	 * a unified unpacking command for both device series.
+	 */
+	packing(buf,     &meta->tstamp,     31, 0, 4, UNPACK, 0);
+	packing(buf + 4, &meta->dmac_byte_4, 7, 0, 1, UNPACK, 0);
+	packing(buf + 5, &meta->dmac_byte_3, 7, 0, 1, UNPACK, 0);
+	packing(buf + 6, &meta->source_port, 7, 0, 1, UNPACK, 0);
+	packing(buf + 7, &meta->switch_id,   7, 0, 1, UNPACK, 0);
+}
+
 static inline bool sja1105_is_meta_frame(const struct sk_buff *skb)
 {
 	const struct ethhdr *hdr = eth_hdr(skb);
@@ -28,6 +58,8 @@ static inline bool sja1105_is_meta_frame(const struct sk_buff *skb)
  */
 static bool sja1105_filter(const struct sk_buff *skb, struct net_device *dev)
 {
+	if (sja1105_is_meta_frame(skb))
+		return true;
 	if (sja1105_is_link_local(skb))
 		return true;
 	if (!dsa_port_is_vlan_filtering(dev->dsa_ptr))
@@ -66,13 +98,18 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 				   struct net_device *netdev,
 				   struct packet_type *pt)
 {
+	struct sja1105_meta meta = {0};
 	int source_port, switch_id;
 	struct sk_buff *nskb;
 	u16 tpid, vid, tci;
+	bool is_link_local;
 	bool is_tagged;
+	bool is_meta;
 
 	nskb = dsa_8021q_rcv(skb, netdev, pt, &tpid, &tci);
 	is_tagged = (nskb && tpid == ETH_P_SJA1105);
+	is_link_local = sja1105_is_link_local(skb);
+	is_meta = sja1105_is_meta_frame(skb);
 
 	skb->offload_fwd_mark = 1;
 
@@ -82,7 +119,7 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 		source_port = dsa_8021q_rx_source_port(vid);
 		switch_id = dsa_8021q_rx_switch_id(vid);
 		skb->priority = (tci & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
-	} else if (sja1105_is_link_local(skb)) {
+	} else if (is_link_local) {
 		struct ethhdr *hdr = eth_hdr(skb);
 
 		/* Management traffic path. Switch embeds the switch ID and
@@ -94,6 +131,10 @@ static struct sk_buff *sja1105_rcv(struct sk_buff *skb,
 		/* Clear the DMAC bytes that were mangled by the switch */
 		hdr->h_dest[3] = 0;
 		hdr->h_dest[4] = 0;
+	} else if (is_meta) {
+		sja1105_meta_unpack(skb, &meta);
+		source_port = meta.source_port;
+		switch_id = meta.switch_id;
 	} else {
 		return NULL;
 	}
Signature · 1 line
2.17.1
author maintainer
#22 lore raw

PATCH v3 14/17 net: dsa: sja1105: Add a global sja1105_tagger_data structure

In reply to Vladimir Oltean

This will be used to keep state for RX timestamping. It is global because the switch serializes timestampable and meta frames when trapping them towards the CPU port (lower port indices have higher priority) and therefore having one state machine per port would create unnecessary complications.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 drivers/net/dsa/sja1105/sja1105.h      |  1 +
 drivers/net/dsa/sja1105/sja1105_main.c |  5 +++++
 include/linux/dsa/sja1105.h            | 15 +++++++++++++++
 3 files changed, 21 insertions(+)
drivers/net/dsa/sja1105/sja1105.h +1 -0
index 5a4f83a3417b..0fc6fe9ada87 100644
@@ -103,6 +103,7 @@ struct sja1105_private {
 	 * the switch doesn't confuse them with one another.
 	 */
 	struct mutex mgmt_lock;
+	struct sja1105_tagger_data tagger_data;
 };
 
 #include "sja1105_dynamic_config.h"
drivers/net/dsa/sja1105/sja1105_main.c +5 -0
index f83f924c0833..9dd2ab2725b6 100644
@@ -1828,6 +1828,7 @@ static int sja1105_check_device_id(struct sja1105_private *priv)
 
 static int sja1105_probe(struct spi_device *spi)
 {
+	struct sja1105_tagger_data *tagger_data;
 	struct device *dev = &spi->dev;
 	struct sja1105_private *priv;
 	struct dsa_switch *ds;
@@ -1882,12 +1883,16 @@ static int sja1105_probe(struct spi_device *spi)
 	ds->priv = priv;
 	priv->ds = ds;
 
+	tagger_data = &priv->tagger_data;
+	skb_queue_head_init(&tagger_data->skb_rxtstamp_queue);
+
 	/* Connections between dsa_port and sja1105_port */
 	for (i = 0; i < SJA1105_NUM_PORTS; i++) {
 		struct sja1105_port *sp = &priv->ports[i];
 
 		ds->ports[i].priv = sp;
 		sp->dp = &ds->ports[i];
+		sp->data = tagger_data;
 	}
 	mutex_init(&priv->mgmt_lock);
 
include/linux/dsa/sja1105.h +15 -0
index b9619d9fca4c..b83f6e428e5d 100644
@@ -31,7 +31,22 @@
 #define SJA1105_META_SMAC			0x222222222222ull
 #define SJA1105_META_DMAC			0x0180C200000Eull
 
+/* Global tagger data: each struct sja1105_port has a reference to
+ * the structure defined in struct sja1105_private.
+ */
+struct sja1105_tagger_data {
+	struct sk_buff_head skb_rxtstamp_queue;
+	struct work_struct rxtstamp_work;
+	struct sk_buff *stampable_skb;
+	/* Protects concurrent access to the meta state machine
+	 * from taggers running on multiple ports on SMP systems
+	 */
+	spinlock_t meta_lock;
+	bool hwts_rx_en;
+};
+
 struct sja1105_port {
+	struct sja1105_tagger_data *data;
 	struct dsa_port *dp;
 	bool hwts_tx_en;
 	int mgmt_slot;
Signature · 1 line
2.17.1
author maintainer
#23 lore raw

PATCH v3 15/17 net: dsa: sja1105: Increase priority of CPU-trapped frames

In reply to Vladimir Oltean

Without noticing any particular issue, this patch ensures that management traffic is treated with the maximum priority on RX by the switch. This is generally desirable, as the driver keeps a state machine that waits for metadata follow-up frames as soon as a management frame is received. Increasing the priority helps expedite the reception (and further reconstruction) of the RX timestamp to the driver after the MAC has generated it.

---
Changes in v3:

None.

Changes in v2:

None.

 drivers/net/dsa/sja1105/sja1105_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/dsa/sja1105/sja1105_main.c +1 -1
index 9dd2ab2725b6..fceca0e5984f 100644
@@ -389,7 +389,7 @@ static int sja1105_init_general_params(struct sja1105_private *priv)
 		.mirr_ptacu = 0,
 		.switchid = priv->ds->index,
 		/* Priority queue for link-local frames trapped to CPU */
-		.hostprio = 0,
+		.hostprio = 7,
 		.mac_fltres1 = SJA1105_LINKLOCAL_FILTER_A,
 		.mac_flt1    = SJA1105_LINKLOCAL_FILTER_A_MASK,
 		.incl_srcpt1 = false,
Signature · 1 line
2.17.1
author maintainer
#24 lore raw

PATCH v3 16/17 net: dsa: sja1105: Add a state machine for RX timestamping

In reply to Vladimir Oltean

Meta frame reception relies on the hardware keeping its promise that it will send no other traffic towards the CPU port between a link-local frame and a meta frame. Otherwise there is no other way to associate the meta frame with the link-local frame it's holding a timestamp of. The receive function is made stateful, and buffers a timestampable frame until its meta frame arrives, then merges the two, drops the meta and releases the link-local frame up the stack.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 drivers/net/dsa/sja1105/sja1105_main.c |  65 +++++++++++++
 include/linux/dsa/sja1105.h            |   7 ++
 net/dsa/tag_sja1105.c                  | 121 ++++++++++++++++++++++++-
 3 files changed, 192 insertions(+), 1 deletion(-)
drivers/net/dsa/sja1105/sja1105_main.c +65 -0
index fceca0e5984f..46a851f54b83 100644
@@ -1600,6 +1600,14 @@ static int sja1105_setup(struct dsa_switch *ds)
 	return sja1105_setup_8021q_tagging(ds, true);
 }
 
+static void sja1105_teardown(struct dsa_switch *ds)
+{
+	struct sja1105_private *priv = ds->priv;
+
+	cancel_work_sync(&priv->tagger_data.rxtstamp_work);
+	skb_queue_purge(&priv->tagger_data.skb_rxtstamp_queue);
+}
+
 static int sja1105_mgmt_xmit(struct dsa_switch *ds, int port, int slot,
 			     struct sk_buff *skb, bool takets)
 {
@@ -1747,6 +1755,60 @@ static int sja1105_set_ageing_time(struct dsa_switch *ds,
 	return sja1105_static_config_reload(priv);
 }
 
+#define to_tagger(d) \
+	container_of((d), struct sja1105_tagger_data, rxtstamp_work)
+#define to_sja1105(d) \
+	container_of((d), struct sja1105_private, tagger_data)
+
+static void sja1105_rxtstamp_work(struct work_struct *work)
+{
+	struct sja1105_tagger_data *data = to_tagger(work);
+	struct sja1105_private *priv = to_sja1105(data);
+	struct sk_buff *skb;
+	u64 now;
+
+	mutex_lock(&priv->ptp_lock);
+
+	now = priv->tstamp_cc.read(&priv->tstamp_cc);
+
+	while ((skb = skb_dequeue(&data->skb_rxtstamp_queue)) != NULL) {
+		struct skb_shared_hwtstamps *shwt = skb_hwtstamps(skb);
+		u64 ts;
+
+		*shwt = (struct skb_shared_hwtstamps) {0};
+
+		ts = SJA1105_SKB_CB(skb)->meta_tstamp;
+		ts = sja1105_tstamp_reconstruct(priv, now, ts);
+		ts = timecounter_cyc2time(&priv->tstamp_tc, ts);
+
+		shwt->hwtstamp = ns_to_ktime(ts);
+		netif_rx_ni(skb);
+	}
+
+	mutex_unlock(&priv->ptp_lock);
+}
+
+/* Called from dsa_skb_defer_rx_timestamp */
+bool sja1105_port_rxtstamp(struct dsa_switch *ds, int port,
+			   struct sk_buff *skb, unsigned int type)
+{
+	struct sja1105_private *priv = ds->priv;
+	struct sja1105_tagger_data *data = &priv->tagger_data;
+
+	if (!data->hwts_rx_en)
+		return false;
+
+	if (!sja1105_is_link_local(skb))
+		return false;
+
+	/* We need to read the full PTP clock to reconstruct the Rx
+	 * timestamp. For that we need a sleepable context.
+	 */
+	skb_queue_tail(&data->skb_rxtstamp_queue, skb);
+	schedule_work(&data->rxtstamp_work);
+	return true;
+}
+
 /* Called from dsa_skb_tx_timestamp. This callback is just to make DSA clone
  * the skb and have it available in DSA_SKB_CB in the .port_deferred_xmit
  * callback, where we will timestamp it synchronously.
@@ -1766,6 +1828,7 @@ bool sja1105_port_txtstamp(struct dsa_switch *ds, int port,
 static const struct dsa_switch_ops sja1105_switch_ops = {
 	.get_tag_protocol	= sja1105_get_tag_protocol,
 	.setup			= sja1105_setup,
+	.teardown		= sja1105_teardown,
 	.set_ageing_time	= sja1105_set_ageing_time,
 	.phylink_validate	= sja1105_phylink_validate,
 	.phylink_mac_config	= sja1105_mac_config,
@@ -1787,6 +1850,7 @@ static const struct dsa_switch_ops sja1105_switch_ops = {
 	.port_mdb_add		= sja1105_mdb_add,
 	.port_mdb_del		= sja1105_mdb_del,
 	.port_deferred_xmit	= sja1105_port_deferred_xmit,
+	.port_rxtstamp		= sja1105_port_rxtstamp,
 	.port_txtstamp		= sja1105_port_txtstamp,
 };
 
@@ -1885,6 +1949,7 @@ static int sja1105_probe(struct spi_device *spi)
 
 	tagger_data = &priv->tagger_data;
 	skb_queue_head_init(&tagger_data->skb_rxtstamp_queue);
+	INIT_WORK(&tagger_data->rxtstamp_work, sja1105_rxtstamp_work);
 
 	/* Connections between dsa_port and sja1105_port */
 	for (i = 0; i < SJA1105_NUM_PORTS; i++) {
include/linux/dsa/sja1105.h +7 -0
index b83f6e428e5d..794ee76aae56 100644
@@ -45,6 +45,13 @@ struct sja1105_tagger_data {
 	bool hwts_rx_en;
 };
 
+struct sja1105_skb_cb {
+	u32 meta_tstamp;
+};
+
+#define SJA1105_SKB_CB(skb) \
+	((struct sja1105_skb_cb *)DSA_SKB_CB_PRIV(skb))
+
 struct sja1105_port {
 	struct sja1105_tagger_data *data;
 	struct dsa_port *dp;
net/dsa/tag_sja1105.c +120 -1

133 lines not on this page · raw message

Signature · 1 line
2.17.1
author maintainer
#25 lore raw

PATCH v3 17/17 net: dsa: sja1105: Expose PTP timestamping ioctls to userspace

In reply to Vladimir Oltean

This enables the PTP support towards userspace applications such as linuxptp.

The switches can timestamp only trapped multicast MAC frames, and therefore only the profiles of 1588 over L2 are supported.

TX timestamping can be enabled per port, but RX timestamping is enabled globally. As long as RX timestamping is enabled, the switch will emit metadata follow-up frames that will be processed by the tagger. It may be a problem that linuxptp does not restore the RX timestamping settings when exiting.

---
Changes in v3:

Split from previous 09/10 patch (no functional changes).

Changes in v2:

None.

 drivers/net/dsa/sja1105/sja1105_main.c | 96 ++++++++++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_ptp.c  |  6 +-
 2 files changed, 100 insertions(+), 2 deletions(-)
drivers/net/dsa/sja1105/sja1105_main.c +96 -0
index 46a851f54b83..b68136c002bb 100644
@@ -1755,6 +1755,100 @@ static int sja1105_set_ageing_time(struct dsa_switch *ds,
 	return sja1105_static_config_reload(priv);
 }
 
+/* Caller must hold priv->tagger_data.meta_lock */
+static int sja1105_change_rxtstamping(struct sja1105_private *priv,
+				      bool on)
+{
+	struct sja1105_general_params_entry *general_params;
+	struct sja1105_table *table;
+	int rc;
+
+	table = &priv->static_config.tables[BLK_IDX_GENERAL_PARAMS];
+	general_params = table->entries;
+	general_params->send_meta1 = on;
+	general_params->send_meta0 = on;
+
+	rc = sja1105_init_avb_params(priv, on);
+	if (rc < 0)
+		return rc;
+
+	/* Initialize the meta state machine to a known state */
+	if (priv->tagger_data.stampable_skb) {
+		kfree_skb(priv->tagger_data.stampable_skb);
+		priv->tagger_data.stampable_skb = NULL;
+	}
+
+	return sja1105_static_config_reload(priv);
+}
+
+static int sja1105_hwtstamp_set(struct dsa_switch *ds, int port,
+				struct ifreq *ifr)
+{
+	struct sja1105_private *priv = ds->priv;
+	struct hwtstamp_config config;
+	bool rx_on;
+	int rc;
+
+	if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
+		return -EFAULT;
+
+	switch (config.tx_type) {
+	case HWTSTAMP_TX_OFF:
+		priv->ports[port].hwts_tx_en = false;
+		break;
+	case HWTSTAMP_TX_ON:
+		priv->ports[port].hwts_tx_en = true;
+		break;
+	default:
+		return -ERANGE;
+	}
+
+	switch (config.rx_filter) {
+	case HWTSTAMP_FILTER_NONE:
+		rx_on = false;
+		break;
+	default:
+		rx_on = true;
+		break;
+	}
+
+	if (rx_on != priv->tagger_data.hwts_rx_en) {
+		spin_lock(&priv->tagger_data.meta_lock);
+		rc = sja1105_change_rxtstamping(priv, rx_on);
+		spin_unlock(&priv->tagger_data.meta_lock);
+		if (rc < 0) {
+			dev_err(ds->dev,
+				"Failed to change RX timestamping: %d\n", rc);
+			return -EFAULT;
+		}
+		priv->tagger_data.hwts_rx_en = rx_on;
+	}
+
+	if (copy_to_user(ifr->ifr_data, &config, sizeof(config)))
+		return -EFAULT;
+	return 0;
+}
+
+static int sja1105_hwtstamp_get(struct dsa_switch *ds, int port,
+				struct ifreq *ifr)
+{
+	struct sja1105_private *priv = ds->priv;
+	struct hwtstamp_config config;
+
+	config.flags = 0;
+	if (priv->ports[port].hwts_tx_en)
+		config.tx_type = HWTSTAMP_TX_ON;
+	else
+		config.tx_type = HWTSTAMP_TX_OFF;
+	if (priv->tagger_data.hwts_rx_en)
+		config.rx_filter = HWTSTAMP_FILTER_PTP_V2_L2_EVENT;
+	else
+		config.rx_filter = HWTSTAMP_FILTER_NONE;
+
+	return copy_to_user(ifr->ifr_data, &config, sizeof(config)) ?
+		-EFAULT : 0;
+}
+
 #define to_tagger(d) \
 	container_of((d), struct sja1105_tagger_data, rxtstamp_work)
 #define to_sja1105(d) \
@@ -1850,6 +1944,8 @@ static const struct dsa_switch_ops sja1105_switch_ops = {
 	.port_mdb_add		= sja1105_mdb_add,
 	.port_mdb_del		= sja1105_mdb_del,
 	.port_deferred_xmit	= sja1105_port_deferred_xmit,
+	.port_hwtstamp_get	= sja1105_hwtstamp_get,
+	.port_hwtstamp_set	= sja1105_hwtstamp_set,
 	.port_rxtstamp		= sja1105_port_rxtstamp,
 	.port_txtstamp		= sja1105_port_txtstamp,
 };
drivers/net/dsa/sja1105/sja1105_ptp.c +4 -2
index 1fce413a3fb4..c7ce1edd8471 100644
@@ -70,8 +70,10 @@ int sja1105_get_ts_info(struct dsa_switch *ds, int port,
 	info->so_timestamping = SOF_TIMESTAMPING_TX_HARDWARE |
 				SOF_TIMESTAMPING_RX_HARDWARE |
 				SOF_TIMESTAMPING_RAW_HARDWARE;
-	info->tx_types = (1 << HWTSTAMP_TX_OFF);
-	info->rx_filters = (1 << HWTSTAMP_FILTER_NONE);
+	info->tx_types = (1 << HWTSTAMP_TX_OFF) |
+			 (1 << HWTSTAMP_TX_ON);
+	info->rx_filters = (1 << HWTSTAMP_FILTER_NONE) |
+			   (1 << HWTSTAMP_FILTER_PTP_V2_L2_EVENT);
 	info->phc_index = ptp_clock_index(priv->clock);
 	return 0;
 }
Signature · 1 line
2.17.1
maintainer
#26 lore raw
In reply to Vladimir Oltean

From: Vladimir Oltean <[email protected]>
Date: Tue, 4 Jun 2019 20:07:39 +0300

Quoted 8 lines

This patchset adds the following:

- A timecounter/cyclecounter based PHC for the free-running timestamping clock of this switch.

- A state machine implemented in the DSA tagger for SJA1105, which keeps track of metadata follow-up Ethernet frames (the switch's way of transmitting RX timestamps).

This series doesn't apply cleanly to net-next, please respin.

Thank you.

author maintainer
#27 lore raw
Quoted 9 lines · David Miller

From: Vladimir Oltean <[email protected]>
Date: Tue, 4 Jun 2019 20:07:39 +0300

Quoted 5 lines

This patchset adds the following:

- A timecounter/cyclecounter based PHC for the free-running timestamping clock of this switch.

Quoted 7 lines
Quoted 3 lines

- A state machine implemented in the DSA tagger for SJA1105, which keeps track of metadata follow-up Ethernet frames (the switch's way of transmitting RX timestamps).

This series doesn't apply cleanly to net-next, please respin.

Thank you.

Hi Dave,

It is conflicting because net-next at the moment lacks this patch that I submitted to net:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=e8d67fa5696e2fcaf956dae36d11e6eff5246101
What would you like me to do: resubmit after you merge net into net-next, add the above patch to this series (which you'll have to skip upon the next merge), or you can just cherry-pick it and then the series will apply?

Thanks!
-Vladimir

author maintainer
#28 lore raw
Quoted 29 lines · Vladimir Oltean
Quoted 9 lines · David Miller

From: Vladimir Oltean <[email protected]>
Date: Tue, 4 Jun 2019 20:07:39 +0300

Quoted 5 lines

This patchset adds the following:

- A timecounter/cyclecounter based PHC for the free-running timestamping clock of this switch.

Quoted 7 lines
Quoted 3 lines

- A state machine implemented in the DSA tagger for SJA1105, which keeps track of metadata follow-up Ethernet frames (the switch's way of transmitting RX timestamps).

This series doesn't apply cleanly to net-next, please respin.

Thank you.

Hi Dave,

It is conflicting because net-next at the moment lacks this patch that I submitted to net:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=e8d67fa5696e2fcaf956dae36d11e6eff5246101
What would you like me to do: resubmit after you merge net into net-next, add the above patch to this series (which you'll have to skip upon the next merge), or you can just cherry-pick it and then the series will apply?

Quoted 2 lines

Thanks!
-Vladimir

In the meantime: Richard, do you have any objections to this patchset? I was wondering whether the path delay difference between E2E and P2P rings any bell to you.

Thank you,
-Vladimir

#29 lore raw
Quoted 1 line · Vladimir Oltean

In the meantime: Richard, do you have any objections to this patchset?

I like the fact that you didn't have to change the dsa or ptp frameworks this time around. I haven't taken a closer look than that yet.

Quoted 2 lines

I was wondering whether the path delay difference between E2E and P2P rings any bell to you.

Can it be that the switch applies corrections in HW?

Thanks,
Richard

author maintainer
#30 lore raw
Quoted 4 lines · Richard Cochran
Quoted 1 line · Vladimir Oltean

In the meantime: Richard, do you have any objections to this patchset?

Quoted 9 lines

I like the fact that you didn't have to change the dsa or ptp frameworks this time around. I haven't taken a closer look than that yet.

Quoted 2 lines

I was wondering whether the path delay difference between E2E and P2P rings any bell to you.

Can it be that the switch applies corrections in HW?

Yes it can be. It was one of the first things I thought of. Normally it updates the correction field with its own residence time in 1-step L2 event messages (but I use 2 step).
It also has a bit called IGNORE2STF (ignore 2-step flag) by which it updates the correction field in all L2 event messages (including sync, thereby violating the spec for a switch, as far as I'm aware). But I'm not setting it.
I also looked at egress frames with wireshark and the correction field is zero.

Quoted 2 lines

Thanks,
Richard

Thanks,
-Vladimir

author maintainer
#31 lore raw
Quoted 16 lines · Vladimir Oltean
Quoted 4 lines · Richard Cochran
Quoted 1 line · Vladimir Oltean

In the meantime: Richard, do you have any objections to this patchset?

Quoted 9 lines

I like the fact that you didn't have to change the dsa or ptp frameworks this time around. I haven't taken a closer look than that yet.

Quoted 2 lines

I was wondering whether the path delay difference between E2E and P2P rings any bell to you.

Can it be that the switch applies corrections in HW?

Quoted 9 lines

Yes it can be. It was one of the first things I thought of. Normally it updates the correction field with its own residence time in 1-step L2 event messages (but I use 2 step).
It also has a bit called IGNORE2STF (ignore 2-step flag) by which it updates the correction field in all L2 event messages (including sync, thereby violating the spec for a switch, as far as I'm aware). But I'm not setting it.
I also looked at egress frames with wireshark and the correction field is zero.

I also changed around the values of ptp_dst_mac and p2p_dst_mac in linuxptp in the hope that I'd throw off whatever hardware parser it has to identify the event frames, but I still get negative path delay with E2E nonetheless. So it's probably not that.

Quoted 5 lines
Quoted 2 lines

Thanks,
Richard

Thanks,
-Vladimir

-Vladimir

author maintainer
#32 lore raw
In reply to Richard Cochran
Quoted 8 lines · Richard Cochran
Quoted 1 line · Vladimir Oltean

In the meantime: Richard, do you have any objections to this patchset?

I like the fact that you didn't have to change the dsa or ptp frameworks this time around. I haven't taken a closer look than that yet.

There's one more thing I wanted to ask you if you get the chance to take a closer look.
Currently I'm using a cyclecounter, but I *will* need actual PHC manipulations for the time-based shaping and policing features that the switch has in hardware. On the other hand I get much tighter sync offset using the free-running counter than with hardware-corrected timestamps. So as far as I see it, I'll need to have two sets of operations.
How should I design such a dual-PHC device driver? Just register two separate clocks, one for the timestamping counter, the other for the scheduling/policing PTP clock, and have phc2sys keep them in sync externally to the driver? Or implement the hardware corrections alongside the timecounter ones, and expose a single PHC (and for clock_gettime, just pick one of the time sources)?

Quoted 7 lines
Quoted 2 lines

I was wondering whether the path delay difference between E2E and P2P rings any bell to you.

Can it be that the switch applies corrections in HW?

Thanks,
Richard

Thanks,
-Vladimir

#33 lore raw
Quoted 3 lines · Vladimir Oltean

Currently I'm using a cyclecounter, but I *will* need actual PHC manipulations for the time-based shaping and policing features that the switch has in hardware.

Okay.

Quoted 3 lines

On the other hand I get much tighter sync
offset using the free-running counter than with hardware-corrected timestamps.

Why? The time stamps come from the very same counter, don't they?

Quoted 2 lines

So as far as I see it, I'll need to have two sets of operations.

I doubt very much that this will work well.

Quoted 4 lines

How should I design such a dual-PHC device driver? Just register two separate clocks, one for the timestamping counter, the other for the scheduling/policing PTP clock, and have phc2sys keep them in sync externally to the driver?

But how would phc2sys do this? By comparing clock_gettime() values? That would surely introduce unnecessary time error.

Quoted 3 lines

Or implement the hardware corrections
alongside the timecounter ones, and expose a single PHC (and for clock_gettime, just pick one of the time sources)?

I would implement the hardware clock and drop the timecounter altogether.

HTH,
Richard

author maintainer
#34 lore raw
Quoted 6 lines · Richard Cochran
Quoted 3 lines · Vladimir Oltean

Currently I'm using a cyclecounter, but I *will* need actual PHC manipulations for the time-based shaping and policing features that the switch has in hardware.

Quoted 8 lines

Okay.

Quoted 3 lines

On the other hand I get much tighter sync
offset using the free-running counter than with hardware-corrected timestamps.

Why? The time stamps come from the very same counter, don't they?

Plain and simply because it doesn't work very well.
Even phc2sys from the system clock to the hardware (no timestamps involved) has trouble staying put (under 1000 ns offset).
And using the hardware-corrected timestamps triggers a lot of clockchecks.

Quoted 22 lines
Quoted 2 lines

So as far as I see it, I'll need to have two sets of operations.

I doubt very much that this will work well.

Quoted 4 lines

How should I design such a dual-PHC device driver? Just register two separate clocks, one for the timestamping counter, the other for the scheduling/policing PTP clock, and have phc2sys keep them in sync externally to the driver?

But how would phc2sys do this? By comparing clock_gettime() values? That would surely introduce unnecessary time error.

Quoted 3 lines

Or implement the hardware corrections
alongside the timecounter ones, and expose a single PHC (and for clock_gettime, just pick one of the time sources)?

I would implement the hardware clock and drop the timecounter altogether.

HTH,
Richard

#35 lore raw
Quoted 4 lines · Vladimir Oltean

Plain and simply because it doesn't work very well.
Even phc2sys from the system clock to the hardware (no timestamps involved) has trouble staying put (under 1000 ns offset).
And using the hardware-corrected timestamps triggers a lot of clockchecks.

It sounds like a bug in reading or adjusting the HW clock. Is the HW clock stable when you don't adjust its frequency?

Thanks,
Richard

author maintainer
#36 lore raw
Quoted 9 lines · Richard Cochran
Quoted 4 lines · Vladimir Oltean

Plain and simply because it doesn't work very well.
Even phc2sys from the system clock to the hardware (no timestamps involved) has trouble staying put (under 1000 ns offset).
And using the hardware-corrected timestamps triggers a lot of clockchecks.

It sounds like a bug in reading or adjusting the HW clock. Is the HW clock stable when you don't adjust its frequency?

How can I tell that for sure?

Quoted 3 lines

Thanks,
Richard

maintainer
#37 lore raw
In reply to Vladimir Oltean

From: Vladimir Oltean <[email protected]>
Date: Wed, 5 Jun 2019 12:13:59 +0300

Quoted 20 lines · David Miller
Quoted 9 lines

From: Vladimir Oltean <[email protected]>
Date: Tue, 4 Jun 2019 20:07:39 +0300

Quoted 5 lines

This patchset adds the following:

- A timecounter/cyclecounter based PHC for the free-running timestamping clock of this switch.

Quoted 7 lines
Quoted 3 lines

- A state machine implemented in the DSA tagger for SJA1105, which keeps track of metadata follow-up Ethernet frames (the switch's way of transmitting RX timestamps).

This series doesn't apply cleanly to net-next, please respin.

Thank you.

Hi Dave,

Quoted 7 lines

It is conflicting because net-next at the moment lacks this patch that I submitted to net:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=e8d67fa5696e2fcaf956dae36d11e6eff5246101
What would you like me to do: resubmit after you merge net into net-next, add the above patch to this series (which you'll have to skip upon the next merge), or you can just cherry-pick it and then the series will apply?

So let me bring this series back to state "Under Review" and I'll apply it after I next merge net into net-next.

Thanks for letting me know.

maintainer
#38 lore raw

From: David Miller <[email protected]>
Date: Wed, 05 Jun 2019 11:44:29 -0700 (PDT)

Quoted 11 lines

From: Vladimir Oltean <[email protected]>
Date: Wed, 5 Jun 2019 12:13:59 +0300

Quoted 7 lines

It is conflicting because net-next at the moment lacks this patch that I submitted to net:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=e8d67fa5696e2fcaf956dae36d11e6eff5246101
What would you like me to do: resubmit after you merge net into net-next, add the above patch to this series (which you'll have to skip upon the next merge), or you can just cherry-pick it and then the series will apply?

Quoted 2 lines

So let me bring this series back to state "Under Review" and I'll apply it after I next merge net into net-next.

So I applied the series but it doesn't even build:

ERROR: "sja1105_unpack" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_spi_send_packed_buf" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_pack" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_spi_send_int" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_get_ts_info" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105pqrs_ptp_cmd" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_clock_unregister" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptpegr_ts_poll" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105et_ptp_cmd" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_reset" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_tstamp_reconstruct" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_clock_register" [drivers/net/dsa/sja1105/sja1105.ko] undefined!

You have to test better with the various modular/non-modular combinations.

Thanks.

author maintainer
#39 lore raw
Quoted 33 lines · David Miller

From: David Miller <[email protected]>
Date: Wed, 05 Jun 2019 11:44:29 -0700 (PDT)

Quoted 11 lines

From: Vladimir Oltean <[email protected]>
Date: Wed, 5 Jun 2019 12:13:59 +0300

Quoted 7 lines

It is conflicting because net-next at the moment lacks this patch that I submitted to net:
https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=e8d67fa5696e2fcaf956dae36d11e6eff5246101
What would you like me to do: resubmit after you merge net into net-next, add the above patch to this series (which you'll have to skip upon the next merge), or you can just cherry-pick it and then the series will apply?

Quoted 2 lines

So let me bring this series back to state "Under Review" and I'll apply it after I next merge net into net-next.

So I applied the series but it doesn't even build:

ERROR: "sja1105_unpack" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_spi_send_packed_buf" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_pack" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_spi_send_int" [drivers/net/dsa/sja1105/sja1105_ptp.ko] undefined!
ERROR: "sja1105_get_ts_info" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105pqrs_ptp_cmd" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_clock_unregister" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptpegr_ts_poll" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105et_ptp_cmd" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_reset" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_tstamp_reconstruct" [drivers/net/dsa/sja1105/sja1105.ko] undefined!
ERROR: "sja1105_ptp_clock_register" [drivers/net/dsa/sja1105/sja1105.ko] undefined!

Quoted 3 lines

You have to test better with the various modular/non-modular combinations.

Thanks.

Ok, my bad, I'll resubmit it tomorrow.

Thanks!
-Vladimir