Merge branch 'sfc-filter-chaining'

Edward Cree says:

====================
sfc: support for cascaded multicast filtering

Recent versions of firmware for SFC9100 adapters add support for filter
 chaining, in which packets matching multiple filters are delivered to all
 filters' recipients, rather than only the highest match-priority filter as was
 previously the case.
This patch series enables this feature and redesigns the filter handling code
 to make use of it; in particular, subscribing to a multicast address on one
 function no longer prevents traffic to that address reaching another function
 which is in promiscuous or allmulti mode.
If the firmware does not support filter chaining, the driver will fall back to
 the old behaviour.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

6 files changed, 2980 insertions, 979 deletions

diff --git a/drivers/net/ethernet/sfc/ef10.c b/drivers/net/ethernet/sfc/ef10.c
index 605cc8948594..8505d82290cb 100644
--- a/drivers/net/ethernet/sfc/ef10.c
+++ b/drivers/net/ethernet/sfc/ef10.c
@@ -49,6 +49,12 @@ enum {
  */
 #define HUNT_FILTER_TBL_ROWS 8192
 
+#define EFX_EF10_FILTER_ID_INVALID 0xffff
+struct efx_ef10_dev_addr {
+	u8 addr[ETH_ALEN];
+	u16 id;
+};
+
 struct efx_ef10_filter_table {
 /* The RX match field masks supported by this fw & hw, in order of priority */
 	enum efx_filter_match_flags rx_match_flags[
@@ -69,13 +75,14 @@ struct efx_ef10_filter_table {
 /* Shadow of net_device address lists, guarded by mac_lock */
 #define EFX_EF10_FILTER_DEV_UC_MAX	32
 #define EFX_EF10_FILTER_DEV_MC_MAX	256
-	struct {
-		u8 addr[ETH_ALEN];
-		u16 id;
-	} dev_uc_list[EFX_EF10_FILTER_DEV_UC_MAX],
-	  dev_mc_list[EFX_EF10_FILTER_DEV_MC_MAX];
-	int dev_uc_count;		/* negative for PROMISC */
-	int dev_mc_count;		/* negative for PROMISC/ALLMULTI */
+	struct efx_ef10_dev_addr dev_uc_list[EFX_EF10_FILTER_DEV_UC_MAX];
+	struct efx_ef10_dev_addr dev_mc_list[EFX_EF10_FILTER_DEV_MC_MAX];
+	int dev_uc_count;
+	int dev_mc_count;
+/* Indices (like efx_ef10_dev_addr.id) for promisc/allmulti filters */
+	u16 ucdef_id;
+	u16 bcast_id;
+	u16 mcdef_id;
 };
 
 /* An arbitrary search limit for the software hash table */
@@ -387,7 +394,7 @@ static int efx_ef10_probe(struct efx_nic *efx)
 	 * First try to enable it, then if we get EPERM, just
 	 * ask if it's already enabled
 	 */
-	rc = efx_mcdi_set_workaround(efx, MC_CMD_WORKAROUND_BUG35388, true);
+	rc = efx_mcdi_set_workaround(efx, MC_CMD_WORKAROUND_BUG35388, true, NULL);
 	if (rc == 0) {
 		nic_data->workaround_35388 = true;
 	} else if (rc == -EPERM) {
@@ -2197,6 +2204,29 @@ static int efx_ef10_ev_probe(struct efx_channel *channel)
 				    GFP_KERNEL);
 }
 
+static void efx_ef10_ev_fini(struct efx_channel *channel)
+{
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_FINI_EVQ_IN_LEN);
+	MCDI_DECLARE_BUF_ERR(outbuf);
+	struct efx_nic *efx = channel->efx;
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, FINI_EVQ_IN_INSTANCE, channel->channel);
+
+	rc = efx_mcdi_rpc_quiet(efx, MC_CMD_FINI_EVQ, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+
+	if (rc && rc != -EALREADY)
+		goto fail;
+
+	return;
+
+fail:
+	efx_mcdi_display_error(efx, MC_CMD_FINI_EVQ, MC_CMD_FINI_EVQ_IN_LEN,
+			       outbuf, outlen, rc);
+}
+
 static int efx_ef10_ev_init(struct efx_channel *channel)
 {
 	MCDI_DECLARE_BUF(inbuf,
@@ -2208,6 +2238,7 @@ static int efx_ef10_ev_init(struct efx_channel *channel)
 	struct efx_ef10_nic_data *nic_data;
 	bool supports_rx_merge;
 	size_t inlen, outlen;
+	unsigned int enabled, implemented;
 	dma_addr_t dma_addr;
 	int rc;
 	int i;
@@ -2248,30 +2279,52 @@ static int efx_ef10_ev_init(struct efx_channel *channel)
 	rc = efx_mcdi_rpc(efx, MC_CMD_INIT_EVQ, inbuf, inlen,
 			  outbuf, sizeof(outbuf), &outlen);
 	/* IRQ return is ignored */
-	return rc;
-}
-
-static void efx_ef10_ev_fini(struct efx_channel *channel)
-{
-	MCDI_DECLARE_BUF(inbuf, MC_CMD_FINI_EVQ_IN_LEN);
-	MCDI_DECLARE_BUF_ERR(outbuf);
-	struct efx_nic *efx = channel->efx;
-	size_t outlen;
-	int rc;
-
-	MCDI_SET_DWORD(inbuf, FINI_EVQ_IN_INSTANCE, channel->channel);
-
-	rc = efx_mcdi_rpc_quiet(efx, MC_CMD_FINI_EVQ, inbuf, sizeof(inbuf),
-			  outbuf, sizeof(outbuf), &outlen);
+	if (channel->channel || rc)
+		return rc;
 
-	if (rc && rc != -EALREADY)
+	/* Successfully created event queue on channel 0 */
+	rc = efx_mcdi_get_workarounds(efx, &implemented, &enabled);
+	if (rc == -ENOSYS) {
+		/* GET_WORKAROUNDS was implemented before the bug26807
+		 * workaround, thus the latter must be unavailable in this fw
+		 */
+		nic_data->workaround_26807 = false;
+		rc = 0;
+	} else if (rc) {
 		goto fail;
+	} else {
+		nic_data->workaround_26807 =
+			!!(enabled & MC_CMD_GET_WORKAROUNDS_OUT_BUG26807);
+
+		if (implemented & MC_CMD_GET_WORKAROUNDS_OUT_BUG26807 &&
+		    !nic_data->workaround_26807) {
+			unsigned int flags;
+
+			rc = efx_mcdi_set_workaround(efx,
+						     MC_CMD_WORKAROUND_BUG26807,
+						     true, &flags);
+
+			if (!rc) {
+				if (flags &
+				    1 << MC_CMD_WORKAROUND_EXT_OUT_FLR_DONE_LBN) {
+					netif_info(efx, drv, efx->net_dev,
+						   "other functions on NIC have been reset\n");
+					/* MC's boot count has incremented */
+					++nic_data->warm_boot_count;
+				}
+				nic_data->workaround_26807 = true;
+			} else if (rc == -EPERM) {
+				rc = 0;
+			}
+		}
+	}
 
-	return;
+	if (!rc)
+		return 0;
 
 fail:
-	efx_mcdi_display_error(efx, MC_CMD_FINI_EVQ, MC_CMD_FINI_EVQ_IN_LEN,
-			       outbuf, outlen, rc);
+	efx_ef10_ev_fini(channel);
+	return rc;
 }
 
 static void efx_ef10_ev_remove(struct efx_channel *channel)
@@ -3225,6 +3278,19 @@ static int efx_ef10_filter_remove_safe(struct efx_nic *efx,
 					       filter_id, false);
 }
 
+static u32 efx_ef10_filter_get_unsafe_id(struct efx_nic *efx, u32 filter_id)
+{
+	return filter_id % HUNT_FILTER_TBL_ROWS;
+}
+
+static int efx_ef10_filter_remove_unsafe(struct efx_nic *efx,
+					 enum efx_filter_priority priority,
+					 u32 filter_id)
+{
+	return efx_ef10_filter_remove_internal(efx, 1U << priority,
+					       filter_id, true);
+}
+
 static int efx_ef10_filter_get_safe(struct efx_nic *efx,
 				    enum efx_filter_priority priority,
 				    u32 filter_id, struct efx_filter_spec *spec)
@@ -3598,6 +3664,10 @@ static int efx_ef10_filter_table_probe(struct efx_nic *efx)
 		goto fail;
 	}
 
+	table->ucdef_id = EFX_EF10_FILTER_ID_INVALID;
+	table->bcast_id = EFX_EF10_FILTER_ID_INVALID;
+	table->mcdef_id = EFX_EF10_FILTER_ID_INVALID;
+
 	efx->filter_state = table;
 	init_waitqueue_head(&table->waitq);
 	return 0;
@@ -3700,145 +3770,233 @@ static void efx_ef10_filter_table_remove(struct efx_nic *efx)
 	kfree(table);
 }
 
-/* Caller must hold efx->filter_sem for read if race against
- * efx_ef10_filter_table_remove() is possible
- */
-static void efx_ef10_filter_sync_rx_mode(struct efx_nic *efx)
+#define EFX_EF10_FILTER_DO_MARK_OLD(id) \
+		if (id != EFX_EF10_FILTER_ID_INVALID) { \
+			filter_idx = efx_ef10_filter_get_unsafe_id(efx, id); \
+			WARN_ON(!table->entry[filter_idx].spec); \
+			table->entry[filter_idx].spec |= EFX_EF10_FILTER_FLAG_AUTO_OLD; \
+		}
+static void efx_ef10_filter_mark_old(struct efx_nic *efx)
 {
 	struct efx_ef10_filter_table *table = efx->filter_state;
-	struct net_device *net_dev = efx->net_dev;
-	struct efx_filter_spec spec;
-	bool remove_failed = false;
-	struct netdev_hw_addr *uc;
-	struct netdev_hw_addr *mc;
-	unsigned int filter_idx;
-	int i, n, rc;
-
-	if (!efx_dev_registered(efx))
-		return;
+	unsigned int filter_idx, i;
 
 	if (!table)
 		return;
 
 	/* Mark old filters that may need to be removed */
 	spin_lock_bh(&efx->filter_lock);
-	n = table->dev_uc_count < 0 ? 1 : table->dev_uc_count;
-	for (i = 0; i < n; i++) {
-		filter_idx = table->dev_uc_list[i].id % HUNT_FILTER_TBL_ROWS;
-		table->entry[filter_idx].spec |= EFX_EF10_FILTER_FLAG_AUTO_OLD;
-	}
-	n = table->dev_mc_count < 0 ? 1 : table->dev_mc_count;
-	for (i = 0; i < n; i++) {
-		filter_idx = table->dev_mc_list[i].id % HUNT_FILTER_TBL_ROWS;
-		table->entry[filter_idx].spec |= EFX_EF10_FILTER_FLAG_AUTO_OLD;
-	}
+	for (i = 0; i < table->dev_uc_count; i++)
+		EFX_EF10_FILTER_DO_MARK_OLD(table->dev_uc_list[i].id);
+	for (i = 0; i < table->dev_mc_count; i++)
+		EFX_EF10_FILTER_DO_MARK_OLD(table->dev_mc_list[i].id);
+	EFX_EF10_FILTER_DO_MARK_OLD(table->ucdef_id);
+	EFX_EF10_FILTER_DO_MARK_OLD(table->bcast_id);
+	EFX_EF10_FILTER_DO_MARK_OLD(table->mcdef_id);
 	spin_unlock_bh(&efx->filter_lock);
+}
+#undef EFX_EF10_FILTER_DO_MARK_OLD
 
-	/* Copy/convert the address lists; add the primary station
-	 * address and broadcast address
-	 */
-	netif_addr_lock_bh(net_dev);
-	if (net_dev->flags & IFF_PROMISC ||
-	    netdev_uc_count(net_dev) >= EFX_EF10_FILTER_DEV_UC_MAX) {
-		table->dev_uc_count = -1;
-	} else {
-		table->dev_uc_count = 1 + netdev_uc_count(net_dev);
-		ether_addr_copy(table->dev_uc_list[0].addr, net_dev->dev_addr);
-		i = 1;
-		netdev_for_each_uc_addr(uc, net_dev) {
-			ether_addr_copy(table->dev_uc_list[i].addr, uc->addr);
-			i++;
+static void efx_ef10_filter_uc_addr_list(struct efx_nic *efx, bool *promisc)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	struct net_device *net_dev = efx->net_dev;
+	struct netdev_hw_addr *uc;
+	int addr_count;
+	unsigned int i;
+
+	table->ucdef_id = EFX_EF10_FILTER_ID_INVALID;
+	addr_count = netdev_uc_count(net_dev);
+	if (net_dev->flags & IFF_PROMISC)
+		*promisc = true;
+	table->dev_uc_count = 1 + addr_count;
+	ether_addr_copy(table->dev_uc_list[0].addr, net_dev->dev_addr);
+	i = 1;
+	netdev_for_each_uc_addr(uc, net_dev) {
+		if (i >= EFX_EF10_FILTER_DEV_UC_MAX) {
+			*promisc = true;
+			break;
 		}
+		ether_addr_copy(table->dev_uc_list[i].addr, uc->addr);
+		table->dev_uc_list[i].id = EFX_EF10_FILTER_ID_INVALID;
+		i++;
 	}
-	if (net_dev->flags & (IFF_PROMISC | IFF_ALLMULTI) ||
-	    netdev_mc_count(net_dev) >= EFX_EF10_FILTER_DEV_MC_MAX) {
-		table->dev_mc_count = -1;
-	} else {
-		table->dev_mc_count = 1 + netdev_mc_count(net_dev);
-		eth_broadcast_addr(table->dev_mc_list[0].addr);
-		i = 1;
-		netdev_for_each_mc_addr(mc, net_dev) {
-			ether_addr_copy(table->dev_mc_list[i].addr, mc->addr);
-			i++;
+}
+
+static void efx_ef10_filter_mc_addr_list(struct efx_nic *efx, bool *promisc)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	struct net_device *net_dev = efx->net_dev;
+	struct netdev_hw_addr *mc;
+	unsigned int i, addr_count;
+
+	table->mcdef_id = EFX_EF10_FILTER_ID_INVALID;
+	table->bcast_id = EFX_EF10_FILTER_ID_INVALID;
+	if (net_dev->flags & (IFF_PROMISC | IFF_ALLMULTI))
+		*promisc = true;
+
+	addr_count = netdev_mc_count(net_dev);
+	i = 0;
+	netdev_for_each_mc_addr(mc, net_dev) {
+		if (i >= EFX_EF10_FILTER_DEV_MC_MAX) {
+			*promisc = true;
+			break;
 		}
+		ether_addr_copy(table->dev_mc_list[i].addr, mc->addr);
+		table->dev_mc_list[i].id = EFX_EF10_FILTER_ID_INVALID;
+		i++;
 	}
-	netif_addr_unlock_bh(net_dev);
 
-	/* Insert/renew unicast filters */
-	if (table->dev_uc_count >= 0) {
-		for (i = 0; i < table->dev_uc_count; i++) {
-			efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
-					   EFX_FILTER_FLAG_RX_RSS,
-					   0);
-			efx_filter_set_eth_local(&spec, EFX_FILTER_VID_UNSPEC,
-						 table->dev_uc_list[i].addr);
-			rc = efx_ef10_filter_insert(efx, &spec, true);
-			if (rc < 0) {
-				/* Fall back to unicast-promisc */
-				while (i--)
-					efx_ef10_filter_remove_safe(
+	table->dev_mc_count = i;
+}
+
+static int efx_ef10_filter_insert_addr_list(struct efx_nic *efx,
+					     bool multicast, bool rollback)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	struct efx_ef10_dev_addr *addr_list;
+	struct efx_filter_spec spec;
+	u8 baddr[ETH_ALEN];
+	unsigned int i, j;
+	int addr_count;
+	int rc;
+
+	if (multicast) {
+		addr_list = table->dev_mc_list;
+		addr_count = table->dev_mc_count;
+	} else {
+		addr_list = table->dev_uc_list;
+		addr_count = table->dev_uc_count;
+	}
+
+	/* Insert/renew filters */
+	for (i = 0; i < addr_count; i++) {
+		efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
+				   EFX_FILTER_FLAG_RX_RSS,
+				   0);
+		efx_filter_set_eth_local(&spec, EFX_FILTER_VID_UNSPEC,
+					 addr_list[i].addr);
+		rc = efx_ef10_filter_insert(efx, &spec, true);
+		if (rc < 0) {
+			if (rollback) {
+				netif_info(efx, drv, efx->net_dev,
+					   "efx_ef10_filter_insert failed rc=%d\n",
+					   rc);
+				/* Fall back to promiscuous */
+				for (j = 0; j < i; j++) {
+					if (addr_list[j].id == EFX_EF10_FILTER_ID_INVALID)
+						continue;
+					efx_ef10_filter_remove_unsafe(
 						efx, EFX_FILTER_PRI_AUTO,
-						table->dev_uc_list[i].id);
-				table->dev_uc_count = -1;
-				break;
+						addr_list[j].id);
+					addr_list[j].id = EFX_EF10_FILTER_ID_INVALID;
+				}
+				return rc;
+			} else {
+				/* mark as not inserted, and carry on */
+				rc = EFX_EF10_FILTER_ID_INVALID;
 			}
-			table->dev_uc_list[i].id = rc;
 		}
+		addr_list[i].id = efx_ef10_filter_get_unsafe_id(efx, rc);
 	}
-	if (table->dev_uc_count < 0) {
+
+	if (multicast && rollback) {
+		/* Also need an Ethernet broadcast filter */
 		efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
 				   EFX_FILTER_FLAG_RX_RSS,
 				   0);
-		efx_filter_set_uc_def(&spec);
+		eth_broadcast_addr(baddr);
+		efx_filter_set_eth_local(&spec, EFX_FILTER_VID_UNSPEC, baddr);
 		rc = efx_ef10_filter_insert(efx, &spec, true);
 		if (rc < 0) {
-			WARN_ON(1);
-			table->dev_uc_count = 0;
+			netif_warn(efx, drv, efx->net_dev,
+				   "Broadcast filter insert failed rc=%d\n", rc);
+			/* Fall back to promiscuous */
+			for (j = 0; j < i; j++) {
+				if (addr_list[j].id == EFX_EF10_FILTER_ID_INVALID)
+					continue;
+				efx_ef10_filter_remove_unsafe(
+					efx, EFX_FILTER_PRI_AUTO,
+					addr_list[j].id);
+				addr_list[j].id = EFX_EF10_FILTER_ID_INVALID;
+			}
+			return rc;
 		} else {
-			table->dev_uc_list[0].id = rc;
+			table->bcast_id = efx_ef10_filter_get_unsafe_id(efx, rc);
 		}
 	}
 
-	/* Insert/renew multicast filters */
-	if (table->dev_mc_count >= 0) {
-		for (i = 0; i < table->dev_mc_count; i++) {
+	return 0;
+}
+
+static int efx_ef10_filter_insert_def(struct efx_nic *efx, bool multicast,
+				      bool rollback)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	struct efx_ef10_nic_data *nic_data = efx->nic_data;
+	struct efx_filter_spec spec;
+	u8 baddr[ETH_ALEN];
+	int rc;
+
+	efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
+			   EFX_FILTER_FLAG_RX_RSS,
+			   0);
+
+	if (multicast)
+		efx_filter_set_mc_def(&spec);
+	else
+		efx_filter_set_uc_def(&spec);
+
+	rc = efx_ef10_filter_insert(efx, &spec, true);
+	if (rc < 0) {
+		netif_warn(efx, drv, efx->net_dev,
+			   "%scast mismatch filter insert failed rc=%d\n",
+			   multicast ? "Multi" : "Uni", rc);
+	} else if (multicast) {
+		table->mcdef_id = efx_ef10_filter_get_unsafe_id(efx, rc);
+		if (!nic_data->workaround_26807) {
+			/* Also need an Ethernet broadcast filter */
 			efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
 					   EFX_FILTER_FLAG_RX_RSS,
 					   0);
+			eth_broadcast_addr(baddr);
 			efx_filter_set_eth_local(&spec, EFX_FILTER_VID_UNSPEC,
-						 table->dev_mc_list[i].addr);
+						 baddr);
 			rc = efx_ef10_filter_insert(efx, &spec, true);
 			if (rc < 0) {
-				/* Fall back to multicast-promisc */
-				while (i--)
-					efx_ef10_filter_remove_safe(
-						efx, EFX_FILTER_PRI_AUTO,
-						table->dev_mc_list[i].id);
-				table->dev_mc_count = -1;
-				break;
+				netif_warn(efx, drv, efx->net_dev,
+					   "Broadcast filter insert failed rc=%d\n",
+					   rc);
+				if (rollback) {
+					/* Roll back the mc_def filter */
+					efx_ef10_filter_remove_unsafe(
+							efx, EFX_FILTER_PRI_AUTO,
+							table->mcdef_id);
+					table->mcdef_id = EFX_EF10_FILTER_ID_INVALID;
+					return rc;
+				}
+			} else {
+				table->bcast_id = efx_ef10_filter_get_unsafe_id(efx, rc);
 			}
-			table->dev_mc_list[i].id = rc;
-		}
-	}
-	if (table->dev_mc_count < 0) {
-		efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO,
-				   EFX_FILTER_FLAG_RX_RSS,
-				   0);
-		efx_filter_set_mc_def(&spec);
-		rc = efx_ef10_filter_insert(efx, &spec, true);
-		if (rc < 0) {
-			WARN_ON(1);
-			table->dev_mc_count = 0;
-		} else {
-			table->dev_mc_list[0].id = rc;
 		}
+		rc = 0;
+	} else {
+		table->ucdef_id = rc;
+		rc = 0;
 	}
+	return rc;
+}
+
+/* Remove filters that weren't renewed.  Since nothing else changes the AUTO_OLD
+ * flag or removes these filters, we don't need to hold the filter_lock while
+ * scanning for these filters.
+ */
+static void efx_ef10_filter_remove_old(struct efx_nic *efx)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	bool remove_failed = false;
+	int i;
 
-	/* Remove filters that weren't renewed.  Since nothing else
-	 * changes the AUTO_OLD flag or removes these filters, we
-	 * don't need to hold the filter_lock while scanning for
-	 * these filters.
-	 */
 	for (i = 0; i < HUNT_FILTER_TBL_ROWS; i++) {
 		if (ACCESS_ONCE(table->entry[i].spec) &
 		    EFX_EF10_FILTER_FLAG_AUTO_OLD) {
@@ -3917,6 +4075,87 @@ reset_nic:
 	return rc ? rc : rc2;
 }
 
+/* Caller must hold efx->filter_sem for read if race against
+ * efx_ef10_filter_table_remove() is possible
+ */
+static void efx_ef10_filter_sync_rx_mode(struct efx_nic *efx)
+{
+	struct efx_ef10_filter_table *table = efx->filter_state;
+	struct efx_ef10_nic_data *nic_data = efx->nic_data;
+	struct net_device *net_dev = efx->net_dev;
+	bool uc_promisc = false, mc_promisc = false;
+
+	if (!efx_dev_registered(efx))
+		return;
+
+	if (!table)
+		return;
+
+	efx_ef10_filter_mark_old(efx);
+
+	/* Copy/convert the address lists; add the primary station
+	 * address and broadcast address
+	 */
+	netif_addr_lock_bh(net_dev);
+	efx_ef10_filter_uc_addr_list(efx, &uc_promisc);
+	efx_ef10_filter_mc_addr_list(efx, &mc_promisc);
+	netif_addr_unlock_bh(net_dev);
+
+	/* Insert/renew unicast filters */
+	if (uc_promisc) {
+		efx_ef10_filter_insert_def(efx, false, false);
+		efx_ef10_filter_insert_addr_list(efx, false, false);
+	} else {
+		/* If any of the filters failed to insert, fall back to
+		 * promiscuous mode - add in the uc_def filter.  But keep
+		 * our individual unicast filters.
+		 */
+		if (efx_ef10_filter_insert_addr_list(efx, false, false))
+			efx_ef10_filter_insert_def(efx, false, false);
+	}
+
+	/* Insert/renew multicast filters */
+	/* If changing promiscuous state with cascaded multicast filters, remove
+	 * old filters first, so that packets are dropped rather than duplicated
+	 */
+	if (nic_data->workaround_26807 && efx->mc_promisc != mc_promisc)
+		efx_ef10_filter_remove_old(efx);
+	if (mc_promisc) {
+		if (nic_data->workaround_26807) {
+			/* If we failed to insert promiscuous filters, rollback
+			 * and fall back to individual multicast filters
+			 */
+			if (efx_ef10_filter_insert_def(efx, true, true)) {
+				/* Changing promisc state, so remove old filters */
+				efx_ef10_filter_remove_old(efx);
+				efx_ef10_filter_insert_addr_list(efx, true, false);
+			}
+		} else {
+			/* If we failed to insert promiscuous filters, don't
+			 * rollback.  Regardless, also insert the mc_list
+			 */
+			efx_ef10_filter_insert_def(efx, true, false);
+			efx_ef10_filter_insert_addr_list(efx, true, false);
+		}
+	} else {
+		/* If any filters failed to insert, rollback and fall back to
+		 * promiscuous mode - mc_def filter and maybe broadcast.  If
+		 * that fails, roll back again and insert as many of our
+		 * individual multicast filters as we can.
+		 */
+		if (efx_ef10_filter_insert_addr_list(efx, true, true)) {
+			/* Changing promisc state, so remove old filters */
+			if (nic_data->workaround_26807)
+				efx_ef10_filter_remove_old(efx);
+			if (efx_ef10_filter_insert_def(efx, true, true))
+				efx_ef10_filter_insert_addr_list(efx, true, false);
+		}
+	}
+
+	efx_ef10_filter_remove_old(efx);
+	efx->mc_promisc = mc_promisc;
+}
+
 static int efx_ef10_set_mac_address(struct efx_nic *efx)
 {
 	MCDI_DECLARE_BUF(inbuf, MC_CMD_VADAPTOR_SET_MAC_IN_LEN);
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index 81640f8bb811..98d172b04f71 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -1779,15 +1779,31 @@ int efx_mcdi_wol_filter_reset(struct efx_nic *efx)
 	return rc;
 }
 
-int efx_mcdi_set_workaround(struct efx_nic *efx, u32 type, bool enabled)
+int efx_mcdi_set_workaround(struct efx_nic *efx, u32 type, bool enabled,
+			    unsigned int *flags)
 {
 	MCDI_DECLARE_BUF(inbuf, MC_CMD_WORKAROUND_IN_LEN);
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_WORKAROUND_EXT_OUT_LEN);
+	size_t outlen;
+	int rc;
 
 	BUILD_BUG_ON(MC_CMD_WORKAROUND_OUT_LEN != 0);
 	MCDI_SET_DWORD(inbuf, WORKAROUND_IN_TYPE, type);
 	MCDI_SET_DWORD(inbuf, WORKAROUND_IN_ENABLED, enabled);
-	return efx_mcdi_rpc(efx, MC_CMD_WORKAROUND, inbuf, sizeof(inbuf),
-			    NULL, 0, NULL);
+	rc = efx_mcdi_rpc(efx, MC_CMD_WORKAROUND, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+
+	if (!flags)
+		return 0;
+
+	if (outlen >= MC_CMD_WORKAROUND_EXT_OUT_LEN)
+		*flags = MCDI_DWORD(outbuf, WORKAROUND_EXT_OUT_FLAGS);
+	else
+		*flags = 0;
+
+	return 0;
 }
 
 int efx_mcdi_get_workarounds(struct efx_nic *efx, unsigned int *impl_out,
@@ -1816,7 +1832,11 @@ int efx_mcdi_get_workarounds(struct efx_nic *efx, unsigned int *impl_out,
 	return 0;
 
 fail:
-	netif_err(efx, hw, efx->net_dev, "%s: failed rc=%d\n", __func__, rc);
+	/* Older firmware lacks GET_WORKAROUNDS and this isn't especially
+	 * terrifying.  The call site will have to deal with it though.
+	 */
+	netif_printk(efx, hw, rc == -ENOSYS ? KERN_DEBUG : KERN_ERR,
+		     efx->net_dev, "%s: failed rc=%d\n", __func__, rc);
 	return rc;
 }
 
diff --git a/drivers/net/ethernet/sfc/mcdi.h b/drivers/net/ethernet/sfc/mcdi.h
index 1838afe2da92..025d504c472b 100644
--- a/drivers/net/ethernet/sfc/mcdi.h
+++ b/drivers/net/ethernet/sfc/mcdi.h
@@ -346,7 +346,8 @@ void efx_mcdi_mac_pull_stats(struct efx_nic *efx);
 bool efx_mcdi_mac_check_fault(struct efx_nic *efx);
 enum reset_type efx_mcdi_map_reset_reason(enum reset_type reason);
 int efx_mcdi_reset(struct efx_nic *efx, enum reset_type method);
-int efx_mcdi_set_workaround(struct efx_nic *efx, u32 type, bool enabled);
+int efx_mcdi_set_workaround(struct efx_nic *efx, u32 type, bool enabled,
+			    unsigned int *flags);
 int efx_mcdi_get_workarounds(struct efx_nic *efx, unsigned int *impl_out,
 			     unsigned int *enabled_out);
 
diff --git a/drivers/net/ethernet/sfc/mcdi_pcol.h b/drivers/net/ethernet/sfc/mcdi_pcol.h
index 45fca9fc66b7..4cc772164a79 100644
--- a/drivers/net/ethernet/sfc/mcdi_pcol.h
+++ b/drivers/net/ethernet/sfc/mcdi_pcol.h
@@ -26,6 +26,10 @@
  * Unlike a warm boot, assume DMEM has been reloaded, so that
  * the MC persistent data must be reinitialised. */
 #define MC_FW_TEPID_BOOT_OK (16)
+/* We have entered the main firmware via recovery mode.  This
+ * means that MC persistent data must be reinitialised, but that
+ * we shouldn't touch PCIe config. */
+#define MC_FW_RECOVERY_MODE_PCIE_INIT_OK (32)
 /* BIST state has been initialized */
 #define MC_FW_BIST_INIT_OK (128)
 
@@ -169,6 +173,8 @@
 #define MC_CMD_ERR_EINTR 4
 /* I/O failure */
 #define MC_CMD_ERR_EIO 5
+/* Already exists */
+#define MC_CMD_ERR_EEXIST 6
 /* Try again */
 #define MC_CMD_ERR_EAGAIN 11
 /* Out of memory */
@@ -181,6 +187,10 @@
 #define MC_CMD_ERR_ENODEV 19
 /* Invalid argument to target */
 #define MC_CMD_ERR_EINVAL 22
+/* Broken pipe */
+#define MC_CMD_ERR_EPIPE 32
+/* Read-only */
+#define MC_CMD_ERR_EROFS 30
 /* Out of range */
 #define MC_CMD_ERR_ERANGE 34
 /* Non-recursive resource is already acquired */
@@ -226,6 +236,43 @@
 #define MC_CMD_ERR_SLAVE_NOT_PRESENT 0x100a
 /* The datapath is disabled. */
 #define MC_CMD_ERR_DATAPATH_DISABLED 0x100b
+/* The requesting client is not a function */
+#define MC_CMD_ERR_CLIENT_NOT_FN  0x100c
+/* The requested operation might require the
+   command to be passed between MCs, and the
+   transport doesn't support that.  Should
+   only ever been seen over the UART. */
+#define MC_CMD_ERR_TRANSPORT_NOPROXY 0x100d
+/* VLAN tag(s) exists */
+#define MC_CMD_ERR_VLAN_EXIST 0x100e
+/* No MAC address assigned to an EVB port */
+#define MC_CMD_ERR_NO_MAC_ADDR 0x100f
+/* Notifies the driver that the request has been relayed
+ * to an admin function for authorization. The driver should
+ * wait for a PROXY_RESPONSE event and then resend its request.
+ * This error code is followed by a 32-bit handle that
+ * helps matching it with the respective PROXY_RESPONSE event. */
+#define MC_CMD_ERR_PROXY_PENDING 0x1010
+#define MC_CMD_ERR_PROXY_PENDING_HANDLE_OFST 4
+/* The request cannot be passed for authorization because
+ * another request from the same function is currently being
+ * authorized. The drvier should try again later. */
+#define MC_CMD_ERR_PROXY_INPROGRESS 0x1011
+/* Returned by MC_CMD_PROXY_COMPLETE if the caller is not the function
+ * that has enabled proxying or BLOCK_INDEX points to a function that
+ * doesn't await an authorization. */
+#define MC_CMD_ERR_PROXY_UNEXPECTED 0x1012
+/* This code is currently only used internally in FW. Its meaning is that
+ * an operation failed due to lack of SR-IOV privilege.
+ * Normally it is translated to EPERM by send_cmd_err(),
+ * but it may also be used to trigger some special mechanism
+ * for handling such case, e.g. to relay the failed request
+ * to a designated admin function for authorization. */
+#define MC_CMD_ERR_NO_PRIVILEGE 0x1013
+/* Workaround 26807 could not be turned on/off because some functions
+ * have already installed filters. See the comment at
+ * MC_CMD_WORKAROUND_BUG26807. */
+#define MC_CMD_ERR_FILTERS_PRESENT 0x1014
 
 #define MC_CMD_ERR_CODE_OFST 0
 
@@ -275,6 +322,11 @@
 	 MC_CMD_DBIWROP_TYPEDEF_VALUE_OFST +		\
 	 (n) * MC_CMD_DBIWROP_TYPEDEF_LEN)
 
+/* This may be ORed with an EVB_PORT_ID_xxx constant to pass a non-default
+ * stack ID (which must be in the range 1-255) along with an EVB port ID.
+ */
+#define EVB_STACK_ID(n)  (((n) & 0xff) << 16)
+
 
 /* Version 2 adds an optional argument to error returns: the errno value
  * may be followed by the (0-based) number of the first argument that
@@ -394,6 +446,8 @@
 #define          MCDI_EVENT_AOE_BYTEBLASTER 0x9
 /* enum: DDR ECC status update */
 #define          MCDI_EVENT_AOE_DDR_ECC_STATUS 0xa
+/* enum: PTP status update */
+#define          MCDI_EVENT_AOE_PTP_STATUS 0xb
 #define        MCDI_EVENT_AOE_ERR_DATA_LBN 8
 #define        MCDI_EVENT_AOE_ERR_DATA_WIDTH 8
 #define        MCDI_EVENT_RX_ERR_RXQ_LBN 0
@@ -408,6 +462,16 @@
 #define        MCDI_EVENT_RX_FLUSH_RXQ_WIDTH 12
 #define        MCDI_EVENT_MC_REBOOT_COUNT_LBN 0
 #define        MCDI_EVENT_MC_REBOOT_COUNT_WIDTH 16
+#define        MCDI_EVENT_MUM_ERR_TYPE_LBN 0
+#define        MCDI_EVENT_MUM_ERR_TYPE_WIDTH 8
+/* enum: MUM failed to load - no valid image? */
+#define          MCDI_EVENT_MUM_NO_LOAD 0x1
+/* enum: MUM f/w reported an exception */
+#define          MCDI_EVENT_MUM_ASSERT 0x2
+/* enum: MUM not kicking watchdog */
+#define          MCDI_EVENT_MUM_WATCHDOG 0x3
+#define        MCDI_EVENT_MUM_ERR_DATA_LBN 8
+#define        MCDI_EVENT_MUM_ERR_DATA_WIDTH 8
 #define       MCDI_EVENT_DATA_LBN 0
 #define       MCDI_EVENT_DATA_WIDTH 32
 #define       MCDI_EVENT_SRC_LBN 36
@@ -416,6 +480,8 @@
 #define       MCDI_EVENT_EV_CODE_WIDTH 4
 #define       MCDI_EVENT_CODE_LBN 44
 #define       MCDI_EVENT_CODE_WIDTH 8
+/* enum: Event generated by host software */
+#define          MCDI_EVENT_SW_EVENT 0x0
 /* enum: Bad assert. */
 #define          MCDI_EVENT_CODE_BADSSERT 0x1
 /* enum: PM Notice. */
@@ -470,6 +536,14 @@
 #define          MCDI_EVENT_CODE_MC_BIST 0x19
 /* enum: PTP tick event providing current NIC time */
 #define          MCDI_EVENT_CODE_PTP_TIME 0x1a
+/* enum: MUM fault */
+#define          MCDI_EVENT_CODE_MUM 0x1b
+/* enum: notify the designated PF of a new authorization request */
+#define          MCDI_EVENT_CODE_PROXY_REQUEST 0x1c
+/* enum: notify a function that awaits an authorization that its request has
+ * been processed and it may now resend the command
+ */
+#define          MCDI_EVENT_CODE_PROXY_RESPONSE 0x1d
 /* enum: Artificial event generated by host and posted via MC for test
  * purposes.
  */
@@ -537,6 +611,33 @@
 /* For CODE_PTP_TIME events, bits 19-26 of the minor value of the PTP clock */
 #define       MCDI_EVENT_PTP_TIME_MINOR_26_19_LBN 36
 #define       MCDI_EVENT_PTP_TIME_MINOR_26_19_WIDTH 8
+/* For CODE_PTP_TIME events where report sync status is enabled, indicates
+ * whether the NIC clock has ever been set
+ */
+#define       MCDI_EVENT_PTP_TIME_NIC_CLOCK_VALID_LBN 36
+#define       MCDI_EVENT_PTP_TIME_NIC_CLOCK_VALID_WIDTH 1
+/* For CODE_PTP_TIME events where report sync status is enabled, indicates
+ * whether the NIC and System clocks are in sync
+ */
+#define       MCDI_EVENT_PTP_TIME_HOST_NIC_IN_SYNC_LBN 37
+#define       MCDI_EVENT_PTP_TIME_HOST_NIC_IN_SYNC_WIDTH 1
+/* For CODE_PTP_TIME events where report sync status is enabled, bits 21-26 of
+ * the minor value of the PTP clock
+ */
+#define       MCDI_EVENT_PTP_TIME_MINOR_26_21_LBN 38
+#define       MCDI_EVENT_PTP_TIME_MINOR_26_21_WIDTH 6
+#define       MCDI_EVENT_PROXY_REQUEST_BUFF_INDEX_OFST 0
+#define       MCDI_EVENT_PROXY_REQUEST_BUFF_INDEX_LBN 0
+#define       MCDI_EVENT_PROXY_REQUEST_BUFF_INDEX_WIDTH 32
+#define       MCDI_EVENT_PROXY_RESPONSE_HANDLE_OFST 0
+#define       MCDI_EVENT_PROXY_RESPONSE_HANDLE_LBN 0
+#define       MCDI_EVENT_PROXY_RESPONSE_HANDLE_WIDTH 32
+/* Zero means that the request has been completed or authorized, and the driver
+ * should resend it. A non-zero value means that the authorization has been
+ * denied, and gives the reason. Typically it will be EPERM.
+ */
+#define       MCDI_EVENT_PROXY_RESPONSE_RC_LBN 36
+#define       MCDI_EVENT_PROXY_RESPONSE_RC_WIDTH 8
 
 /* FCDI_EVENT structuredef */
 #define    FCDI_EVENT_LEN 8
@@ -581,6 +682,10 @@
 #define          FCDI_EVENT_CODE_PTP_TICK 0x7
 /* enum: ECC error counters */
 #define          FCDI_EVENT_CODE_DDR_ECC_STATUS 0x8
+/* enum: Current status of PTP */
+#define          FCDI_EVENT_CODE_PTP_STATUS 0x9
+/* enum: Port id config to map MC-FC port idx */
+#define          FCDI_EVENT_CODE_PORT_CONFIG 0xa
 #define       FCDI_EVENT_ASSERT_INSTR_ADDRESS_OFST 0
 #define       FCDI_EVENT_ASSERT_INSTR_ADDRESS_LBN 0
 #define       FCDI_EVENT_ASSERT_INSTR_ADDRESS_WIDTH 32
@@ -594,11 +699,24 @@
 #define       FCDI_EVENT_LINK_STATE_DATA_OFST 0
 #define       FCDI_EVENT_LINK_STATE_DATA_LBN 0
 #define       FCDI_EVENT_LINK_STATE_DATA_WIDTH 32
+#define       FCDI_EVENT_PTP_STATE_OFST 0
+#define          FCDI_EVENT_PTP_UNDEFINED 0x0 /* enum */
+#define          FCDI_EVENT_PTP_SETUP_FAILED 0x1 /* enum */
+#define          FCDI_EVENT_PTP_OPERATIONAL 0x2 /* enum */
+#define       FCDI_EVENT_PTP_STATE_LBN 0
+#define       FCDI_EVENT_PTP_STATE_WIDTH 32
 #define       FCDI_EVENT_DDR_ECC_STATUS_BANK_ID_LBN 36
 #define       FCDI_EVENT_DDR_ECC_STATUS_BANK_ID_WIDTH 8
 #define       FCDI_EVENT_DDR_ECC_STATUS_STATUS_OFST 0
 #define       FCDI_EVENT_DDR_ECC_STATUS_STATUS_LBN 0
 #define       FCDI_EVENT_DDR_ECC_STATUS_STATUS_WIDTH 32
+/* Index of MC port being referred to */
+#define       FCDI_EVENT_PORT_CONFIG_SRC_LBN 36
+#define       FCDI_EVENT_PORT_CONFIG_SRC_WIDTH 8
+/* FC Port index that matches the MC port index in SRC */
+#define       FCDI_EVENT_PORT_CONFIG_DATA_OFST 0
+#define       FCDI_EVENT_PORT_CONFIG_DATA_LBN 0
+#define       FCDI_EVENT_PORT_CONFIG_DATA_WIDTH 32
 
 /* FCDI_EXTENDED_EVENT_PPS structuredef: Extended FCDI event to send PPS events
  * to the MC. Note that this structure | is overlayed over a normal FCDI event
@@ -631,6 +749,90 @@
 #define       FCDI_EXTENDED_EVENT_PPS_TIMESTAMPS_LBN 64
 #define       FCDI_EXTENDED_EVENT_PPS_TIMESTAMPS_WIDTH 64
 
+/* MUM_EVENT structuredef */
+#define    MUM_EVENT_LEN 8
+#define       MUM_EVENT_CONT_LBN 32
+#define       MUM_EVENT_CONT_WIDTH 1
+#define       MUM_EVENT_LEVEL_LBN 33
+#define       MUM_EVENT_LEVEL_WIDTH 3
+/* enum: Info. */
+#define          MUM_EVENT_LEVEL_INFO  0x0
+/* enum: Warning. */
+#define          MUM_EVENT_LEVEL_WARN 0x1
+/* enum: Error. */
+#define          MUM_EVENT_LEVEL_ERR 0x2
+/* enum: Fatal. */
+#define          MUM_EVENT_LEVEL_FATAL 0x3
+#define       MUM_EVENT_DATA_OFST 0
+#define        MUM_EVENT_SENSOR_ID_LBN 0
+#define        MUM_EVENT_SENSOR_ID_WIDTH 8
+/*             Enum values, see field(s): */
+/*                MC_CMD_SENSOR_INFO/MC_CMD_SENSOR_INFO_OUT/MASK */
+#define        MUM_EVENT_SENSOR_STATE_LBN 8
+#define        MUM_EVENT_SENSOR_STATE_WIDTH 8
+#define        MUM_EVENT_PORT_PHY_READY_LBN 0
+#define        MUM_EVENT_PORT_PHY_READY_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_LINK_UP_LBN 1
+#define        MUM_EVENT_PORT_PHY_LINK_UP_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_TX_LOL_LBN 2
+#define        MUM_EVENT_PORT_PHY_TX_LOL_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_RX_LOL_LBN 3
+#define        MUM_EVENT_PORT_PHY_RX_LOL_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_TX_LOS_LBN 4
+#define        MUM_EVENT_PORT_PHY_TX_LOS_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_RX_LOS_LBN 5
+#define        MUM_EVENT_PORT_PHY_RX_LOS_WIDTH 1
+#define        MUM_EVENT_PORT_PHY_TX_FAULT_LBN 6
+#define        MUM_EVENT_PORT_PHY_TX_FAULT_WIDTH 1
+#define       MUM_EVENT_DATA_LBN 0
+#define       MUM_EVENT_DATA_WIDTH 32
+#define       MUM_EVENT_SRC_LBN 36
+#define       MUM_EVENT_SRC_WIDTH 8
+#define       MUM_EVENT_EV_CODE_LBN 60
+#define       MUM_EVENT_EV_CODE_WIDTH 4
+#define       MUM_EVENT_CODE_LBN 44
+#define       MUM_EVENT_CODE_WIDTH 8