[dpdk-dev] [PATCH 3/5] net/mlx5: use Netlink to add/remove MAC addresses

[Adrien Mazarguil]

On Tue, Mar 13, 2018 at 01:50:37PM +0100, Nelio Laranjeiro wrote:
> VF devices are not able to receive traffic unless it fully requests it
> though Netlink.  This will cause the request to be processed by the PF
> which will add/remove the MAC address to the VF table if the VF is trusted.
> 
> Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro at 6wind.com>

Same basic comments as on the previous patch:

- Check random capitalization in documentation and comments.
- "vf" => "nl" for functions, objects and files that aren't really
   VF-specific but involve Netlink.

More below.

> ---
>  drivers/net/mlx5/Makefile   |   3 +-
>  drivers/net/mlx5/mlx5.c     |   7 +
>  drivers/net/mlx5/mlx5.h     |   6 +
>  drivers/net/mlx5/mlx5_mac.c |  25 +++-
>  drivers/net/mlx5/mlx5_vf.c  | 315 ++++++++++++++++++++++++++++++++++++++++++++
>  5 files changed, 353 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/mlx5/Makefile b/drivers/net/mlx5/Makefile
> index afda4118f..bbda0d1ff 100644
> --- a/drivers/net/mlx5/Makefile
> +++ b/drivers/net/mlx5/Makefile
> @@ -59,6 +59,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_rss.c
>  SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_mr.c
>  SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_flow.c
>  SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_socket.c
> +SRCS-$(CONFIG_RTE_LIBRTE_MLX5_PMD) += mlx5_vf.c
>  
>  ifeq ($(CONFIG_RTE_LIBRTE_MLX5_DLOPEN_DEPS),y)
>  INSTALL-$(CONFIG_RTE_LIBRTE_MLX5_PMD)-lib += $(LIB_GLUE)
> @@ -84,7 +85,7 @@ LDLIBS += -libverbs -lmlx5
>  endif
>  LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
>  LDLIBS += -lrte_ethdev -lrte_net -lrte_kvargs
> -LDLIBS += -lrte_bus_pci
> +LDLIBS += -lrte_bus_pci -lrte_netlink
>  
>  # A few warnings cannot be avoided in external headers.
>  CFLAGS += -Wno-error=cast-qual
> diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
> index d5cc85d19..e966884bd 100644
> --- a/drivers/net/mlx5/mlx5.c
> +++ b/drivers/net/mlx5/mlx5.c
> @@ -205,6 +205,13 @@ mlx5_dev_close(struct rte_eth_dev *dev)
>  		rte_free(priv->reta_idx);
>  	if (priv->primary_socket)
>  		mlx5_socket_uninit(dev);
> +	if (priv->config.vf) {
> +		ret = mlx5_vf_mac_addr_flush(dev);
> +		if (ret)
> +			DRV_LOG(WARNING,
> +				"port %u some MAC address remains configured",
> +				dev->data->port_id);
> +	}
>  	ret = mlx5_hrxq_ibv_verify(dev);
>  	if (ret)
>  		DRV_LOG(WARNING, "port %u some hash Rx queue still remain",
> diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
> index 9191b2949..a4333e6a5 100644
> --- a/drivers/net/mlx5/mlx5.h
> +++ b/drivers/net/mlx5/mlx5.h
> @@ -298,4 +298,10 @@ struct mlx5_mr *mlx5_mr_get(struct rte_eth_dev *dev, struct rte_mempool *mp);
>  int mlx5_mr_release(struct mlx5_mr *mr);
>  int mlx5_mr_verify(struct rte_eth_dev *dev);
>  
> +/* mlx5_vf.c */
> +
> +int mlx5_vf_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac);
> +int mlx5_vf_mac_addr_remove(struct rte_eth_dev *dev, struct ether_addr *mac);
> +int mlx5_vf_mac_addr_flush(struct rte_eth_dev *dev);
> +
>  #endif /* RTE_PMD_MLX5_H_ */
> diff --git a/drivers/net/mlx5/mlx5_mac.c b/drivers/net/mlx5/mlx5_mac.c
> index 01c7ba17a..5137ad11d 100644
> --- a/drivers/net/mlx5/mlx5_mac.c
> +++ b/drivers/net/mlx5/mlx5_mac.c
> @@ -67,11 +67,24 @@ mlx5_get_mac(struct rte_eth_dev *dev, uint8_t (*mac)[ETHER_ADDR_LEN])
>  void
>  mlx5_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
>  {
> +	struct priv *priv = dev->data->dev_private;
> +	const int vf = priv->config.vf;
> +	int ret;
> +
>  	assert(index < MLX5_MAX_MAC_ADDRESSES);
> +	if (index && vf) {

I don't think checking index is necessary here. mlx5_vf_mac_addr_remove()
will fail if the MAC in question happens to be the last one configured on
the netdevice, but depending on external configuration, it's not necessarily
the same as mac_addrs[0].

Checking "vf" only seems more reliable.

> +		ret = mlx5_vf_mac_addr_remove(dev,
> +					      &dev->data->mac_addrs[index]);
> +		if (ret) {
> +			DRV_LOG(WARNING,
> +				"port %u cannot remove mac address at index %d",
> +				dev->data->port_id, index);
> +			return;
> +		}
> +	}
>  	memset(&dev->data->mac_addrs[index], 0, sizeof(struct ether_addr));
>  	if (!dev->data->promiscuous) {
> -		int ret = mlx5_traffic_restart(dev);
> -
> +		ret = mlx5_traffic_restart(dev);
>  		if (ret)
>  			DRV_LOG(ERR, "port %u cannot remove mac address: %s",
>  				dev->data->port_id, strerror(rte_errno));
> @@ -97,6 +110,8 @@ int
>  mlx5_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac,
>  		  uint32_t index, uint32_t vmdq __rte_unused)
>  {
> +	struct priv *priv = dev->data->dev_private;
> +	const int vf = priv->config.vf;
>  	unsigned int i;
>  
>  	assert(index < MLX5_MAX_MAC_ADDRESSES);
> @@ -111,6 +126,12 @@ mlx5_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac,
>  		rte_errno = EADDRINUSE;
>  		return -rte_errno;
>  	}
> +	if (index && vf) {

Same comment as mlx5_mac_addr_remove() regarding index.

> +		int ret = mlx5_vf_mac_addr_add(dev, mac);
> +
> +		if (ret)
> +			return ret;
> +	}
>  	dev->data->mac_addrs[index] = *mac;
>  	if (!dev->data->promiscuous)
>  		return mlx5_traffic_restart(dev);
> diff --git a/drivers/net/mlx5/mlx5_vf.c b/drivers/net/mlx5/mlx5_vf.c
> index 357407f56..3db8ee0eb 100644
> --- a/drivers/net/mlx5/mlx5_vf.c
> +++ b/drivers/net/mlx5/mlx5_vf.c
> @@ -11,12 +11,29 @@
>  #include "mlx5.h"
>  #include "mlx5_utils.h"
>  
> +/*
> + * Define NDA_RTA as defined in iproute2 sources.
> + *
> + * see in iproute2 sources file include/libnetlink.h
> + */
> +#ifndef NDA_RTA
> +#define NDA_RTA(r) \
> +	((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
> +#endif
> +

I suggest an internal MLX5_* macro instead, just in case.

>  /* Data exchanged between Netlink library and PMD. */
>  struct mlx5_vf_iface {
>  	struct rte_eth_dev *dev; /**< Pointer to Ethernet device. */
>  	int iface_idx; /**< Network Interface index. */
>  };
>  
> +/* Add/Remove mac address through Metlink */

Remove => remove
Metlink => Netlink

> +struct mlx5_vf_mac_addr {
> +	struct ether_addr (*mac)[];
> +	/**< MAC Address handled by the device. */

Address => address

> +	int mac_n; /**< Number of address in the array. */

address => addresses

(Please check the remaining typos and extra caps.)

> +};
> +
>  /**
>   * Parse Netlink message to retrieve the interface index.
>   *
> @@ -132,3 +149,301 @@ mlx5_vf_iface_idx(struct rte_eth_dev *dev)
>  		dev->data->port_id, strerror(rte_errno));
>  	return -rte_errno;
>  }
> +
> +/**
> + * Parse Netlink message to retrieve the bridge MAC address.
> + *
> + * @param nh
> + *   Pointer to Netlink Message Header.
> + * @param arg
> + *   PMD data register with this callback.
> + *
> + * @return
> + *   0 on success, -1 otherwise.

How about a negative errno value in case of error for consistency?

> + */
> +static int
> +mlx5_vf_mac_addr_cb(struct nlmsghdr *nh, void *arg)
> +{
> +	struct mlx5_vf_mac_addr *data = arg;
> +	struct ndmsg *r = NLMSG_DATA(nh);
> +	struct rtattr *attribute;
> +	int len;
> +
> +	len = nh->nlmsg_len - NLMSG_LENGTH(sizeof(*r));
> +	for (attribute = NDA_RTA(r);
> +	     RTA_OK(attribute, len);
> +	     attribute = RTA_NEXT(attribute, len)) {
> +		if (attribute->rta_type == NDA_LLADDR) {
> +			if (data->mac_n == MLX5_MAX_MAC_ADDRESSES) {
> +				DRV_LOG(WARNING,
> +					"not enough room to finalise the"
> +					" request");
> +				return -1;

How about -ENOMEM?

> +			}
> +#ifndef NDEBUG
> +			struct ether_addr m;
> +
> +			memcpy(&m, RTA_DATA(attribute), ETHER_ADDR_LEN);
> +			DRV_LOG(DEBUG,
> +				"brige MAC address"

brige => bridge

> +				" %02X:%02X:%02X:%02X:%02X:%02X",
> +				m.addr_bytes[0], m.addr_bytes[1],
> +				m.addr_bytes[2], m.addr_bytes[3],
> +				m.addr_bytes[4], m.addr_bytes[5]);
> +#endif
> +			memcpy(&(*data->mac)[data->mac_n++],
> +			       RTA_DATA(attribute), ETHER_ADDR_LEN);
> +		}
> +	}
> +	return 0;
> +}
> +
> +/**
> + * Get bridge MAC addresses.
> + *
> + * @param dev
> + *   Pointer to Ethernet device.
> + * @param mac[out]
> + *   Pointer to the array table of MAC addresses to fill.
> + *   Its size should be of MLX5_MAX_MAC_ADDRESSES.
> + * @param mac_n[out]
> + *   Number of entries filled in MAC array.
> + *
> + * @return
> + *   0 on success, a negative errno value otherwise and rte_errno is set.
> + */

Since the mac_n is only an output value and the size of mac is fixed to
MLX5_MAX_MAC_ADDRESSES, I suggest to return the value of mac_n (positive in
case of success) and remove it from the parameter list.

> +static int
> +mlx5_vf_mac_addr_list(struct rte_eth_dev *dev, struct ether_addr (*mac)[],

Another suggestion for clarity: (*mac)[MLX5_MAX_MAC_ADDRESSES]

> +		      int *mac_n)
> +{
> +	int iface_idx = mlx5_vf_iface_idx(dev);
> +	struct {
> +		struct nlmsghdr	hdr;
> +		struct ifinfomsg ifm;
> +	} req = {
> +		.hdr = {
> +			.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
> +			.nlmsg_type = RTM_GETNEIGH,
> +			.nlmsg_flags = NLM_F_DUMP | NLM_F_REQUEST,
> +		},
> +		.ifm = {
> +			.ifi_family = PF_BRIDGE,
> +			.ifi_index = iface_idx,
> +		},
> +	};
> +	struct mlx5_vf_mac_addr data = {
> +		.mac = mac,
> +		.mac_n = 0,
> +	};
> +	int fd;
> +	int ret;
> +
> +	fd = rte_nl_init(RTMGRP_LINK);
> +	if (fd < 0) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	ret = rte_nl_request(fd, &req.hdr, &req.ifm, sizeof(struct ifinfomsg));
> +	if (ret < 0) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	ret = rte_nl_recv(fd, mlx5_vf_mac_addr_cb, &data);
> +	if (ret < 0) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	rte_nl_final(fd);
> +	*mac_n = data.mac_n;
> +	return 0;
> +error:
> +	if (fd >= 0)
> +		rte_nl_final(fd);
> +	DRV_LOG(DEBUG, "port %u cannot retrieve MAC address list %s",
> +		dev->data->port_id, strerror(rte_errno));
> +	return -rte_errno;
> +}
> +
> +/**
> + * Modify the mac address neighbour table with Netlink.

mac => MAC
Metlink => Netlink

> + *
> + * @param dev
> + *   Pointer to Ethernet device.
> + * @param mac
> + *   MAC address to consider.
> + * @param add
> + *   1 to add the MAC address, 0 to remove the MAC address.
> + *
> + * @return
> + *   0 on success, a negative errno value otherwise and rte_errno is set.
> + */
> +static int
> +mlx5_vf_mac_addr_modify(struct rte_eth_dev *dev, struct ether_addr *mac,
> +			int add)
> +{
> +	int iface_idx = mlx5_vf_iface_idx(dev);
> +	struct {
> +		struct nlmsghdr hdr;
> +		struct ndmsg ndm;
> +		struct rtattr rta;
> +	} req = {
> +		.hdr = {
> +			.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg)),
> +			.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE |
> +				NLM_F_EXCL | NLM_F_ACK,
> +			.nlmsg_type = add ? RTM_NEWNEIGH : RTM_DELNEIGH,
> +		},
> +		.ndm = {
> +			.ndm_family = PF_BRIDGE,
> +			.ndm_state = NUD_NOARP | NUD_PERMANENT,
> +			.ndm_ifindex = iface_idx,
> +			.ndm_flags = NTF_SELF,
> +		},
> +		.rta = {
> +			.rta_type = NDA_LLADDR,
> +			.rta_len = RTA_LENGTH(ETHER_ADDR_LEN),
> +		},
> +	};
> +	int fd;
> +	int ret;
> +
> +	memcpy(RTA_DATA(&req.rta), mac, ETHER_ADDR_LEN);
> +	req.hdr.nlmsg_len = NLMSG_ALIGN(req.hdr.nlmsg_len) +
> +		RTA_ALIGN(req.rta.rta_len);
> +	fd = rte_nl_init(RTMGRP_LINK);
> +	if (fd < 0) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	ret = rte_nl_send(fd, &req.hdr);
> +	if (ret == -1) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	ret = rte_nl_recv_ack(fd);
> +	if (ret == -1) {
> +		rte_errno = errno;
> +		goto error;
> +	}
> +	rte_nl_final(fd);
> +	return 0;
> +error:
> +	if (fd >= 0)
> +		rte_nl_final(fd);
> +	DRV_LOG(DEBUG,
> +		"port %u cannot %s MAC address %02X:%02X:%02X:%02X:%02X:%02X"
> +		" %s",
> +		dev->data->port_id,
> +		add ? "add" : "remove",
> +		mac->addr_bytes[0], mac->addr_bytes[1],
> +		mac->addr_bytes[2], mac->addr_bytes[3],
> +		mac->addr_bytes[4], mac->addr_bytes[5],
> +		strerror(rte_errno));
> +	return -rte_errno;
> +}
> +
> +/**
> + * add a MAC address.

add => Add (can't stop!)

> + *
> + * @param dev
> + *   Pointer to Ethernet device.
> + * @param mac_addr
> + *   MAC address to register.
> + *
> + * @return
> + *   0 on success, a negative errno value otherwise and rte_errno is set.
> + */
> +int
> +mlx5_vf_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac)
> +{
> +	struct ether_addr macs[MLX5_MAX_MAC_ADDRESSES];
> +	int macs_n = 0;
> +	int i;
> +	int mac_index = MLX5_MAX_MAC_ADDRESSES;
> +	int ret;
> +
> +	ret = mlx5_vf_mac_addr_list(dev, &macs, &macs_n);
> +	if (ret)
> +		return ret;
> +	for (i = 0; i != macs_n; ++i) {
> +		if (!memcmp(&macs[i], mac, ETHER_ADDR_LEN)) {
> +			mac_index = i;
> +			break;
> +		}
> +	}
> +	if (mac_index != MLX5_MAX_MAC_ADDRESSES) {
> +		DRV_LOG(INFO, "port %u MAC address already added",

added => present

> +			dev->data->port_id);
> +		rte_errno = EADDRINUSE;
> +		return -rte_errno;
> +	}
> +	return mlx5_vf_mac_addr_modify(dev, mac, 1);
> +}
> +
> +/**
> + * Remove a MAC address.
> + *
> + * @param dev
> + *   Pointer to Ethernet device.
> + * @param index
> + *   MAC address to remove.
> + *
> + * @return
> + *   0 on success, a negative errno value otherwise and rte_errno is set.
> + */
> +int
> +mlx5_vf_mac_addr_remove(struct rte_eth_dev *dev, struct ether_addr *mac)
> +{
> +	struct ether_addr macs[MLX5_MAX_MAC_ADDRESSES];
> +	int macs_n = 0;
> +	int i;
> +	int mac_index = MLX5_MAX_MAC_ADDRESSES;
> +	int ret;
> +
> +	ret = mlx5_vf_mac_addr_list(dev, &macs, &macs_n);
> +	if (ret)
> +		return ret;
> +	for (i = 0; i != macs_n; ++i) {
> +		if (!memcmp(&macs[i], mac, ETHER_ADDR_LEN)) {
> +			mac_index = i;
> +			break;
> +		}
> +	}
> +	if (mac_index == MLX5_MAX_MAC_ADDRESSES) {
> +		DRV_LOG(INFO, "port %u MAC address not found",
> +			dev->data->port_id);
> +		rte_errno = EINVAL;

How about EADDRNOTAVAIL (a better counterpart for EADDRINUSE)?

> +		return -rte_errno;
> +	}
> +	return mlx5_vf_mac_addr_modify(dev, mac, 0);
> +}
> +
> +/**
> + * Flush all added MAC addresses.
> + *
> + * @param dev
> + *   Pointer to Ethernet device.
> + *
> + * @return
> + *   0 on success, a negative errno value otherwise and rte_errno is set.
> + */
> +int
> +mlx5_vf_mac_addr_flush(struct rte_eth_dev *dev)
> +{
> +	int i;
> +	const struct ether_addr mac_null = { .addr_bytes = { 0 }, };
> +
> +	/* Skip the default mac at index 0. */

I'm not sure, what if index 0 happens to be yet another secondary MAC
address of the netdevice at this point? I think the PMD should attempt to
flush them all starting from the end of the MAC array and not care about
errors; this function should return void.

> +	for (i = 1; i != MLX5_MAX_MAC_ADDRESSES; ++i) {
> +		struct ether_addr *m = &dev->data->mac_addrs[i];
> +
> +		if (memcmp(&mac_null, m, ETHER_ADDR_LEN)) {
> +			int ret;
> +
> +			ret = mlx5_vf_mac_addr_remove(dev, m);
> +			if (ret)
> +				return ret;
> +		}
> +	}
> +	return 0;
> +}

-- 
Adrien Mazarguil
6WIND