[PATCH v4 net-next 0/6] tc-taprio offload for SJA1105 DSA

[PATCH v4 net-next 0/6] tc-taprio offload for SJA1105 DSA

[Vladimir Oltean]

This is the third attempt to submit the tc-taprio offload model for
inclusion in the networking tree. The sja1105 switch driver will provide
the first implementation of the offload. Only the bare minimum is added:

- The offload model and a DSA pass-through
- The hardware implementation
- The interaction with the netdev queues in the tagger code
- Documentation

What has been removed from previous attempts is support for
PTP-as-clocksource in sja1105, as well as configuring the traffic class
for management traffic.  These will be added as soon as the offload
model is settled.

Vinicius Costa Gomes (1):
  taprio: Add support for hardware offloading

Vladimir Oltean (5):
  net: dsa: Pass ndo_setup_tc slave callback to drivers
  net: dsa: sja1105: Add static config tables for scheduling
  net: dsa: sja1105: Advertise the 8 TX queues
  net: dsa: sja1105: Configure the Time-Aware Scheduler via tc-taprio
    offload
  docs: net: dsa: sja1105: Add info about the Time-Aware Scheduler

 Documentation/networking/dsa/sja1105.rst      |  90 ++++
 drivers/net/dsa/sja1105/Kconfig               |   8 +
 drivers/net/dsa/sja1105/Makefile              |   4 +
 drivers/net/dsa/sja1105/sja1105.h             |   6 +
 .../net/dsa/sja1105/sja1105_dynamic_config.c  |   8 +
 drivers/net/dsa/sja1105/sja1105_main.c        |  26 +-
 .../net/dsa/sja1105/sja1105_static_config.c   | 167 +++++++
 .../net/dsa/sja1105/sja1105_static_config.h   |  48 +-
 drivers/net/dsa/sja1105/sja1105_tas.c         | 423 ++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_tas.h         |  41 ++
 include/linux/netdevice.h                     |   1 +
 include/net/dsa.h                             |   2 +
 include/net/pkt_sched.h                       |  23 +
 include/uapi/linux/pkt_sched.h                |   3 +-
 net/dsa/slave.c                               |  12 +-
 net/dsa/tag_sja1105.c                         |   3 +-
 net/sched/sch_taprio.c                        | 409 +++++++++++++++--
 17 files changed, 1222 insertions(+), 52 deletions(-)
 create mode 100644 drivers/net/dsa/sja1105/sja1105_tas.c
 create mode 100644 drivers/net/dsa/sja1105/sja1105_tas.h

-- 

For those who want to follow along with the hardware implementation, the
manual is here:
https://www.nxp.com/docs/en/user-guide/UM10944.pdf

Changes in v3 (actually v4 which is what v3 should have been):
- Removed "[PATCH v2 net-next 5/7] net: dsa: sja1105: Make HOSTPRIO a
  kernel config", which was deemed non-critical at this point.
- A bit of cleanup in "[PATCH v3 net-next 5/6] net: dsa: sja1105:
  Configure the Time-Aware Scheduler via tc-taprio offload".

Notable changes in v2:
- Made the series independent from PTP (which is temporarily removed)
- Changed the meaning of the gate_mask - it is now acting on traffic
  classes even in the view exposed by taprio to drivers.
- Removed the next_sched hrtimer.
- Summarized one of the responses given to Vinicius into a new
  documentation section.

The first version of the net-next patch series can be found here:

https://www.spinics.net/lists/netdev/msg597214.html

Changes in the first version of the net-next series compared to RFC v2:
- Made "flags 1" and "flags 2" mutually exclusive in the taprio qdisc
- Moved taprio_enable_offload and taprio_disable_offload out of atomic
  context - spin_lock_bh(qdisc_lock(sch)). This allows drivers that
  implement the ndo_setup_tc to sleep and for taprio memory to be
  allocated with GFP_KERNEL. The only thing that was kept under the
  spinlock is the assignment of the q->dequeue and q->peek pointers.
- Finally making proper use of own API - added a taprio_alloc helper to
  avoid passing stack memory to drivers.

The second version of the RFC is at:
https://www.spinics.net/lists/netdev/msg596663.html

Changes in v2 of the RFC since v1:
- Adapted the taprio offload patch to work by specifying "flags 2" to
  the iproute2-next tc. At the moment I don't clearly understand whether
  the full offload and the txtime assist ("flags 1") are mutually
  exclusive or not (i.e. whether a "flags 3" mode should be rejected,
  which it currently isn't).
- Added reference counting to the taprio offload structure. Maybe the
  function names and placement could have been better though. As for the
  other complaint (cycle time calculation) it got fixed in the taprio
  parser in the meantime.
- Converted sja1105 to use the hardware PTP registers, and save/restore
  the PTP time across resets.
- Made the DSA callback for ndo_setup_tc a bit more generic, but I don't
  know whether it fulfills expectations. Drivers still can't do blocking
  operations in its execution context.
- Added a state machine for starting/stopping the scheduler based on the
  last command run on the PTP clock.

The first RFC from July can be seen at:
https://lists.openwall.net/netdev/2019/07/07/81

Original cover letter:

Using Vinicius Costa Gomes' configuration interface for 802.1Qbv (later
resent by Voon Weifeng for the stmmac driver), I am submitting for
review a draft implementation of this offload for a DSA switch.

I don't want to insist too much on the hardware specifics of SJA1105
which isn't otherwise very compliant to the IEEE spec.

In order to be able to test with Vedang Patel's iproute2 patch for
taprio offload (https://www.spinics.net/lists/netdev/msg573072.html)
I had to actually revert the txtime-assist branch as it had changed the
iproute2 interface.

In terms of impact for DSA drivers, I would like to point out that:

- Maybe somebody should pre-populate qopt->cycle_time in case the user
  does not provide one. Otherwise each driver needs to iterate over the
  GCL once, just to set the cycle time (right now stmmac does as well).

- Configuring the switch over SPI cannot apparently be done from this
  ndo_setup_tc callback because it runs in atomic context. I also have
  some downstream patches to offload tc clsact matchall with mirred
  action, but in that case it looks like the atomic context restriction
  does not apply.

- I had to copy the struct tc_taprio_qopt_offload to driver private
  memory because a static config needs to be constructed every time a
  change takes place, and there are up to 4 switch ports that may take a
  TAS configuration. I have created a private
  tc_taprio_qopt_offload_copy() helper for this - I don't know whether
  it's of any help in the general case.

There is more to be done however. The TAS needs to be integrated with
the PTP driver. This is because with a PTP clock source, the base time
is written dynamically to the PTPSCHTM (PTP schedule time) register and
must be a time in the future. Then the "real" base time of each port's
TAS config can be offset by at most ~50 ms (the DELTA field from the
Schedule Entry Points Table) relative to PTPSCHTM.
Because base times in the past are completely ignored by this hardware,
we need to decide if it's ok behaviorally for a driver to "roll" a past
base time into the immediate future by incrementally adding the cycle
time (so the phase doesn't change). If it is, then decide by how long in
the future it is ok to do so. Or alternatively, is it preferable if the
driver errors out if the user-supplied base time is in the past and the
hardware doesn't like it? But even then, there might be fringe cases
when the base time becomes a past PTP time right as the driver tries to
apply the config.
Also applying a tc-taprio offload to a second SJA1105 switch port will
inevitably need to roll the first port's (now past) base time into an
equivalent future time.
All of this is going to be complicated even further by the fact that
resetting the switch (to apply the tc-taprio offload) makes it reset its
PTP time.

2.17.1

[PATCH v4 net-next 1/6] taprio: Add support for hardware offloading

[Vladimir Oltean]

From: Vinicius Costa Gomes <vinicius.gomes@intel.com>

This allows taprio to offload the schedule enforcement to capable
network cards, resulting in more precise windows and less CPU usage.

The gate mask acts on traffic classes (groups of queues of same
priority), as specified in IEEE 802.1Q-2018, and following the existing
taprio and mqprio semantics.
It is up to the driver to perform conversion between tc and individual
netdev queues if for some reason it needs to make that distinction.

Full offload is requested from the network interface by specifying
"flags 2" in the tc qdisc creation command, which in turn corresponds to
the TCA_TAPRIO_ATTR_FLAG_FULL_OFFLOAD bit.

The important detail here is the clockid which is implicitly /dev/ptpN
for full offload, and hence not configurable.

A reference counting API is added to support the use case where Ethernet
drivers need to keep the taprio offload structure locally (i.e. they are
a multi-port switch driver, and configuring a port depends on the
settings of other ports as well). The refcount_t variable is kept in a
private structure (__tc_taprio_qopt_offload) and not exposed to drivers.

In the future, the private structure might also be expanded with a
backpointer to taprio_sched *q, to implement the notification system
described in the patch (of when admin became oper, or an error occurred,
etc, so the offload can be monitored with 'tc qdisc show').

Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Voon Weifeng <weifeng.voon@intel.com>
Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
---
Changes since v2:
- None.

Changes since v1:
- Turned the next_sched hrtimer function into a simple
  taprio_offload_config_changed function called synchronously (for now)
  from taprio_enable_offload. But the idea is that the driver may have a
  lot more means to figure out when the admin schedule is no longer
  pending (perhaps even an interrupt), so leave an open window for
  implementing a notification system from the driver.
- Made it an error to specify 'clockid' with full offload.
- Created a wrapper __tc_taprio_qopt_offload structure which holds the
  refcount_t (for now) and maybe a backpointer to the qdisc_priv in the
  future.
- Renamed taprio_get and taprio_free to taprio_offload_get and
  taprio_offload_free. Renamed the "taprio" variable to "offload".
- Moved the reference counting helper implementations to sch_taprio.c.
- Removed the tc_mask_to_queue_mask manipulation done to the gate_mask
  before passing it on to drivers. Instead of netdev queue gates, they
  now see a mask of traffic class gates, which:
  - They need to care about anyway, if they have a multi-queue device
    and they need to configure the queue-to-tc hardware mapping.
  - Makes no difference to them if the hardware makes no distinction
    between queue and traffic class (there is only one egress queue per
    tc, having a fixed priority). The sja1105 hw is in this situation.

Changes since RFC:
- Made the combination of FULL_OFFLOAD and TXTIME_ASSIST invalid.
- Made ndo_setup_tc be called from sleepable context.
- Added a taprio_alloc helper to avoid passing stack memory to drivers.
- Made taprio_disable_offload take the extack as well.
- Conditioned the setup of the software (and txtime-assisted)
  implementation of taprio on there not being a full offload in place.
- Fixed a lockdep-related compilation bug.

 include/linux/netdevice.h      |   1 +
 include/net/pkt_sched.h        |  23 ++
 include/uapi/linux/pkt_sched.h |   3 +-
 net/sched/sch_taprio.c         | 409 +++++++++++++++++++++++++++++----
 4 files changed, 392 insertions(+), 44 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d7d5626002e9..9eda1c31d1f7 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -847,6 +847,7 @@ enum tc_setup_type {
 	TC_SETUP_QDISC_ETF,
 	TC_SETUP_ROOT_QDISC,
 	TC_SETUP_QDISC_GRED,
+	TC_SETUP_QDISC_TAPRIO,
 };
 
 /* These structures hold the attributes of bpf state that are being passed
diff --git a/include/net/pkt_sched.h b/include/net/pkt_sched.h
index a16fbe9a2a67..d1632979622e 100644
--- a/include/net/pkt_sched.h
+++ b/include/net/pkt_sched.h
@@ -161,4 +161,27 @@ struct tc_etf_qopt_offload {
 	s32 queue;
 };
 
+struct tc_taprio_sched_entry {
+	u8 command; /* TC_TAPRIO_CMD_* */
+
+	/* The gate_mask in the offloading side refers to traffic classes */
+	u32 gate_mask;
+	u32 interval;
+};
+
+struct tc_taprio_qopt_offload {
+	u8 enable;
+	ktime_t base_time;
+	u64 cycle_time;
+	u64 cycle_time_extension;
+
+	size_t num_entries;
+	struct tc_taprio_sched_entry entries[0];
+};
+
+/* Reference counting */
+struct tc_taprio_qopt_offload *taprio_offload_get(struct tc_taprio_qopt_offload
+						  *offload);
+void taprio_offload_free(struct tc_taprio_qopt_offload *offload);
+
 #endif
diff --git a/include/uapi/linux/pkt_sched.h b/include/uapi/linux/pkt_sched.h
index 18f185299f47..5011259b8f67 100644
--- a/include/uapi/linux/pkt_sched.h
+++ b/include/uapi/linux/pkt_sched.h
@@ -1160,7 +1160,8 @@ enum {
  *       [TCA_TAPRIO_ATTR_SCHED_ENTRY_INTERVAL]
  */
 
-#define TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST 0x1
+#define TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST	BIT(0)
+#define TCA_TAPRIO_ATTR_FLAG_FULL_OFFLOAD	BIT(1)
 
 enum {
 	TCA_TAPRIO_ATTR_UNSPEC,
diff --git a/net/sched/sch_taprio.c b/net/sched/sch_taprio.c
index 84b863e2bdbd..2f7b34205c82 100644
--- a/net/sched/sch_taprio.c
+++ b/net/sched/sch_taprio.c
@@ -29,8 +29,8 @@ static DEFINE_SPINLOCK(taprio_list_lock);
 
 #define TAPRIO_ALL_GATES_OPEN -1
 
-#define FLAGS_VALID(flags) (!((flags) & ~TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST))
 #define TXTIME_ASSIST_IS_ENABLED(flags) ((flags) & TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST)
+#define FULL_OFFLOAD_IS_ENABLED(flags) ((flags) & TCA_TAPRIO_ATTR_FLAG_FULL_OFFLOAD)
 
 struct sched_entry {
 	struct list_head list;
@@ -75,9 +75,16 @@ struct taprio_sched {
 	struct sched_gate_list __rcu *admin_sched;
 	struct hrtimer advance_timer;
 	struct list_head taprio_list;
+	struct sk_buff *(*dequeue)(struct Qdisc *sch);
+	struct sk_buff *(*peek)(struct Qdisc *sch);
 	u32 txtime_delay;
 };
 
+struct __tc_taprio_qopt_offload {
+	refcount_t users;
+	struct tc_taprio_qopt_offload offload;
+};
+
 static ktime_t sched_base_time(const struct sched_gate_list *sched)
 {
 	if (!sched)
@@ -268,6 +275,19 @@ static bool is_valid_interval(struct sk_buff *skb, struct Qdisc *sch)
 	return entry;
 }
 
+static bool taprio_flags_valid(u32 flags)
+{
+	/* Make sure no other flag bits are set. */
+	if (flags & ~(TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST |
+		      TCA_TAPRIO_ATTR_FLAG_FULL_OFFLOAD))
+		return false;
+	/* txtime-assist and full offload are mutually exclusive */
+	if ((flags & TCA_TAPRIO_ATTR_FLAG_TXTIME_ASSIST) &&
+	    (flags & TCA_TAPRIO_ATTR_FLAG_FULL_OFFLOAD))
+		return false;
+	return true;
+}
+
 /* This returns the tstamp value set by TCP in terms of the set clock. */
 static ktime_t get_tcp_tstamp(struct taprio_sched *q, struct sk_buff *skb)
 {
@@ -417,7 +437,7 @@ static int taprio_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	return qdisc_enqueue(skb, child, to_free);
 }
 
-static struct sk_buff *taprio_peek(struct Qdisc *sch)
+static struct sk_buff *taprio_peek_soft(struct Qdisc *sch)
 {
 	struct taprio_sched *q = qdisc_priv(sch);
 	struct net_device *dev = qdisc_dev(sch);
@@ -461,6 +481,36 @@ static struct sk_buff *taprio_peek(struct Qdisc *sch)
 	return NULL;
 }
 
+static struct sk_buff *taprio_peek_offload(struct Qdisc *sch)
+{
+	struct taprio_sched *q = qdisc_priv(sch);
+	struct net_device *dev = qdisc_dev(sch);
+	struct sk_buff *skb;
+	int i;
+
+	for (i = 0; i < dev->num_tx_queues; i++) {
+		struct Qdisc *child = q->qdiscs[i];
+
+		if (unlikely(!child))
+			continue;
+
+		skb = child->ops->peek(child);
+		if (!skb)
+			continue;
+
+		return skb;
+	}
+
+	return NULL;
+}
+
+static struct sk_buff *taprio_peek(struct Qdisc *sch)
+{
+	struct taprio_sched *q = qdisc_priv(sch);
+
+	return q->peek(sch);
+}
+
 static void taprio_set_budget(struct taprio_sched *q, struct sched_entry *entry)
 {
 	atomic_set(&entry->budget,
@@ -468,7 +518,7 @@ static void taprio_set_budget(struct taprio_sched *q, struct sched_entry *entry)
 			     atomic64_read(&q->picos_per_byte)));
 }
 
-static struct sk_buff *taprio_dequeue(struct Qdisc *sch)
+static struct sk_buff *taprio_dequeue_soft(struct Qdisc *sch)
 {
 	struct taprio_sched *q = qdisc_priv(sch);
 	struct net_device *dev = qdisc_dev(sch);
@@ -550,6 +600,40 @@ static struct sk_buff *taprio_dequeue(struct Qdisc *sch)
 	return skb;
 }
 
+static struct sk_buff *taprio_dequeue_offload(struct Qdisc *sch)
+{
+	struct taprio_sched *q = qdisc_priv(sch);
+	struct net_device *dev = qdisc_dev(sch);
+	struct sk_buff *skb;
+	int i;
+
+	for (i = 0; i < dev->num_tx_queues; i++) {
+		struct Qdisc *child = q->qdiscs[i];
+
+		if (unlikely(!child))
+			continue;
+
+		skb = child->ops->dequeue(child);
+		if (unlikely(!skb))
+			continue;
+
+		qdisc_bstats_update(sch, skb);
+		qdisc_qstats_backlog_dec(sch, skb);
+		sch->q.qlen--;
+
+		return skb;
+	}
+
+	return NULL;
+}
+
+static struct sk_buff *taprio_dequeue(struct Qdisc *sch)
+{
+	struct taprio_sched *q = qdisc_priv(sch);
+
+	return q->dequeue(sch);
+}
+
 static bool should_restart_cycle(const struct sched_gate_list *oper,
 				 const struct sched_entry *entry)
 {
@@ -932,6 +1016,9 @@ static void taprio_start_sched(struct Qdisc *sch,
 	struct taprio_sched *q = qdisc_priv(sch);
 	ktime_t expires;
 
+	if (FULL_OFFLOAD_IS_ENABLED(q->flags))
+		return;
+
 	expires = hrtimer_get_expires(&q->advance_timer);
 	if (expires == 0)
 		expires = KTIME_MAX;
@@ -1011,6 +1098,254 @@ static void setup_txtime(struct taprio_sched *q,
 	}
 }
 
+static struct tc_taprio_qopt_offload *taprio_offload_alloc(int num_entries)
+{
+	size_t size = sizeof(struct tc_taprio_sched_entry) * num_entries +
+		      sizeof(struct __tc_taprio_qopt_offload);
+	struct __tc_taprio_qopt_offload *__offload;
+
+	__offload = kzalloc(size, GFP_KERNEL);
+	if (!__offload)
+		return NULL;
+
+	refcount_set(&__offload->users, 1);
+
+	return &__offload->offload;
+}
+
+struct tc_taprio_qopt_offload *taprio_offload_get(struct tc_taprio_qopt_offload
+						  *offload)
+{
+	struct __tc_taprio_qopt_offload *__offload;
+
+	__offload = container_of(offload, struct __tc_taprio_qopt_offload,
+				 offload);
+
+	refcount_inc(&__offload->users);
+
+	return offload;
+}
+EXPORT_SYMBOL_GPL(taprio_offload_get);
+
+void taprio_offload_free(struct tc_taprio_qopt_offload *offload)
+{
+	struct __tc_taprio_qopt_offload *__offload;
+
+	__offload = container_of(offload, struct __tc_taprio_qopt_offload,
+				 offload);
+
+	if (!refcount_dec_and_test(&__offload->users))
+		return;
+
+	kfree(__offload);
+}
+EXPORT_SYMBOL_GPL(taprio_offload_free);
+
+/* The function will only serve to keep the pointers to the "oper" and "admin"
+ * schedules valid in relation to their base times, so when calling dump() the
+ * users looks at the right schedules.
+ * When using full offload, the admin configuration is promoted to oper at the
+ * base_time in the PHC time domain.  But because the system time is not
+ * necessarily in sync with that, we can't just trigger a hrtimer to call
+ * switch_schedules at the right hardware time.
+ * At the moment we call this by hand right away from taprio, but in the future
+ * it will be useful to create a mechanism for drivers to notify taprio of the
+ * offload state (PENDING, ACTIVE, INACTIVE) so it can be visible in dump().
+ * This is left as TODO.
+ */
+void taprio_offload_config_changed(struct taprio_sched *q)
+{
+	struct sched_gate_list *oper, *admin;
+
+	spin_lock(&q->current_entry_lock);
+
+	oper = rcu_dereference_protected(q->oper_sched,
+					 lockdep_is_held(&q->current_entry_lock));
+	admin = rcu_dereference_protected(q->admin_sched,
+					  lockdep_is_held(&q->current_entry_lock));
+
+	switch_schedules(q, &admin, &oper);
+
+	spin_unlock(&q->current_entry_lock);
+}
+
+static void taprio_sched_to_offload(struct taprio_sched *q,
+				    struct sched_gate_list *sched,
+				    const struct tc_mqprio_qopt *mqprio,
+				    struct tc_taprio_qopt_offload *offload)
+{
+	struct sched_entry *entry;
+	int i = 0;
+
+	offload->base_time = sched->base_time;
+	offload->cycle_time = sched->cycle_time;
+	offload->cycle_time_extension = sched->cycle_time_extension;
+
+	list_for_each_entry(entry, &sched->entries, list) {
+		struct tc_taprio_sched_entry *e = &offload->entries[i];
+
+		e->command = entry->command;
+		e->interval = entry->interval;
+		e->gate_mask = entry->gate_mask;
+		i++;
+	}
+
+	offload->num_entries = i;
+}
+
+static int taprio_enable_offload(struct net_device *dev,
+				 struct tc_mqprio_qopt *mqprio,
+				 struct taprio_sched *q,
+				 struct sched_gate_list *sched,
+				 struct netlink_ext_ack *extack)
+{
+	const struct net_device_ops *ops = dev->netdev_ops;
+	struct tc_taprio_qopt_offload *offload;
+	int err = 0;
+
+	if (!ops->ndo_setup_tc) {
+		NL_SET_ERR_MSG(extack,
+			       "Device does not support taprio offload");
+		return -EOPNOTSUPP;
+	}
+
+	offload = taprio_offload_alloc(sched->num_entries);
+	if (!offload) {
+		NL_SET_ERR_MSG(extack,
+			       "Not enough memory for enabling offload mode");
+		return -ENOMEM;
+	}
+	offload->enable = 1;
+	taprio_sched_to_offload(q, sched, mqprio, offload);
+
+	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_TAPRIO, offload);
+	if (err < 0) {
+		NL_SET_ERR_MSG(extack,
+			       "Device failed to setup taprio offload");
+		goto done;
+	}
+
+	taprio_offload_config_changed(q);
+
+done:
+	taprio_offload_free(offload);
+
+	return err;
+}
+
+static int taprio_disable_offload(struct net_device *dev,
+				  struct taprio_sched *q,
+				  struct netlink_ext_ack *extack)
+{
+	const struct net_device_ops *ops = dev->netdev_ops;
+	struct tc_taprio_qopt_offload *offload;
+	int err;
+
+	if (!FULL_OFFLOAD_IS_ENABLED(q->flags))
+		return 0;
+
+	if (!ops->ndo_setup_tc)
+		return -EOPNOTSUPP;
+
+	offload = taprio_offload_alloc(0);
+	if (!offload) {
+		NL_SET_ERR_MSG(extack,
+			       "Not enough memory to disable offload mode");
+		return -ENOMEM;
+	}
+	offload->enable = 0;
+
+	err = ops->ndo_setup_tc(dev, TC_SETUP_QDISC_TAPRIO, offload);
+	if (err < 0) {
+		NL_SET_ERR_MSG(extack,
+			       "Device failed to disable offload");
+		goto out;
+	}
+
+out:
+	taprio_offload_free(offload);
+
+	return err;
+}
+
+/* If full offload is enabled, the only possible clockid is the net device's
+ * PHC. For that reason, specifying a clockid through netlink is incorrect.
+ * For txtime-assist, it is implicitly assumed that the device's PHC is kept
+ * in sync with the specified clockid via a user space daemon such as phc2sys.
+ * For both software taprio and txtime-assist, the clockid is used for the
+ * hrtimer that advances the schedule and hence mandatory.
+ */
+static int taprio_parse_clockid(struct Qdisc *sch, struct nlattr **tb,
+				struct netlink_ext_ack *extack)
+{
+	struct taprio_sched *q = qdisc_priv(sch);
+	struct net_device *dev = qdisc_dev(sch);
+	int err = -EINVAL;
+
+	if (FULL_OFFLOAD_IS_ENABLED(q->flags)) {
+		const struct ethtool_ops *ops = dev->ethtool_ops;
+		struct ethtool_ts_info info = {
+			.cmd = ETHTOOL_GET_TS_INFO,
+			.phc_index = -1,
+		};
+
+		if (tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]) {
+			NL_SET_ERR_MSG(extack,
+				       "The 'clockid' cannot be specified for full offload");
+			goto out;
+		}
+
+		if (ops && ops->get_ts_info)
+			err = ops->get_ts_info(dev, &info);
+
+		if (err || info.phc_index < 0) {
+			NL_SET_ERR_MSG(extack,
+				       "Device does not have a PTP clock");
+			err = -ENOTSUPP;
+			goto out;
+		}
+	} else if (tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]) {
+		int clockid = nla_get_s32(tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]);
+
+		/* We only support static clockids and we don't allow
+		 * for it to be modified after the first init.
+		 */
+		if (clockid < 0 ||
+		    (q->clockid != -1 && q->clockid != clockid)) {
+			NL_SET_ERR_MSG(extack,
+				       "Changing the 'clockid' of a running schedule is not supported");
+			err = -ENOTSUPP;
+			goto out;
+		}
+
+		switch (clockid) {
+		case CLOCK_REALTIME:
+			q->tk_offset = TK_OFFS_REAL;
+			break;
+		case CLOCK_MONOTONIC:
+			q->tk_offset = TK_OFFS_MAX;
+			break;
+		case CLOCK_BOOTTIME:
+			q->tk_offset = TK_OFFS_BOOT;
+			break;
+		case CLOCK_TAI:
+			q->tk_offset = TK_OFFS_TAI;
+			break;
+		default:
+			NL_SET_ERR_MSG(extack, "Invalid 'clockid'");
+			err = -EINVAL;
+			goto out;
+		}
+
+		q->clockid = clockid;
+	} else {
+		NL_SET_ERR_MSG(extack, "Specifying a 'clockid' is mandatory");
+		goto out;
+	}
+out:
+	return err;
+}
+
 static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 			 struct netlink_ext_ack *extack)
 {
@@ -1020,9 +1355,9 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 	struct net_device *dev = qdisc_dev(sch);
 	struct tc_mqprio_qopt *mqprio = NULL;
 	u32 taprio_flags = 0;
-	int i, err, clockid;
 	unsigned long flags;
 	ktime_t start;
+	int i, err;
 
 	err = nla_parse_nested_deprecated(tb, TCA_TAPRIO_ATTR_MAX, opt,
 					  taprio_policy, extack);
@@ -1038,7 +1373,7 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 		if (q->flags != 0 && q->flags != taprio_flags) {
 			NL_SET_ERR_MSG_MOD(extack, "Changing 'flags' of a running schedule is not supported");
 			return -EOPNOTSUPP;
-		} else if (!FLAGS_VALID(taprio_flags)) {
+		} else if (!taprio_flags_valid(taprio_flags)) {
 			NL_SET_ERR_MSG_MOD(extack, "Specified 'flags' are not valid");
 			return -EINVAL;
 		}
@@ -1078,30 +1413,19 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 		goto free_sched;
 	}
 
-	if (tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]) {
-		clockid = nla_get_s32(tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]);
-
-		/* We only support static clockids and we don't allow
-		 * for it to be modified after the first init.
-		 */
-		if (clockid < 0 ||
-		    (q->clockid != -1 && q->clockid != clockid)) {
-			NL_SET_ERR_MSG(extack, "Changing the 'clockid' of a running schedule is not supported");
-			err = -ENOTSUPP;
-			goto free_sched;
-		}
-
-		q->clockid = clockid;
-	}
-
-	if (q->clockid == -1 && !tb[TCA_TAPRIO_ATTR_SCHED_CLOCKID]) {
-		NL_SET_ERR_MSG(extack, "Specifying a 'clockid' is mandatory");
-		err = -EINVAL;
+	err = taprio_parse_clockid(sch, tb, extack);
+	if (err < 0)
 		goto free_sched;
-	}
 
 	taprio_set_picos_per_byte(dev, q);
 
+	if (FULL_OFFLOAD_IS_ENABLED(taprio_flags))
+		err = taprio_enable_offload(dev, mqprio, q, new_admin, extack);
+	else
+		err = taprio_disable_offload(dev, q, extack);
+	if (err)
+		goto free_sched;
+
 	/* Protects against enqueue()/dequeue() */
 	spin_lock_bh(qdisc_lock(sch));
 
@@ -1116,6 +1440,7 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 	}
 
 	if (!TXTIME_ASSIST_IS_ENABLED(taprio_flags) &&
+	    !FULL_OFFLOAD_IS_ENABLED(taprio_flags) &&
 	    !hrtimer_active(&q->advance_timer)) {
 		hrtimer_init(&q->advance_timer, q->clockid, HRTIMER_MODE_ABS);
 		q->advance_timer.function = advance_sched;
@@ -1134,23 +1459,15 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
 					       mqprio->prio_tc_map[i]);
 	}
 
-	switch (q->clockid) {
-	case CLOCK_REALTIME:
-		q->tk_offset = TK_OFFS_REAL;
-		break;
-	case CLOCK_MONOTONIC:
-		q->tk_offset = TK_OFFS_MAX;
-		break;
-	case CLOCK_BOOTTIME:
-		q->tk_offset = TK_OFFS_BOOT;
-		break;
-	case CLOCK_TAI:
-		q->tk_offset = TK_OFFS_TAI;
-		break;
-	default:
-		NL_SET_ERR_MSG(extack, "Invalid 'clockid'");
-		err = -EINVAL;
-		goto unlock;
+	if (FULL_OFFLOAD_IS_ENABLED(taprio_flags)) {
+		q->dequeue = taprio_dequeue_offload;
+		q->peek = taprio_peek_offload;
+	} else {
+		/* Be sure to always keep the function pointers
+		 * in a consistent state.
+		 */
+		q->dequeue = taprio_dequeue_soft;
+		q->peek = taprio_peek_soft;
 	}
 
 	err = taprio_get_start_time(sch, new_admin, &start);
@@ -1212,6 +1529,8 @@ static void taprio_destroy(struct Qdisc *sch)
 
 	hrtimer_cancel(&q->advance_timer);
 
+	taprio_disable_offload(dev, q, NULL);
+
 	if (q->qdiscs) {
 		for (i = 0; i < dev->num_tx_queues && q->qdiscs[i]; i++)
 			qdisc_put(q->qdiscs[i]);
@@ -1241,6 +1560,9 @@ static int taprio_init(struct Qdisc *sch, struct nlattr *opt,
 	hrtimer_init(&q->advance_timer, CLOCK_TAI, HRTIMER_MODE_ABS);
 	q->advance_timer.function = advance_sched;
 
+	q->dequeue = taprio_dequeue_soft;
+	q->peek = taprio_peek_soft;
+
 	q->root = sch;
 
 	/* We only support static clockids. Use an invalid value as default
@@ -1423,7 +1745,8 @@ static int taprio_dump(struct Qdisc *sch, struct sk_buff *skb)
 	if (nla_put(skb, TCA_TAPRIO_ATTR_PRIOMAP, sizeof(opt), &opt))
 		goto options_error;
 
-	if (nla_put_s32(skb, TCA_TAPRIO_ATTR_SCHED_CLOCKID, q->clockid))
+	if (!FULL_OFFLOAD_IS_ENABLED(q->flags) &&
+	    nla_put_s32(skb, TCA_TAPRIO_ATTR_SCHED_CLOCKID, q->clockid))
 		goto options_error;
 
 	if (q->flags && nla_put_u32(skb, TCA_TAPRIO_ATTR_FLAGS, q->flags))
-- 
2.17.1

[PATCH v4 net-next 2/6] net: dsa: Pass ndo_setup_tc slave callback to drivers

[Vladimir Oltean]

DSA currently handles shared block filters (for the classifier-action
qdisc) in the core due to what I believe are simply pragmatic reasons -
hiding the complexity from drivers and offerring a simple API for port
mirroring.

Extend the dsa_slave_setup_tc function by passing all other qdisc
offloads to the driver layer, where the driver may choose what it
implements and how. DSA is simply a pass-through in this case.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
---
Changes since v2:
- Added Florian Fainelli's Reviewed-by.

Changes since v1:
- Added Kurt Kanzenbach's Acked-by.

Changes since RFC:
- Removed the unused declaration of struct tc_taprio_qopt_offload.

 include/net/dsa.h |  2 ++
 net/dsa/slave.c   | 12 ++++++++----
 2 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/include/net/dsa.h b/include/net/dsa.h
index 96acb14ec1a8..541fb514e31d 100644
--- a/include/net/dsa.h
+++ b/include/net/dsa.h
@@ -515,6 +515,8 @@ struct dsa_switch_ops {
 				   bool ingress);
 	void	(*port_mirror_del)(struct dsa_switch *ds, int port,
 				   struct dsa_mall_mirror_tc_entry *mirror);
+	int	(*port_setup_tc)(struct dsa_switch *ds, int port,
+				 enum tc_setup_type type, void *type_data);
 
 	/*
 	 * Cross-chip operations
diff --git a/net/dsa/slave.c b/net/dsa/slave.c
index 9a88035517a6..75d58229a4bd 100644
--- a/net/dsa/slave.c
+++ b/net/dsa/slave.c
@@ -1035,12 +1035,16 @@ static int dsa_slave_setup_tc_block(struct net_device *dev,
 static int dsa_slave_setup_tc(struct net_device *dev, enum tc_setup_type type,
 			      void *type_data)
 {
-	switch (type) {
-	case TC_SETUP_BLOCK:
+	struct dsa_port *dp = dsa_slave_to_port(dev);
+	struct dsa_switch *ds = dp->ds;
+
+	if (type == TC_SETUP_BLOCK)
 		return dsa_slave_setup_tc_block(dev, type_data);
-	default:
+
+	if (!ds->ops->port_setup_tc)
 		return -EOPNOTSUPP;
-	}
+
+	return ds->ops->port_setup_tc(ds, dp->index, type, type_data);
 }
 
 static void dsa_slave_get_stats64(struct net_device *dev,
-- 
2.17.1

[PATCH v4 net-next 3/6] net: dsa: sja1105: Add static config tables for scheduling

[Vladimir Oltean]

In order to support tc-taprio offload, the TTEthernet egress scheduling
core registers must be made visible through the static interface.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
---
Changes since v2:
- None.

Changes since v1:
- None.

Changes since RFC:
- None.

 .../net/dsa/sja1105/sja1105_dynamic_config.c  |   8 +
 .../net/dsa/sja1105/sja1105_static_config.c   | 167 ++++++++++++++++++
 .../net/dsa/sja1105/sja1105_static_config.h   |  48 ++++-
 3 files changed, 222 insertions(+), 1 deletion(-)

diff --git a/drivers/net/dsa/sja1105/sja1105_dynamic_config.c b/drivers/net/dsa/sja1105/sja1105_dynamic_config.c
index 9988c9d18567..91da430045ff 100644
--- a/drivers/net/dsa/sja1105/sja1105_dynamic_config.c
+++ b/drivers/net/dsa/sja1105/sja1105_dynamic_config.c
@@ -488,6 +488,8 @@ sja1105et_general_params_entry_packing(void *buf, void *entry_ptr,
 
 /* SJA1105E/T: First generation */
 struct sja1105_dynamic_table_ops sja1105et_dyn_ops[BLK_IDX_MAX_DYN] = {
+	[BLK_IDX_SCHEDULE] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {0},
 	[BLK_IDX_L2_LOOKUP] = {
 		.entry_packing = sja1105et_dyn_l2_lookup_entry_packing,
 		.cmd_packing = sja1105et_l2_lookup_cmd_packing,
@@ -529,6 +531,8 @@ struct sja1105_dynamic_table_ops sja1105et_dyn_ops[BLK_IDX_MAX_DYN] = {
 		.packed_size = SJA1105ET_SIZE_MAC_CONFIG_DYN_CMD,
 		.addr = 0x36,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {0},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.entry_packing = sja1105et_l2_lookup_params_entry_packing,
 		.cmd_packing = sja1105et_l2_lookup_params_cmd_packing,
@@ -552,6 +556,8 @@ struct sja1105_dynamic_table_ops sja1105et_dyn_ops[BLK_IDX_MAX_DYN] = {
 
 /* SJA1105P/Q/R/S: Second generation */
 struct sja1105_dynamic_table_ops sja1105pqrs_dyn_ops[BLK_IDX_MAX_DYN] = {
+	[BLK_IDX_SCHEDULE] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {0},
 	[BLK_IDX_L2_LOOKUP] = {
 		.entry_packing = sja1105pqrs_dyn_l2_lookup_entry_packing,
 		.cmd_packing = sja1105pqrs_l2_lookup_cmd_packing,
@@ -593,6 +599,8 @@ struct sja1105_dynamic_table_ops sja1105pqrs_dyn_ops[BLK_IDX_MAX_DYN] = {
 		.packed_size = SJA1105PQRS_SIZE_MAC_CONFIG_DYN_CMD,
 		.addr = 0x4B,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {0},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.entry_packing = sja1105et_l2_lookup_params_entry_packing,
 		.cmd_packing = sja1105et_l2_lookup_params_cmd_packing,
diff --git a/drivers/net/dsa/sja1105/sja1105_static_config.c b/drivers/net/dsa/sja1105/sja1105_static_config.c
index b31c737dc560..0d03e13e9909 100644
--- a/drivers/net/dsa/sja1105/sja1105_static_config.c
+++ b/drivers/net/dsa/sja1105/sja1105_static_config.c
@@ -371,6 +371,63 @@ size_t sja1105pqrs_mac_config_entry_packing(void *buf, void *entry_ptr,
 	return size;
 }
 
+static size_t
+sja1105_schedule_entry_points_params_entry_packing(void *buf, void *entry_ptr,
+						   enum packing_op op)
+{
+	struct sja1105_schedule_entry_points_params_entry *entry = entry_ptr;
+	const size_t size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_PARAMS_ENTRY;
+
+	sja1105_packing(buf, &entry->clksrc,    31, 30, size, op);
+	sja1105_packing(buf, &entry->actsubsch, 29, 27, size, op);
+	return size;
+}
+
+static size_t
+sja1105_schedule_entry_points_entry_packing(void *buf, void *entry_ptr,
+					    enum packing_op op)
+{
+	struct sja1105_schedule_entry_points_entry *entry = entry_ptr;
+	const size_t size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_ENTRY;
+
+	sja1105_packing(buf, &entry->subschindx, 31, 29, size, op);
+	sja1105_packing(buf, &entry->delta,      28, 11, size, op);
+	sja1105_packing(buf, &entry->address,    10, 1,  size, op);
+	return size;
+}
+
+static size_t sja1105_schedule_params_entry_packing(void *buf, void *entry_ptr,
+						    enum packing_op op)
+{
+	const size_t size = SJA1105_SIZE_SCHEDULE_PARAMS_ENTRY;
+	struct sja1105_schedule_params_entry *entry = entry_ptr;
+	int offset, i;
+
+	for (i = 0, offset = 16; i < 8; i++, offset += 10)
+		sja1105_packing(buf, &entry->subscheind[i],
+				offset + 9, offset + 0, size, op);
+	return size;
+}
+
+static size_t sja1105_schedule_entry_packing(void *buf, void *entry_ptr,
+					     enum packing_op op)
+{
+	const size_t size = SJA1105_SIZE_SCHEDULE_ENTRY;
+	struct sja1105_schedule_entry *entry = entry_ptr;
+
+	sja1105_packing(buf, &entry->winstindex,  63, 54, size, op);
+	sja1105_packing(buf, &entry->winend,      53, 53, size, op);
+	sja1105_packing(buf, &entry->winst,       52, 52, size, op);
+	sja1105_packing(buf, &entry->destports,   51, 47, size, op);
+	sja1105_packing(buf, &entry->setvalid,    46, 46, size, op);
+	sja1105_packing(buf, &entry->txen,        45, 45, size, op);
+	sja1105_packing(buf, &entry->resmedia_en, 44, 44, size, op);
+	sja1105_packing(buf, &entry->resmedia,    43, 36, size, op);
+	sja1105_packing(buf, &entry->vlindex,     35, 26, size, op);
+	sja1105_packing(buf, &entry->delta,       25, 8,  size, op);
+	return size;
+}
+
 size_t sja1105_vlan_lookup_entry_packing(void *buf, void *entry_ptr,
 					 enum packing_op op)
 {
@@ -447,11 +504,15 @@ static void sja1105_table_write_crc(u8 *table_start, u8 *crc_ptr)
  * before blindly indexing kernel memory with the blk_idx.
  */
 static u64 blk_id_map[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = BLKID_SCHEDULE,
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = BLKID_SCHEDULE_ENTRY_POINTS,
 	[BLK_IDX_L2_LOOKUP] = BLKID_L2_LOOKUP,
 	[BLK_IDX_L2_POLICING] = BLKID_L2_POLICING,
 	[BLK_IDX_VLAN_LOOKUP] = BLKID_VLAN_LOOKUP,
 	[BLK_IDX_L2_FORWARDING] = BLKID_L2_FORWARDING,
 	[BLK_IDX_MAC_CONFIG] = BLKID_MAC_CONFIG,
+	[BLK_IDX_SCHEDULE_PARAMS] = BLKID_SCHEDULE_PARAMS,
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = BLKID_SCHEDULE_ENTRY_POINTS_PARAMS,
 	[BLK_IDX_L2_LOOKUP_PARAMS] = BLKID_L2_LOOKUP_PARAMS,
 	[BLK_IDX_L2_FORWARDING_PARAMS] = BLKID_L2_FORWARDING_PARAMS,
 	[BLK_IDX_AVB_PARAMS] = BLKID_AVB_PARAMS,
@@ -461,6 +522,13 @@ static u64 blk_id_map[BLK_IDX_MAX] = {
 
 const char *sja1105_static_config_error_msg[] = {
 	[SJA1105_CONFIG_OK] = "",
+	[SJA1105_TTETHERNET_NOT_SUPPORTED] =
+		"schedule-table present, but TTEthernet is "
+		"only supported on T and Q/S",
+	[SJA1105_INCORRECT_TTETHERNET_CONFIGURATION] =
+		"schedule-table present, but one of "
+		"schedule-entry-points-table, schedule-parameters-table or "
+		"schedule-entry-points-parameters table is empty",
 	[SJA1105_MISSING_L2_POLICING_TABLE] =
 		"l2-policing-table needs to have at least one entry",
 	[SJA1105_MISSING_L2_FORWARDING_TABLE] =
@@ -508,6 +576,21 @@ sja1105_static_config_check_valid(const struct sja1105_static_config *config)
 #define IS_FULL(blk_idx) \
 	(tables[blk_idx].entry_count == tables[blk_idx].ops->max_entry_count)
 
+	if (tables[BLK_IDX_SCHEDULE].entry_count) {
+		if (config->device_id != SJA1105T_DEVICE_ID &&
+		    config->device_id != SJA1105QS_DEVICE_ID)
+			return SJA1105_TTETHERNET_NOT_SUPPORTED;
+
+		if (tables[BLK_IDX_SCHEDULE_ENTRY_POINTS].entry_count == 0)
+			return SJA1105_INCORRECT_TTETHERNET_CONFIGURATION;
+
+		if (!IS_FULL(BLK_IDX_SCHEDULE_PARAMS))
+			return SJA1105_INCORRECT_TTETHERNET_CONFIGURATION;
+
+		if (!IS_FULL(BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS))
+			return SJA1105_INCORRECT_TTETHERNET_CONFIGURATION;
+	}
+
 	if (tables[BLK_IDX_L2_POLICING].entry_count == 0)
 		return SJA1105_MISSING_L2_POLICING_TABLE;
 
@@ -614,6 +697,8 @@ sja1105_static_config_get_length(const struct sja1105_static_config *config)
 
 /* SJA1105E: First generation, no TTEthernet */
 struct sja1105_table_ops sja1105e_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {0},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105et_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -644,6 +729,8 @@ struct sja1105_table_ops sja1105e_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105ET_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {0},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105et_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
@@ -678,6 +765,18 @@ struct sja1105_table_ops sja1105e_table_ops[BLK_IDX_MAX] = {
 
 /* SJA1105T: First generation, TTEthernet */
 struct sja1105_table_ops sja1105t_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {
+		.packing = sja1105_schedule_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {
+		.packing = sja1105_schedule_entry_points_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105et_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -708,6 +807,18 @@ struct sja1105_table_ops sja1105t_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105ET_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {
+		.packing = sja1105_schedule_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_PARAMS_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {
+		.packing = sja1105_schedule_entry_points_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105et_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
@@ -742,6 +853,8 @@ struct sja1105_table_ops sja1105t_table_ops[BLK_IDX_MAX] = {
 
 /* SJA1105P: Second generation, no TTEthernet, no SGMII */
 struct sja1105_table_ops sja1105p_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {0},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105pqrs_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -772,6 +885,8 @@ struct sja1105_table_ops sja1105p_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105PQRS_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {0},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105pqrs_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
@@ -806,6 +921,18 @@ struct sja1105_table_ops sja1105p_table_ops[BLK_IDX_MAX] = {
 
 /* SJA1105Q: Second generation, TTEthernet, no SGMII */
 struct sja1105_table_ops sja1105q_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {
+		.packing = sja1105_schedule_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {
+		.packing = sja1105_schedule_entry_points_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105pqrs_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -836,6 +963,18 @@ struct sja1105_table_ops sja1105q_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105PQRS_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {
+		.packing = sja1105_schedule_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_PARAMS_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {
+		.packing = sja1105_schedule_entry_points_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105pqrs_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
@@ -870,6 +1009,8 @@ struct sja1105_table_ops sja1105q_table_ops[BLK_IDX_MAX] = {
 
 /* SJA1105R: Second generation, no TTEthernet, SGMII */
 struct sja1105_table_ops sja1105r_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {0},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105pqrs_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -900,6 +1041,8 @@ struct sja1105_table_ops sja1105r_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105PQRS_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {0},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {0},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105pqrs_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
@@ -934,6 +1077,18 @@ struct sja1105_table_ops sja1105r_table_ops[BLK_IDX_MAX] = {
 
 /* SJA1105S: Second generation, TTEthernet, SGMII */
 struct sja1105_table_ops sja1105s_table_ops[BLK_IDX_MAX] = {
+	[BLK_IDX_SCHEDULE] = {
+		.packing = sja1105_schedule_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS] = {
+		.packing = sja1105_schedule_entry_points_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP] = {
 		.packing = sja1105pqrs_l2_lookup_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_entry),
@@ -964,6 +1119,18 @@ struct sja1105_table_ops sja1105s_table_ops[BLK_IDX_MAX] = {
 		.packed_entry_size = SJA1105PQRS_SIZE_MAC_CONFIG_ENTRY,
 		.max_entry_count = SJA1105_MAX_MAC_CONFIG_COUNT,
 	},
+	[BLK_IDX_SCHEDULE_PARAMS] = {
+		.packing = sja1105_schedule_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_PARAMS_COUNT,
+	},
+	[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS] = {
+		.packing = sja1105_schedule_entry_points_params_entry_packing,
+		.unpacked_entry_size = sizeof(struct sja1105_schedule_entry_points_params_entry),
+		.packed_entry_size = SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_PARAMS_ENTRY,
+		.max_entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT,
+	},
 	[BLK_IDX_L2_LOOKUP_PARAMS] = {
 		.packing = sja1105pqrs_l2_lookup_params_entry_packing,
 		.unpacked_entry_size = sizeof(struct sja1105_l2_lookup_params_entry),
diff --git a/drivers/net/dsa/sja1105/sja1105_static_config.h b/drivers/net/dsa/sja1105/sja1105_static_config.h
index 684465fc0882..7f87022a2d61 100644
--- a/drivers/net/dsa/sja1105/sja1105_static_config.h
+++ b/drivers/net/dsa/sja1105/sja1105_static_config.h
@@ -11,11 +11,15 @@
 
 #define SJA1105_SIZE_DEVICE_ID				4
 #define SJA1105_SIZE_TABLE_HEADER			12
+#define SJA1105_SIZE_SCHEDULE_ENTRY			8
+#define SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_ENTRY	4
 #define SJA1105_SIZE_L2_POLICING_ENTRY			8
 #define SJA1105_SIZE_VLAN_LOOKUP_ENTRY			8
 #define SJA1105_SIZE_L2_FORWARDING_ENTRY		8
 #define SJA1105_SIZE_L2_FORWARDING_PARAMS_ENTRY		12
 #define SJA1105_SIZE_XMII_PARAMS_ENTRY			4
+#define SJA1105_SIZE_SCHEDULE_PARAMS_ENTRY		12
+#define SJA1105_SIZE_SCHEDULE_ENTRY_POINTS_PARAMS_ENTRY	4
 #define SJA1105ET_SIZE_L2_LOOKUP_ENTRY			12
 #define SJA1105ET_SIZE_MAC_CONFIG_ENTRY			28
 #define SJA1105ET_SIZE_L2_LOOKUP_PARAMS_ENTRY		4
@@ -29,11 +33,15 @@
 
 /* UM10944.pdf Page 11, Table 2. Configuration Blocks */
 enum {
+	BLKID_SCHEDULE					= 0x00,
+	BLKID_SCHEDULE_ENTRY_POINTS			= 0x01,
 	BLKID_L2_LOOKUP					= 0x05,
 	BLKID_L2_POLICING				= 0x06,
 	BLKID_VLAN_LOOKUP				= 0x07,
 	BLKID_L2_FORWARDING				= 0x08,
 	BLKID_MAC_CONFIG				= 0x09,
+	BLKID_SCHEDULE_PARAMS				= 0x0A,
+	BLKID_SCHEDULE_ENTRY_POINTS_PARAMS		= 0x0B,
 	BLKID_L2_LOOKUP_PARAMS				= 0x0D,
 	BLKID_L2_FORWARDING_PARAMS			= 0x0E,
 	BLKID_AVB_PARAMS				= 0x10,
@@ -42,11 +50,15 @@ enum {
 };
 
 enum sja1105_blk_idx {
-	BLK_IDX_L2_LOOKUP = 0,
+	BLK_IDX_SCHEDULE = 0,
+	BLK_IDX_SCHEDULE_ENTRY_POINTS,
+	BLK_IDX_L2_LOOKUP,
 	BLK_IDX_L2_POLICING,
 	BLK_IDX_VLAN_LOOKUP,
 	BLK_IDX_L2_FORWARDING,
 	BLK_IDX_MAC_CONFIG,
+	BLK_IDX_SCHEDULE_PARAMS,
+	BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS,
 	BLK_IDX_L2_LOOKUP_PARAMS,
 	BLK_IDX_L2_FORWARDING_PARAMS,
 	BLK_IDX_AVB_PARAMS,
@@ -59,11 +71,15 @@ enum sja1105_blk_idx {
 	BLK_IDX_INVAL = -1,
 };
 
+#define SJA1105_MAX_SCHEDULE_COUNT			1024
+#define SJA1105_MAX_SCHEDULE_ENTRY_POINTS_COUNT		2048
 #define SJA1105_MAX_L2_LOOKUP_COUNT			1024
 #define SJA1105_MAX_L2_POLICING_COUNT			45
 #define SJA1105_MAX_VLAN_LOOKUP_COUNT			4096
 #define SJA1105_MAX_L2_FORWARDING_COUNT			13
 #define SJA1105_MAX_MAC_CONFIG_COUNT			5
+#define SJA1105_MAX_SCHEDULE_PARAMS_COUNT		1
+#define SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT	1
 #define SJA1105_MAX_L2_LOOKUP_PARAMS_COUNT		1
 #define SJA1105_MAX_L2_FORWARDING_PARAMS_COUNT		1
 #define SJA1105_MAX_GENERAL_PARAMS_COUNT		1
@@ -83,6 +99,23 @@ enum sja1105_blk_idx {
 #define SJA1105R_PART_NO				0x9A86
 #define SJA1105S_PART_NO				0x9A87
 
+struct sja1105_schedule_entry {
+	u64 winstindex;
+	u64 winend;
+	u64 winst;
+	u64 destports;
+	u64 setvalid;
+	u64 txen;
+	u64 resmedia_en;
+	u64 resmedia;
+	u64 vlindex;
+	u64 delta;
+};
+
+struct sja1105_schedule_params_entry {
+	u64 subscheind[8];
+};
+
 struct sja1105_general_params_entry {
 	u64 vllupformat;
 	u64 mirr_ptacu;
@@ -112,6 +145,17 @@ struct sja1105_general_params_entry {
 	u64 replay_port;
 };
 
+struct sja1105_schedule_entry_points_entry {
+	u64 subschindx;
+	u64 delta;
+	u64 address;
+};
+
+struct sja1105_schedule_entry_points_params_entry {
+	u64 clksrc;
+	u64 actsubsch;
+};
+
 struct sja1105_vlan_lookup_entry {
 	u64 ving_mirr;
 	u64 vegr_mirr;
@@ -256,6 +300,8 @@ sja1105_static_config_get_length(const struct sja1105_static_config *config);
 
 typedef enum {
 	SJA1105_CONFIG_OK = 0,
+	SJA1105_TTETHERNET_NOT_SUPPORTED,
+	SJA1105_INCORRECT_TTETHERNET_CONFIGURATION,
 	SJA1105_MISSING_L2_POLICING_TABLE,
 	SJA1105_MISSING_L2_FORWARDING_TABLE,
 	SJA1105_MISSING_L2_FORWARDING_PARAMS_TABLE,
-- 
2.17.1

[PATCH v4 net-next 4/6] net: dsa: sja1105: Advertise the 8 TX queues

[Vladimir Oltean]

This is a preparation patch for the tc-taprio offload (and potentially
for other future offloads such as tc-mqprio).

Instead of looking directly at skb->priority during xmit, let's get the
netdev queue and the queue-to-traffic-class mapping, and put the
resulting traffic class into the dsa_8021q PCP field. The switch is
configured with a 1-to-1 PCP-to-ingress-queue-to-egress-queue mapping
(see vlan_pmap in sja1105_main.c), so the effect is that we can inject
into a front-panel's egress traffic class through VLAN tagging from
Linux, completely transparently.

Unfortunately the switch doesn't look at the VLAN PCP in the case of
management traffic to/from the CPU (link-local frames at
01-80-C2-xx-xx-xx or 01-1B-19-xx-xx-xx) so we can't alter the
transmission queue of this type of traffic on a frame-by-frame basis. It
is only selected through the "hostprio" setting which ATM is harcoded in
the driver to 7.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
---
Changes since v2:
- None.

Changes since v1:
- None, but the use of netdev_txq_to_tc is now finally correct after
  adjusting the gate_mask meaning in the taprio offload structure.

Changes since RFC:
- None.

 drivers/net/dsa/sja1105/sja1105_main.c | 7 ++++++-
 net/dsa/tag_sja1105.c                  | 3 ++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/net/dsa/sja1105/sja1105_main.c b/drivers/net/dsa/sja1105/sja1105_main.c
index d8cff0107ec4..108f62c27c28 100644
--- a/drivers/net/dsa/sja1105/sja1105_main.c
+++ b/drivers/net/dsa/sja1105/sja1105_main.c
@@ -384,7 +384,9 @@ static int sja1105_init_general_params(struct sja1105_private *priv)
 		/* Disallow dynamic changing of the mirror port */
 		.mirr_ptacu = 0,
 		.switchid = priv->ds->index,
-		/* Priority queue for link-local frames trapped to CPU */
+		/* Priority queue for link-local management frames
+		 * (both ingress to and egress from CPU - PTP, STP etc)
+		 */
 		.hostprio = 7,
 		.mac_fltres1 = SJA1105_LINKLOCAL_FILTER_A,
 		.mac_flt1    = SJA1105_LINKLOCAL_FILTER_A_MASK,
@@ -1711,6 +1713,9 @@ static int sja1105_setup(struct dsa_switch *ds)
 	 */
 	ds->vlan_filtering_is_global = true;
 
+	/* Advertise the 8 egress queues */
+	ds->num_tx_queues = SJA1105_NUM_TC;
+
 	/* The DSA/switchdev model brings up switch ports in standalone mode by
 	 * default, and that means vlan_filtering is 0 since they're not under
 	 * a bridge, so it's safe to set up switch tagging at this time.
diff --git a/net/dsa/tag_sja1105.c b/net/dsa/tag_sja1105.c
index 47ee88163a9d..9c9aff3e52cf 100644
--- a/net/dsa/tag_sja1105.c
+++ b/net/dsa/tag_sja1105.c
@@ -89,7 +89,8 @@ static struct sk_buff *sja1105_xmit(struct sk_buff *skb,
 	struct dsa_port *dp = dsa_slave_to_port(netdev);
 	struct dsa_switch *ds = dp->ds;
 	u16 tx_vid = dsa_8021q_tx_vid(ds, dp->index);
-	u8 pcp = skb->priority;
+	u16 queue_mapping = skb_get_queue_mapping(skb);
+	u8 pcp = netdev_txq_to_tc(netdev, queue_mapping);
 
 	/* Transmitting management traffic does not rely upon switch tagging,
 	 * but instead SPI-installed management routes. Part 2 of this
-- 
2.17.1

[PATCH v4 net-next 5/6] net: dsa: sja1105: Configure the Time-Aware Scheduler via tc-taprio offload

[Vladimir Oltean]

This qdisc offload is the closest thing to what the SJA1105 supports in
hardware for time-based egress shaping. The switch core really is built
around SAE AS6802/TTEthernet (a TTTech standard) but can be made to
operate similarly to IEEE 802.1Qbv with some constraints:

- The gate control list is a global list for all ports. There are 8
  execution threads that iterate through this global list in parallel.
  I don't know why 8, there are only 4 front-panel ports.

- Care must be taken by the user to make sure that two execution threads
  never get to execute a GCL entry simultaneously. I created a O(n^4)
  checker for this hardware limitation, prior to accepting a taprio
  offload configuration as valid.

- The spec says that if a GCL entry's interval is shorter than the frame
  length, you shouldn't send it (and end up in head-of-line blocking).
  Well, this switch does anyway.

- The switch has no concept of ADMIN and OPER configurations. Because
  it's so simple, the TAS settings are loaded through the static config
  tables interface, so there isn't even place for any discussion about
  'graceful switchover between ADMIN and OPER'. You just reset the
  switch and upload a new OPER config.

- The switch accepts multiple time sources for the gate events. Right
  now I am using the standalone clock source as opposed to PTP. So the
  base time parameter doesn't really do much. Support for the PTP clock
  source will be added in a future series.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
---
Changes since v2:
- Made all functions that are entry points into sja1105_tas.c take a
  dsa_struct *ds argument instead of sja1105_private *priv. This also
  happens to avoid a build error reported by the Kbuild test robot.
- Moved the iteration over switch ports outside of
  sja1105_tas_check_conflicts, to address some checkpatch complaints.
- Renamed "new" -> "admin".

Changes since v1:
- Adapted to the naming convention changes in 01/07 (taprio_get ->
  taprio_offload_get, tas_config -> offload, etc).

Changes since RFC:
- Removed the sja1105_tas_config_work workqueue.
- Allocating memory with GFP_KERNEL.
- Made the ASCII art drawing fit in < 80 characters.
- Made most of the time-holding variables s64 instead of u64 (for fear
  of them not holding the result of signed arithmetics properly).

 drivers/net/dsa/sja1105/Kconfig        |   8 +
 drivers/net/dsa/sja1105/Makefile       |   4 +
 drivers/net/dsa/sja1105/sja1105.h      |   6 +
 drivers/net/dsa/sja1105/sja1105_main.c |  19 +-
 drivers/net/dsa/sja1105/sja1105_tas.c  | 423 +++++++++++++++++++++++++
 drivers/net/dsa/sja1105/sja1105_tas.h  |  41 +++
 6 files changed, 500 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/dsa/sja1105/sja1105_tas.c
 create mode 100644 drivers/net/dsa/sja1105/sja1105_tas.h

diff --git a/drivers/net/dsa/sja1105/Kconfig b/drivers/net/dsa/sja1105/Kconfig
index 770134a66e48..55424f39cb0d 100644
--- a/drivers/net/dsa/sja1105/Kconfig
+++ b/drivers/net/dsa/sja1105/Kconfig
@@ -23,3 +23,11 @@ config NET_DSA_SJA1105_PTP
 	help
 	  This enables support for timestamping and PTP clock manipulations in
 	  the SJA1105 DSA driver.
+
+config NET_DSA_SJA1105_TAS
+	bool "Support for the Time-Aware Scheduler on NXP SJA1105"
+	depends on NET_DSA_SJA1105
+	help
+	  This enables support for the TTEthernet-based egress scheduling
+	  engine in the SJA1105 DSA driver, which is controlled using a
+	  hardware offload of the tc-tqprio qdisc.
diff --git a/drivers/net/dsa/sja1105/Makefile b/drivers/net/dsa/sja1105/Makefile
index 4483113e6259..66161e874344 100644
--- a/drivers/net/dsa/sja1105/Makefile
+++ b/drivers/net/dsa/sja1105/Makefile
@@ -12,3 +12,7 @@ sja1105-objs := \
 ifdef CONFIG_NET_DSA_SJA1105_PTP
 sja1105-objs += sja1105_ptp.o
 endif
+
+ifdef CONFIG_NET_DSA_SJA1105_TAS
+sja1105-objs += sja1105_tas.o
+endif
diff --git a/drivers/net/dsa/sja1105/sja1105.h b/drivers/net/dsa/sja1105/sja1105.h
index 78094db32622..e53e494c22e0 100644
--- a/drivers/net/dsa/sja1105/sja1105.h
+++ b/drivers/net/dsa/sja1105/sja1105.h
@@ -20,6 +20,8 @@
  */
 #define SJA1105_AGEING_TIME_MS(ms)	((ms) / 10)
 
+#include "sja1105_tas.h"
+
 /* Keeps the different addresses between E/T and P/Q/R/S */
 struct sja1105_regs {
 	u64 device_id;
@@ -104,6 +106,7 @@ struct sja1105_private {
 	 */
 	struct mutex mgmt_lock;
 	struct sja1105_tagger_data tagger_data;
+	struct sja1105_tas_data tas_data;
 };
 
 #include "sja1105_dynamic_config.h"
@@ -120,6 +123,9 @@ typedef enum {
 	SPI_WRITE = 1,
 } sja1105_spi_rw_mode_t;
 
+/* From sja1105_main.c */
+int sja1105_static_config_reload(struct sja1105_private *priv);
+
 /* From sja1105_spi.c */
 int sja1105_spi_send_packed_buf(const struct sja1105_private *priv,
 				sja1105_spi_rw_mode_t rw, u64 reg_addr,
diff --git a/drivers/net/dsa/sja1105/sja1105_main.c b/drivers/net/dsa/sja1105/sja1105_main.c
index 108f62c27c28..b9def744bcb3 100644
--- a/drivers/net/dsa/sja1105/sja1105_main.c
+++ b/drivers/net/dsa/sja1105/sja1105_main.c
@@ -22,6 +22,7 @@
 #include <linux/if_ether.h>
 #include <linux/dsa/8021q.h>
 #include "sja1105.h"
+#include "sja1105_tas.h"
 
 static void sja1105_hw_reset(struct gpio_desc *gpio, unsigned int pulse_len,
 			     unsigned int startup_delay)
@@ -1382,7 +1383,7 @@ static void sja1105_bridge_leave(struct dsa_switch *ds, int port,
  * modify at runtime (currently only MAC) and restore them after uploading,
  * such that this operation is relatively seamless.
  */
-static int sja1105_static_config_reload(struct sja1105_private *priv)
+int sja1105_static_config_reload(struct sja1105_private *priv)
 {
 	struct sja1105_mac_config_entry *mac;
 	int speed_mbps[SJA1105_NUM_PORTS];
@@ -1727,6 +1728,7 @@ static void sja1105_teardown(struct dsa_switch *ds)
 {
 	struct sja1105_private *priv = ds->priv;
 
+	sja1105_tas_teardown(ds);
 	cancel_work_sync(&priv->tagger_data.rxtstamp_work);
 	skb_queue_purge(&priv->tagger_data.skb_rxtstamp_queue);
 	sja1105_ptp_clock_unregister(priv);
@@ -2056,6 +2058,18 @@ static bool sja1105_port_txtstamp(struct dsa_switch *ds, int port,
 	return true;
 }
 
+static int sja1105_port_setup_tc(struct dsa_switch *ds, int port,
+				 enum tc_setup_type type,
+				 void *type_data)
+{
+	switch (type) {
+	case TC_SETUP_QDISC_TAPRIO:
+		return sja1105_setup_tc_taprio(ds, port, type_data);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static const struct dsa_switch_ops sja1105_switch_ops = {
 	.get_tag_protocol	= sja1105_get_tag_protocol,
 	.setup			= sja1105_setup,
@@ -2088,6 +2102,7 @@ static const struct dsa_switch_ops sja1105_switch_ops = {
 	.port_hwtstamp_set	= sja1105_hwtstamp_set,
 	.port_rxtstamp		= sja1105_port_rxtstamp,
 	.port_txtstamp		= sja1105_port_txtstamp,
+	.port_setup_tc		= sja1105_port_setup_tc,
 };
 
 static int sja1105_check_device_id(struct sja1105_private *priv)
@@ -2197,6 +2212,8 @@ static int sja1105_probe(struct spi_device *spi)
 	}
 	mutex_init(&priv->mgmt_lock);
 
+	sja1105_tas_setup(ds);
+
 	return dsa_register_switch(priv->ds);
 }
 
diff --git a/drivers/net/dsa/sja1105/sja1105_tas.c b/drivers/net/dsa/sja1105/sja1105_tas.c
new file mode 100644
index 000000000000..33eca6a82ec5
--- /dev/null
+++ b/drivers/net/dsa/sja1105/sja1105_tas.c
@@ -0,0 +1,423 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019, Vladimir Oltean <olteanv@gmail.com>
+ */
+#include "sja1105.h"
+
+#define SJA1105_TAS_CLKSRC_DISABLED	0
+#define SJA1105_TAS_CLKSRC_STANDALONE	1
+#define SJA1105_TAS_CLKSRC_AS6802	2
+#define SJA1105_TAS_CLKSRC_PTP		3
+#define SJA1105_TAS_MAX_DELTA		BIT(19)
+#define SJA1105_GATE_MASK		GENMASK_ULL(SJA1105_NUM_TC - 1, 0)
+
+/* This is not a preprocessor macro because the "ns" argument may or may not be
+ * s64 at caller side. This ensures it is properly type-cast before div_s64.
+ */
+static s64 ns_to_sja1105_delta(s64 ns)
+{
+	return div_s64(ns, 200);
+}
+
+/* Lo and behold: the egress scheduler from hell.
+ *
+ * At the hardware level, the Time-Aware Shaper holds a global linear arrray of
+ * all schedule entries for all ports. These are the Gate Control List (GCL)
+ * entries, let's call them "timeslots" for short. This linear array of
+ * timeslots is held in BLK_IDX_SCHEDULE.
+ *
+ * Then there are a maximum of 8 "execution threads" inside the switch, which
+ * iterate cyclically through the "schedule". Each "cycle" has an entry point
+ * and an exit point, both being timeslot indices in the schedule table. The
+ * hardware calls each cycle a "subschedule".
+ *
+ * Subschedule (cycle) i starts when
+ *   ptpclkval >= ptpschtm + BLK_IDX_SCHEDULE_ENTRY_POINTS[i].delta.
+ *
+ * The hardware scheduler iterates BLK_IDX_SCHEDULE with a k ranging from
+ *   k = BLK_IDX_SCHEDULE_ENTRY_POINTS[i].address to
+ *   k = BLK_IDX_SCHEDULE_PARAMS.subscheind[i]
+ *
+ * For each schedule entry (timeslot) k, the engine executes the gate control
+ * list entry for the duration of BLK_IDX_SCHEDULE[k].delta.
+ *
+ *         +---------+
+ *         |         | BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS
+ *         +---------+
+ *              |
+ *              +-----------------+
+ *                                | .actsubsch
+ *  BLK_IDX_SCHEDULE_ENTRY_POINTS v
+ *                 +-------+-------+
+ *                 |cycle 0|cycle 1|
+ *                 +-------+-------+
+ *                   |  |      |  |
+ *  +----------------+  |      |  +-------------------------------------+
+ *  |   .subschindx     |      |             .subschindx                |
+ *  |                   |      +---------------+                        |
+ *  |          .address |        .address      |                        |
+ *  |                   |                      |                        |
+ *  |                   |                      |                        |
+ *  |  BLK_IDX_SCHEDULE v                      v                        |
+ *  |              +-------+-------+-------+-------+-------+------+     |
+ *  |              |entry 0|entry 1|entry 2|entry 3|entry 4|entry5|     |
+ *  |              +-------+-------+-------+-------+-------+------+     |
+ *  |                                  ^                    ^  ^  ^     |
+ *  |                                  |                    |  |  |     |
+ *  |        +-------------------------+                    |  |  |     |
+ *  |        |              +-------------------------------+  |  |     |
+ *  |        |              |              +-------------------+  |     |
+ *  |        |              |              |                      |     |
+ *  | +---------------------------------------------------------------+ |
+ *  | |subscheind[0]<=subscheind[1]<=subscheind[2]<=...<=subscheind[7]| |
+ *  | +---------------------------------------------------------------+ |
+ *  |        ^              ^                BLK_IDX_SCHEDULE_PARAMS    |
+ *  |        |              |                                           |
+ *  +--------+              +-------------------------------------------+
+ *
+ *  In the above picture there are two subschedules (cycles):
+ *
+ *  - cycle 0: iterates the schedule table from 0 to 2 (and back)
+ *  - cycle 1: iterates the schedule table from 3 to 5 (and back)
+ *
+ *  All other possible execution threads must be marked as unused by making
+ *  their "subschedule end index" (subscheind) equal to the last valid
+ *  subschedule's end index (in this case 5).
+ */
+static int sja1105_init_scheduling(struct sja1105_private *priv)
+{
+	struct sja1105_schedule_entry_points_entry *schedule_entry_points;
+	struct sja1105_schedule_entry_points_params_entry
+					*schedule_entry_points_params;
+	struct sja1105_schedule_params_entry *schedule_params;
+	struct sja1105_tas_data *tas_data = &priv->tas_data;
+	struct sja1105_schedule_entry *schedule;
+	struct sja1105_table *table;
+	int schedule_start_idx;
+	s64 entry_point_delta;
+	int schedule_end_idx;
+	int num_entries = 0;
+	int num_cycles = 0;
+	int cycle = 0;
+	int i, k = 0;
+	int port;
+
+	/* Discard previous Schedule Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE];
+	if (table->entry_count) {
+		kfree(table->entries);
+		table->entry_count = 0;
+	}
+
+	/* Discard previous Schedule Entry Points Parameters Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS];
+	if (table->entry_count) {
+		kfree(table->entries);
+		table->entry_count = 0;
+	}
+
+	/* Discard previous Schedule Parameters Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_PARAMS];
+	if (table->entry_count) {
+		kfree(table->entries);
+		table->entry_count = 0;
+	}
+
+	/* Discard previous Schedule Entry Points Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_ENTRY_POINTS];
+	if (table->entry_count) {
+		kfree(table->entries);
+		table->entry_count = 0;
+	}
+
+	/* Figure out the dimensioning of the problem */
+	for (port = 0; port < SJA1105_NUM_PORTS; port++) {
+		if (tas_data->offload[port]) {
+			num_entries += tas_data->offload[port]->num_entries;
+			num_cycles++;
+		}
+	}
+
+	/* Nothing to do */
+	if (!num_cycles)
+		return 0;
+
+	/* Pre-allocate space in the static config tables */
+
+	/* Schedule Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE];
+	table->entries = kcalloc(num_entries, table->ops->unpacked_entry_size,
+				 GFP_KERNEL);
+	if (!table->entries)
+		return -ENOMEM;
+	table->entry_count = num_entries;
+	schedule = table->entries;
+
+	/* Schedule Points Parameters Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_ENTRY_POINTS_PARAMS];
+	table->entries = kcalloc(SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT,
+				 table->ops->unpacked_entry_size, GFP_KERNEL);
+	if (!table->entries)
+		/* Previously allocated memory will be freed automatically in
+		 * sja1105_static_config_free. This is true for all early
+		 * returns below.
+		 */
+		return -ENOMEM;
+	table->entry_count = SJA1105_MAX_SCHEDULE_ENTRY_POINTS_PARAMS_COUNT;
+	schedule_entry_points_params = table->entries;
+
+	/* Schedule Parameters Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_PARAMS];
+	table->entries = kcalloc(SJA1105_MAX_SCHEDULE_PARAMS_COUNT,
+				 table->ops->unpacked_entry_size, GFP_KERNEL);
+	if (!table->entries)
+		return -ENOMEM;
+	table->entry_count = SJA1105_MAX_SCHEDULE_PARAMS_COUNT;
+	schedule_params = table->entries;
+
+	/* Schedule Entry Points Table */
+	table = &priv->static_config.tables[BLK_IDX_SCHEDULE_ENTRY_POINTS];
+	table->entries = kcalloc(num_cycles, table->ops->unpacked_entry_size,
+				 GFP_KERNEL);
+	if (!table->entries)
+		return -ENOMEM;
+	table->entry_count = num_cycles;
+	schedule_entry_points = table->entries;
+
+	/* Finally start populating the static config tables */
+	schedule_entry_points_params->clksrc = SJA1105_TAS_CLKSRC_STANDALONE;
+	schedule_entry_points_params->actsubsch = num_cycles - 1;
+
+	for (port = 0; port < SJA1105_NUM_PORTS; port++) {
+		const struct tc_taprio_qopt_offload *offload;
+
+		offload = tas_data->offload[port];
+		if (!offload)
+			continue;
+
+		schedule_start_idx = k;
+		schedule_end_idx = k + offload->num_entries - 1;
+		/* TODO this is the base time for the port's subschedule,
+		 * relative to PTPSCHTM. But as we're using the standalone
+		 * clock source and not PTP clock as time reference, there's
+		 * little point in even trying to put more logic into this,
+		 * like preserving the phases between the subschedules of
+		 * different ports. We'll get all of that when switching to the
+		 * PTP clock source.
+		 */
+		entry_point_delta = 1;
+
+		schedule_entry_points[cycle].subschindx = cycle;
+		schedule_entry_points[cycle].delta = entry_point_delta;
+		schedule_entry_points[cycle].address = schedule_start_idx;
+
+		/* The subschedule end indices need to be
+		 * monotonically increasing.
+		 */
+		for (i = cycle; i < 8; i++)
+			schedule_params->subscheind[i] = schedule_end_idx;
+
+		for (i = 0; i < offload->num_entries; i++, k++) {
+			s64 delta_ns = offload->entries[i].interval;
+
+			schedule[k].delta = ns_to_sja1105_delta(delta_ns);
+			schedule[k].destports = BIT(port);
+			schedule[k].resmedia_en = true;
+			schedule[k].resmedia = SJA1105_GATE_MASK &
+					~offload->entries[i].gate_mask;
+		}
+		cycle++;
+	}
+
+	return 0;
+}
+
+/* Be there 2 port subschedules, each executing an arbitrary number of gate
+ * open/close events cyclically.
+ * None of those gate events must ever occur at the exact same time, otherwise
+ * the switch is known to act in exotically strange ways.
+ * However the hardware doesn't bother performing these integrity checks.
+ * So here we are with the task of validating whether the new @admin offload
+ * has any conflict with the already established TAS configuration in
+ * tas_data->offload.  We already know the other ports are in harmony with one
+ * another, otherwise we wouldn't have saved them.
+ * Each gate event executes periodically, with a period of @cycle_time and a
+ * phase given by its cycle's @base_time plus its offset within the cycle
+ * (which in turn is given by the length of the events prior to it).
+ * There are two aspects to possible collisions:
+ * - Collisions within one cycle's (actually the longest cycle's) time frame.
+ *   For that, we need to compare the cartesian product of each possible
+ *   occurrence of each event within one cycle time.
+ * - Collisions in the future. Events may not collide within one cycle time,
+ *   but if two port schedules don't have the same periodicity (aka the cycle
+ *   times aren't multiples of one another), they surely will some time in the
+ *   future (actually they will collide an infinite amount of times).
+ */
+static bool
+sja1105_tas_check_conflicts(struct sja1105_private *priv, int port,
+			    const struct tc_taprio_qopt_offload *admin)
+{
+	struct sja1105_tas_data *tas_data = &priv->tas_data;
+	const struct tc_taprio_qopt_offload *offload;
+	s64 max_cycle_time, min_cycle_time;
+	s64 delta1, delta2;
+	s64 rbt1, rbt2;
+	s64 stop_time;
+	s64 t1, t2;
+	int i, j;
+	s32 rem;
+
+	offload = tas_data->offload[port];
+	if (!offload)
+		return false;
+
+	/* Check if the two cycle times are multiples of one another.
+	 * If they aren't, then they will surely collide.
+	 */
+	max_cycle_time = max(offload->cycle_time, admin->cycle_time);
+	min_cycle_time = min(offload->cycle_time, admin->cycle_time);
+	div_s64_rem(max_cycle_time, min_cycle_time, &rem);
+	if (rem)
+		return true;
+
+	/* Calculate the "reduced" base time of each of the two cycles
+	 * (transposed back as close to 0 as possible) by dividing to
+	 * the cycle time.
+	 */
+	div_s64_rem(offload->base_time, offload->cycle_time, &rem);
+	rbt1 = rem;
+
+	div_s64_rem(admin->base_time, admin->cycle_time, &rem);
+	rbt2 = rem;
+
+	stop_time = max_cycle_time + max(rbt1, rbt2);
+
+	/* delta1 is the relative base time of each GCL entry within
+	 * the established ports' TAS config.
+	 */
+	for (i = 0, delta1 = 0;
+	     i < offload->num_entries;
+	     delta1 += offload->entries[i].interval, i++) {
+		/* delta2 is the relative base time of each GCL entry
+		 * within the newly added TAS config.
+		 */
+		for (j = 0, delta2 = 0;
+		     j < admin->num_entries;
+		     delta2 += admin->entries[j].interval, j++) {
+			/* t1 follows all possible occurrences of the
+			 * established ports' GCL entry i within the
+			 * first cycle time.
+			 */
+			for (t1 = rbt1 + delta1;
+			     t1 <= stop_time;
+			     t1 += offload->cycle_time) {
+				/* t2 follows all possible occurrences
+				 * of the newly added GCL entry j
+				 * within the first cycle time.
+				 */
+				for (t2 = rbt2 + delta2;
+				     t2 <= stop_time;
+				     t2 += admin->cycle_time) {
+					if (t1 == t2) {
+						dev_warn(priv->ds->dev,
+							 "GCL entry %d collides with entry %d of port %d\n",
+							 j, i, port);
+						return true;
+					}
+				}
+			}
+		}
+	}
+
+	return false;
+}
+
+int sja1105_setup_tc_taprio(struct dsa_switch *ds, int port,
+			    struct tc_taprio_qopt_offload *admin)
+{
+	struct sja1105_private *priv = ds->priv;
+	struct sja1105_tas_data *tas_data = &priv->tas_data;
+	int other_port, rc, i;
+
+	/* Can't change an already configured port (must delete qdisc first).
+	 * Can't delete the qdisc from an unconfigured port.
+	 */
+	if (!!tas_data->offload[port] == admin->enable)
+		return -EINVAL;
+
+	if (!admin->enable) {
+		taprio_offload_free(tas_data->offload[port]);
+		tas_data->offload[port] = NULL;
+
+		rc = sja1105_init_scheduling(priv);
+		if (rc < 0)
+			return rc;
+
+		return sja1105_static_config_reload(priv);
+	}
+
+	/* The cycle time extension is the amount of time the last cycle from
+	 * the old OPER needs to be extended in order to phase-align with the
+	 * base time of the ADMIN when that becomes the new OPER.
+	 * But of course our switch needs to be reset to switch-over between
+	 * the ADMIN and the OPER configs - so much for a seamless transition.
+	 * So don't add insult over injury and just say we don't support cycle
+	 * time extension.
+	 */
+	if (admin->cycle_time_extension)
+		return -ENOTSUPP;
+
+	if (!ns_to_sja1105_delta(admin->base_time)) {
+		dev_err(ds->dev, "A base time of zero is not hardware-allowed\n");
+		return -ERANGE;
+	}
+
+	for (i = 0; i < admin->num_entries; i++) {
+		s64 delta_ns = admin->entries[i].interval;
+		s64 delta_cycles = ns_to_sja1105_delta(delta_ns);
+		bool too_long, too_short;
+
+		too_long = (delta_cycles >= SJA1105_TAS_MAX_DELTA);
+		too_short = (delta_cycles == 0);
+		if (too_long || too_short) {
+			dev_err(priv->ds->dev,
+				"Interval %llu too %s for GCL entry %d\n",
+				delta_ns, too_long ? "long" : "short", i);
+			return -ERANGE;
+		}
+	}
+
+	for (other_port = 0; other_port < SJA1105_NUM_PORTS; other_port++) {
+		if (other_port == port)
+			continue;
+
+		if (sja1105_tas_check_conflicts(priv, other_port, admin))
+			return -ERANGE;
+	}
+
+	tas_data->offload[port] = taprio_offload_get(admin);
+
+	rc = sja1105_init_scheduling(priv);
+	if (rc < 0)
+		return rc;
+
+	return sja1105_static_config_reload(priv);
+}
+
+void sja1105_tas_setup(struct dsa_switch *ds)
+{
+}
+
+void sja1105_tas_teardown(struct dsa_switch *ds)
+{
+	struct sja1105_private *priv = ds->priv;
+	struct tc_taprio_qopt_offload *offload;
+	int port;
+
+	for (port = 0; port < SJA1105_NUM_PORTS; port++) {
+		offload = priv->tas_data.offload[port];
+		if (!offload)
+			continue;
+
+		taprio_offload_free(offload);
+	}
+}
diff --git a/drivers/net/dsa/sja1105/sja1105_tas.h b/drivers/net/dsa/sja1105/sja1105_tas.h
new file mode 100644
index 000000000000..0b803c30e640
--- /dev/null
+++ b/drivers/net/dsa/sja1105/sja1105_tas.h
@@ -0,0 +1,41 @@
+/* SPDX-License-Identifier: GPL-2.0
+ * Copyright (c) 2019, Vladimir Oltean <olteanv@gmail.com>
+ */
+#ifndef _SJA1105_TAS_H
+#define _SJA1105_TAS_H
+
+#include <net/pkt_sched.h>
+
+#if IS_ENABLED(CONFIG_NET_DSA_SJA1105_TAS)
+
+struct sja1105_tas_data {
+	struct tc_taprio_qopt_offload *offload[SJA1105_NUM_PORTS];
+};
+
+int sja1105_setup_tc_taprio(struct dsa_switch *ds, int port,
+			    struct tc_taprio_qopt_offload *admin);
+
+void sja1105_tas_setup(struct dsa_switch *ds);
+
+void sja1105_tas_teardown(struct dsa_switch *ds);
+
+#else
+
+/* C doesn't allow empty structures, bah! */
+struct sja1105_tas_data {
+	u8 dummy;
+};
+
+static inline int sja1105_setup_tc_taprio(struct dsa_switch *ds, int port,
+					  struct tc_taprio_qopt_offload *admin)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline void sja1105_tas_setup(struct dsa_switch *ds) { }
+
+static inline void sja1105_tas_teardown(struct dsa_switch *ds) { }
+
+#endif /* IS_ENABLED(CONFIG_NET_DSA_SJA1105_TAS) */
+
+#endif /* _SJA1105_TAS_H */
-- 
2.17.1

[PATCH v4 net-next 6/6] docs: net: dsa: sja1105: Add info about the Time-Aware Scheduler

[Vladimir Oltean]

While not an exhaustive usage tutorial, this describes the details
needed to build more complex scenarios.

Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
---
Changes since v2:
- Specified that the HOSTPRIO switch setting is not configurable at the
  moment, since I dropped the patch for configuring it.

Changes since v1:
- Patch is new.

 Documentation/networking/dsa/sja1105.rst | 90 ++++++++++++++++++++++++
 1 file changed, 90 insertions(+)

diff --git a/Documentation/networking/dsa/sja1105.rst b/Documentation/networking/dsa/sja1105.rst
index cb2858dece93..eef20d0bcf7c 100644
--- a/Documentation/networking/dsa/sja1105.rst
+++ b/Documentation/networking/dsa/sja1105.rst
@@ -146,6 +146,96 @@ enslaves eth0 and eth1 (the DSA master of the switch ports). This is because in
 this mode, the switch ports beneath br0 are not capable of regular traffic, and
 are only used as a conduit for switchdev operations.
 
+Offloads
+========
+
+Time-aware scheduling
+---------------------
+
+The switch supports a variation of the enhancements for scheduled traffic
+specified in IEEE 802.1Q-2018 (formerly 802.1Qbv). This means it can be used to
+ensure deterministic latency for priority traffic that is sent in-band with its
+gate-open event in the network schedule.
+
+This capability can be managed through the tc-taprio offload ('flags 2'). The
+difference compared to the software implementation of taprio is that the latter
+would only be able to shape traffic originated from the CPU, but not
+autonomously forwarded flows.
+
+The device has 8 traffic classes, and maps incoming frames to one of them based
+on the VLAN PCP bits (if no VLAN is present, the port-based default is used).
+As described in the previous sections, depending on the value of
+``vlan_filtering``, the EtherType recognized by the switch as being VLAN can
+either be the typical 0x8100 or a custom value used internally by the driver
+for tagging. Therefore, the switch ignores the VLAN PCP if used in standalone
+or bridge mode with ``vlan_filtering=0``, as it will not recognize the 0x8100
+EtherType. In these modes, injecting into a particular TX queue can only be
+done by the DSA net devices, which populate the PCP field of the tagging header
+on egress. Using ``vlan_filtering=1``, the behavior is the other way around:
+offloaded flows can be steered to TX queues based on the VLAN PCP, but the DSA
+net devices are no longer able to do that. To inject frames into a hardware TX
+queue with VLAN awareness active, it is necessary to create a VLAN
+sub-interface on the DSA master port, and send normal (0x8100) VLAN-tagged
+towards the switch, with the VLAN PCP bits set appropriately.
+
+Management traffic (having DMAC 01-80-C2-xx-xx-xx or 01-19-1B-xx-xx-xx) is the
+notable exception: the switch always treats it with a fixed priority and
+disregards any VLAN PCP bits even if present. The traffic class for management
+traffic has a value of 7 (highest priority) at the moment, which is not
+configurable in the driver.
+
+Below is an example of configuring a 500 us cyclic schedule on egress port
+``swp5``. The traffic class gate for management traffic (7) is open for 100 us,
+and the gates for all other traffic classes are open for 400 us::
+
+  #!/bin/bash
+
+  set -e -u -o pipefail
+
+  NSEC_PER_SEC="1000000000"
+
+  gatemask() {
+          local tc_list="$1"
+          local mask=0
+
+          for tc in ${tc_list}; do
+                  mask=$((${mask} | (1 << ${tc})))
+          done
+
+          printf "%02x" ${mask}
+  }
+
+  if ! systemctl is-active --quiet ptp4l; then
+          echo "Please start the ptp4l service"
+          exit
+  fi
+
+  now=$(phc_ctl /dev/ptp1 get | gawk '/clock time is/ { print $5; }')
+  # Phase-align the base time to the start of the next second.
+  sec=$(echo "${now}" | gawk -F. '{ print $1; }')
+  base_time="$(((${sec} + 1) * ${NSEC_PER_SEC}))"
+
+  tc qdisc add dev swp5 parent root handle 100 taprio \
+          num_tc 8 \
+          map 0 1 2 3 5 6 7 \
+          queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
+          base-time ${base_time} \
+          sched-entry S $(gatemask 7) 100000 \
+          sched-entry S $(gatemask "0 1 2 3 4 5 6") 400000 \
+          flags 2
+
+It is possible to apply the tc-taprio offload on multiple egress ports. There
+are hardware restrictions related to the fact that no gate event may trigger
+simultaneously on two ports. The driver checks the consistency of the schedules
+against this restriction and errors out when appropriate. Schedule analysis is
+needed to avoid this, which is outside the scope of the document.
+
+At the moment, the time-aware scheduler can only be triggered based on a
+standalone clock and not based on PTP time. This means the base-time argument
+from tc-taprio is ignored and the schedule starts right away. It also means it
+is more difficult to phase-align the scheduler with the other devices in the
+network.
+
 Device Tree bindings and board design
 =====================================
 
-- 
2.17.1

Re: [PATCH v4 net-next 4/6] net: dsa: sja1105: Advertise the 8 TX queues

[Florian Fainelli]

On 9/14/2019 7:00 PM, Vladimir Oltean wrote:
> This is a preparation patch for the tc-taprio offload (and potentially
> for other future offloads such as tc-mqprio).
> 
> Instead of looking directly at skb->priority during xmit, let's get the
> netdev queue and the queue-to-traffic-class mapping, and put the
> resulting traffic class into the dsa_8021q PCP field. The switch is
> configured with a 1-to-1 PCP-to-ingress-queue-to-egress-queue mapping
> (see vlan_pmap in sja1105_main.c), so the effect is that we can inject
> into a front-panel's egress traffic class through VLAN tagging from
> Linux, completely transparently.
> 
> Unfortunately the switch doesn't look at the VLAN PCP in the case of
> management traffic to/from the CPU (link-local frames at
> 01-80-C2-xx-xx-xx or 01-1B-19-xx-xx-xx) so we can't alter the
> transmission queue of this type of traffic on a frame-by-frame basis. It
> is only selected through the "hostprio" setting which ATM is harcoded in
> the driver to 7.
> 
> Signed-off-by: Vladimir Oltean <olteanv@gmail.com>

Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
-- 
Florian

RE: [PATCH v4 net-next 4/6] net: dsa: sja1105: Advertise the 8 TX queues

[Jose Abreu]

From: Vladimir Oltean <olteanv@gmail.com>
Date: Sep/15/2019, 03:00:01 (UTC+00:00)

> Instead of looking directly at skb->priority during xmit, let's get the
> netdev queue and the queue-to-traffic-class mapping, and put the
> resulting traffic class into the dsa_8021q PCP field. The switch is
> configured with a 1-to-1 PCP-to-ingress-queue-to-egress-queue mapping
> (see vlan_pmap in sja1105_main.c), so the effect is that we can inject
> into a front-panel's egress traffic class through VLAN tagging from
> Linux, completely transparently.

Wouldn't it be better to just rely on skb queue mapping as per userspace 
settings ? I mean this way you cant create some complex scenarios 
without having the VLAN ID need.

I generally use u32 filter and skbedit queue_mapping action to achieve 
this.

---
Thanks,
Jose Miguel Abreu

Re: [PATCH v4 net-next 4/6] net: dsa: sja1105: Advertise the 8 TX queues

[Vladimir Oltean]

Hi Jose,

On Sun, 15 Sep 2019 at 19:41, Jose Abreu <Jose.Abreu@synopsys.com> wrote:
>
> From: Vladimir Oltean <olteanv@gmail.com>
> Date: Sep/15/2019, 03:00:01 (UTC+00:00)
>
> > Instead of looking directly at skb->priority during xmit, let's get the
> > netdev queue and the queue-to-traffic-class mapping, and put the
> > resulting traffic class into the dsa_8021q PCP field. The switch is
> > configured with a 1-to-1 PCP-to-ingress-queue-to-egress-queue mapping
> > (see vlan_pmap in sja1105_main.c), so the effect is that we can inject
> > into a front-panel's egress traffic class through VLAN tagging from
> > Linux, completely transparently.
>
> Wouldn't it be better to just rely on skb queue mapping as per userspace
> settings ? I mean this way you cant create some complex scenarios
> without having the VLAN ID need.
>

I think I need to clarify the "without having the VLAN ID need" portion.
This is a DSA switch, and the thing that all these switches have in
common is that their data path is connected to the host system through
an Ethernet port pair (as opposed to e.g. DMA and real queues).
When you send a frame from the host through that Ethernet port, the
switch's natural tendency is to forward/flood it according to FDB
rules, which may not be what you want to achieve (i.e. you may want to
xmit on a specific front-panel port only, to achieve a 'port
multiplier' behavior out of it). Switch vendors achieve that by using
a proprietary frame header which indicates stuff such as: egress port,
QoS priority, etc. The host adds this tag to frames sent towards the
switch, and the switch consumes it.
In the case of sja1105, the frame tagging format is actually
'open-spec', e.g. it is an actual VLAN header, just with a custom
EtherType. Hence the QoS priority in this tagging format is the PCP
field. Other DSA switch vendors still have a 'priority' field in their
tagging format, even though it is not PCP.
So to answer your question, I do need VLAN tagging either way, for
even more basic reasons. But it is VLAN only from the switch's
hardware view. From the operating system's view it is no different
than populating the 'priority' field of any other DSA frame tagger.

> I generally use u32 filter and skbedit queue_mapping action to achieve
> this.
>

The trouble with relying on 'queue_mapping' only is that it's too
open-ended: what does a queue really mean for an Ethernet-managed
switch?
In my interpretation of the mqprio/taprio offloads, the netdev queue
is only an intermediary representation between the qdisc and the NIC's
traffic classes. Now that makes sense, because even though the switch
xmit procedure isn't multi-queue, there is still strict egress
prioritization based on the traffic class mapping (inferred from VLAN
PCP).
I have no idea how people are managing multi-queue net devices without
an mqprio-type offload, if those netdev queues are not of equal
priority. Technically, if you use the 'skbedit queue_mapping' action
with this switch, it is your choice how you manage the
net_device->tc_to_txq array, keeping in mind that tc 7 will have
higher priority over tc 6. Alternatively, you can still use an
mqprio-type offload, specify "num_tc 8 map 0 1 2 3 5 6 7 queues 1@0
1@1 1@2 1@3 1@4 1@5 1@6 1@7", and then just use the 'skbedit priority'
action. Then things would just work.
So I am just using the 'queues' term for this switch to express the
strict priority scheduler. There isn't really any benefit to having
more than one netdev queue exposed per traffic class.

> ---
> Thanks,
> Jose Miguel Abreu

Hope this clarifies,
-Vladimir

Re: [PATCH v4 net-next 2/6] net: dsa: Pass ndo_setup_tc slave callback to drivers

[Ilias Apalodimas]

On Sun, Sep 15, 2019 at 04:59:59AM +0300, Vladimir Oltean wrote:
> DSA currently handles shared block filters (for the classifier-action
> qdisc) in the core due to what I believe are simply pragmatic reasons -
> hiding the complexity from drivers and offerring a simple API for port
> mirroring.
> 
> Extend the dsa_slave_setup_tc function by passing all other qdisc
> offloads to the driver layer, where the driver may choose what it
> implements and how. DSA is simply a pass-through in this case.
> 
> Signed-off-by: Vladimir Oltean <olteanv@gmail.com>
> Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
> Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
> Changes since v2:
> - Added Florian Fainelli's Reviewed-by.
> 
> Changes since v1:
> - Added Kurt Kanzenbach's Acked-by.
> 
> Changes since RFC:
> - Removed the unused declaration of struct tc_taprio_qopt_offload.
> 
>  include/net/dsa.h |  2 ++
>  net/dsa/slave.c   | 12 ++++++++----
>  2 files changed, 10 insertions(+), 4 deletions(-)
> 
> diff --git a/include/net/dsa.h b/include/net/dsa.h
> index 96acb14ec1a8..541fb514e31d 100644
> --- a/include/net/dsa.h
> +++ b/include/net/dsa.h
> @@ -515,6 +515,8 @@ struct dsa_switch_ops {
>  				   bool ingress);
>  	void	(*port_mirror_del)(struct dsa_switch *ds, int port,
>  				   struct dsa_mall_mirror_tc_entry *mirror);
> +	int	(*port_setup_tc)(struct dsa_switch *ds, int port,
> +				 enum tc_setup_type type, void *type_data);
>  
>  	/*
>  	 * Cross-chip operations
> diff --git a/net/dsa/slave.c b/net/dsa/slave.c
> index 9a88035517a6..75d58229a4bd 100644
> --- a/net/dsa/slave.c
> +++ b/net/dsa/slave.c
> @@ -1035,12 +1035,16 @@ static int dsa_slave_setup_tc_block(struct net_device *dev,
>  static int dsa_slave_setup_tc(struct net_device *dev, enum tc_setup_type type,
>  			      void *type_data)
>  {
> -	switch (type) {
> -	case TC_SETUP_BLOCK:
> +	struct dsa_port *dp = dsa_slave_to_port(dev);
> +	struct dsa_switch *ds = dp->ds;
> +
> +	if (type == TC_SETUP_BLOCK)
>  		return dsa_slave_setup_tc_block(dev, type_data);
> -	default:
> +
> +	if (!ds->ops->port_setup_tc)
>  		return -EOPNOTSUPP;
> -	}
> +
> +	return ds->ops->port_setup_tc(ds, dp->index, type, type_data);
>  }
>  
>  static void dsa_slave_get_stats64(struct net_device *dev,
> -- 
> 2.17.1
> 

Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>

Re: [PATCH v4 net-next 0/6] tc-taprio offload for SJA1105 DSA

[David Miller]

From: Vladimir Oltean <olteanv@gmail.com>
Date: Sun, 15 Sep 2019 04:59:57 +0300

> This is the third attempt to submit the tc-taprio offload model for
> inclusion in the networking tree. The sja1105 switch driver will provide
> the first implementation of the offload. Only the bare minimum is added:
> 
> - The offload model and a DSA pass-through
> - The hardware implementation
> - The interaction with the netdev queues in the tagger code
> - Documentation
> 
> What has been removed from previous attempts is support for
> PTP-as-clocksource in sja1105, as well as configuring the traffic class
> for management traffic.  These will be added as soon as the offload
> model is settled.

Series applied, thanks.

Re: [PATCH v4 net-next 0/6] tc-taprio offload for SJA1105 DSA

[Vladimir Oltean]

On Mon, 16 Sep 2019 at 22:36, David Miller <davem@davemloft.net> wrote:
>
> From: Vladimir Oltean <olteanv@gmail.com>
> Date: Sun, 15 Sep 2019 04:59:57 +0300
>
> > This is the third attempt to submit the tc-taprio offload model for
> > inclusion in the networking tree. The sja1105 switch driver will provide
> > the first implementation of the offload. Only the bare minimum is added:
> >
> > - The offload model and a DSA pass-through
> > - The hardware implementation
> > - The interaction with the netdev queues in the tagger code
> > - Documentation
> >
> > What has been removed from previous attempts is support for
> > PTP-as-clocksource in sja1105, as well as configuring the traffic class
> > for management traffic.  These will be added as soon as the offload
> > model is settled.
>
> Series applied, thanks.

Thanks a lot, that's what I call cutting it close!

-Vladimir