sfc: use a dynamic m-port for representor RX and set it promisc

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>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

diff --git a/drivers/net/ethernet/sfc/ef100_nic.c b/drivers/net/ethernet/sfc/ef100_nic.c
index 5fe18b383e2058c29020d602cfe51ed01098e871..8061efdaf82ce184f2a38e398609c39cbdf39d9a 100644 (file)
--- 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 @@ fail_unspec:
 
 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 40b2af8bfb81c69ff4f1d55f45e147adb5f66aed..4d928839d2922e3647aac606f46f2c15b4accd70 100644 (file)
--- 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 ea87ec83e6184ac5d18d93733f1d52f8f740fbea..97627f5e36741f029b2b9788e93c262df3a5dae5 100644 (file)
--- 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 e9651f611750513a7e86e9df890a5a96ae7262f9..0369be4d8983f32476d04cca51caf1d9d58915be 100644 (file)
--- 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 1523be77b9db6c509e0492e23c60b52f5409de0b..4ff6586116ee134d423e98b25b22c78cd3ca234e 100644 (file)
--- 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 0fb01f73c56e9315873688c7e8104ae424884a96..0c0aeb91f5009dc4a1731f0215a07f9647635359 100644 (file)
--- 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 46c5101eaa8d9a34b69b9bc8e78d9fdddd4257e8..309123c6b3867059e56c1d768f3d9c5dd1e20feb 100644 (file)
--- 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);