@@ -134,6 +134,8 @@ static const struct eth_dev_ops mlx5_dev_ops = {
.tx_queue_setup = mlx5_tx_queue_setup,
.rx_queue_release = mlx5_rx_queue_release,
.tx_queue_release = mlx5_tx_queue_release,
+ .mac_addr_remove = mlx5_mac_addr_remove,
+ .mac_addr_add = mlx5_mac_addr_add,
};
static struct {
@@ -390,7 +392,7 @@ mlx5_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev)
claim_zero(priv_mac_addr_add(priv, 0,
(const uint8_t (*)[ETHER_ADDR_LEN])
mac.addr_bytes));
- claim_zero(priv_mac_addr_add(priv, 1,
+ claim_zero(priv_mac_addr_add(priv, (RTE_DIM(priv->mac) - 1),
&(const uint8_t [ETHER_ADDR_LEN])
{ "\xff\xff\xff\xff\xff\xff" }));
#ifndef NDEBUG
@@ -173,7 +173,12 @@ int mlx5_ibv_device_to_pci_addr(const struct ibv_device *,
/* mlx5_mac.c */
int priv_get_mac(struct priv *, uint8_t (*)[ETHER_ADDR_LEN]);
+void rxq_mac_addrs_del(struct rxq *);
+void mlx5_mac_addr_remove(struct rte_eth_dev *, uint32_t);
+int rxq_mac_addrs_add(struct rxq *);
int priv_mac_addr_add(struct priv *, unsigned int,
const uint8_t (*)[ETHER_ADDR_LEN]);
+void mlx5_mac_addr_add(struct rte_eth_dev *, struct ether_addr *, uint32_t,
+ uint32_t);
#endif /* RTE_PMD_MLX5_H_ */
@@ -65,6 +65,8 @@
#include "mlx5.h"
#include "mlx5_utils.h"
+#include "mlx5_rxtx.h"
+#include "mlx5_defs.h"
/**
* Get MAC address by querying netdevice.
@@ -89,8 +91,86 @@ priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN])
}
/**
+ * Delete flow steering rule.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param mac_index
+ * MAC address index.
+ * @param vlan_index
+ * VLAN index.
+ */
+static void
+rxq_del_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index)
+{
+#ifndef NDEBUG
+ struct priv *priv = rxq->priv;
+ const uint8_t (*mac)[ETHER_ADDR_LEN] =
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ priv->mac[mac_index].addr_bytes;
+#endif
+ assert(rxq->mac_flow[mac_index][vlan_index] != NULL);
+ DEBUG("%p: removing MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u"
+ " (VLAN ID %" PRIu16 ")",
+ (void *)rxq,
+ (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5],
+ mac_index, priv->vlan_filter[vlan_index].id);
+ claim_zero(ibv_destroy_flow(rxq->mac_flow[mac_index][vlan_index]));
+ rxq->mac_flow[mac_index][vlan_index] = NULL;
+}
+
+/**
+ * Unregister a MAC address from a RX queue.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param mac_index
+ * MAC address index.
+ */
+static void
+rxq_mac_addr_del(struct rxq *rxq, unsigned int mac_index)
+{
+ struct priv *priv = rxq->priv;
+ unsigned int i;
+ unsigned int vlans = 0;
+
+ assert(mac_index < RTE_DIM(priv->mac));
+ if (!BITFIELD_ISSET(rxq->mac_configured, mac_index))
+ return;
+ for (i = 0; (i != RTE_DIM(priv->vlan_filter)); ++i) {
+ if (!priv->vlan_filter[i].enabled)
+ continue;
+ rxq_del_flow(rxq, mac_index, i);
+ vlans++;
+ }
+ if (!vlans) {
+ rxq_del_flow(rxq, mac_index, 0);
+ }
+ BITFIELD_RESET(rxq->mac_configured, mac_index);
+}
+
+/**
+ * Unregister all MAC addresses from a RX queue.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ */
+void
+rxq_mac_addrs_del(struct rxq *rxq)
+{
+ struct priv *priv = rxq->priv;
+ unsigned int i;
+
+ for (i = 0; (i != RTE_DIM(priv->mac)); ++i)
+ rxq_mac_addr_del(rxq, i);
+}
+
+/**
* Unregister a MAC address.
*
+ * In RSS mode, the MAC address is unregistered from the parent queue,
+ * otherwise it is unregistered from each queue directly.
+ *
* @param priv
* Pointer to private structure.
* @param mac_index
@@ -99,15 +179,217 @@ priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN])
static void
priv_mac_addr_del(struct priv *priv, unsigned int mac_index)
{
+ unsigned int i;
+
assert(mac_index < RTE_DIM(priv->mac));
if (!BITFIELD_ISSET(priv->mac_configured, mac_index))
return;
+ if (priv->rss) {
+ rxq_mac_addr_del(&priv->rxq_parent, mac_index);
+ goto end;
+ }
+ for (i = 0; (i != priv->dev->data->nb_rx_queues); ++i)
+ rxq_mac_addr_del((*priv->rxqs)[i], mac_index);
+end:
BITFIELD_RESET(priv->mac_configured, mac_index);
}
/**
+ * DPDK callback to remove a MAC address.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param index
+ * MAC address index.
+ */
+void
+mlx5_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ priv_lock(priv);
+ DEBUG("%p: removing MAC address from index %" PRIu32,
+ (void *)dev, index);
+ /* Last array entry is reserved for broadcast. */
+ if (index >= (RTE_DIM(priv->mac) - 1))
+ goto end;
+ priv_mac_addr_del(priv, index);
+end:
+ priv_unlock(priv);
+}
+
+/**
+ * Add single flow steering rule.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param mac_index
+ * MAC address index to register.
+ * @param vlan_index
+ * VLAN index. Use -1 for a flow without VLAN.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+rxq_add_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index)
+{
+ struct ibv_flow *flow;
+ struct priv *priv = rxq->priv;
+ const uint8_t (*mac)[ETHER_ADDR_LEN] =
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ priv->mac[mac_index].addr_bytes;
+
+ /* Allocate flow specification on the stack. */
+ struct __attribute__((packed)) {
+ struct ibv_flow_attr attr;
+ struct ibv_flow_spec_eth spec;
+ } data;
+ struct ibv_flow_attr *attr = &data.attr;
+ struct ibv_flow_spec_eth *spec = &data.spec;
+
+ assert(mac_index < RTE_DIM(priv->mac));
+ assert((vlan_index < RTE_DIM(priv->vlan_filter)) || (vlan_index == -1u));
+ /*
+ * No padding must be inserted by the compiler between attr and spec.
+ * This layout is expected by libibverbs.
+ */
+ assert(((uint8_t *)attr + sizeof(*attr)) == (uint8_t *)spec);
+ *attr = (struct ibv_flow_attr){
+ .type = IBV_FLOW_ATTR_NORMAL,
+ .num_of_specs = 1,
+ .port = priv->port,
+ .flags = 0
+ };
+ *spec = (struct ibv_flow_spec_eth){
+ .type = IBV_FLOW_SPEC_ETH,
+ .size = sizeof(*spec),
+ .val = {
+ .dst_mac = {
+ (*mac)[0], (*mac)[1], (*mac)[2],
+ (*mac)[3], (*mac)[4], (*mac)[5]
+ },
+ .vlan_tag = ((vlan_index != -1u) ?
+ htons(priv->vlan_filter[vlan_index].id) :
+ 0),
+ },
+ .mask = {
+ .dst_mac = "\xff\xff\xff\xff\xff\xff",
+ .vlan_tag = ((vlan_index != -1u) ? htons(0xfff) : 0),
+ }
+ };
+ DEBUG("%p: adding MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u"
+ " (VLAN %s %" PRIu16 ")",
+ (void *)rxq,
+ (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5],
+ mac_index,
+ ((vlan_index != -1u) ? "ID" : "index"),
+ ((vlan_index != -1u) ? priv->vlan_filter[vlan_index].id : -1u));
+ /* Create related flow. */
+ errno = 0;
+ flow = ibv_create_flow(rxq->qp, attr);
+ if (flow == NULL) {
+ /* It's not clear whether errno is always set in this case. */
+ ERROR("%p: flow configuration failed, errno=%d: %s",
+ (void *)rxq, errno,
+ (errno ? strerror(errno) : "Unknown error"));
+ if (errno)
+ return errno;
+ return EINVAL;
+ }
+ if (vlan_index == -1u)
+ vlan_index = 0;
+ assert(rxq->mac_flow[mac_index][vlan_index] == NULL);
+ rxq->mac_flow[mac_index][vlan_index] = flow;
+ return 0;
+}
+
+/**
+ * Register a MAC address in a RX queue.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ * @param mac_index
+ * MAC address index to register.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+static int
+rxq_mac_addr_add(struct rxq *rxq, unsigned int mac_index)
+{
+ struct priv *priv = rxq->priv;
+ unsigned int i;
+ unsigned int vlans = 0;
+ int ret;
+
+ assert(mac_index < RTE_DIM(priv->mac));
+ if (BITFIELD_ISSET(rxq->mac_configured, mac_index))
+ rxq_mac_addr_del(rxq, mac_index);
+ /* Fill VLAN specifications. */
+ for (i = 0; (i != RTE_DIM(priv->vlan_filter)); ++i) {
+ if (!priv->vlan_filter[i].enabled)
+ continue;
+ /* Create related flow. */
+ ret = rxq_add_flow(rxq, mac_index, i);
+ if (!ret) {
+ vlans++;
+ continue;
+ }
+ /* Failure, rollback. */
+ while (i != 0)
+ if (priv->vlan_filter[--i].enabled)
+ rxq_del_flow(rxq, mac_index, i);
+ assert(ret > 0);
+ return ret;
+ }
+ /* In case there is no VLAN filter. */
+ if (!vlans) {
+ ret = rxq_add_flow(rxq, mac_index, -1);
+ if (ret)
+ return ret;
+ }
+ BITFIELD_SET(rxq->mac_configured, mac_index);
+ return 0;
+}
+
+/**
+ * Register all MAC addresses in a RX queue.
+ *
+ * @param rxq
+ * Pointer to RX queue structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+rxq_mac_addrs_add(struct rxq *rxq)
+{
+ struct priv *priv = rxq->priv;
+ unsigned int i;
+ int ret;
+
+ for (i = 0; (i != RTE_DIM(priv->mac)); ++i) {
+ if (!BITFIELD_ISSET(priv->mac_configured, i))
+ continue;
+ ret = rxq_mac_addr_add(rxq, i);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0)
+ rxq_mac_addr_del(rxq, --i);
+ assert(ret > 0);
+ return ret;
+ }
+ return 0;
+}
+
+/**
* Register a MAC address.
*
+ * In RSS mode, the MAC address is registered in the parent queue,
+ * otherwise it is registered in each queue directly.
+ *
* @param priv
* Pointer to private structure.
* @param mac_index
@@ -123,6 +405,7 @@ priv_mac_addr_add(struct priv *priv, unsigned int mac_index,
const uint8_t (*mac)[ETHER_ADDR_LEN])
{
unsigned int i;
+ int ret;
assert(mac_index < RTE_DIM(priv->mac));
/* First, make sure this address isn't already configured. */
@@ -145,6 +428,70 @@ priv_mac_addr_add(struct priv *priv, unsigned int mac_index,
(*mac)[3], (*mac)[4], (*mac)[5]
}
};
+ /* If device isn't started, this is all we need to do. */
+ if (!priv->started) {
+#ifndef NDEBUG
+ /* Verify that all queues have this index disabled. */
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ if ((*priv->rxqs)[i] == NULL)
+ continue;
+ assert(!BITFIELD_ISSET
+ ((*priv->rxqs)[i]->mac_configured, mac_index));
+ }
+#endif
+ goto end;
+ }
+ if (priv->rss) {
+ ret = rxq_mac_addr_add(&priv->rxq_parent, mac_index);
+ if (ret)
+ return ret;
+ goto end;
+ }
+ for (i = 0; (i != priv->rxqs_n); ++i) {
+ if ((*priv->rxqs)[i] == NULL)
+ continue;
+ ret = rxq_mac_addr_add((*priv->rxqs)[i], mac_index);
+ if (!ret)
+ continue;
+ /* Failure, rollback. */
+ while (i != 0)
+ if ((*priv->rxqs)[(--i)] != NULL)
+ rxq_mac_addr_del((*priv->rxqs)[i], mac_index);
+ return ret;
+ }
+end:
BITFIELD_SET(priv->mac_configured, mac_index);
return 0;
}
+
+/**
+ * DPDK callback to add a MAC address.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param mac_addr
+ * MAC address to register.
+ * @param index
+ * MAC address index.
+ * @param vmdq
+ * VMDq pool index to associate address with (ignored).
+ */
+void
+mlx5_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
+ uint32_t index, uint32_t vmdq)
+{
+ struct priv *priv = dev->data->dev_private;
+
+ (void)vmdq;
+ priv_lock(priv);
+ DEBUG("%p: adding MAC address at index %" PRIu32,
+ (void *)dev, index);
+ /* Last array entry is reserved for broadcast. */
+ if (index >= (RTE_DIM(priv->mac) - 1))
+ goto end;
+ priv_mac_addr_add(priv, index,
+ (const uint8_t (*)[ETHER_ADDR_LEN])
+ mac_addr->addr_bytes);
+end:
+ priv_unlock(priv);
+}
@@ -248,6 +248,7 @@ rxq_cleanup(struct rxq *rxq)
¶ms));
}
if (rxq->qp != NULL) {
+ rxq_mac_addrs_del(rxq);
claim_zero(ibv_destroy_qp(rxq->qp));
}
if (rxq->cq != NULL)
@@ -515,6 +516,15 @@ skip_mr:
(void *)dev, strerror(ret));
goto error;
}
+ if ((parent) || (!priv->rss)) {
+ /* Configure MAC and broadcast addresses. */
+ ret = rxq_mac_addrs_add(&tmpl);
+ if (ret) {
+ ERROR("%p: QP flow attachment failed: %s",
+ (void *)dev, strerror(ret));
+ goto error;
+ }
+ }
/* Allocate descriptors for RX queues, except for the RSS parent. */
if (parent)
goto skip_alloc;