[PATCH v4 10/10] net/mlx5: accept more unicast MAC addresses
David Marchand
david.marchand at redhat.com
Thu Jul 9 18:02:46 CEST 2026
Starting firmware version 22.49.1014, the number of mac addresses
per VF is not capped to 128 anymore.
The value can be increased via devlink:
$ devlink dev param set pci/0000:3b:00.2 name max_macs value 4096 \
cmode driverinit
$ devlink dev reload pci/0000:3b:00.2
On the DPDK side, we must retrieve the maximum number of unicast
and multicast addresses supported with a query to the firmware.
Then, dynamically allocate the mac addresses arrays and report the
limit instead of the previous hardcoded value.
Signed-off-by: David Marchand <david.marchand at redhat.com>
---
drivers/common/mlx5/mlx5_devx_cmds.c | 4 +++
drivers/common/mlx5/mlx5_devx_cmds.h | 2 ++
drivers/net/mlx5/linux/mlx5_os.c | 42 ++++++++++++++++++++++------
drivers/net/mlx5/mlx5.c | 9 ++----
drivers/net/mlx5/mlx5.h | 8 ++++--
drivers/net/mlx5/mlx5_ethdev.c | 2 +-
drivers/net/mlx5/mlx5_mac.c | 22 +++++++++------
drivers/net/mlx5/mlx5_trigger.c | 10 +++----
drivers/net/mlx5/windows/mlx5_os.c | 40 ++++++++++++++++++++------
9 files changed, 99 insertions(+), 40 deletions(-)
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.c b/drivers/common/mlx5/mlx5_devx_cmds.c
index 140b057ab4..e5d9c92779 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.c
+++ b/drivers/common/mlx5/mlx5_devx_cmds.c
@@ -1110,6 +1110,10 @@ mlx5_devx_cmd_query_hca_attr(void *ctx,
attr->log_max_pd = MLX5_GET(cmd_hca_cap, hcattr, log_max_pd);
attr->log_max_srq = MLX5_GET(cmd_hca_cap, hcattr, log_max_srq);
attr->log_max_srq_sz = MLX5_GET(cmd_hca_cap, hcattr, log_max_srq_sz);
+ attr->log_max_current_uc_list = MLX5_GET(cmd_hca_cap, hcattr,
+ log_max_current_uc_list);
+ attr->log_max_current_mc_list = MLX5_GET(cmd_hca_cap, hcattr,
+ log_max_current_mc_list);
attr->reg_c_preserve =
MLX5_GET(cmd_hca_cap, hcattr, reg_c_preserve);
attr->mmo_regex_qp_en = MLX5_GET(cmd_hca_cap, hcattr, regexp_mmo_qp);
diff --git a/drivers/common/mlx5/mlx5_devx_cmds.h b/drivers/common/mlx5/mlx5_devx_cmds.h
index 90beb2e9e6..7fe89bc6a4 100644
--- a/drivers/common/mlx5/mlx5_devx_cmds.h
+++ b/drivers/common/mlx5/mlx5_devx_cmds.h
@@ -356,6 +356,8 @@ struct mlx5_hca_attr {
uint8_t tx_sw_owner_v2:1;
uint8_t esw_sw_owner:1;
uint8_t esw_sw_owner_v2:1;
+ uint32_t log_max_current_uc_list:5;
+ uint32_t log_max_current_mc_list:5;
};
/* LAG Context. */
diff --git a/drivers/net/mlx5/linux/mlx5_os.c b/drivers/net/mlx5/linux/mlx5_os.c
index 1cad6e1091..23632f55af 100644
--- a/drivers/net/mlx5/linux/mlx5_os.c
+++ b/drivers/net/mlx5/linux/mlx5_os.c
@@ -389,6 +389,16 @@ mlx5_os_capabilities_prepare(struct mlx5_dev_ctx_shared *sh)
sh->dev_cap.esw_info.regc_mask = 0;
#endif
sh->dev_cap.esw_info.is_set = 1;
+ if (hca_attr->log_max_current_uc_list > 0)
+ sh->dev_cap.max_uc_mac_addrs = 1u << hca_attr->log_max_current_uc_list;
+ else
+ sh->dev_cap.max_uc_mac_addrs = MLX5_MAX_UC_MAC_ADDRESSES;
+ if (hca_attr->log_max_current_mc_list > 0)
+ sh->dev_cap.max_mc_mac_addrs = 1u << hca_attr->log_max_current_mc_list;
+ else
+ sh->dev_cap.max_mc_mac_addrs = MLX5_MAX_MC_MAC_ADDRESSES;
+ sh->dev_cap.max_mac_addrs =
+ sh->dev_cap.max_uc_mac_addrs + sh->dev_cap.max_mc_mac_addrs;
return 0;
}
@@ -1464,6 +1474,22 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
priv->sh = sh;
priv->dev_port = spawn->phys_port;
priv->pci_dev = spawn->pci_dev;
+ priv->mac = mlx5_malloc(MLX5_MEM_ZERO | MLX5_MEM_RTE,
+ sizeof(*priv->mac) * sh->dev_cap.max_mac_addrs,
+ RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+ if (priv->mac == NULL) {
+ DRV_LOG(ERR, "Failed to allocate MAC address array.");
+ err = ENOMEM;
+ goto error;
+ }
+ priv->mac_own = mlx5_malloc(MLX5_MEM_ZERO | MLX5_MEM_RTE,
+ RTE_BITSET_SIZE(sh->dev_cap.max_mac_addrs),
+ RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+ if (priv->mac_own == NULL) {
+ DRV_LOG(ERR, "Failed to allocate MAC ownership bitmap.");
+ err = ENOMEM;
+ goto error;
+ }
/* Some internal functions rely on Netlink sockets, open them now. */
priv->nl_socket_rdma = nl_rdma;
priv->nl_socket_route = mlx5_nl_init(NETLINK_ROUTE, 0);
@@ -1762,8 +1788,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
mlx5_nl_mac_addr_sync(priv->nl_socket_route,
mlx5_ifindex(eth_dev),
eth_dev->data->mac_addrs,
- MLX5_MAX_UC_MAC_ADDRESSES,
- MLX5_MAX_MAC_ADDRESSES);
+ sh->dev_cap.max_mac_addrs,
+ sh->dev_cap.max_uc_mac_addrs);
priv->ctrl_flows = 0;
rte_spinlock_init(&priv->flow_list_lock);
TAILQ_INIT(&priv->flow_meters);
@@ -1963,17 +1989,15 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
mlx5_flex_item_port_cleanup(eth_dev);
mlx5_free(priv->ext_rxqs);
mlx5_free(priv->ext_txqs);
+ mlx5_free(priv->mac);
+ eth_dev->data->mac_addrs = NULL;
+ mlx5_free(priv->mac_own);
mlx5_free(priv);
if (eth_dev != NULL)
eth_dev->data->dev_private = NULL;
}
- if (eth_dev != NULL) {
- /* mac_addrs must not be freed alone because part of
- * dev_private
- **/
- eth_dev->data->mac_addrs = NULL;
+ if (eth_dev != NULL)
rte_eth_dev_release_port(eth_dev);
- }
if (sh)
mlx5_free_shared_dev_ctx(sh);
if (nl_rdma >= 0)
@@ -3497,7 +3521,7 @@ mlx5_os_mac_addr_flush(struct rte_eth_dev *dev)
const int vf = priv->sh->dev_cap.vf;
int i;
- for (i = MLX5_MAX_MAC_ADDRESSES - 1; i >= 0; --i) {
+ for (i = priv->sh->dev_cap.max_mac_addrs - 1; i >= 0; --i) {
if (rte_bitset_test(priv->mac_own, i)) {
if (vf)
mlx5_nl_mac_addr_remove(priv->nl_socket_route,
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index 61c26d1206..1323ba49ac 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -2546,6 +2546,9 @@ mlx5_dev_close(struct rte_eth_dev *dev)
mlx5_list_destroy(priv->hrxqs);
mlx5_free(priv->ext_rxqs);
mlx5_free(priv->ext_txqs);
+ mlx5_free(priv->mac);
+ dev->data->mac_addrs = NULL;
+ mlx5_free(priv->mac_own);
sh->port[priv->dev_port - 1].nl_ih_port_id = RTE_MAX_ETHPORTS;
/*
* The interrupt handler port id must be reset before priv is reset
@@ -2580,12 +2583,6 @@ mlx5_dev_close(struct rte_eth_dev *dev)
mlx5_flow_pools_destroy(priv);
memset(priv, 0, sizeof(*priv));
priv->domain_id = RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID;
- /*
- * Reset mac_addrs to NULL such that it is not freed as part of
- * rte_eth_dev_release_port(). mac_addrs is part of dev_private so
- * it is freed when dev_private is freed.
- */
- dev->data->mac_addrs = NULL;
return 0;
}
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index f7ba8df108..1fcc40fade 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -217,6 +217,9 @@ struct mlx5_dev_cap {
} mprq; /* Capability for Multi-Packet RQ. */
char fw_ver[64]; /* Firmware version of this device. */
struct flow_hw_port_info esw_info; /* E-switch manager reg_c0. */
+ uint16_t max_uc_mac_addrs; /* Maximum unicast MAC addresses. */
+ uint16_t max_mc_mac_addrs; /* Maximum multicast MAC addresses. */
+ uint16_t max_mac_addrs; /* Total maximum MAC addresses. */
};
#define MLX5_MPESW_PORT_INVALID (-1)
@@ -2018,9 +2021,8 @@ struct mlx5_priv {
struct mlx5_dev_ctx_shared *sh; /* Shared device context. */
uint32_t dev_port; /* Device port number. */
struct rte_pci_device *pci_dev; /* Backend PCI device. */
- struct rte_ether_addr mac[MLX5_MAX_MAC_ADDRESSES]; /* MAC addresses. */
- RTE_BITSET_DECLARE(mac_own, MLX5_MAX_MAC_ADDRESSES);
- /* Bit-field of MAC addresses owned by the PMD. */
+ struct rte_ether_addr *mac; /* MAC addresses. */
+ uint64_t *mac_own; /* Bit-field of MAC addresses owned by the PMD. */
uint16_t vlan_filter[MLX5_MAX_VLAN_IDS]; /* VLAN filters table. */
unsigned int vlan_filter_n; /* Number of configured VLAN filters. */
/* Device properties. */
diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index 8160d10e7e..306c1cd734 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -392,7 +392,7 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
max = RTE_MIN(max, (unsigned int)UINT16_MAX);
info->max_rx_queues = max;
info->max_tx_queues = max;
- info->max_mac_addrs = MLX5_MAX_UC_MAC_ADDRESSES;
+ info->max_mac_addrs = priv->sh->dev_cap.max_uc_mac_addrs;
info->rx_queue_offload_capa = mlx5_get_rx_queue_offloads(dev);
info->rx_seg_capa.max_nseg = MLX5_MAX_RXQ_NSEG;
info->rx_seg_capa.multi_pools = !priv->config.mprq.enabled;
diff --git a/drivers/net/mlx5/mlx5_mac.c b/drivers/net/mlx5/mlx5_mac.c
index 0e5d2be530..2ce7cfd407 100644
--- a/drivers/net/mlx5/mlx5_mac.c
+++ b/drivers/net/mlx5/mlx5_mac.c
@@ -36,7 +36,9 @@ mlx5_internal_mac_addr_remove(struct rte_eth_dev *dev,
uint32_t index,
struct rte_ether_addr *addr)
{
- MLX5_ASSERT(index < MLX5_MAX_MAC_ADDRESSES);
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+ MLX5_ASSERT(index < priv->sh->dev_cap.max_mac_addrs);
if (rte_is_zero_ether_addr(&dev->data->mac_addrs[index]))
return false;
mlx5_os_mac_addr_remove(dev, index);
@@ -63,16 +65,17 @@ static int
mlx5_internal_mac_addr_add(struct rte_eth_dev *dev, struct rte_ether_addr *mac,
uint32_t index)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
unsigned int i;
int ret;
- MLX5_ASSERT(index < MLX5_MAX_MAC_ADDRESSES);
+ MLX5_ASSERT(index < priv->sh->dev_cap.max_mac_addrs);
if (rte_is_zero_ether_addr(mac)) {
rte_errno = EINVAL;
return -rte_errno;
}
/* First, make sure this address isn't already configured. */
- for (i = 0; (i != MLX5_MAX_MAC_ADDRESSES); ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
/* Skip this index, it's going to be reconfigured. */
if (i == index)
continue;
@@ -101,10 +104,11 @@ mlx5_internal_mac_addr_add(struct rte_eth_dev *dev, struct rte_ether_addr *mac,
void
mlx5_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
struct rte_ether_addr addr = { 0 };
int ret;
- if (index >= MLX5_MAX_UC_MAC_ADDRESSES)
+ if (index >= priv->sh->dev_cap.max_uc_mac_addrs)
return;
if (mlx5_internal_mac_addr_remove(dev, index, &addr)) {
ret = mlx5_traffic_mac_remove(dev, &addr);
@@ -133,9 +137,10 @@ int
mlx5_mac_addr_add(struct rte_eth_dev *dev, struct rte_ether_addr *mac,
uint32_t index, uint32_t vmdq __rte_unused)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
int ret;
- if (index >= MLX5_MAX_UC_MAC_ADDRESSES) {
+ if (index >= priv->sh->dev_cap.max_uc_mac_addrs) {
rte_errno = EINVAL;
return -rte_errno;
}
@@ -217,16 +222,17 @@ int
mlx5_set_mc_addr_list(struct rte_eth_dev *dev,
struct rte_ether_addr *mc_addr_set, uint32_t nb_mc_addr)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
uint32_t i;
int ret;
- if (nb_mc_addr >= MLX5_MAX_MC_MAC_ADDRESSES) {
+ if (nb_mc_addr >= priv->sh->dev_cap.max_mc_mac_addrs) {
rte_errno = ENOSPC;
return -rte_errno;
}
- for (i = MLX5_MAX_UC_MAC_ADDRESSES; i != MLX5_MAX_MAC_ADDRESSES; ++i)
+ for (i = priv->sh->dev_cap.max_uc_mac_addrs; i != priv->sh->dev_cap.max_mac_addrs; ++i)
mlx5_internal_mac_addr_remove(dev, i, NULL);
- i = MLX5_MAX_UC_MAC_ADDRESSES;
+ i = priv->sh->dev_cap.max_uc_mac_addrs;
while (nb_mc_addr--) {
ret = mlx5_internal_mac_addr_add(dev, mc_addr_set++, i++);
if (ret)
diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
index e41e4643e0..e535e3e5be 100644
--- a/drivers/net/mlx5/mlx5_trigger.c
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -1902,7 +1902,7 @@ mlx5_traffic_enable(struct rte_eth_dev *dev)
}
}
/* Add MAC address flows. */
- for (i = 0; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
struct rte_ether_addr *mac = &dev->data->mac_addrs[i];
/* Add flows for unicast and multicast mac addresses added by API. */
@@ -2172,7 +2172,7 @@ mlx5_traffic_vlan_add(struct rte_eth_dev *dev, const uint16_t vid)
return 0;
/* Add all unicast DMAC flow rules with new VLAN attached. */
- for (i = 0; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
struct rte_ether_addr *mac = &dev->data->mac_addrs[i];
if (rte_is_zero_ether_addr(mac))
@@ -2189,7 +2189,7 @@ mlx5_traffic_vlan_add(struct rte_eth_dev *dev, const uint16_t vid)
* Removing after creating VLAN rules so that traffic "gap" is not introduced.
*/
- for (i = 0; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
struct rte_ether_addr *mac = &dev->data->mac_addrs[i];
if (rte_is_zero_ether_addr(mac))
@@ -2227,7 +2227,7 @@ mlx5_traffic_vlan_remove(struct rte_eth_dev *dev, const uint16_t vid)
* Recreating first to ensure no traffic "gap".
*/
- for (i = 0; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
struct rte_ether_addr *mac = &dev->data->mac_addrs[i];
if (rte_is_zero_ether_addr(mac))
@@ -2240,7 +2240,7 @@ mlx5_traffic_vlan_remove(struct rte_eth_dev *dev, const uint16_t vid)
}
/* Remove all unicast DMAC flow rules with this VLAN. */
- for (i = 0; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
+ for (i = 0; i != priv->sh->dev_cap.max_mac_addrs; ++i) {
struct rte_ether_addr *mac = &dev->data->mac_addrs[i];
if (rte_is_zero_ether_addr(mac))
diff --git a/drivers/net/mlx5/windows/mlx5_os.c b/drivers/net/mlx5/windows/mlx5_os.c
index 15de5c22a9..30d99b8b64 100644
--- a/drivers/net/mlx5/windows/mlx5_os.c
+++ b/drivers/net/mlx5/windows/mlx5_os.c
@@ -261,6 +261,16 @@ mlx5_os_capabilities_prepare(struct mlx5_dev_ctx_shared *sh)
MLX5_GET(initial_seg, pv_iseg, fw_rev_subminor));
DRV_LOG(DEBUG, "Packet pacing is not supported.");
mlx5_rt_timestamp_config(sh, hca_attr);
+ if (hca_attr->log_max_current_uc_list > 0)
+ sh->dev_cap.max_uc_mac_addrs = 1u << hca_attr->log_max_current_uc_list;
+ else
+ sh->dev_cap.max_uc_mac_addrs = MLX5_MAX_UC_MAC_ADDRESSES;
+ if (hca_attr->log_max_current_mc_list > 0)
+ sh->dev_cap.max_mc_mac_addrs = 1u << hca_attr->log_max_current_mc_list;
+ else
+ sh->dev_cap.max_mc_mac_addrs = MLX5_MAX_MC_MAC_ADDRESSES;
+ sh->dev_cap.max_mac_addrs =
+ sh->dev_cap.max_uc_mac_addrs + sh->dev_cap.max_mc_mac_addrs;
return 0;
}
@@ -396,6 +406,22 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
priv->sh = sh;
priv->dev_port = spawn->phys_port;
priv->pci_dev = spawn->pci_dev;
+ priv->mac = mlx5_malloc(MLX5_MEM_ZERO | MLX5_MEM_RTE,
+ sizeof(*priv->mac) * sh->dev_cap.max_mac_addrs,
+ RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+ if (priv->mac == NULL) {
+ DRV_LOG(ERR, "Failed to allocate MAC address array.");
+ err = ENOMEM;
+ goto error;
+ }
+ priv->mac_own = mlx5_malloc(MLX5_MEM_ZERO | MLX5_MEM_RTE,
+ RTE_BITSET_SIZE(sh->dev_cap.max_mac_addrs),
+ RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+ if (priv->mac_own == NULL) {
+ DRV_LOG(ERR, "Failed to allocate MAC ownership bitmap.");
+ err = ENOMEM;
+ goto error;
+ }
priv->mp_id.port_id = port_id;
strlcpy(priv->mp_id.name, MLX5_MP_NAME, RTE_MP_MAX_NAME_LEN);
priv->representor = !!switch_info->representor;
@@ -612,17 +638,15 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
mlx5_l3t_destroy(priv->mtr_profile_tbl);
if (own_domain_id)
claim_zero(rte_eth_switch_domain_free(priv->domain_id));
+ mlx5_free(priv->mac);
+ eth_dev->data->mac_addrs = NULL;
+ mlx5_free(priv->mac_own);
mlx5_free(priv);
if (eth_dev != NULL)
eth_dev->data->dev_private = NULL;
}
- if (eth_dev != NULL) {
- /* mac_addrs must not be freed alone because part of
- * dev_private
- **/
- eth_dev->data->mac_addrs = NULL;
+ if (eth_dev != NULL)
rte_eth_dev_release_port(eth_dev);
- }
if (sh)
mlx5_free_shared_dev_ctx(sh);
MLX5_ASSERT(err > 0);
@@ -698,7 +722,7 @@ mlx5_os_mac_addr_flush(struct rte_eth_dev *dev)
struct mlx5_priv *priv = dev->data->dev_private;
int i;
- for (i = MLX5_MAX_MAC_ADDRESSES - 1; i >= 0; --i) {
+ for (i = priv->sh->dev_cap.max_mac_addrs - 1; i >= 0; --i) {
if (rte_bitset_test(priv->mac_own, i))
rte_bitset_clear(priv->mac_own, i);
}
@@ -718,7 +742,7 @@ mlx5_os_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
{
struct mlx5_priv *priv = dev->data->dev_private;
- if (index < MLX5_MAX_MAC_ADDRESSES)
+ if (index < priv->sh->dev_cap.max_mac_addrs)
rte_bitset_clear(priv->mac_own, index);
}
--
2.54.0
More information about the dev
mailing list