[PATCH net-next v2 00/14] sfc: VF representors for EF100 - RX side

[PATCH net-next v2 00/14] sfc: VF representors for EF100 - RX side

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

This series adds the receive path for EF100 VF representors, plus other
 minor features such as statistics and MAC address setting.

Changes in v2: fixed build failure on CONFIG_SFC_SRIOV=n (kernel test robot).

Edward Cree (14):
  sfc: plumb ef100 representor stats
  sfc: ef100 representor RX NAPI poll
  sfc: ef100 representor RX top half
  sfc: determine wire m-port at EF100 PF probe time
  sfc: check ef100 RX packets are from the wire
  sfc: receive packets from EF100 VFs into representors
  sfc: insert default MAE rules to connect VFs to representors
  sfc: move table locking into filter_table_{probe,remove} methods
  sfc: use a dynamic m-port for representor RX and set it promisc
  sfc: look up VF's client ID when creating representor
  sfc: fetch existing assigned MAC address from FW when creating VF rep
  sfc: set EF100 VF MAC address through representor
  sfc: get provisioned MAC address on EF100 VF probe
  sfc: implement ethtool get/set RX ring size for EF100 reps

 drivers/net/ethernet/sfc/Makefile         |   3 +-
 drivers/net/ethernet/sfc/ef10.c           |  26 +-
 drivers/net/ethernet/sfc/ef100.c          |   3 +
 drivers/net/ethernet/sfc/ef100_netdev.c   |  14 +
 drivers/net/ethernet/sfc/ef100_nic.c      | 151 +++++++++--
 drivers/net/ethernet/sfc/ef100_nic.h      |   5 +
 drivers/net/ethernet/sfc/ef100_rep.c      | 250 +++++++++++++++++-
 drivers/net/ethernet/sfc/ef100_rep.h      |  22 ++
 drivers/net/ethernet/sfc/ef100_rx.c       |  46 +++-
 drivers/net/ethernet/sfc/ef10_sriov.c     |  16 +-
 drivers/net/ethernet/sfc/ethtool_common.c |   1 +
 drivers/net/ethernet/sfc/filter.h         |  18 ++
 drivers/net/ethernet/sfc/mae.c            | 304 +++++++++++++++++++++-
 drivers/net/ethernet/sfc/mae.h            |  20 ++
 drivers/net/ethernet/sfc/mcdi.h           |   4 +
 drivers/net/ethernet/sfc/mcdi_filters.c   |   6 +-
 drivers/net/ethernet/sfc/mcdi_filters.h   |   1 +
 drivers/net/ethernet/sfc/mcdi_pcol_mae.h  |  24 ++
 drivers/net/ethernet/sfc/net_driver.h     |   5 +
 drivers/net/ethernet/sfc/rx_common.c      |   4 -
 drivers/net/ethernet/sfc/tc.c             | 252 ++++++++++++++++++
 drivers/net/ethernet/sfc/tc.h             |  85 ++++++
 22 files changed, 1194 insertions(+), 66 deletions(-)
 create mode 100644 drivers/net/ethernet/sfc/mcdi_pcol_mae.h
 create mode 100644 drivers/net/ethernet/sfc/tc.c
 create mode 100644 drivers/net/ethernet/sfc/tc.h

[PATCH net-next v2 01/14] sfc: plumb ef100 representor stats

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Implement .ndo_get_stats64() method to read values out of struct
 efx_rep_sw_stats.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index d07539f091b8..102071ed051b 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -79,10 +79,24 @@ static int efx_ef100_rep_get_phys_port_name(struct net_device *dev,
 	return 0;
 }
 
+static void efx_ef100_rep_get_stats64(struct net_device *dev,
+				      struct rtnl_link_stats64 *stats)
+{
+	struct efx_rep *efv = netdev_priv(dev);
+
+	stats->rx_packets = atomic64_read(&efv->stats.rx_packets);
+	stats->tx_packets = atomic64_read(&efv->stats.tx_packets);
+	stats->rx_bytes = atomic64_read(&efv->stats.rx_bytes);
+	stats->tx_bytes = atomic64_read(&efv->stats.tx_bytes);
+	stats->rx_dropped = atomic64_read(&efv->stats.rx_dropped);
+	stats->tx_errors = atomic64_read(&efv->stats.tx_errors);
+}
+
 static const struct net_device_ops efx_ef100_rep_netdev_ops = {
 	.ndo_start_xmit		= efx_ef100_rep_xmit,
 	.ndo_get_port_parent_id	= efx_ef100_rep_get_port_parent_id,
 	.ndo_get_phys_port_name	= efx_ef100_rep_get_phys_port_name,
+	.ndo_get_stats64	= efx_ef100_rep_get_stats64,
 };
 
 static void efx_ef100_rep_get_drvinfo(struct net_device *dev,

[PATCH net-next v2 02/14] sfc: ef100 representor RX NAPI poll

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

This patch adds the 'bottom half' napi->poll routine for representor RX.
See the next patch (with the top half) for an explanation of the 'fake
 interrupt' scheme used to drive this NAPI context.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 64 ++++++++++++++++++++++++++++
 drivers/net/ethernet/sfc/ef100_rep.h | 11 +++++
 2 files changed, 75 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index 102071ed051b..fe45ae963391 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -16,12 +16,16 @@
 
 #define EFX_EF100_REP_DRIVER	"efx_ef100_rep"
 
+static int efx_ef100_rep_poll(struct napi_struct *napi, int weight);
+
 static int efx_ef100_rep_init_struct(struct efx_nic *efx, struct efx_rep *efv,
 				     unsigned int i)
 {
 	efv->parent = efx;
 	efv->idx = i;
 	INIT_LIST_HEAD(&efv->list);
+	INIT_LIST_HEAD(&efv->rx_list);
+	spin_lock_init(&efv->rx_lock);
 	efv->msg_enable = NETIF_MSG_DRV | NETIF_MSG_PROBE |
 			  NETIF_MSG_LINK | NETIF_MSG_IFDOWN |
 			  NETIF_MSG_IFUP | NETIF_MSG_RX_ERR |
@@ -29,6 +33,25 @@ static int efx_ef100_rep_init_struct(struct efx_nic *efx, struct efx_rep *efv,
 	return 0;
 }
 
+static int efx_ef100_rep_open(struct net_device *net_dev)
+{
+	struct efx_rep *efv = netdev_priv(net_dev);
+
+	netif_napi_add(net_dev, &efv->napi, efx_ef100_rep_poll,
+		       NAPI_POLL_WEIGHT);
+	napi_enable(&efv->napi);
+	return 0;
+}
+
+static int efx_ef100_rep_close(struct net_device *net_dev)
+{
+	struct efx_rep *efv = netdev_priv(net_dev);
+
+	napi_disable(&efv->napi);
+	netif_napi_del(&efv->napi);
+	return 0;
+}
+
 static netdev_tx_t efx_ef100_rep_xmit(struct sk_buff *skb,
 				      struct net_device *dev)
 {
@@ -93,6 +116,8 @@ static void efx_ef100_rep_get_stats64(struct net_device *dev,
 }
 
 static const struct net_device_ops efx_ef100_rep_netdev_ops = {
+	.ndo_open		= efx_ef100_rep_open,
+	.ndo_stop		= efx_ef100_rep_close,
 	.ndo_start_xmit		= efx_ef100_rep_xmit,
 	.ndo_get_port_parent_id	= efx_ef100_rep_get_port_parent_id,
 	.ndo_get_phys_port_name	= efx_ef100_rep_get_phys_port_name,
@@ -256,3 +281,42 @@ void efx_ef100_fini_vfreps(struct efx_nic *efx)
 	list_for_each_entry_safe(efv, next, &efx->vf_reps, list)
 		efx_ef100_vfrep_destroy(efx, efv);
 }
+
+static int efx_ef100_rep_poll(struct napi_struct *napi, int weight)
+{
+	struct efx_rep *efv = container_of(napi, struct efx_rep, napi);
+	unsigned int read_index;
+	struct list_head head;
+	struct sk_buff *skb;
+	bool need_resched;
+	int spent = 0;
+
+	INIT_LIST_HEAD(&head);
+	/* Grab up to 'weight' pending SKBs */
+	spin_lock_bh(&efv->rx_lock);
+	read_index = efv->write_index;
+	while (spent < weight && !list_empty(&efv->rx_list)) {
+		skb = list_first_entry(&efv->rx_list, struct sk_buff, list);
+		list_del(&skb->list);
+		list_add_tail(&skb->list, &head);
+		spent++;
+	}
+	spin_unlock_bh(&efv->rx_lock);
+	/* Receive them */
+	netif_receive_skb_list(&head);
+	if (spent < weight)
+		if (napi_complete_done(napi, spent)) {
+			spin_lock_bh(&efv->rx_lock);
+			efv->read_index = read_index;
+			/* If write_index advanced while we were doing the
+			 * RX, then storing our read_index won't re-prime the
+			 * fake-interrupt.  In that case, we need to schedule
+			 * NAPI again to consume the additional packet(s).
+			 */
+			need_resched = efv->write_index != read_index;
+			spin_unlock_bh(&efv->rx_lock);
+			if (need_resched)
+				napi_schedule(&efv->napi);
+		}
+	return spent;
+}
diff --git a/drivers/net/ethernet/sfc/ef100_rep.h b/drivers/net/ethernet/sfc/ef100_rep.h
index d47fd8ff6220..77037ab22052 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.h
+++ b/drivers/net/ethernet/sfc/ef100_rep.h
@@ -29,7 +29,13 @@ struct efx_rep_sw_stats {
  * @msg_enable: log message enable flags
  * @mport: m-port ID of corresponding VF
  * @idx: VF index
+ * @write_index: number of packets enqueued to @rx_list
+ * @read_index: number of packets consumed from @rx_list
+ * @rx_pring_size: max length of RX list
  * @list: entry on efx->vf_reps
+ * @rx_list: list of SKBs queued for receive in NAPI poll
+ * @rx_lock: protects @rx_list
+ * @napi: NAPI control structure
  * @stats: software traffic counters for netdev stats
  */
 struct efx_rep {
@@ -38,7 +44,12 @@ struct efx_rep {
 	u32 msg_enable;
 	u32 mport;
 	unsigned int idx;
+	unsigned int write_index, read_index;
+	unsigned int rx_pring_size;
 	struct list_head list;
+	struct list_head rx_list;
+	spinlock_t rx_lock;
+	struct napi_struct napi;
 	struct efx_rep_sw_stats stats;
 };
 

[PATCH net-next v2 03/14] sfc: ef100 representor RX top half

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Representor RX uses a NAPI context driven by a 'fake interrupt': when
 the parent PF receives a packet destined for the representor, it adds
 it to an SKB list (efv->rx_list), and schedules NAPI if the 'fake
 interrupt' is primed.  The NAPI poll then pulls packets off this list
 and feeds them to the stack with netif_receive_skb_list().
This scheme allows us to decouple representor RX from the parent PF's
 RX fast-path.
This patch implements the 'top half', which builds an SKB, copies data
 into it from the RX buffer (which can then be released), adds it to
 the queue and fires the 'fake interrupt' if necessary.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 55 ++++++++++++++++++++++++++++
 drivers/net/ethernet/sfc/ef100_rep.h |  1 +
 2 files changed, 56 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index fe45ae963391..e6c6e9e764b2 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -13,9 +13,12 @@
 #include "ef100_netdev.h"
 #include "ef100_nic.h"
 #include "mae.h"
+#include "rx_common.h"
 
 #define EFX_EF100_REP_DRIVER	"efx_ef100_rep"
 
+#define EFX_REP_DEFAULT_PSEUDO_RING_SIZE	64
+
 static int efx_ef100_rep_poll(struct napi_struct *napi, int weight);
 
 static int efx_ef100_rep_init_struct(struct efx_nic *efx, struct efx_rep *efv,
@@ -198,6 +201,7 @@ static int efx_ef100_configure_rep(struct efx_rep *efv)
 	u32 selector;
 	int rc;
 
+	efv->rx_pring_size = EFX_REP_DEFAULT_PSEUDO_RING_SIZE;
 	/* Construct mport selector for corresponding VF */
 	efx_mae_mport_vf(efx, efv->idx, &selector);
 	/* Look up actual mport ID */
@@ -320,3 +324,54 @@ static int efx_ef100_rep_poll(struct napi_struct *napi, int weight)
 		}
 	return spent;
 }
+
+void efx_ef100_rep_rx_packet(struct efx_rep *efv, struct efx_rx_buffer *rx_buf)
+{
+	u8 *eh = efx_rx_buf_va(rx_buf);
+	struct sk_buff *skb;
+	bool primed;
+
+	/* Don't allow too many queued SKBs to build up, as they consume
+	 * GFP_ATOMIC memory.  If we overrun, just start dropping.
+	 */
+	if (efv->write_index - READ_ONCE(efv->read_index) > efv->rx_pring_size) {
+		atomic64_inc(&efv->stats.rx_dropped);
+		if (net_ratelimit())
+			netif_dbg(efv->parent, rx_err, efv->net_dev,
+				  "nodesc-dropped packet of length %u\n",
+				  rx_buf->len);
+		return;
+	}
+
+	skb = netdev_alloc_skb(efv->net_dev, rx_buf->len);
+	if (!skb) {
+		atomic64_inc(&efv->stats.rx_dropped);
+		if (net_ratelimit())
+			netif_dbg(efv->parent, rx_err, efv->net_dev,
+				  "noskb-dropped packet of length %u\n",
+				  rx_buf->len);
+		return;
+	}
+	memcpy(skb->data, eh, rx_buf->len);
+	__skb_put(skb, rx_buf->len);
+
+	skb_record_rx_queue(skb, 0); /* rep is single-queue */
+
+	/* Move past the ethernet header */
+	skb->protocol = eth_type_trans(skb, efv->net_dev);
+
+	skb_checksum_none_assert(skb);
+
+	atomic64_inc(&efv->stats.rx_packets);
+	atomic64_add(rx_buf->len, &efv->stats.rx_bytes);
+
+	/* Add it to the rx list */
+	spin_lock_bh(&efv->rx_lock);
+	primed = efv->read_index == efv->write_index;
+	list_add_tail(&skb->list, &efv->rx_list);
+	efv->write_index++;
+	spin_unlock_bh(&efv->rx_lock);
+	/* Trigger rx work */
+	if (primed)
+		napi_schedule(&efv->napi);
+}
diff --git a/drivers/net/ethernet/sfc/ef100_rep.h b/drivers/net/ethernet/sfc/ef100_rep.h
index 77037ab22052..7d2f15cee8d1 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.h
+++ b/drivers/net/ethernet/sfc/ef100_rep.h
@@ -57,4 +57,5 @@ int efx_ef100_vfrep_create(struct efx_nic *efx, unsigned int i);
 void efx_ef100_vfrep_destroy(struct efx_nic *efx, struct efx_rep *efv);
 void efx_ef100_fini_vfreps(struct efx_nic *efx);
 
+void efx_ef100_rep_rx_packet(struct efx_rep *efv, struct efx_rx_buffer *rx_buf);
 #endif /* EF100_REP_H */

[PATCH net-next v2 04/14] sfc: determine wire m-port at EF100 PF probe time

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Traffic delivered to the (MAE admin) PF could be from either the wire
 or a VF.  The INGRESS_MPORT field of the RX prefix distinguishes these;
 base_mport is the value this field will have for traffic from the wire
 (which should be delivered to the PF's netdevice, not a representor).

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_nic.c | 37 ++++++++++++++++++++++++++++
 drivers/net/ethernet/sfc/ef100_nic.h |  2 ++
 drivers/net/ethernet/sfc/mae.c       | 10 ++++++++
 drivers/net/ethernet/sfc/mae.h       |  1 +
 4 files changed, 50 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 4625d35269e6..393d6ca4525c 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -24,6 +24,7 @@
 #include "ef100_tx.h"
 #include "ef100_sriov.h"
 #include "ef100_netdev.h"
+#include "mae.h"
 #include "rx_common.h"
 
 #define EF100_MAX_VIS 4096
@@ -704,6 +705,31 @@ static unsigned int efx_ef100_recycle_ring_size(const struct efx_nic *efx)
 	return 10 * EFX_RECYCLE_RING_SIZE_10G;
 }
 
+#ifdef CONFIG_SFC_SRIOV
+static int efx_ef100_get_base_mport(struct efx_nic *efx)
+{
+	struct ef100_nic_data *nic_data = efx->nic_data;
+	u32 selector, id;
+	int rc;
+
+	/* Construct mport selector for "physical network port" */
+	efx_mae_mport_wire(efx, &selector);
+	/* Look up actual mport ID */
+	rc = efx_mae_lookup_mport(efx, selector, &id);
+	if (rc)
+		return rc;
+	/* The ID should always fit in 16 bits, because that's how wide the
+	 * corresponding fields in the RX prefix & TX override descriptor are
+	 */
+	if (id >> 16)
+		netif_warn(efx, probe, efx->net_dev, "Bad base m-port id %#x\n",
+			   id);
+	nic_data->base_mport = id;
+	nic_data->have_mport = true;
+	return 0;
+}
+#endif
+
 static int compare_versions(const char *a, const char *b)
 {
 	int a_major, a_minor, a_point, a_patch;
@@ -1064,6 +1090,17 @@ int ef100_probe_netdev_pf(struct efx_nic *efx)
 	eth_hw_addr_set(net_dev, net_dev->perm_addr);
 	memcpy(nic_data->port_id, net_dev->perm_addr, ETH_ALEN);
 
+	if (!nic_data->grp_mae)
+		return 0;
+
+#ifdef CONFIG_SFC_SRIOV
+	rc = efx_ef100_get_base_mport(efx);
+	if (rc) {
+		netif_warn(efx, probe, net_dev,
+			   "Failed to probe base mport rc %d; representors will not function\n",
+			   rc);
+	}
+#endif
 	return 0;
 
 fail:
diff --git a/drivers/net/ethernet/sfc/ef100_nic.h b/drivers/net/ethernet/sfc/ef100_nic.h
index 40f84a275057..0295933145fa 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.h
+++ b/drivers/net/ethernet/sfc/ef100_nic.h
@@ -72,6 +72,8 @@ struct ef100_nic_data {
 	u8 port_id[ETH_ALEN];
 	DECLARE_BITMAP(evq_phases, EFX_MAX_CHANNELS);
 	u64 stats[EF100_STAT_COUNT];
+	u32 base_mport;
+	bool have_mport; /* base_mport was populated successfully */
 	bool grp_mae; /* MAE Privilege */
 	u16 tso_max_hdr_len;
 	u16 tso_max_payload_num_segs;
diff --git a/drivers/net/ethernet/sfc/mae.c b/drivers/net/ethernet/sfc/mae.c
index 011ebd46ada5..0cbcadde6677 100644
--- a/drivers/net/ethernet/sfc/mae.c
+++ b/drivers/net/ethernet/sfc/mae.c
@@ -13,6 +13,16 @@
 #include "mcdi.h"
 #include "mcdi_pcol.h"
 
+void efx_mae_mport_wire(struct efx_nic *efx, u32 *out)
+{
+	efx_dword_t mport;
+
+	EFX_POPULATE_DWORD_2(mport,
+			     MAE_MPORT_SELECTOR_TYPE, MAE_MPORT_SELECTOR_TYPE_PPORT,
+			     MAE_MPORT_SELECTOR_PPORT_ID, efx->port_num);
+	*out = EFX_DWORD_VAL(mport);
+}
+
 void efx_mae_mport_vf(struct efx_nic *efx __always_unused, u32 vf_id, u32 *out)
 {
 	efx_dword_t mport;
diff --git a/drivers/net/ethernet/sfc/mae.h b/drivers/net/ethernet/sfc/mae.h
index 27e69e8a54b6..25c2fd94e158 100644
--- a/drivers/net/ethernet/sfc/mae.h
+++ b/drivers/net/ethernet/sfc/mae.h
@@ -15,6 +15,7 @@
 
 #include "net_driver.h"
 
+void efx_mae_mport_wire(struct efx_nic *efx, u32 *out);
 void efx_mae_mport_vf(struct efx_nic *efx, u32 vf_id, u32 *out);
 
 int efx_mae_lookup_mport(struct efx_nic *efx, u32 selector, u32 *id);

[PATCH net-next v2 05/14] sfc: check ef100 RX packets are from the wire

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

If not, for now drop them and warn.  A subsequent patch will look up
 the source m-port to try and find a representor to deliver them to.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rx.c       | 28 ++++++++++++++++++-----
 drivers/net/ethernet/sfc/ethtool_common.c |  1 +
 drivers/net/ethernet/sfc/net_driver.h     |  3 +++
 3 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef100_rx.c b/drivers/net/ethernet/sfc/ef100_rx.c
index 85207acf7dee..b8da9e3b7bf2 100644
--- a/drivers/net/ethernet/sfc/ef100_rx.c
+++ b/drivers/net/ethernet/sfc/ef100_rx.c
@@ -55,10 +55,14 @@ static bool ef100_has_fcs_error(struct efx_channel *channel, u32 *prefix)
 
 void __ef100_rx_packet(struct efx_channel *channel)
 {
-	struct efx_rx_buffer *rx_buf = efx_rx_buffer(&channel->rx_queue, channel->rx_pkt_index);
+	struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
+	struct efx_rx_buffer *rx_buf = efx_rx_buffer(rx_queue,
+						     channel->rx_pkt_index);
 	struct efx_nic *efx = channel->efx;
+	struct ef100_nic_data *nic_data;
 	u8 *eh = efx_rx_buf_va(rx_buf);
 	__wsum csum = 0;
+	u16 ing_port;
 	u32 *prefix;
 
 	prefix = (u32 *)(eh - ESE_GZ_RX_PKT_PREFIX_LEN);
@@ -76,6 +80,19 @@ void __ef100_rx_packet(struct efx_channel *channel)
 		goto out;
 	}
 
+	ing_port = le16_to_cpu((__force __le16) PREFIX_FIELD(prefix, INGRESS_MPORT));
+
+	nic_data = efx->nic_data;
+
+	if (nic_data->have_mport && ing_port != nic_data->base_mport) {
+		if (net_ratelimit())
+			netif_warn(efx, drv, efx->net_dev,
+				   "Unrecognised ing_port %04x (base %04x), dropping\n",
+				   ing_port, nic_data->base_mport);
+		channel->n_rx_mport_bad++;
+		goto free_rx_buffer;
+	}
+
 	if (likely(efx->net_dev->features & NETIF_F_RXCSUM)) {
 		if (PREFIX_FIELD(prefix, NT_OR_INNER_L3_CLASS) == 1) {
 			++channel->n_rx_ip_hdr_chksum_err;
@@ -87,17 +104,16 @@ void __ef100_rx_packet(struct efx_channel *channel)
 	}
 
 	if (channel->type->receive_skb) {
-		struct efx_rx_queue *rx_queue =
-			efx_channel_get_rx_queue(channel);
-
 		/* no support for special channels yet, so just discard */
 		WARN_ON_ONCE(1);
-		efx_free_rx_buffers(rx_queue, rx_buf, 1);
-		goto out;
+		goto free_rx_buffer;
 	}
 
 	efx_rx_packet_gro(channel, rx_buf, channel->rx_pkt_n_frags, eh, csum);
+	goto out;
 
+free_rx_buffer:
+	efx_free_rx_buffers(rx_queue, rx_buf, 1);
 out:
 	channel->rx_pkt_n_frags = 0;
 }
diff --git a/drivers/net/ethernet/sfc/ethtool_common.c b/drivers/net/ethernet/sfc/ethtool_common.c
index 58ad9d665805..bc840ede3053 100644
--- a/drivers/net/ethernet/sfc/ethtool_common.c
+++ b/drivers/net/ethernet/sfc/ethtool_common.c
@@ -91,6 +91,7 @@ static const struct efx_sw_stat_desc efx_sw_stat_desc[] = {
 	EFX_ETHTOOL_UINT_CHANNEL_STAT(rx_xdp_bad_drops),
 	EFX_ETHTOOL_UINT_CHANNEL_STAT(rx_xdp_tx),
 	EFX_ETHTOOL_UINT_CHANNEL_STAT(rx_xdp_redirect),
+	EFX_ETHTOOL_UINT_CHANNEL_STAT(rx_mport_bad),
 #ifdef CONFIG_RFS_ACCEL
 	EFX_ETHTOOL_UINT_CHANNEL_STAT_NO_N(rfs_filter_count),
 	EFX_ETHTOOL_UINT_CHANNEL_STAT(rfs_succeeded),
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index 4cde54cf77b9..6b64ba3a7d36 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -478,6 +478,8 @@ enum efx_sync_events_state {
  * @n_rx_xdp_bad_drops: Count of RX packets dropped due to XDP errors
  * @n_rx_xdp_tx: Count of RX packets retransmitted due to XDP
  * @n_rx_xdp_redirect: Count of RX packets redirected to a different NIC by XDP
+ * @n_rx_mport_bad: Count of RX packets dropped because their ingress mport was
+ *	not recognised
  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
  *	__efx_rx_packet(), or zero if there is none
  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
@@ -540,6 +542,7 @@ struct efx_channel {
 	unsigned int n_rx_xdp_bad_drops;
 	unsigned int n_rx_xdp_tx;
 	unsigned int n_rx_xdp_redirect;
+	unsigned int n_rx_mport_bad;
 
 	unsigned int rx_pkt_n_frags;
 	unsigned int rx_pkt_index;

[PATCH net-next v2 06/14] sfc: receive packets from EF100 VFs into representors

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

If the source m-port of a packet in __ef100_rx_packet() is a VF,
 hand off the packet to the corresponding representor with
 efx_ef100_rep_rx_packet().

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 19 +++++++++++++++++++
 drivers/net/ethernet/sfc/ef100_rep.h |  5 +++++
 drivers/net/ethernet/sfc/ef100_rx.c  | 18 ++++++++++++++++++
 3 files changed, 42 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index e6c6e9e764b2..c0bc12b9e348 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -224,6 +224,7 @@ static void efx_ef100_rep_destroy_netdev(struct efx_rep *efv)
 	list_del(&efv->list);
 	spin_unlock_bh(&efx->vf_reps_lock);
 	rtnl_unlock();
+	synchronize_rcu();
 	free_netdev(efv->net_dev);
 }
 
@@ -375,3 +376,21 @@ void efx_ef100_rep_rx_packet(struct efx_rep *efv, struct efx_rx_buffer *rx_buf)
 	if (primed)
 		napi_schedule(&efv->napi);
 }
+
+struct efx_rep *efx_ef100_find_rep_by_mport(struct efx_nic *efx, u16 mport)
+{
+	struct efx_rep *efv, *out = NULL;
+
+	/* spinlock guards against list mutation while we're walking it;
+	 * but caller must also hold rcu_read_lock() to ensure the netdev
+	 * isn't freed after we drop the spinlock.
+	 */
+	spin_lock_bh(&efx->vf_reps_lock);
+	list_for_each_entry(efv, &efx->vf_reps, list)
+		if (efv->mport == mport) {
+			out = efv;
+			break;
+		}
+	spin_unlock_bh(&efx->vf_reps_lock);
+	return out;
+}
diff --git a/drivers/net/ethernet/sfc/ef100_rep.h b/drivers/net/ethernet/sfc/ef100_rep.h
index 7d2f15cee8d1..f3787133f793 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.h
+++ b/drivers/net/ethernet/sfc/ef100_rep.h
@@ -58,4 +58,9 @@ void efx_ef100_vfrep_destroy(struct efx_nic *efx, struct efx_rep *efv);
 void efx_ef100_fini_vfreps(struct efx_nic *efx);
 
 void efx_ef100_rep_rx_packet(struct efx_rep *efv, struct efx_rx_buffer *rx_buf);
+/* Returns the representor corresponding to a VF m-port, or NULL
+ * @mport is an m-port label, *not* an m-port ID!
+ * Caller must hold rcu_read_lock().
+ */
+struct efx_rep *efx_ef100_find_rep_by_mport(struct efx_nic *efx, u16 mport);
 #endif /* EF100_REP_H */
diff --git a/drivers/net/ethernet/sfc/ef100_rx.c b/drivers/net/ethernet/sfc/ef100_rx.c
index b8da9e3b7bf2..65bbe37753e6 100644
--- a/drivers/net/ethernet/sfc/ef100_rx.c
+++ b/drivers/net/ethernet/sfc/ef100_rx.c
@@ -85,6 +85,24 @@ void __ef100_rx_packet(struct efx_channel *channel)
 	nic_data = efx->nic_data;
 
 	if (nic_data->have_mport && ing_port != nic_data->base_mport) {
+#ifdef CONFIG_SFC_SRIOV
+		struct efx_rep *efv;
+
+		rcu_read_lock();
+		efv = efx_ef100_find_rep_by_mport(efx, ing_port);
+		if (efv) {
+			if (efv->net_dev->flags & IFF_UP)
+				efx_ef100_rep_rx_packet(efv, rx_buf);
+			rcu_read_unlock();
+			/* Representor Rx doesn't care about PF Rx buffer
+			 * ownership, it just makes a copy. So, we are done
+			 * with the Rx buffer from PF point of view and should
+			 * free it.
+			 */
+			goto free_rx_buffer;
+		}
+		rcu_read_unlock();
+#endif
 		if (net_ratelimit())
 			netif_warn(efx, drv, efx->net_dev,
 				   "Unrecognised ing_port %04x (base %04x), dropping\n",

[PATCH net-next v2 07/14] sfc: insert default MAE rules to connect VFs to representors

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Default rules are low-priority switching rules which the hardware uses
 in the absence of higher-priority rules.  Each representor requires a
 corresponding rule matching traffic from its representee VF and
 delivering to the PF (where a check on INGRESS_MPORT in
 __ef100_rx_packet() will direct it to the representor).  No rule is
 required in the reverse direction, because representor TX uses a TX
 override descriptor to bypass the MAE and deliver directly to the VF.
Since inserting any rule into the MAE disables the firmware's own
 default rules, also insert a pair of rules to connect the PF to the
 physical network port and vice-versa.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/Makefile        |   3 +-
 drivers/net/ethernet/sfc/ef100.c         |   3 +
 drivers/net/ethernet/sfc/ef100_netdev.c  |   4 +
 drivers/net/ethernet/sfc/ef100_nic.c     |  17 ++
 drivers/net/ethernet/sfc/ef100_rep.c     |  20 +-
 drivers/net/ethernet/sfc/ef100_rep.h     |   3 +
 drivers/net/ethernet/sfc/mae.c           | 257 ++++++++++++++++++++++-
 drivers/net/ethernet/sfc/mae.h           |  16 ++
 drivers/net/ethernet/sfc/mcdi.h          |   4 +
 drivers/net/ethernet/sfc/mcdi_pcol_mae.h |  24 +++
 drivers/net/ethernet/sfc/net_driver.h    |   2 +
 drivers/net/ethernet/sfc/tc.c            | 183 ++++++++++++++++
 drivers/net/ethernet/sfc/tc.h            |  76 +++++++
 13 files changed, 606 insertions(+), 6 deletions(-)
 create mode 100644 drivers/net/ethernet/sfc/mcdi_pcol_mae.h
 create mode 100644 drivers/net/ethernet/sfc/tc.c
 create mode 100644 drivers/net/ethernet/sfc/tc.h

diff --git a/drivers/net/ethernet/sfc/Makefile b/drivers/net/ethernet/sfc/Makefile
index 4c759488fc77..bb06fa228367 100644
--- a/drivers/net/ethernet/sfc/Makefile
+++ b/drivers/net/ethernet/sfc/Makefile
@@ -8,7 +8,8 @@ sfc-y			+= efx.o efx_common.o efx_channels.o nic.o \
 			   ef100.o ef100_nic.o ef100_netdev.o \
 			   ef100_ethtool.o ef100_rx.o ef100_tx.o
 sfc-$(CONFIG_SFC_MTD)	+= mtd.o
-sfc-$(CONFIG_SFC_SRIOV)	+= sriov.o ef10_sriov.o ef100_sriov.o ef100_rep.o mae.o
+sfc-$(CONFIG_SFC_SRIOV)	+= sriov.o ef10_sriov.o ef100_sriov.o ef100_rep.o \
+                           mae.o tc.o
 
 obj-$(CONFIG_SFC)	+= sfc.o
 
diff --git a/drivers/net/ethernet/sfc/ef100.c b/drivers/net/ethernet/sfc/ef100.c
index 425017fbcb25..71aab3d0480f 100644
--- a/drivers/net/ethernet/sfc/ef100.c
+++ b/drivers/net/ethernet/sfc/ef100.c
@@ -431,6 +431,9 @@ static void ef100_pci_remove(struct pci_dev *pci_dev)
 
 	probe_data = container_of(efx, struct efx_probe_data, efx);
 	ef100_remove_netdev(probe_data);
+#ifdef CONFIG_SFC_SRIOV
+	efx_fini_struct_tc(efx);
+#endif
 
 	ef100_remove(efx);
 	efx_fini_io(efx);
diff --git a/drivers/net/ethernet/sfc/ef100_netdev.c b/drivers/net/ethernet/sfc/ef100_netdev.c
index 9e65de1ab889..17b9d37218cb 100644
--- a/drivers/net/ethernet/sfc/ef100_netdev.c
+++ b/drivers/net/ethernet/sfc/ef100_netdev.c
@@ -329,6 +329,10 @@ void ef100_remove_netdev(struct efx_probe_data *probe_data)
 
 	ef100_unregister_netdev(efx);
 
+#ifdef CONFIG_SFC_SRIOV
+	efx_fini_tc(efx);
+#endif
+
 	down_write(&efx->filter_sem);
 	efx_mcdi_filter_table_remove(efx);
 	up_write(&efx->filter_sem);
diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 393d6ca4525c..25cd43e3fcf7 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -1094,12 +1094,29 @@ int ef100_probe_netdev_pf(struct efx_nic *efx)
 		return 0;
 
 #ifdef CONFIG_SFC_SRIOV
+	rc = efx_init_struct_tc(efx);
+	if (rc)
+		return rc;
+
 	rc = efx_ef100_get_base_mport(efx);
 	if (rc) {
 		netif_warn(efx, probe, net_dev,
 			   "Failed to probe base mport rc %d; representors will not function\n",
 			   rc);
 	}
+
+	rc = efx_init_tc(efx);
+	if (rc) {
+		/* Either we don't have an MAE at all (i.e. legacy v-switching),
+		 * or we do but we failed to probe it.  In the latter case, we
+		 * may not have set up default rules, in which case we won't be
+		 * able to pass any traffic.  However, we don't fail the probe,
+		 * because the user might need to use the netdevice to apply
+		 * configuration changes to fix whatever's wrong with the MAE.
+		 */
+		netif_warn(efx, probe, net_dev, "Failed to probe MAE rc %d\n",
+			   rc);
+	}
 #endif
 	return 0;
 
diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index c0bc12b9e348..eac932710c63 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -27,6 +27,8 @@ static int efx_ef100_rep_init_struct(struct efx_nic *efx, struct efx_rep *efv,
 	efv->parent = efx;
 	efv->idx = i;
 	INIT_LIST_HEAD(&efv->list);
+	efv->dflt.fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
+	INIT_LIST_HEAD(&efv->dflt.acts.list);
 	INIT_LIST_HEAD(&efv->rx_list);
 	spin_lock_init(&efv->rx_lock);
 	efv->msg_enable = NETIF_MSG_DRV | NETIF_MSG_PROBE |
@@ -212,7 +214,14 @@ static int efx_ef100_configure_rep(struct efx_rep *efv)
 	/* mport label should fit in 16 bits */
 	WARN_ON(efv->mport >> 16);
 
-	return 0;
+	return efx_tc_configure_default_rule_rep(efv);
+}
+
+static void efx_ef100_deconfigure_rep(struct efx_rep *efv)
+{
+	struct efx_nic *efx = efv->parent;
+
+	efx_tc_deconfigure_default_rule(efx, &efv->dflt);
 }
 
 static void efx_ef100_rep_destroy_netdev(struct efx_rep *efv)
@@ -246,19 +255,21 @@ int efx_ef100_vfrep_create(struct efx_nic *efx, unsigned int i)
 		pci_err(efx->pci_dev,
 			"Failed to configure representor for VF %d, rc %d\n",
 			i, rc);
-		goto fail;
+		goto fail1;
 	}
 	rc = register_netdev(efv->net_dev);
 	if (rc) {
 		pci_err(efx->pci_dev,
 			"Failed to register representor for VF %d, rc %d\n",
 			i, rc);
-		goto fail;
+		goto fail2;
 	}
 	pci_dbg(efx->pci_dev, "Representor for VF %d is %s\n", i,
 		efv->net_dev->name);
 	return 0;
-fail:
+fail2:
+	efx_ef100_deconfigure_rep(efv);
+fail1:
 	efx_ef100_rep_destroy_netdev(efv);
 	return rc;
 }
@@ -272,6 +283,7 @@ void efx_ef100_vfrep_destroy(struct efx_nic *efx, struct efx_rep *efv)
 		return;
 	netif_dbg(efx, drv, rep_dev, "Removing VF representor\n");
 	unregister_netdev(rep_dev);
+	efx_ef100_deconfigure_rep(efv);
 	efx_ef100_rep_destroy_netdev(efv);
 }
 
diff --git a/drivers/net/ethernet/sfc/ef100_rep.h b/drivers/net/ethernet/sfc/ef100_rep.h
index f3787133f793..070f700893c1 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.h
+++ b/drivers/net/ethernet/sfc/ef100_rep.h
@@ -14,6 +14,7 @@
 #define EF100_REP_H
 
 #include "net_driver.h"
+#include "tc.h"
 
 struct efx_rep_sw_stats {
 	atomic64_t rx_packets, tx_packets;
@@ -32,6 +33,7 @@ struct efx_rep_sw_stats {
  * @write_index: number of packets enqueued to @rx_list
  * @read_index: number of packets consumed from @rx_list
  * @rx_pring_size: max length of RX list
+ * @dflt: default-rule for MAE switching
  * @list: entry on efx->vf_reps
  * @rx_list: list of SKBs queued for receive in NAPI poll
  * @rx_lock: protects @rx_list
@@ -46,6 +48,7 @@ struct efx_rep {
 	unsigned int idx;
 	unsigned int write_index, read_index;
 	unsigned int rx_pring_size;
+	struct efx_tc_flow_rule dflt;
 	struct list_head list;
 	struct list_head rx_list;
 	spinlock_t rx_lock;
diff --git a/drivers/net/ethernet/sfc/mae.c b/drivers/net/ethernet/sfc/mae.c
index 0cbcadde6677..ea87ec83e618 100644
--- a/drivers/net/ethernet/sfc/mae.c
+++ b/drivers/net/ethernet/sfc/mae.c
@@ -11,7 +11,7 @@
 
 #include "mae.h"
 #include "mcdi.h"
-#include "mcdi_pcol.h"
+#include "mcdi_pcol_mae.h"
 
 void efx_mae_mport_wire(struct efx_nic *efx, u32 *out)
 {
@@ -23,6 +23,17 @@ void efx_mae_mport_wire(struct efx_nic *efx, u32 *out)
 	*out = EFX_DWORD_VAL(mport);
 }
 
+void efx_mae_mport_uplink(struct efx_nic *efx __always_unused, u32 *out)
+{
+	efx_dword_t mport;
+
+	EFX_POPULATE_DWORD_3(mport,
+			     MAE_MPORT_SELECTOR_TYPE, MAE_MPORT_SELECTOR_TYPE_FUNC,
+			     MAE_MPORT_SELECTOR_FUNC_PF_ID, MAE_MPORT_SELECTOR_FUNC_PF_ID_CALLER,
+			     MAE_MPORT_SELECTOR_FUNC_VF_ID, MAE_MPORT_SELECTOR_FUNC_VF_ID_NULL);
+	*out = EFX_DWORD_VAL(mport);
+}
+
 void efx_mae_mport_vf(struct efx_nic *efx __always_unused, u32 vf_id, u32 *out)
 {
 	efx_dword_t mport;
@@ -34,6 +45,17 @@ void efx_mae_mport_vf(struct efx_nic *efx __always_unused, u32 vf_id, u32 *out)
 	*out = EFX_DWORD_VAL(mport);
 }
 
+/* Constructs an mport selector from an mport ID, because they're not the same */
+void efx_mae_mport_mport(struct efx_nic *efx __always_unused, u32 mport_id, u32 *out)
+{
+	efx_dword_t mport;
+
+	EFX_POPULATE_DWORD_2(mport,
+			     MAE_MPORT_SELECTOR_TYPE, MAE_MPORT_SELECTOR_TYPE_MPORT_ID,
+			     MAE_MPORT_SELECTOR_MPORT_ID, mport_id);
+	*out = EFX_DWORD_VAL(mport);
+}
+
 /* id is really only 24 bits wide */
 int efx_mae_lookup_mport(struct efx_nic *efx, u32 selector, u32 *id)
 {
@@ -52,3 +74,236 @@ int efx_mae_lookup_mport(struct efx_nic *efx, u32 selector, u32 *id)
 	*id = MCDI_DWORD(outbuf, MAE_MPORT_LOOKUP_OUT_MPORT_ID);
 	return 0;
 }
+
+static bool efx_mae_asl_id(u32 id)
+{
+	return !!(id & BIT(31));
+}
+
+int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_SET_ALLOC_OUT_LEN);
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_ACTION_SET_ALLOC_IN_LEN);
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_SRC_MAC_ID,
+		       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_DST_MAC_ID,
+		       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_COUNTER_ID,
+		       MC_CMD_MAE_COUNTER_ALLOC_OUT_COUNTER_ID_NULL);
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_COUNTER_LIST_ID,
+		       MC_CMD_MAE_COUNTER_LIST_ALLOC_OUT_COUNTER_LIST_ID_NULL);
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_ENCAP_HEADER_ID,
+		       MC_CMD_MAE_ENCAP_HEADER_ALLOC_OUT_ENCAP_HEADER_ID_NULL);
+	if (act->deliver)
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_DELIVER,
+			       act->dest_mport);
+	BUILD_BUG_ON(MAE_MPORT_SELECTOR_NULL);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_SET_ALLOC, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	act->fw_id = MCDI_DWORD(outbuf, MAE_ACTION_SET_ALLOC_OUT_AS_ID);
+	/* We rely on the high bit of AS IDs always being clear.
+	 * The firmware API guarantees this, but let's check it ourselves.
+	 */
+	if (WARN_ON_ONCE(efx_mae_asl_id(act->fw_id))) {
+		efx_mae_free_action_set(efx, act->fw_id);
+		return -EIO;
+	}
+	return 0;
+}
+
+int efx_mae_free_action_set(struct efx_nic *efx, u32 fw_id)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_SET_FREE_OUT_LEN(1));
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_ACTION_SET_FREE_IN_LEN(1));
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_FREE_IN_AS_ID, fw_id);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_SET_FREE, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	/* FW freed a different ID than we asked for, should never happen.
+	 * Warn because it means we've now got a different idea to the FW of
+	 * what action-sets exist, which could cause mayhem later.
+	 */
+	if (WARN_ON(MCDI_DWORD(outbuf, MAE_ACTION_SET_FREE_OUT_FREED_AS_ID) != fw_id))
+		return -EIO;
+	return 0;
+}
+
+int efx_mae_alloc_action_set_list(struct efx_nic *efx,
+				  struct efx_tc_action_set_list *acts)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_SET_LIST_ALLOC_OUT_LEN);
+	struct efx_tc_action_set *act;
+	size_t inlen, outlen, i = 0;
+	efx_dword_t *inbuf;
+	int rc;
+
+	list_for_each_entry(act, &acts->list, list)
+		i++;
+	if (i == 0)
+		return -EINVAL;
+	if (i == 1) {
+		/* Don't wrap an ASL around a single AS, just use the AS_ID
+		 * directly.  ASLs are a more limited resource.
+		 */
+		act = list_first_entry(&acts->list, struct efx_tc_action_set, list);
+		acts->fw_id = act->fw_id;
+		return 0;
+	}
+	if (i > MC_CMD_MAE_ACTION_SET_LIST_ALLOC_IN_AS_IDS_MAXNUM_MCDI2)
+		return -EOPNOTSUPP; /* Too many actions */
+	inlen = MC_CMD_MAE_ACTION_SET_LIST_ALLOC_IN_LEN(i);
+	inbuf = kzalloc(inlen, GFP_KERNEL);
+	if (!inbuf)
+		return -ENOMEM;
+	i = 0;
+	list_for_each_entry(act, &acts->list, list) {
+		MCDI_SET_ARRAY_DWORD(inbuf, MAE_ACTION_SET_LIST_ALLOC_IN_AS_IDS,
+				     i, act->fw_id);
+		i++;
+	}
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_LIST_ALLOC_IN_COUNT, i);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_SET_LIST_ALLOC, inbuf, inlen,
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		goto out_free;
+	if (outlen < sizeof(outbuf)) {
+		rc = -EIO;
+		goto out_free;
+	}
+	acts->fw_id = MCDI_DWORD(outbuf, MAE_ACTION_SET_LIST_ALLOC_OUT_ASL_ID);
+	/* We rely on the high bit of ASL IDs always being set.
+	 * The firmware API guarantees this, but let's check it ourselves.
+	 */
+	if (WARN_ON_ONCE(!efx_mae_asl_id(acts->fw_id))) {
+		efx_mae_free_action_set_list(efx, acts);
+		rc = -EIO;
+	}
+out_free:
+	kfree(inbuf);
+	return rc;
+}
+
+int efx_mae_free_action_set_list(struct efx_nic *efx,
+				 struct efx_tc_action_set_list *acts)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_SET_LIST_FREE_OUT_LEN(1));
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_ACTION_SET_LIST_FREE_IN_LEN(1));
+	size_t outlen;
+	int rc;
+
+	/* If this is just an AS_ID with no ASL wrapper, then there is
+	 * nothing for us to free.  (The AS will be freed later.)
+	 */
+	if (efx_mae_asl_id(acts->fw_id)) {
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_LIST_FREE_IN_ASL_ID,
+			       acts->fw_id);
+		rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_SET_LIST_FREE, inbuf,
+				  sizeof(inbuf), outbuf, sizeof(outbuf), &outlen);
+		if (rc)
+			return rc;
+		if (outlen < sizeof(outbuf))
+			return -EIO;
+		/* FW freed a different ID than we asked for, should never happen.
+		 * Warn because it means we've now got a different idea to the FW of
+		 * what action-set-lists exist, which could cause mayhem later.
+		 */
+		if (WARN_ON(MCDI_DWORD(outbuf, MAE_ACTION_SET_LIST_FREE_OUT_FREED_ASL_ID) != acts->fw_id))
+			return -EIO;
+	}
+	/* We're probably about to free @acts, but let's just make sure its
+	 * fw_id is blatted so that it won't look valid if it leaks out.
+	 */
+	acts->fw_id = MC_CMD_MAE_ACTION_SET_LIST_ALLOC_OUT_ACTION_SET_LIST_ID_NULL;
+	return 0;
+}
+
+static int efx_mae_populate_match_criteria(MCDI_DECLARE_STRUCT_PTR(match_crit),
+					   const struct efx_tc_match *match)
+{
+	if (match->mask.ingress_port) {
+		if (~match->mask.ingress_port)
+			return -EOPNOTSUPP;
+		MCDI_STRUCT_SET_DWORD(match_crit,
+				      MAE_FIELD_MASK_VALUE_PAIRS_V2_INGRESS_MPORT_SELECTOR,
+				      match->value.ingress_port);
+	}
+	MCDI_STRUCT_SET_DWORD(match_crit, MAE_FIELD_MASK_VALUE_PAIRS_V2_INGRESS_MPORT_SELECTOR_MASK,
+			      match->mask.ingress_port);
+	return 0;
+}
+
+int efx_mae_insert_rule(struct efx_nic *efx, const struct efx_tc_match *match,
+			u32 prio, u32 acts_id, u32 *id)
+{
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_ACTION_RULE_INSERT_IN_LEN(MAE_FIELD_MASK_VALUE_PAIRS_V2_LEN));
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_RULE_INSERT_OUT_LEN);
+	MCDI_DECLARE_STRUCT_PTR(match_crit);
+	MCDI_DECLARE_STRUCT_PTR(response);
+	size_t outlen;
+	int rc;
+
+	if (!id)
+		return -EINVAL;
+
+	match_crit = _MCDI_DWORD(inbuf, MAE_ACTION_RULE_INSERT_IN_MATCH_CRITERIA);
+	response = _MCDI_DWORD(inbuf, MAE_ACTION_RULE_INSERT_IN_RESPONSE);
+	if (efx_mae_asl_id(acts_id)) {
+		MCDI_STRUCT_SET_DWORD(response, MAE_ACTION_RULE_RESPONSE_ASL_ID, acts_id);
+		MCDI_STRUCT_SET_DWORD(response, MAE_ACTION_RULE_RESPONSE_AS_ID,
+				      MC_CMD_MAE_ACTION_SET_ALLOC_OUT_ACTION_SET_ID_NULL);
+	} else {
+		/* We only had one AS, so we didn't wrap it in an ASL */
+		MCDI_STRUCT_SET_DWORD(response, MAE_ACTION_RULE_RESPONSE_ASL_ID,
+				      MC_CMD_MAE_ACTION_SET_LIST_ALLOC_OUT_ACTION_SET_LIST_ID_NULL);
+		MCDI_STRUCT_SET_DWORD(response, MAE_ACTION_RULE_RESPONSE_AS_ID, acts_id);
+	}
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_RULE_INSERT_IN_PRIO, prio);
+	rc = efx_mae_populate_match_criteria(match_crit, match);
+	if (rc)
+		return rc;
+
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_RULE_INSERT, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	*id = MCDI_DWORD(outbuf, MAE_ACTION_RULE_INSERT_OUT_AR_ID);
+	return 0;
+}
+
+int efx_mae_delete_rule(struct efx_nic *efx, u32 id)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_RULE_DELETE_OUT_LEN(1));
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_ACTION_RULE_DELETE_IN_LEN(1));
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, MAE_ACTION_RULE_DELETE_IN_AR_ID, id);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_ACTION_RULE_DELETE, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	/* FW freed a different ID than we asked for, should also never happen.
+	 * Warn because it means we've now got a different idea to the FW of
+	 * what rules exist, which could cause mayhem later.
+	 */
+	if (WARN_ON(MCDI_DWORD(outbuf, MAE_ACTION_RULE_DELETE_OUT_DELETED_AR_ID) != id))
+		return -EIO;
+	return 0;
+}
diff --git a/drivers/net/ethernet/sfc/mae.h b/drivers/net/ethernet/sfc/mae.h
index 25c2fd94e158..e9651f611750 100644
--- a/drivers/net/ethernet/sfc/mae.h
+++ b/drivers/net/ethernet/sfc/mae.h
@@ -14,10 +14,26 @@
 /* MCDI interface for the ef100 Match-Action Engine */
 
 #include "net_driver.h"
+#include "tc.h"
+#include "mcdi_pcol.h" /* needed for various MC_CMD_MAE_*_NULL defines */
 
 void efx_mae_mport_wire(struct efx_nic *efx, u32 *out);
+void efx_mae_mport_uplink(struct efx_nic *efx, u32 *out);
 void efx_mae_mport_vf(struct efx_nic *efx, u32 vf_id, u32 *out);
+void efx_mae_mport_mport(struct efx_nic *efx, u32 mport_id, u32 *out);
 
 int efx_mae_lookup_mport(struct efx_nic *efx, u32 selector, u32 *id);
 
+int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act);
+int efx_mae_free_action_set(struct efx_nic *efx, u32 fw_id);
+
+int efx_mae_alloc_action_set_list(struct efx_nic *efx,
+				  struct efx_tc_action_set_list *acts);
+int efx_mae_free_action_set_list(struct efx_nic *efx,
+				 struct efx_tc_action_set_list *acts);
+
+int efx_mae_insert_rule(struct efx_nic *efx, const struct efx_tc_match *match,
+			u32 prio, u32 acts_id, u32 *id);
+int efx_mae_delete_rule(struct efx_nic *efx, u32 id);
+
 #endif /* EF100_MAE_H */
diff --git a/drivers/net/ethernet/sfc/mcdi.h b/drivers/net/ethernet/sfc/mcdi.h
index f74f6ce8b27d..26bc69f76801 100644
--- a/drivers/net/ethernet/sfc/mcdi.h
+++ b/drivers/net/ethernet/sfc/mcdi.h
@@ -205,6 +205,8 @@ void efx_mcdi_sensor_event(struct efx_nic *efx, efx_qword_t *ev);
 	((_ofst) + BUILD_BUG_ON_ZERO((_ofst) & (_align - 1)))
 #define _MCDI_DWORD(_buf, _field)					\
 	((_buf) + (_MCDI_CHECK_ALIGN(MC_CMD_ ## _field ## _OFST, 4) >> 2))
+#define _MCDI_STRUCT_DWORD(_buf, _field)				\
+	((_buf) + (_MCDI_CHECK_ALIGN(_field ## _OFST, 4) >> 2))
 
 #define MCDI_BYTE(_buf, _field)						\
 	((void)BUILD_BUG_ON_ZERO(MC_CMD_ ## _field ## _LEN != 1),	\
@@ -214,6 +216,8 @@ void efx_mcdi_sensor_event(struct efx_nic *efx, efx_qword_t *ev);
 	 le16_to_cpu(*(__force const __le16 *)MCDI_PTR(_buf, _field)))
 #define MCDI_SET_DWORD(_buf, _field, _value)				\
 	EFX_POPULATE_DWORD_1(*_MCDI_DWORD(_buf, _field), EFX_DWORD_0, _value)
+#define MCDI_STRUCT_SET_DWORD(_buf, _field, _value)			\
+	EFX_POPULATE_DWORD_1(*_MCDI_STRUCT_DWORD(_buf, _field), EFX_DWORD_0, _value)
 #define MCDI_DWORD(_buf, _field)					\
 	EFX_DWORD_FIELD(*_MCDI_DWORD(_buf, _field), EFX_DWORD_0)
 #define MCDI_POPULATE_DWORD_1(_buf, _field, _name1, _value1)		\
diff --git a/drivers/net/ethernet/sfc/mcdi_pcol_mae.h b/drivers/net/ethernet/sfc/mcdi_pcol_mae.h
new file mode 100644
index 000000000000..ff6d80c8e486
--- /dev/null
+++ b/drivers/net/ethernet/sfc/mcdi_pcol_mae.h
@@ -0,0 +1,24 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/****************************************************************************
+ * Driver for Solarflare network controllers and boards
+ * Copyright 2019 Solarflare Communications Inc.
+ * Copyright 2019-2022 Xilinx, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation, incorporated herein by reference.
+ */
+
+#ifndef MCDI_PCOL_MAE_H
+#define MCDI_PCOL_MAE_H
+/* MCDI definitions for Match-Action Engine functionality, that are
+ * missing from the main mcdi_pcol.h
+ */
+
+/* MC_CMD_MAE_COUNTER_LIST_ALLOC is not (yet) a released API, but the
+ * following value is needed as an argument to MC_CMD_MAE_ACTION_SET_ALLOC.
+ */
+/* enum: A counter ID that is guaranteed never to represent a real counter */
+#define          MC_CMD_MAE_COUNTER_LIST_ALLOC_OUT_COUNTER_LIST_ID_NULL 0xffffffff
+
+#endif /* MCDI_PCOL_MAE_H */
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index 6b64ba3a7d36..7ef823d7a89a 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -978,6 +978,7 @@ enum efx_xdp_tx_queues_mode {
  * @xdp_rxq_info_failed: Have any of the rx queues failed to initialise their
  *      xdp_rxq_info structures?
  * @netdev_notifier: Netdevice notifier.
+ * @tc: state for TC offload (EF100).
  * @mem_bar: The BAR that is mapped into membase.
  * @reg_base: Offset from the start of the bar to the function control window.
  * @monitor_work: Hardware monitor workitem
@@ -1161,6 +1162,7 @@ struct efx_nic {
 	bool xdp_rxq_info_failed;
 
 	struct notifier_block netdev_notifier;
+	struct efx_tc_state *tc;
 
 	unsigned int mem_bar;
 	u32 reg_base;
diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
new file mode 100644
index 000000000000..0fb01f73c56e
--- /dev/null
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -0,0 +1,183 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/****************************************************************************
+ * Driver for Solarflare network controllers and boards
+ * Copyright 2019 Solarflare Communications Inc.
+ * Copyright 2020-2022 Xilinx Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation, incorporated herein by reference.
+ */
+
+#include "tc.h"
+#include "mae.h"
+#include "ef100_rep.h"
+
+static void efx_tc_free_action_set(struct efx_nic *efx,
+				   struct efx_tc_action_set *act, bool in_hw)
+{
+	/* Failure paths calling this on the 'running action' set in_hw=false,
+	 * because if the alloc had succeeded we'd've put it in acts.list and
+	 * not still have it in act.
+	 */
+	if (in_hw) {
+		efx_mae_free_action_set(efx, act->fw_id);
+		/* in_hw is true iff we are on an acts.list; make sure to
+		 * remove ourselves from that list before we are freed.
+		 */
+		list_del(&act->list);
+	}
+	kfree(act);
+}
+
+static void efx_tc_free_action_set_list(struct efx_nic *efx,
+					struct efx_tc_action_set_list *acts,
+					bool in_hw)
+{
+	struct efx_tc_action_set *act, *next;
+
+	/* Failure paths set in_hw=false, because usually the acts didn't get
+	 * to efx_mae_alloc_action_set_list(); if they did, the failure tree
+	 * has a separate efx_mae_free_action_set_list() before calling us.
+	 */
+	if (in_hw)
+		efx_mae_free_action_set_list(efx, acts);
+	/* Any act that's on the list will be in_hw even if the list isn't */
+	list_for_each_entry_safe(act, next, &acts->list, list)
+		efx_tc_free_action_set(efx, act, true);
+	/* Don't kfree, as acts is embedded inside a struct efx_tc_flow_rule */
+}
+
+static void efx_tc_delete_rule(struct efx_nic *efx, struct efx_tc_flow_rule *rule)
+{
+	efx_mae_delete_rule(efx, rule->fw_id);
+
+	/* Release entries in subsidiary tables */
+	efx_tc_free_action_set_list(efx, &rule->acts, true);
+	rule->fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
+}
+
+static int efx_tc_configure_default_rule(struct efx_nic *efx, u32 ing_port,
+					 u32 eg_port, struct efx_tc_flow_rule *rule)
+{
+	struct efx_tc_action_set_list *acts = &rule->acts;
+	struct efx_tc_match *match = &rule->match;
+	struct efx_tc_action_set *act;
+	int rc;
+
+	match->value.ingress_port = ing_port;
+	match->mask.ingress_port = ~0;
+	act = kzalloc(sizeof(*act), GFP_KERNEL);
+	if (!act)
+		return -ENOMEM;
+	act->deliver = 1;
+	act->dest_mport = eg_port;
+	rc = efx_mae_alloc_action_set(efx, act);
+	if (rc)
+		goto fail1;
+	EFX_WARN_ON_PARANOID(!list_empty(&acts->list));
+	list_add_tail(&act->list, &acts->list);
+	rc = efx_mae_alloc_action_set_list(efx, acts);
+	if (rc)
+		goto fail2;
+	rc = efx_mae_insert_rule(efx, match, EFX_TC_PRIO_DFLT,
+				 acts->fw_id, &rule->fw_id);
+	if (rc)
+		goto fail3;
+	return 0;
+fail3:
+	efx_mae_free_action_set_list(efx, acts);
+fail2:
+	list_del(&act->list);
+	efx_mae_free_action_set(efx, act->fw_id);
+fail1:
+	kfree(act);
+	return rc;
+}
+
+static int efx_tc_configure_default_rule_pf(struct efx_nic *efx)
+{
+	struct efx_tc_flow_rule *rule = &efx->tc->dflt.pf;
+	u32 ing_port, eg_port;
+
+	efx_mae_mport_uplink(efx, &ing_port);
+	efx_mae_mport_wire(efx, &eg_port);
+	return efx_tc_configure_default_rule(efx, ing_port, eg_port, rule);
+}
+
+static int efx_tc_configure_default_rule_wire(struct efx_nic *efx)
+{
+	struct efx_tc_flow_rule *rule = &efx->tc->dflt.wire;
+	u32 ing_port, eg_port;
+
+	efx_mae_mport_wire(efx, &ing_port);
+	efx_mae_mport_uplink(efx, &eg_port);
+	return efx_tc_configure_default_rule(efx, ing_port, eg_port, rule);
+}
+
+int efx_tc_configure_default_rule_rep(struct efx_rep *efv)
+{
+	struct efx_tc_flow_rule *rule = &efv->dflt;
+	struct efx_nic *efx = efv->parent;
+	u32 ing_port, eg_port;
+
+	efx_mae_mport_mport(efx, efv->mport, &ing_port);
+	efx_mae_mport_uplink(efx, &eg_port);
+	return efx_tc_configure_default_rule(efx, ing_port, eg_port, rule);
+}
+
+void efx_tc_deconfigure_default_rule(struct efx_nic *efx,
+				     struct efx_tc_flow_rule *rule)
+{
+	if (rule->fw_id != MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL)
+		efx_tc_delete_rule(efx, rule);
+	rule->fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
+}
+
+int efx_init_tc(struct efx_nic *efx)
+{
+	int rc;
+
+	rc = efx_tc_configure_default_rule_pf(efx);
+	if (rc)
+		return rc;
+	return efx_tc_configure_default_rule_wire(efx);
+}
+
+void efx_fini_tc(struct efx_nic *efx)
+{
+	/* We can get called even if efx_init_struct_tc() failed */
+	if (!efx->tc)
+		return;
+	efx_tc_deconfigure_default_rule(efx, &efx->tc->dflt.pf);
+	efx_tc_deconfigure_default_rule(efx, &efx->tc->dflt.wire);
+}
+
+int efx_init_struct_tc(struct efx_nic *efx)
+{
+	if (efx->type->is_vf)
+		return 0;
+
+	efx->tc = kzalloc(sizeof(*efx->tc), GFP_KERNEL);
+	if (!efx->tc)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&efx->tc->dflt.pf.acts.list);
+	efx->tc->dflt.pf.fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
+	INIT_LIST_HEAD(&efx->tc->dflt.wire.acts.list);
+	efx->tc->dflt.wire.fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
+	return 0;
+}
+
+void efx_fini_struct_tc(struct efx_nic *efx)
+{
+	if (!efx->tc)
+		return;
+
+	EFX_WARN_ON_PARANOID(efx->tc->dflt.pf.fw_id !=
+			     MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL);
+	EFX_WARN_ON_PARANOID(efx->tc->dflt.wire.fw_id !=
+			     MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL);
+	kfree(efx->tc);
+	efx->tc = NULL;
+}
diff --git a/drivers/net/ethernet/sfc/tc.h b/drivers/net/ethernet/sfc/tc.h
new file mode 100644
index 000000000000..46c5101eaa8d
--- /dev/null
+++ b/drivers/net/ethernet/sfc/tc.h
@@ -0,0 +1,76 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/****************************************************************************
+ * Driver for Solarflare network controllers and boards
+ * Copyright 2019 Solarflare Communications Inc.
+ * Copyright 2020-2022 Xilinx Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published
+ * by the Free Software Foundation, incorporated herein by reference.
+ */
+
+#ifndef EFX_TC_H
+#define EFX_TC_H
+#include "net_driver.h"
+
+struct efx_tc_action_set {
+	u16 deliver:1;
+	u32 dest_mport;
+	u32 fw_id; /* index of this entry in firmware actions table */
+	struct list_head list;
+};
+
+struct efx_tc_match_fields {
+	/* L1 */
+	u32 ingress_port;
+};
+
+struct efx_tc_match {
+	struct efx_tc_match_fields value;
+	struct efx_tc_match_fields mask;
+};
+
+struct efx_tc_action_set_list {
+	struct list_head list;
+	u32 fw_id;
+};
+
+struct efx_tc_flow_rule {
+	struct efx_tc_match match;
+	struct efx_tc_action_set_list acts;
+	u32 fw_id;
+};
+
+enum efx_tc_rule_prios {
+	EFX_TC_PRIO_DFLT, /* Default switch rule; one of efx_tc_default_rules */
+	EFX_TC_PRIO__NUM
+};
+
+/**
+ * struct efx_tc_state - control plane data for TC offload
+ *
+ * @dflt: Match-action rules for default switching; at priority
+ *	%EFX_TC_PRIO_DFLT.  Named by *ingress* port
+ * @dflt.pf: rule for traffic ingressing from PF (egresses to wire)
+ * @dflt.wire: rule for traffic ingressing from wire (egresses to PF)
+ */
+struct efx_tc_state {
+	struct {
+		struct efx_tc_flow_rule pf;
+		struct efx_tc_flow_rule wire;
+	} dflt;
+};
+
+struct efx_rep;
+
+int efx_tc_configure_default_rule_rep(struct efx_rep *efv);
+void efx_tc_deconfigure_default_rule(struct efx_nic *efx,
+				     struct efx_tc_flow_rule *rule);
+
+int efx_init_tc(struct efx_nic *efx);
+void efx_fini_tc(struct efx_nic *efx);
+
+int efx_init_struct_tc(struct efx_nic *efx);
+void efx_fini_struct_tc(struct efx_nic *efx);
+
+#endif /* EFX_TC_H */

[PATCH net-next v2 08/14] sfc: move table locking into filter_table_{probe,remove} methods

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

We need to be able to drop the efx->filter_sem in ef100_filter_table_up()
 so that we can call functions that insert filters (and thus take that
 rwsem for read), which means the efx->type->filter_table_probe method
 needs to be responsible for taking the lock in the first place.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef10.c         | 26 ++++++++++++++-----------
 drivers/net/ethernet/sfc/ef100_nic.c    | 22 +++++++++++++--------
 drivers/net/ethernet/sfc/ef10_sriov.c   | 16 +++------------
 drivers/net/ethernet/sfc/mcdi_filters.h |  1 +
 drivers/net/ethernet/sfc/rx_common.c    |  4 ----
 5 files changed, 33 insertions(+), 36 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef10.c b/drivers/net/ethernet/sfc/ef10.c
index ab979fd11133..ee734b69150f 100644
--- a/drivers/net/ethernet/sfc/ef10.c
+++ b/drivers/net/ethernet/sfc/ef10.c
@@ -2538,23 +2538,33 @@ static int efx_ef10_filter_table_probe(struct efx_nic *efx)
 
 	if (rc)
 		return rc;
+	down_write(&efx->filter_sem);
 	rc = efx_mcdi_filter_table_probe(efx, nic_data->workaround_26807);
 
 	if (rc)
-		return rc;
+		goto out_unlock;
 
 	list_for_each_entry(vlan, &nic_data->vlan_list, list) {
 		rc = efx_mcdi_filter_add_vlan(efx, vlan->vid);
 		if (rc)
 			goto fail_add_vlan;
 	}
-	return 0;
+	goto out_unlock;
 
 fail_add_vlan:
 	efx_mcdi_filter_table_remove(efx);
+out_unlock:
+	up_write(&efx->filter_sem);
 	return rc;
 }
 
+static void efx_ef10_filter_table_remove(struct efx_nic *efx)
+{
+	down_write(&efx->filter_sem);
+	efx_mcdi_filter_table_remove(efx);
+	up_write(&efx->filter_sem);
+}
+
 /* This creates an entry in the RX descriptor queue */
 static inline void
 efx_ef10_build_rx_desc(struct efx_rx_queue *rx_queue, unsigned int index)
@@ -3211,9 +3221,7 @@ static int efx_ef10_vport_set_mac_address(struct efx_nic *efx)
 
 	efx_device_detach_sync(efx);
 	efx_net_stop(efx->net_dev);
-	down_write(&efx->filter_sem);
-	efx_mcdi_filter_table_remove(efx);
-	up_write(&efx->filter_sem);
+	efx_ef10_filter_table_remove(efx);
 
 	rc = efx_ef10_vadaptor_free(efx, efx->vport_id);
 	if (rc)
@@ -3243,9 +3251,7 @@ static int efx_ef10_vport_set_mac_address(struct efx_nic *efx)
 	if (rc2)
 		goto reset_nic;
 restore_filters:
-	down_write(&efx->filter_sem);
 	rc2 = efx_ef10_filter_table_probe(efx);
-	up_write(&efx->filter_sem);
 	if (rc2)
 		goto reset_nic;
 
@@ -3275,8 +3281,7 @@ static int efx_ef10_set_mac_address(struct efx_nic *efx)
 	efx_net_stop(efx->net_dev);
 
 	mutex_lock(&efx->mac_lock);
-	down_write(&efx->filter_sem);
-	efx_mcdi_filter_table_remove(efx);
+	efx_ef10_filter_table_remove(efx);
 
 	ether_addr_copy(MCDI_PTR(inbuf, VADAPTOR_SET_MAC_IN_MACADDR),
 			efx->net_dev->dev_addr);
@@ -3286,7 +3291,6 @@ static int efx_ef10_set_mac_address(struct efx_nic *efx)
 				sizeof(inbuf), NULL, 0, NULL);
 
 	efx_ef10_filter_table_probe(efx);
-	up_write(&efx->filter_sem);
 	mutex_unlock(&efx->mac_lock);
 
 	if (was_enabled)
@@ -4092,7 +4096,7 @@ const struct efx_nic_type efx_hunt_a0_vf_nic_type = {
 	.ev_test_generate = efx_ef10_ev_test_generate,
 	.filter_table_probe = efx_ef10_filter_table_probe,
 	.filter_table_restore = efx_mcdi_filter_table_restore,
-	.filter_table_remove = efx_mcdi_filter_table_remove,
+	.filter_table_remove = efx_ef10_filter_table_remove,
 	.filter_update_rx_scatter = efx_mcdi_update_rx_scatter,
 	.filter_insert = efx_mcdi_filter_insert,
 	.filter_remove_safe = efx_mcdi_filter_remove_safe,
diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 25cd43e3fcf7..5fe18b383e20 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -375,26 +375,32 @@ static int ef100_filter_table_up(struct efx_nic *efx)
 {
 	int rc;
 
+	down_write(&efx->filter_sem);
 	rc = efx_mcdi_filter_add_vlan(efx, EFX_FILTER_VID_UNSPEC);
-	if (rc) {
-		efx_mcdi_filter_table_down(efx);
-		return rc;
-	}
+	if (rc)
+		goto fail_unspec;
 
 	rc = efx_mcdi_filter_add_vlan(efx, 0);
-	if (rc) {
-		efx_mcdi_filter_del_vlan(efx, EFX_FILTER_VID_UNSPEC);
-		efx_mcdi_filter_table_down(efx);
-	}
+	if (rc)
+		goto fail_vlan0;
+	up_write(&efx->filter_sem);
+	return 0;
 
+fail_vlan0:
+	efx_mcdi_filter_del_vlan(efx, EFX_FILTER_VID_UNSPEC);
+fail_unspec:
+	efx_mcdi_filter_table_down(efx);
+	up_write(&efx->filter_sem);
 	return rc;
 }
 
 static void ef100_filter_table_down(struct efx_nic *efx)
 {
+	down_write(&efx->filter_sem);
 	efx_mcdi_filter_del_vlan(efx, 0);
 	efx_mcdi_filter_del_vlan(efx, EFX_FILTER_VID_UNSPEC);
 	efx_mcdi_filter_table_down(efx);
+	up_write(&efx->filter_sem);
 }
 
 /*	Other
diff --git a/drivers/net/ethernet/sfc/ef10_sriov.c b/drivers/net/ethernet/sfc/ef10_sriov.c
index 92550c7e85ce..9aae0d8b713f 100644
--- a/drivers/net/ethernet/sfc/ef10_sriov.c
+++ b/drivers/net/ethernet/sfc/ef10_sriov.c
@@ -501,14 +501,11 @@ int efx_ef10_sriov_set_vf_mac(struct efx_nic *efx, int vf_i, const u8 *mac)
 		efx_device_detach_sync(vf->efx);
 		efx_net_stop(vf->efx->net_dev);
 
-		down_write(&vf->efx->filter_sem);
 		vf->efx->type->filter_table_remove(vf->efx);
 
 		rc = efx_ef10_vadaptor_free(vf->efx, EVB_PORT_ID_ASSIGNED);
-		if (rc) {
-			up_write(&vf->efx->filter_sem);
+		if (rc)
 			return rc;
-		}
 	}
 
 	rc = efx_ef10_evb_port_assign(efx, EVB_PORT_ID_NULL, vf_i);
@@ -539,12 +536,9 @@ int efx_ef10_sriov_set_vf_mac(struct efx_nic *efx, int vf_i, const u8 *mac)
 	if (vf->efx) {
 		/* VF cannot use the vport_id that the PF created */
 		rc = efx_ef10_vadaptor_alloc(vf->efx, EVB_PORT_ID_ASSIGNED);
-		if (rc) {
-			up_write(&vf->efx->filter_sem);
+		if (rc)
 			return rc;
-		}
 		vf->efx->type->filter_table_probe(vf->efx);
-		up_write(&vf->efx->filter_sem);
 		efx_net_open(vf->efx->net_dev);
 		efx_device_attach_if_not_resetting(vf->efx);
 	}
@@ -580,7 +574,6 @@ int efx_ef10_sriov_set_vf_vlan(struct efx_nic *efx, int vf_i, u16 vlan,
 		efx_net_stop(vf->efx->net_dev);
 
 		mutex_lock(&vf->efx->mac_lock);
-		down_write(&vf->efx->filter_sem);
 		vf->efx->type->filter_table_remove(vf->efx);
 
 		rc = efx_ef10_vadaptor_free(vf->efx, EVB_PORT_ID_ASSIGNED);
@@ -654,7 +647,6 @@ int efx_ef10_sriov_set_vf_vlan(struct efx_nic *efx, int vf_i, u16 vlan,
 		if (rc2)
 			goto reset_nic_up_write;
 
-		up_write(&vf->efx->filter_sem);
 		mutex_unlock(&vf->efx->mac_lock);
 
 		rc2 = efx_net_open(vf->efx->net_dev);
@@ -666,10 +658,8 @@ int efx_ef10_sriov_set_vf_vlan(struct efx_nic *efx, int vf_i, u16 vlan,
 	return rc;
 
 reset_nic_up_write:
-	if (vf->efx) {
-		up_write(&vf->efx->filter_sem);
+	if (vf->efx)
 		mutex_unlock(&vf->efx->mac_lock);
-	}
 reset_nic:
 	if (vf->efx) {
 		netif_err(efx, drv, efx->net_dev,
diff --git a/drivers/net/ethernet/sfc/mcdi_filters.h b/drivers/net/ethernet/sfc/mcdi_filters.h
index 06426aa9f2f3..c0d6558b9fd2 100644
--- a/drivers/net/ethernet/sfc/mcdi_filters.h
+++ b/drivers/net/ethernet/sfc/mcdi_filters.h
@@ -89,6 +89,7 @@ struct efx_mcdi_filter_table {
 	 */
 	bool mc_chaining;
 	bool vlan_filter;
+	/* Entries on the vlan_list are added/removed under filter_sem */
 	struct list_head vlan_list;
 };
 
diff --git a/drivers/net/ethernet/sfc/rx_common.c b/drivers/net/ethernet/sfc/rx_common.c
index bd21d6ac778a..4826e6a7e4ce 100644
--- a/drivers/net/ethernet/sfc/rx_common.c
+++ b/drivers/net/ethernet/sfc/rx_common.c
@@ -793,7 +793,6 @@ int efx_probe_filters(struct efx_nic *efx)
 	int rc;
 
 	mutex_lock(&efx->mac_lock);
-	down_write(&efx->filter_sem);
 	rc = efx->type->filter_table_probe(efx);
 	if (rc)
 		goto out_unlock;
@@ -830,7 +829,6 @@ int efx_probe_filters(struct efx_nic *efx)
 	}
 #endif
 out_unlock:
-	up_write(&efx->filter_sem);
 	mutex_unlock(&efx->mac_lock);
 	return rc;
 }
@@ -846,9 +844,7 @@ void efx_remove_filters(struct efx_nic *efx)
 		channel->rps_flow_id = NULL;
 	}
 #endif
-	down_write(&efx->filter_sem);
 	efx->type->filter_table_remove(efx);
-	up_write(&efx->filter_sem);
 }
 
 #ifdef CONFIG_RFS_ACCEL

[PATCH net-next v2 09/14] sfc: use a dynamic m-port for representor RX and set it promisc

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Representors do not want to be subject to the PF's Ethernet address
 filters, since traffic from VFs will typically have a destination
 either elsewhere on the link segment or on an overlay network.
So, create a dynamic m-port with promiscuous and all-multicast
 filters, and set it as the egress port of representor default rules.
 Since the m-port is an alias of the calling PF's own m-port, traffic
 will still be delivered to the PF's RXQs, but it will be subject to
 the VNRX filter rules installed on the dynamic m-port (specified by
 the v-port ID field of the filter spec).

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_nic.c    | 15 +++++
 drivers/net/ethernet/sfc/filter.h       | 18 ++++++
 drivers/net/ethernet/sfc/mae.c          | 37 +++++++++++++
 drivers/net/ethernet/sfc/mae.h          |  3 +
 drivers/net/ethernet/sfc/mcdi_filters.c |  6 +-
 drivers/net/ethernet/sfc/tc.c           | 73 ++++++++++++++++++++++++-
 drivers/net/ethernet/sfc/tc.h           |  9 +++
 7 files changed, 158 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 5fe18b383e20..8061efdaf82c 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -24,6 +24,7 @@
 #include "ef100_tx.h"
 #include "ef100_sriov.h"
 #include "ef100_netdev.h"
+#include "tc.h"
 #include "mae.h"
 #include "rx_common.h"
 
@@ -383,7 +384,18 @@ static int ef100_filter_table_up(struct efx_nic *efx)
 	rc = efx_mcdi_filter_add_vlan(efx, 0);
 	if (rc)
 		goto fail_vlan0;
+	/* Drop the lock: we've finished altering table existence, and
+	 * filter insertion will need to take the lock for read.
+	 */
 	up_write(&efx->filter_sem);
+#ifdef CONFIG_SFC_SRIOV
+	rc = efx_tc_insert_rep_filters(efx);
+	/* Rep filter failure is nonfatal */
+	if (rc)
+		netif_warn(efx, drv, efx->net_dev,
+			   "Failed to insert representor filters, rc %d\n",
+			   rc);
+#endif
 	return 0;
 
 fail_vlan0:
@@ -396,6 +408,9 @@ static int ef100_filter_table_up(struct efx_nic *efx)
 
 static void ef100_filter_table_down(struct efx_nic *efx)
 {
+#ifdef CONFIG_SFC_SRIOV
+	efx_tc_remove_rep_filters(efx);
+#endif
 	down_write(&efx->filter_sem);
 	efx_mcdi_filter_del_vlan(efx, 0);
 	efx_mcdi_filter_del_vlan(efx, EFX_FILTER_VID_UNSPEC);
diff --git a/drivers/net/ethernet/sfc/filter.h b/drivers/net/ethernet/sfc/filter.h
index 40b2af8bfb81..4d928839d292 100644
--- a/drivers/net/ethernet/sfc/filter.h
+++ b/drivers/net/ethernet/sfc/filter.h
@@ -88,6 +88,7 @@ enum efx_filter_priority {
  *	the automatic filter in its place.
  * @EFX_FILTER_FLAG_RX: Filter is for RX
  * @EFX_FILTER_FLAG_TX: Filter is for TX
+ * @EFX_FILTER_FLAG_VPORT_ID: Virtual port ID for adapter switching.
  */
 enum efx_filter_flags {
 	EFX_FILTER_FLAG_RX_RSS = 0x01,
@@ -95,6 +96,7 @@ enum efx_filter_flags {
 	EFX_FILTER_FLAG_RX_OVER_AUTO = 0x04,
 	EFX_FILTER_FLAG_RX = 0x08,
 	EFX_FILTER_FLAG_TX = 0x10,
+	EFX_FILTER_FLAG_VPORT_ID = 0x20,
 };
 
 /** enum efx_encap_type - types of encapsulation
@@ -127,6 +129,9 @@ enum efx_encap_type {
  *	MCFW context_id.
  * @dmaq_id: Source/target queue index, or %EFX_FILTER_RX_DMAQ_ID_DROP for
  *	an RX drop filter
+ * @vport_id: Virtual port ID associated with RX queue, for adapter switching,
+ *	if %EFX_FILTER_FLAG_VPORT_ID is set.  This is an MCFW vport_id, or on
+ *	EF100 an mport selector.
  * @outer_vid: Outer VLAN ID to match, if %EFX_FILTER_MATCH_OUTER_VID is set
  * @inner_vid: Inner VLAN ID to match, if %EFX_FILTER_MATCH_INNER_VID is set
  * @loc_mac: Local MAC address to match, if %EFX_FILTER_MATCH_LOC_MAC or
@@ -156,6 +161,7 @@ struct efx_filter_spec {
 	u32	priority:2;
 	u32	flags:6;
 	u32	dmaq_id:12;
+	u32	vport_id;
 	u32	rss_context;
 	__be16	outer_vid __aligned(4); /* allow jhash2() of match values */
 	__be16	inner_vid;
@@ -292,6 +298,18 @@ static inline int efx_filter_set_mc_def(struct efx_filter_spec *spec)
 	return 0;
 }
 
+/**
+ * efx_filter_set_vport_id - override virtual port id relating to filter
+ * @spec: Specification to initialise
+ * @vport_id: firmware ID of the virtual port
+ */
+static inline void efx_filter_set_vport_id(struct efx_filter_spec *spec,
+					   u32 vport_id)
+{
+	spec->flags |= EFX_FILTER_FLAG_VPORT_ID;
+	spec->vport_id = vport_id;
+}
+
 static inline void efx_filter_set_encap_type(struct efx_filter_spec *spec,
 					     enum efx_encap_type encap_type)
 {
diff --git a/drivers/net/ethernet/sfc/mae.c b/drivers/net/ethernet/sfc/mae.c
index ea87ec83e618..97627f5e3674 100644
--- a/drivers/net/ethernet/sfc/mae.c
+++ b/drivers/net/ethernet/sfc/mae.c
@@ -13,6 +13,43 @@
 #include "mcdi.h"
 #include "mcdi_pcol_mae.h"
 
+int efx_mae_allocate_mport(struct efx_nic *efx, u32 *id, u32 *label)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_MPORT_ALLOC_ALIAS_OUT_LEN);
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_MPORT_ALLOC_ALIAS_IN_LEN);
+	size_t outlen;
+	int rc;
+
+	if (WARN_ON_ONCE(!id))
+		return -EINVAL;
+	if (WARN_ON_ONCE(!label))
+		return -EINVAL;
+
+	MCDI_SET_DWORD(inbuf, MAE_MPORT_ALLOC_ALIAS_IN_TYPE,
+		       MC_CMD_MAE_MPORT_ALLOC_ALIAS_IN_MPORT_TYPE_ALIAS);
+	MCDI_SET_DWORD(inbuf, MAE_MPORT_ALLOC_ALIAS_IN_DELIVER_MPORT,
+		       MAE_MPORT_SELECTOR_ASSIGNED);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_MPORT_ALLOC, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	*id = MCDI_DWORD(outbuf, MAE_MPORT_ALLOC_ALIAS_OUT_MPORT_ID);
+	*label = MCDI_DWORD(outbuf, MAE_MPORT_ALLOC_ALIAS_OUT_LABEL);
+	return 0;
+}
+
+int efx_mae_free_mport(struct efx_nic *efx, u32 id)
+{
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_MPORT_FREE_IN_LEN);
+
+	BUILD_BUG_ON(MC_CMD_MAE_MPORT_FREE_OUT_LEN);
+	MCDI_SET_DWORD(inbuf, MAE_MPORT_FREE_IN_MPORT_ID, id);
+	return efx_mcdi_rpc(efx, MC_CMD_MAE_MPORT_FREE, inbuf, sizeof(inbuf),
+			    NULL, 0, NULL);
+}
+
 void efx_mae_mport_wire(struct efx_nic *efx, u32 *out)
 {
 	efx_dword_t mport;
diff --git a/drivers/net/ethernet/sfc/mae.h b/drivers/net/ethernet/sfc/mae.h
index e9651f611750..0369be4d8983 100644
--- a/drivers/net/ethernet/sfc/mae.h
+++ b/drivers/net/ethernet/sfc/mae.h
@@ -17,6 +17,9 @@
 #include "tc.h"
 #include "mcdi_pcol.h" /* needed for various MC_CMD_MAE_*_NULL defines */
 
+int efx_mae_allocate_mport(struct efx_nic *efx, u32 *id, u32 *label);
+int efx_mae_free_mport(struct efx_nic *efx, u32 id);
+
 void efx_mae_mport_wire(struct efx_nic *efx, u32 *out);
 void efx_mae_mport_uplink(struct efx_nic *efx, u32 *out);
 void efx_mae_mport_vf(struct efx_nic *efx, u32 vf_id, u32 *out);
diff --git a/drivers/net/ethernet/sfc/mcdi_filters.c b/drivers/net/ethernet/sfc/mcdi_filters.c
index 1523be77b9db..4ff6586116ee 100644
--- a/drivers/net/ethernet/sfc/mcdi_filters.c
+++ b/drivers/net/ethernet/sfc/mcdi_filters.c
@@ -221,7 +221,10 @@ static void efx_mcdi_filter_push_prep(struct efx_nic *efx,
 		efx_mcdi_filter_push_prep_set_match_fields(efx, spec, inbuf);
 	}
 
-	MCDI_SET_DWORD(inbuf, FILTER_OP_IN_PORT_ID, efx->vport_id);
+	if (flags & EFX_FILTER_FLAG_VPORT_ID)
+		MCDI_SET_DWORD(inbuf, FILTER_OP_IN_PORT_ID, spec->vport_id);
+	else
+		MCDI_SET_DWORD(inbuf, FILTER_OP_IN_PORT_ID, efx->vport_id);
 	MCDI_SET_DWORD(inbuf, FILTER_OP_IN_RX_DEST,
 		       spec->dmaq_id == EFX_FILTER_RX_DMAQ_ID_DROP ?
 		       MC_CMD_FILTER_OP_IN_RX_DEST_DROP :
@@ -488,6 +491,7 @@ static s32 efx_mcdi_filter_insert_locked(struct efx_nic *efx,
 			saved_spec->flags |= spec->flags;
 			saved_spec->rss_context = spec->rss_context;
 			saved_spec->dmaq_id = spec->dmaq_id;
+			saved_spec->vport_id = spec->vport_id;
 		}
 	} else if (!replacing) {
 		kfree(saved_spec);
diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 0fb01f73c56e..0c0aeb91f500 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -12,6 +12,7 @@
 #include "tc.h"
 #include "mae.h"
 #include "ef100_rep.h"
+#include "efx.h"
 
 static void efx_tc_free_action_set(struct efx_nic *efx,
 				   struct efx_tc_action_set *act, bool in_hw)
@@ -122,7 +123,7 @@ int efx_tc_configure_default_rule_rep(struct efx_rep *efv)
 	u32 ing_port, eg_port;
 
 	efx_mae_mport_mport(efx, efv->mport, &ing_port);
-	efx_mae_mport_uplink(efx, &eg_port);
+	efx_mae_mport_mport(efx, efx->tc->reps_mport_id, &eg_port);
 	return efx_tc_configure_default_rule(efx, ing_port, eg_port, rule);
 }
 
@@ -134,6 +135,68 @@ void efx_tc_deconfigure_default_rule(struct efx_nic *efx,
 	rule->fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
 }
 
+static int efx_tc_configure_rep_mport(struct efx_nic *efx)
+{
+	u32 rep_mport_label;
+	int rc;
+
+	rc = efx_mae_allocate_mport(efx, &efx->tc->reps_mport_id, &rep_mport_label);
+	if (rc)
+		return rc;
+	pci_dbg(efx->pci_dev, "created rep mport 0x%08x (0x%04x)\n",
+		efx->tc->reps_mport_id, rep_mport_label);
+	/* Use mport *selector* as vport ID */
+	efx_mae_mport_mport(efx, efx->tc->reps_mport_id,
+			    &efx->tc->reps_mport_vport_id);
+	return 0;
+}
+
+static void efx_tc_deconfigure_rep_mport(struct efx_nic *efx)
+{
+	efx_mae_free_mport(efx, efx->tc->reps_mport_id);
+	efx->tc->reps_mport_id = MAE_MPORT_SELECTOR_NULL;
+}
+
+int efx_tc_insert_rep_filters(struct efx_nic *efx)
+{
+	struct efx_filter_spec promisc, allmulti;
+	int rc;
+
+	if (efx->type->is_vf)
+		return 0;
+	if (!efx->tc)
+		return 0;
+	efx_filter_init_rx(&promisc, EFX_FILTER_PRI_REQUIRED, 0, 0);
+	efx_filter_set_uc_def(&promisc);
+	efx_filter_set_vport_id(&promisc, efx->tc->reps_mport_vport_id);
+	rc = efx_filter_insert_filter(efx, &promisc, false);
+	if (rc < 0)
+		return rc;
+	efx->tc->reps_filter_uc = rc;
+	efx_filter_init_rx(&allmulti, EFX_FILTER_PRI_REQUIRED, 0, 0);
+	efx_filter_set_mc_def(&allmulti);
+	efx_filter_set_vport_id(&allmulti, efx->tc->reps_mport_vport_id);
+	rc = efx_filter_insert_filter(efx, &allmulti, false);
+	if (rc < 0)
+		return rc;
+	efx->tc->reps_filter_mc = rc;
+	return 0;
+}
+
+void efx_tc_remove_rep_filters(struct efx_nic *efx)
+{
+	if (efx->type->is_vf)
+		return;
+	if (!efx->tc)
+		return;
+	if (efx->tc->reps_filter_mc >= 0)
+		efx_filter_remove_id_safe(efx, EFX_FILTER_PRI_REQUIRED, efx->tc->reps_filter_mc);
+	efx->tc->reps_filter_mc = -1;
+	if (efx->tc->reps_filter_uc >= 0)
+		efx_filter_remove_id_safe(efx, EFX_FILTER_PRI_REQUIRED, efx->tc->reps_filter_uc);
+	efx->tc->reps_filter_uc = -1;
+}
+
 int efx_init_tc(struct efx_nic *efx)
 {
 	int rc;
@@ -141,7 +204,10 @@ int efx_init_tc(struct efx_nic *efx)
 	rc = efx_tc_configure_default_rule_pf(efx);
 	if (rc)
 		return rc;
-	return efx_tc_configure_default_rule_wire(efx);
+	rc = efx_tc_configure_default_rule_wire(efx);
+	if (rc)
+		return rc;
+	return efx_tc_configure_rep_mport(efx);
 }
 
 void efx_fini_tc(struct efx_nic *efx)
@@ -149,6 +215,7 @@ void efx_fini_tc(struct efx_nic *efx)
 	/* We can get called even if efx_init_struct_tc() failed */
 	if (!efx->tc)
 		return;
+	efx_tc_deconfigure_rep_mport(efx);
 	efx_tc_deconfigure_default_rule(efx, &efx->tc->dflt.pf);
 	efx_tc_deconfigure_default_rule(efx, &efx->tc->dflt.wire);
 }
@@ -162,6 +229,8 @@ int efx_init_struct_tc(struct efx_nic *efx)
 	if (!efx->tc)
 		return -ENOMEM;
 
+	efx->tc->reps_filter_uc = -1;
+	efx->tc->reps_filter_mc = -1;
 	INIT_LIST_HEAD(&efx->tc->dflt.pf.acts.list);
 	efx->tc->dflt.pf.fw_id = MC_CMD_MAE_ACTION_RULE_INSERT_OUT_ACTION_RULE_ID_NULL;
 	INIT_LIST_HEAD(&efx->tc->dflt.wire.acts.list);
diff --git a/drivers/net/ethernet/sfc/tc.h b/drivers/net/ethernet/sfc/tc.h
index 46c5101eaa8d..309123c6b386 100644
--- a/drivers/net/ethernet/sfc/tc.h
+++ b/drivers/net/ethernet/sfc/tc.h
@@ -49,12 +49,18 @@ enum efx_tc_rule_prios {
 /**
  * struct efx_tc_state - control plane data for TC offload
  *
+ * @reps_mport_id: MAE port allocated for representor RX
+ * @reps_filter_uc: VNIC filter for representor unicast RX (promisc)
+ * @reps_filter_mc: VNIC filter for representor multicast RX (allmulti)
+ * @reps_mport_vport_id: vport_id for representor RX filters
  * @dflt: Match-action rules for default switching; at priority
  *	%EFX_TC_PRIO_DFLT.  Named by *ingress* port
  * @dflt.pf: rule for traffic ingressing from PF (egresses to wire)
  * @dflt.wire: rule for traffic ingressing from wire (egresses to PF)
  */
 struct efx_tc_state {
+	u32 reps_mport_id, reps_mport_vport_id;
+	s32 reps_filter_uc, reps_filter_mc;
 	struct {
 		struct efx_tc_flow_rule pf;
 		struct efx_tc_flow_rule wire;
@@ -67,6 +73,9 @@ int efx_tc_configure_default_rule_rep(struct efx_rep *efv);
 void efx_tc_deconfigure_default_rule(struct efx_nic *efx,
 				     struct efx_tc_flow_rule *rule);
 
+int efx_tc_insert_rep_filters(struct efx_nic *efx);
+void efx_tc_remove_rep_filters(struct efx_nic *efx);
+
 int efx_init_tc(struct efx_nic *efx);
 void efx_fini_tc(struct efx_nic *efx);
 

[PATCH net-next v2 10/14] sfc: look up VF's client ID when creating representor

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Firmware gives us a handle which we will later use to administrate the
 VF's virtual MAC configuration.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_nic.c | 23 +++++++++++++++++++++++
 drivers/net/ethernet/sfc/ef100_nic.h |  1 +
 drivers/net/ethernet/sfc/ef100_rep.c | 17 +++++++++++++++++
 drivers/net/ethernet/sfc/ef100_rep.h |  2 ++
 4 files changed, 43 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 8061efdaf82c..f79587a2f4ab 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -1098,6 +1098,29 @@ static int ef100_probe_main(struct efx_nic *efx)
 	return rc;
 }
 
+int efx_ef100_lookup_client_id(struct efx_nic *efx, efx_qword_t pciefn, u32 *id)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_GET_CLIENT_HANDLE_OUT_LEN);
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_GET_CLIENT_HANDLE_IN_LEN);
+	u64 pciefn_flat = le64_to_cpu(pciefn.u64[0]);
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, GET_CLIENT_HANDLE_IN_TYPE,
+		       MC_CMD_GET_CLIENT_HANDLE_IN_TYPE_FUNC);
+	MCDI_SET_QWORD(inbuf, GET_CLIENT_HANDLE_IN_FUNC,
+		       pciefn_flat);
+
+	rc = efx_mcdi_rpc(efx, MC_CMD_GET_CLIENT_HANDLE, inbuf,
+			  sizeof(inbuf), outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	*id = MCDI_DWORD(outbuf, GET_CLIENT_HANDLE_OUT_HANDLE);
+	return 0;
+}
+
 int ef100_probe_netdev_pf(struct efx_nic *efx)
 {
 	struct ef100_nic_data *nic_data = efx->nic_data;
diff --git a/drivers/net/ethernet/sfc/ef100_nic.h b/drivers/net/ethernet/sfc/ef100_nic.h
index 0295933145fa..cf78a5a2b7d6 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.h
+++ b/drivers/net/ethernet/sfc/ef100_nic.h
@@ -17,6 +17,7 @@
 extern const struct efx_nic_type ef100_pf_nic_type;
 extern const struct efx_nic_type ef100_vf_nic_type;
 
+int efx_ef100_lookup_client_id(struct efx_nic *efx, efx_qword_t pciefn, u32 *id);
 int ef100_probe_netdev_pf(struct efx_nic *efx);
 int ef100_probe_vf(struct efx_nic *efx);
 void ef100_remove(struct efx_nic *efx);
diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index eac932710c63..3c5f22a6c3b5 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -200,6 +200,7 @@ static struct efx_rep *efx_ef100_rep_create_netdev(struct efx_nic *efx,
 static int efx_ef100_configure_rep(struct efx_rep *efv)
 {
 	struct efx_nic *efx = efv->parent;
+	efx_qword_t pciefn;
 	u32 selector;
 	int rc;
 
@@ -214,6 +215,22 @@ static int efx_ef100_configure_rep(struct efx_rep *efv)
 	/* mport label should fit in 16 bits */
 	WARN_ON(efv->mport >> 16);
 
+	/* Construct PCIE_FUNCTION structure for the representee */
+	EFX_POPULATE_QWORD_3(pciefn,
+			     PCIE_FUNCTION_PF, PCIE_FUNCTION_PF_NULL,
+			     PCIE_FUNCTION_VF, efv->idx,
+			     PCIE_FUNCTION_INTF, PCIE_INTERFACE_CALLER);
+	/* look up representee's client ID */
+	rc = efx_ef100_lookup_client_id(efx, pciefn, &efv->clid);
+	if (rc) {
+		efv->clid = CLIENT_HANDLE_NULL;
+		pci_dbg(efx->pci_dev, "Failed to get VF %u client ID, rc %d\n",
+			efv->idx, rc);
+	} else {
+		pci_dbg(efx->pci_dev, "VF %u client ID %#x\n",
+			efv->idx, efv->clid);
+	}
+
 	return efx_tc_configure_default_rule_rep(efv);
 }
 
diff --git a/drivers/net/ethernet/sfc/ef100_rep.h b/drivers/net/ethernet/sfc/ef100_rep.h
index 070f700893c1..0fa6ad6b2c92 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.h
+++ b/drivers/net/ethernet/sfc/ef100_rep.h
@@ -29,6 +29,7 @@ struct efx_rep_sw_stats {
  * @net_dev: representor netdevice
  * @msg_enable: log message enable flags
  * @mport: m-port ID of corresponding VF
+ * @clid: client ID of corresponding VF
  * @idx: VF index
  * @write_index: number of packets enqueued to @rx_list
  * @read_index: number of packets consumed from @rx_list
@@ -45,6 +46,7 @@ struct efx_rep {
 	struct net_device *net_dev;
 	u32 msg_enable;
 	u32 mport;
+	u32 clid;
 	unsigned int idx;
 	unsigned int write_index, read_index;
 	unsigned int rx_pring_size;

[PATCH net-next v2 11/14] sfc: fetch existing assigned MAC address from FW when creating VF rep

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

For the sake of code commonality, this changes the PF probe path to also
 use MC_CMD_GET_CLIENT_MAC_ADDRESSES instead of the old
 MC_CMD_GET_MAC_ADDRESSES.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_nic.c | 32 +++++++++++++++++++---------
 drivers/net/ethernet/sfc/ef100_nic.h |  2 ++
 drivers/net/ethernet/sfc/ef100_rep.c |  6 ++++++
 3 files changed, 30 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index f79587a2f4ab..2d244a425821 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -130,23 +130,34 @@ static void ef100_mcdi_reboot_detected(struct efx_nic *efx)
 
 /*	MCDI calls
  */
-static int ef100_get_mac_address(struct efx_nic *efx, u8 *mac_address)
+int ef100_get_mac_address(struct efx_nic *efx, u8 *mac_address,
+			  int client_handle, bool empty_ok)
 {
-	MCDI_DECLARE_BUF(outbuf, MC_CMD_GET_MAC_ADDRESSES_OUT_LEN);
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_GET_CLIENT_MAC_ADDRESSES_OUT_LEN(1));
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_GET_CLIENT_MAC_ADDRESSES_IN_LEN);
 	size_t outlen;
 	int rc;
 
-	BUILD_BUG_ON(MC_CMD_GET_MAC_ADDRESSES_IN_LEN != 0);
+	MCDI_SET_DWORD(inbuf, GET_CLIENT_MAC_ADDRESSES_IN_CLIENT_HANDLE,
+		       client_handle);
 
-	rc = efx_mcdi_rpc(efx, MC_CMD_GET_MAC_ADDRESSES, NULL, 0,
-			  outbuf, sizeof(outbuf), &outlen);
+	rc = efx_mcdi_rpc(efx, MC_CMD_GET_CLIENT_MAC_ADDRESSES, inbuf,
+			  sizeof(inbuf), outbuf, sizeof(outbuf), &outlen);
 	if (rc)
 		return rc;
-	if (outlen < MC_CMD_GET_MAC_ADDRESSES_OUT_LEN)
-		return -EIO;
 
-	ether_addr_copy(mac_address,
-			MCDI_PTR(outbuf, GET_MAC_ADDRESSES_OUT_MAC_ADDR_BASE));
+	if (outlen >= MC_CMD_GET_CLIENT_MAC_ADDRESSES_OUT_LEN(1)) {
+		ether_addr_copy(mac_address,
+				MCDI_PTR(outbuf,
+					 GET_CLIENT_MAC_ADDRESSES_OUT_MAC_ADDRS));
+	} else if (empty_ok) {
+		pci_warn(efx->pci_dev,
+			 "No MAC address provisioned for client ID %#x.\n",
+			 client_handle);
+		eth_zero_addr(mac_address);
+	} else {
+		return -ENOENT;
+	}
 	return 0;
 }
 
@@ -1127,7 +1138,8 @@ int ef100_probe_netdev_pf(struct efx_nic *efx)
 	struct net_device *net_dev = efx->net_dev;
 	int rc;
 
-	rc = ef100_get_mac_address(efx, net_dev->perm_addr);
+	rc = ef100_get_mac_address(efx, net_dev->perm_addr, CLIENT_HANDLE_SELF,
+				   false);
 	if (rc)
 		goto fail;
 	/* Assign MAC address */
diff --git a/drivers/net/ethernet/sfc/ef100_nic.h b/drivers/net/ethernet/sfc/ef100_nic.h
index cf78a5a2b7d6..5241fc1d928e 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.h
+++ b/drivers/net/ethernet/sfc/ef100_nic.h
@@ -17,6 +17,8 @@
 extern const struct efx_nic_type ef100_pf_nic_type;
 extern const struct efx_nic_type ef100_vf_nic_type;
 
+int ef100_get_mac_address(struct efx_nic *efx, u8 *mac_address,
+			  int client_handle, bool empty_ok);
 int efx_ef100_lookup_client_id(struct efx_nic *efx, efx_qword_t pciefn, u32 *id);
 int ef100_probe_netdev_pf(struct efx_nic *efx);
 int ef100_probe_vf(struct efx_nic *efx);
diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index 3c5f22a6c3b5..ebab4579e63b 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -199,6 +199,7 @@ static struct efx_rep *efx_ef100_rep_create_netdev(struct efx_nic *efx,
 
 static int efx_ef100_configure_rep(struct efx_rep *efv)
 {
+	struct net_device *net_dev = efv->net_dev;
 	struct efx_nic *efx = efv->parent;
 	efx_qword_t pciefn;
 	u32 selector;
@@ -229,6 +230,11 @@ static int efx_ef100_configure_rep(struct efx_rep *efv)
 	} else {
 		pci_dbg(efx->pci_dev, "VF %u client ID %#x\n",
 			efv->idx, efv->clid);
+
+		/* Get the assigned MAC address */
+		(void)ef100_get_mac_address(efx, net_dev->perm_addr, efv->clid,
+					    true);
+		eth_hw_addr_set(net_dev, net_dev->perm_addr);
 	}
 
 	return efx_tc_configure_default_rule_rep(efv);

[PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

When setting the VF rep's MAC address, set the provisioned MAC address
 for the VF through MC_CMD_SET_CLIENT_MAC_ADDRESSES.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index ebab4579e63b..58365a4c7c6a 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -107,6 +107,33 @@ static int efx_ef100_rep_get_phys_port_name(struct net_device *dev,
 	return 0;
 }
 
+static int efx_ef100_rep_set_mac_address(struct net_device *net_dev, void *data)
+{
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_SET_CLIENT_MAC_ADDRESSES_IN_LEN(1));
+	struct efx_rep *efv = netdev_priv(net_dev);
+	struct efx_nic *efx = efv->parent;
+	struct sockaddr *addr = data;
+	const u8 *new_addr = addr->sa_data;
+	int rc;
+
+	if (efv->clid == CLIENT_HANDLE_NULL) {
+		netif_info(efx, drv, net_dev, "Unable to set representee MAC address (client ID is null)\n");
+	} else {
+		BUILD_BUG_ON(MC_CMD_SET_CLIENT_MAC_ADDRESSES_OUT_LEN);
+		MCDI_SET_DWORD(inbuf, SET_CLIENT_MAC_ADDRESSES_IN_CLIENT_HANDLE,
+			       efv->clid);
+		ether_addr_copy(MCDI_PTR(inbuf, SET_CLIENT_MAC_ADDRESSES_IN_MAC_ADDRS),
+				new_addr);
+		rc = efx_mcdi_rpc(efx, MC_CMD_SET_CLIENT_MAC_ADDRESSES, inbuf,
+				  sizeof(inbuf), NULL, 0, NULL);
+		if (rc)
+			return rc;
+	}
+
+	eth_hw_addr_set(net_dev, new_addr);
+	return 0;
+}
+
 static void efx_ef100_rep_get_stats64(struct net_device *dev,
 				      struct rtnl_link_stats64 *stats)
 {
@@ -126,6 +153,7 @@ static const struct net_device_ops efx_ef100_rep_netdev_ops = {
 	.ndo_start_xmit		= efx_ef100_rep_xmit,
 	.ndo_get_port_parent_id	= efx_ef100_rep_get_port_parent_id,
 	.ndo_get_phys_port_name	= efx_ef100_rep_get_phys_port_name,
+	.ndo_set_mac_address    = efx_ef100_rep_set_mac_address,
 	.ndo_get_stats64	= efx_ef100_rep_get_stats64,
 };
 

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Jakub Kicinski]

On Wed, 27 Jul 2022 18:46:02 +0100 ecree@xilinx.com wrote:
> From: Edward Cree <ecree.xilinx@gmail.com>
> 
> When setting the VF rep's MAC address, set the provisioned MAC address
>  for the VF through MC_CMD_SET_CLIENT_MAC_ADDRESSES.

Wait.. hm? The VF rep is not the VF. It's the other side of the wire.
Are you passing the VF rep's MAC on the VF? Ethernet packets between
the hypervisor and the VF would have the same SA and DA.

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Edward Cree]

On 28/07/2022 04:10, Jakub Kicinski wrote:
> On Wed, 27 Jul 2022 18:46:02 +0100 ecree@xilinx.com wrote:
>> From: Edward Cree <ecree.xilinx@gmail.com>
>>
>> When setting the VF rep's MAC address, set the provisioned MAC address
>>  for the VF through MC_CMD_SET_CLIENT_MAC_ADDRESSES.
> 
> Wait.. hm? The VF rep is not the VF. It's the other side of the wire.
> Are you passing the VF rep's MAC on the VF? Ethernet packets between
> the hypervisor and the VF would have the same SA and DA.
> 

Yes (but only if there's an IP stack on the repr; I think it's fine if
 the repr is plugged straight into a bridge so any ARP picks up a
 different DA?).
I thought that was weird but I also thought that was 'how it's done'
 with reps — properties of the VF are set by applying them to the rep.
Is there some other way to configure VF MAC?  (Are we supposed to still
 be using the legacy SR-IOV interface, .ndo_set_vf_mac()?  I thought
 that was deprecated in favour of more switchdev-flavoured stuff…)

-ed

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Jakub Kicinski]

On Thu, 28 Jul 2022 16:47:36 +0100 Edward Cree wrote:
> On 28/07/2022 04:10, Jakub Kicinski wrote:
> > On Wed, 27 Jul 2022 18:46:02 +0100 ecree@xilinx.com wrote:  
> >> When setting the VF rep's MAC address, set the provisioned MAC address
> >>  for the VF through MC_CMD_SET_CLIENT_MAC_ADDRESSES.  
> > 
> > Wait.. hm? The VF rep is not the VF. It's the other side of the wire.
> > Are you passing the VF rep's MAC on the VF? Ethernet packets between
> > the hypervisor and the VF would have the same SA and DA.
> 
> Yes (but only if there's an IP stack on the repr; I think it's fine if
>  the repr is plugged straight into a bridge so any ARP picks up a
>  different DA?).
> I thought that was weird but I also thought that was 'how it's done'
>  with reps — properties of the VF are set by applying them to the rep.
> Is there some other way to configure VF MAC?  (Are we supposed to still
>  be using the legacy SR-IOV interface, .ndo_set_vf_mac()?  I thought
>  that was deprecated in favour of more switchdev-flavoured stuff…)

It's set thru

 devlink port function set DEV/PORT_INDEX hw_addr ADDR

"port functions" is a weird object representing something 
in Mellanox FW. Hopefully it makes more sense to you than
it does to me.

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Edward Cree]

On 28/07/2022 17:20, Jakub Kicinski wrote:
> It's set thru
> 
>  devlink port function set DEV/PORT_INDEX hw_addr ADDR
> 
> "port functions" is a weird object representing something 
> in Mellanox FW. Hopefully it makes more sense to you than
> it does to me.
Hmm that does look weird, looks like it acts on a PCI device
 (DEV is a PCI address) and then I'm not sure what PORT_INDEX
 is meant to mean (the man page doesn't describe it at all).
 Possibly it doesn't have semantics as such and is just a
 synthetic index into a list of ports…
I can't say it makes sense to me either :shrug:

We did take a look at what nfp does, as well; they use the
 old .ndo_set_vf_mac(), but they appear to support it both on
 the PF and on the VF reprs — meaning that (AFAICT) it allows
 to set the MAC address of VF 0 through the repr for VF 1.
(There is no check that I can see in nfp_app_set_vf_mac()
 that the value of `int vf` matches the caller.)

Our (SN1000) approach to the problem of configuring 'remote'
 functions (VFs in VMs, PFs on the embedded SoC) is to use
 representors for them all (VF reps as added in this & prev
 series, PF reps coming in the future.  Similarly, if we
 were ever to add Subfunctions, each SF would have a
 corresponding SF representor that would work in much the
 same way as VF reps).  At which point you should always be
 able to configure an object through its associated rep,
 and there should never be a need for an 'index' parameter
 (be that 'VF index' or 'port index').
While .ndo_set_mac_address() might be the Wrong Thing (if
 we want to be able to set VF and VF-rep addresses
 independently to different things), the Right Thing ought
 to have the same signature (i.e. just taking a netdev and
 a hwaddr).  Devlink seems to me like a needless
 complication here.

Anyway, since the proper direction is unclear, I'll respin
 the series without patches 10-13 in the hope of getting
 the rest of it in before the merge window.

-ed

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Jakub Kicinski]

On Thu, 28 Jul 2022 19:12:34 +0100 Edward Cree wrote:
> On 28/07/2022 17:20, Jakub Kicinski wrote:
> > It's set thru
> > 
> >  devlink port function set DEV/PORT_INDEX hw_addr ADDR
> > 
> > "port functions" is a weird object representing something 
> > in Mellanox FW. Hopefully it makes more sense to you than
> > it does to me.  
> Hmm that does look weird, looks like it acts on a PCI device
>  (DEV is a PCI address) and then I'm not sure what PORT_INDEX
>  is meant to mean (the man page doesn't describe it at all).
>  Possibly it doesn't have semantics as such and is just a
>  synthetic index into a list of ports…
> I can't say it makes sense to me either :shrug:
> 
> We did take a look at what nfp does, as well; they use the
>  old .ndo_set_vf_mac(), but they appear to support it both on
>  the PF and on the VF reprs — meaning that (AFAICT) it allows
>  to set the MAC address of VF 0 through the repr for VF 1.
> (There is no check that I can see in nfp_app_set_vf_mac()
>  that the value of `int vf` matches the caller.)

IIRC the reprs are all linked to the PCI device of the PF in sysfs,
and OpenStack would pick a device linked to the PCI parent almost
at random. So the VF reprs needed the legacy NDOs. At least that's
what I remember being told.

I think the legacy NDOs are acceptable, devlink way is preferred
(devlink way did not exist when NFP code was written).

> Our (SN1000) approach to the problem of configuring 'remote'
>  functions (VFs in VMs, PFs on the embedded SoC) is to use
>  representors for them all (VF reps as added in this & prev
>  series, PF reps coming in the future.  Similarly, if we
>  were ever to add Subfunctions, each SF would have a
>  corresponding SF representor that would work in much the
>  same way as VF reps).  At which point you should always be
>  able to configure an object through its associated rep,
>  and there should never be a need for an 'index' parameter
>  (be that 'VF index' or 'port index').

How do you map reprs to VFs? The PCI devices of the VF may be on 
a different system.

> While .ndo_set_mac_address() might be the Wrong Thing (if
>  we want to be able to set VF and VF-rep addresses
>  independently to different things), the Right Thing ought
>  to have the same signature (i.e. just taking a netdev and
>  a hwaddr).  Devlink seems to me like a needless
>  complication here.

But reps are like switch ports in a switch ASIC, and the PCI
device is the other side of the virtual wire. You would not be
configuring the MAC address of a peer to peer link by setting 
the local address.

> Anyway, since the proper direction is unclear, I'll respin
>  the series without patches 10-13 in the hope of getting
>  the rest of it in before the merge window.

SG

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Edward Cree]

On 28/07/2022 19:32, Jakub Kicinski wrote:
> IIRC the reprs are all linked to the PCI device of the PF in sysfs,
You mean /sys/class/net/$VFREP/device?  On sfc that's (intentionally)
 nonexistent; the only visible connection between repr and its owning
 PF is /sys/class/net/$VFREP/phys_switch_id which holds the same
 value as /sys/class/net/$PF/phys_port_id.

> How do you map reprs to VFs? The PCI devices of the VF may be on 
> a different system.
That's what the client ID from patch #10 is for.  We ask the FW for
 a handle to "caller's PCIe controller → caller's PF → VF number
 efv->idx", and that handle is what we store in efv->clid, and later
 pass to MC_CMD_SET_CLIENT_MAC_ADDRESSES in patch #12.

The user determines which repr corresponds to which VF by looking in
 /sys/class/net/$VFREP/phys_port_name (e.g. "p0pf0vf0").

> But reps are like switch ports in a switch ASIC, and the PCI
> device is the other side of the virtual wire. You would not be
> configuring the MAC address of a peer to peer link by setting 
> the local address.
Indeed.  I agree that .ndo_set_mac_address() is the wrong interface.
But the interface I have in mind would be something like
    int (*ndo_set_partner_mac_address)(struct net_device *, void *);
 and would only be implemented by representor netdevs.
Idk what the uAPI/UI for that would be; probably a new `ip link set`
 parameter.

-ed

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Jakub Kicinski]

On Thu, 28 Jul 2022 19:54:21 +0100 Edward Cree wrote:
> On 28/07/2022 19:32, Jakub Kicinski wrote:
> > How do you map reprs to VFs? The PCI devices of the VF may be on 
> > a different system.  
> That's what the client ID from patch #10 is for.  We ask the FW for
>  a handle to "caller's PCIe controller → caller's PF → VF number
>  efv->idx", and that handle is what we store in efv->clid, and later
>  pass to MC_CMD_SET_CLIENT_MAC_ADDRESSES in patch #12.
> 
> The user determines which repr corresponds to which VF by looking in
>  /sys/class/net/$VFREP/phys_port_name (e.g. "p0pf0vf0").

.. and that would also most likely be what the devlink port ID would be.

> > But reps are like switch ports in a switch ASIC, and the PCI
> > device is the other side of the virtual wire. You would not be
> > configuring the MAC address of a peer to peer link by setting 
> > the local address.  
> Indeed.  I agree that .ndo_set_mac_address() is the wrong interface.
> But the interface I have in mind would be something like
>     int (*ndo_set_partner_mac_address)(struct net_device *, void *);
>  and would only be implemented by representor netdevs.
> Idk what the uAPI/UI for that would be; probably a new `ip link set`
>  parameter.

Yup... If only you were there during the fight over this uAPI.
Now it's the devlink "port function" thing.

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Edward Cree]

On 28/07/2022 20:27, Jakub Kicinski wrote:
> On Thu, 28 Jul 2022 19:54:21 +0100 Edward Cree wrote:
>> The user determines which repr corresponds to which VF by looking in
>>  /sys/class/net/$VFREP/phys_port_name (e.g. "p0pf0vf0").
> 
> .. and that would also most likely be what the devlink port ID would be.
AFAICT the devlink port index is just an integer.  The example in the
 man page is
    devlink port function set pci/0000:01:00.0/1 hw_addr 00:00:00:11:22:33
Moreover, struct devlink_port has `unsigned int index`.
Though it does also have `struct devlink_port_attrs` which appears to
 encode the PF and VF numbers; I think those can be read with `devlink
 port show`.
But the whole devlink port abstraction is unnecessary when we already
 *have* an object to represent the port.

>> Indeed.  I agree that .ndo_set_mac_address() is the wrong interface.
>> But the interface I have in mind would be something like
>>     int (*ndo_set_partner_mac_address)(struct net_device *, void *);
>>  and would only be implemented by representor netdevs.
>> Idk what the uAPI/UI for that would be; probably a new `ip link set`
>>  parameter.
> 
> Yup... If only you were there during the fight over this uAPI.
> Now it's the devlink "port function" thing.
Sadly I was too busy with EF100 bring-up, and naïvely assumed that I
 could safely ignore devlink port stuff as it was so obviously going
 to be a classic Mellanox design: tasteless, overweight, and not
 cleanly mappable onto any other vendor.  Which seems to have been
 true but they've managed to make it the standard anyway by virtue
 of being there first, as usual :'(
(Yeah, I probably shouldn't publicly say things like that about
 another vendor's devs.  But I'm getting frustrated at this recurring
 pattern.)

Devlink port function *would* be useful for administering functions
 that don't have a representor.  I just can't see any good reason
 why such things should ever exist.
Maybe it's not too late to introduce my API to exist alongside it…
 though I have no idea how much work it would take to teach the
 orchestration frameworks to use it :/

-ed

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Jakub Kicinski]

On Thu, 28 Jul 2022 21:23:23 +0100 Edward Cree wrote:
> Sadly I was too busy with EF100 bring-up, and naïvely assumed that I
>  could safely ignore devlink port stuff as it was so obviously going
>  to be a classic Mellanox design: tasteless, overweight, and not
>  cleanly mappable onto any other vendor.  Which seems to have been
>  true but they've managed to make it the standard anyway by virtue
>  of being there first, as usual :'(
> (Yeah, I probably shouldn't publicly say things like that about
>  another vendor's devs.  But I'm getting frustrated at this recurring
>  pattern.)

I spend an unhealthy amount of time thinking about the problem 
of vendors not paying attention when new uAPIs are forged.
Happy to try things.

> Devlink port function *would* be useful for administering functions
>  that don't have a representor.  I just can't see any good reason
>  why such things should ever exist.

The SmartNIC/DPU/IPU/isolated hv+IO CPU can expose storage functions
to the peer. nVidia is working on extending the devlink rate limit API
to cover such cases.

Re: [PATCH net-next v2 12/14] sfc: set EF100 VF MAC address through representor

[Edward Cree]

On 29/07/2022 02:45, Jakub Kicinski wrote:
>> Devlink port function *would* be useful for administering functions
>>  that don't have a representor.  I just can't see any good reason
>>  why such things should ever exist.
> 
> The SmartNIC/DPU/IPU/isolated hv+IO CPU can expose storage functions
> to the peer. nVidia is working on extending the devlink rate limit API
> to cover such cases.

All the storage-on-SmartNIC setups I can imagine involve the storage
 function (e.g. a virtio-blk PF) being connected to the network switch,
 either to access remote network storage or to export the local storage
 over the network.  (I'm not quite sure why you'd bother combining
 storage and networking functionality onto a single device if they
 _weren't_ connected in this way.)
Which means that your storage function has a v-switch port, and thus
 should have a representor netdevice so you can e.g. use tc rules to
 define its access to the physical network.
Arguably any network rate limiting you then want to apply to that
 function's v-switch port should be in the form of a tc police action.
(Which is far more flexible than devlink rate, because you can have
 different policers for traffic matching different tc filters, e.g.
 separate rate limits for control and data traffic of the dFS.)

Idk, maybe I'm being crazy in assuming that hardware has sane design
 semantics.  But the obvious way to build a SmartNIC maps very cleanly
 onto representors, without any need for devlink port function, and I
 think it makes more sense to say that maybe some weird device might
 end up having representors to control some objects that don't have
 network access, than that everyone has to implement this whole
 parallel structure of devlink objects for things that already have
 representors.

-ed

[PATCH net-next v2 13/14] sfc: get provisioned MAC address on EF100 VF probe

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

Move the implementation from the PF-specific probe path to that part
 which is common to both PFs and VFs.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_netdev.c | 10 ++++++++++
 drivers/net/ethernet/sfc/ef100_nic.c    | 13 ++-----------
 2 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/sfc/ef100_netdev.c b/drivers/net/ethernet/sfc/ef100_netdev.c
index 17b9d37218cb..4694417a4054 100644
--- a/drivers/net/ethernet/sfc/ef100_netdev.c
+++ b/drivers/net/ethernet/sfc/ef100_netdev.c
@@ -349,6 +349,7 @@ int ef100_probe_netdev(struct efx_probe_data *probe_data)
 {
 	struct efx_nic *efx = &probe_data->efx;
 	struct efx_probe_data **probe_ptr;
+	struct ef100_nic_data *nic_data;
 	struct net_device *net_dev;
 	int rc;
 
@@ -400,6 +401,15 @@ int ef100_probe_netdev(struct efx_probe_data *probe_data)
 	/* Don't fail init if RSS setup doesn't work. */
 	efx_mcdi_push_default_indir_table(efx, efx->n_rx_channels);
 
+	rc = ef100_get_mac_address(efx, net_dev->perm_addr, CLIENT_HANDLE_SELF,
+				   efx->type->is_vf);
+	if (rc)
+		goto fail;
+	/* Assign MAC address */
+	eth_hw_addr_set(net_dev, net_dev->perm_addr);
+	nic_data = efx->nic_data;
+	memcpy(nic_data->port_id, net_dev->perm_addr, ETH_ALEN);
+
 	rc = ef100_register_netdev(efx);
 	if (rc)
 		goto fail;
diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 2d244a425821..b5c08023c1fb 100644
--- a/drivers/net/ethernet/sfc/ef100_nic.c
+++ b/drivers/net/ethernet/sfc/ef100_nic.c
@@ -1135,16 +1135,10 @@ int efx_ef100_lookup_client_id(struct efx_nic *efx, efx_qword_t pciefn, u32 *id)
 int ef100_probe_netdev_pf(struct efx_nic *efx)
 {
 	struct ef100_nic_data *nic_data = efx->nic_data;
+#ifdef CONFIG_SFC_SRIOV
 	struct net_device *net_dev = efx->net_dev;
 	int rc;
-
-	rc = ef100_get_mac_address(efx, net_dev->perm_addr, CLIENT_HANDLE_SELF,
-				   false);
-	if (rc)
-		goto fail;
-	/* Assign MAC address */
-	eth_hw_addr_set(net_dev, net_dev->perm_addr);
-	memcpy(nic_data->port_id, net_dev->perm_addr, ETH_ALEN);
+#endif
 
 	if (!nic_data->grp_mae)
 		return 0;
@@ -1175,9 +1169,6 @@ int ef100_probe_netdev_pf(struct efx_nic *efx)
 	}
 #endif
 	return 0;
-
-fail:
-	return rc;
 }
 
 int ef100_probe_vf(struct efx_nic *efx)

[PATCH net-next v2 14/14] sfc: implement ethtool get/set RX ring size for EF100 reps

[ecree]

From: Edward Cree <ecree.xilinx@gmail.com>

It's not truly a ring, but the maximum length of the list of queued RX
 SKBs is analogous to an RX ring size, so use that API to configure it.

Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
---
 drivers/net/ethernet/sfc/ef100_rep.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/drivers/net/ethernet/sfc/ef100_rep.c b/drivers/net/ethernet/sfc/ef100_rep.c
index 58365a4c7c6a..efbe7d482d7b 100644
--- a/drivers/net/ethernet/sfc/ef100_rep.c
+++ b/drivers/net/ethernet/sfc/ef100_rep.c
@@ -178,10 +178,37 @@ static void efx_ef100_rep_ethtool_set_msglevel(struct net_device *net_dev,
 	efv->msg_enable = msg_enable;
 }
 
+static void efx_ef100_rep_ethtool_get_ringparam(struct net_device *net_dev,
+						struct ethtool_ringparam *ring,
+						struct kernel_ethtool_ringparam *kring,
+						struct netlink_ext_ack *ext_ack)
+{
+	struct efx_rep *efv = netdev_priv(net_dev);
+
+	ring->rx_max_pending = U32_MAX;
+	ring->rx_pending = efv->rx_pring_size;
+}
+
+static int efx_ef100_rep_ethtool_set_ringparam(struct net_device *net_dev,
+					       struct ethtool_ringparam *ring,
+					       struct kernel_ethtool_ringparam *kring,
+					       struct netlink_ext_ack *ext_ack)
+{
+	struct efx_rep *efv = netdev_priv(net_dev);
+
+	if (ring->rx_mini_pending || ring->rx_jumbo_pending || ring->tx_pending)
+		return -EINVAL;
+
+	efv->rx_pring_size = ring->rx_pending;
+	return 0;
+}
+
 static const struct ethtool_ops efx_ef100_rep_ethtool_ops = {
 	.get_drvinfo		= efx_ef100_rep_get_drvinfo,
 	.get_msglevel		= efx_ef100_rep_ethtool_get_msglevel,
 	.set_msglevel		= efx_ef100_rep_ethtool_set_msglevel,
+	.get_ringparam		= efx_ef100_rep_ethtool_get_ringparam,
+	.set_ringparam		= efx_ef100_rep_ethtool_set_ringparam,
 };
 
 static struct efx_rep *efx_ef100_rep_create_netdev(struct efx_nic *efx,