net/tap: add new macpair option for split MAC address

Normally, the MAC address of the kernel interface is the same as in the interface in dpdk, as they represent the same interface. It is useful to allow viewing them as separate connected interfaces (like ip's veth).

This solves a problem I have running a freebsd-based IPv6 stack on top of dpdk and using the tap interface, as both the kernel and freebsd stacks configure the MAC derived IPv6 address on the interface (as can be seen with ifconfig for the kernel), and they both complain about duplicate IPv6 address and the freebsd disables IPv6 as a result.

Signed-off-by Isaac Boukris <[email protected]>

---
 doc/guides/nics/tap.rst       | 18 +++++++++++++++++
 drivers/net/tap/rte_eth_tap.c | 37 ++++++++++++++++++++++++++++++-----
 drivers/net/tap/rte_eth_tap.h |  1 +
 3 files changed, 51 insertions(+), 5 deletions(-)

doc/guides/nics/tap.rst +18 -0

index f01663c700..fd936c6084 100644
@@ -71,6 +71,24 @@ But this behavior can be overridden by the use of the persist flag, example::
 
   --vdev=net_tap0,iface=tap0,persist ...
 
+Normally, the MAC address of the kernel interface is the same as in the
+interface in dpdk, as they represent the same interface. If a MAC address is set
+by the dpdk application using ``rte_eth_dev_default_mac_addr_set()`` then the
+kernel interface changes accordingly (but not vice versa).
+
+If the ``macpair`` option is given, then the MAC address of the kernel
+interface is initially set to a different random MAC address and it is no
+longer altered by dpdk. This allows to view the two interfaces as separate
+connected interfaces, similar to linux's veth pair interfaces. For instance,
+you can configure an IP address on the kernel interface with the ``ip`` command
+in the same subnet as the one used in dpdk and use it as next gateway.
+
+The ``macpair`` options is incompatible with the ``remote`` option (see above).
+
+Example::
+
+    --vdev=net_tap0,iface=tap0,macpair
+
 
 TUN devices
 -----------

drivers/net/tap/rte_eth_tap.c +32 -5

index c5af5751f6..3cdd130092 100644
@@ -56,6 +56,7 @@
 #define ETH_TAP_MAC_ARG         "mac"
 #define ETH_TAP_MAC_FIXED       "fixed"
 #define ETH_TAP_PERSIST_ARG     "persist"
+#define ETH_TAP_MACPAIR_ARG     "macpair"
 
 #define ETH_TAP_USR_MAC_FMT     "xx:xx:xx:xx:xx:xx"
 #define ETH_TAP_CMP_MAC_FMT     "0123456789ABCDEFabcdef"

@@ -95,6 +96,7 @@ static const char *valid_arguments[] = {
        ETH_TAP_REMOTE_ARG,
        ETH_TAP_MAC_ARG,
        ETH_TAP_PERSIST_ARG,
+       ETH_TAP_MACPAIR_ARG,
        NULL
 };
 

@@ -1391,6 +1393,12 @@ tap_mac_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
                        dev->device->name);
                return -EINVAL;
        }
+
+       if (pmd->mac_pair) {
+               rte_ether_addr_copy(&pmd->eth_addr, mac_addr);
+               return 0;
+       }
+
        /* Check the actual current MAC address on the tap netdevice */
        ret = tap_ioctl(pmd, SIOCGIFHWADDR, &ifr, 0, LOCAL_ONLY);
        if (ret < 0)

@@ -1915,7 +1923,7 @@ static const struct eth_dev_ops ops = {
 static int
 eth_dev_tap_create(struct rte_vdev_device *vdev, const char *tap_name,
                   char *remote_iface, struct rte_ether_addr *mac_addr,
-                  enum rte_tuntap_type type, int persist)
+                  enum rte_tuntap_type type, int persist, int mac_pair)
 {
        int numa_node = rte_socket_id();
        struct rte_eth_dev *dev;

@@ -2023,12 +2031,20 @@ eth_dev_tap_create(struct rte_vdev_device *vdev, const char *tap_name,
        if (pmd->type == ETH_TUNTAP_TYPE_TAP) {
                memset(&ifr, 0, sizeof(struct ifreq));
                ifr.ifr_hwaddr.sa_family = AF_LOCAL;
-               rte_memcpy(ifr.ifr_hwaddr.sa_data, &pmd->eth_addr,
-                               RTE_ETHER_ADDR_LEN);
+
+               if (mac_pair) {
+                       rte_eth_random_addr((uint8_t *)ifr.ifr_hwaddr.sa_data);
+               } else {
+                       memcpy(ifr.ifr_hwaddr.sa_data, &pmd->eth_addr,
+                              RTE_ETHER_ADDR_LEN);
+               }
+
                if (tap_ioctl(pmd, SIOCSIFHWADDR, &ifr, 0, LOCAL_ONLY) < 0)
                        goto error_exit;
        }
 
+       pmd->mac_pair = mac_pair;
+
        /* Make network device persist after application exit */
        pmd->persist = persist;
 

@@ -2306,7 +2322,7 @@ rte_pmd_tun_probe(struct rte_vdev_device *dev)
        TAP_LOG(DEBUG, "Initializing pmd_tun for %s", name);
 
        ret = eth_dev_tap_create(dev, tun_name, remote_iface, 0,
-                                ETH_TUNTAP_TYPE_TUN, 0);
+                                ETH_TUNTAP_TYPE_TUN, 0, 0);
 
 leave:
        if (ret == -1) {

@@ -2429,6 +2445,7 @@ rte_pmd_tap_probe(struct rte_vdev_device *dev)
        struct rte_eth_dev *eth_dev;
        int tap_devices_count_increased = 0;
        int persist = 0;
+       int mac_pair = 0;
 
        name = rte_vdev_device_name(dev);
        params = rte_vdev_device_args(dev);

@@ -2504,6 +2521,16 @@ rte_pmd_tap_probe(struct rte_vdev_device *dev)
                                        goto leave;
                        }
 
+                       if (rte_kvargs_count(kvlist, ETH_TAP_MACPAIR_ARG) == 1) {
+                               if (strlen(remote_iface)) {
+                                       TAP_LOG(ERR, "mac pair not supported "
+                                                    "with remote interface.");
+                                       ret = -1;
+                                       goto leave;
+                               }
+                               mac_pair = 1;
+                       }
+
                        if (rte_kvargs_count(kvlist, ETH_TAP_MAC_ARG) == 1) {
                                ret = rte_kvargs_process(kvlist,
                                                         ETH_TAP_MAC_ARG,

@@ -2533,7 +2560,7 @@ rte_pmd_tap_probe(struct rte_vdev_device *dev)
        tap_devices_count++;
        tap_devices_count_increased = 1;
        ret = eth_dev_tap_create(dev, tap_name, remote_iface, &user_mac,
-                                ETH_TUNTAP_TYPE_TAP, persist);
+                                ETH_TUNTAP_TYPE_TAP, persist, mac_pair);
 
 leave:
        if (ret == -1) {

drivers/net/tap/rte_eth_tap.h +1 -1

index ce4322ad04..5a33698f76 100644
@@ -72,6 +72,7 @@ struct pmd_internals {
        char name[RTE_ETH_NAME_MAX_LEN];  /* Internal Tap device name */
        int type;                         /* Type field - TUN|TAP */
        int persist;                      /* 1 if keep link up, else 0 */
+       int mac_pair;                     /* 1 if mac pair enabled, else 0 */
        struct rte_ether_addr eth_addr;   /* Mac address of the device port */
        struct ifreq remote_initial_flags;/* Remote netdevice flags on init */
        int remote_if_index;              /* remote netdevice IF_INDEX */
--

2.45.0

Normally, the MAC address of the kernel interface is the same as in the interface in dpdk, as they represent the same interface. It is useful to allow viewing them as separate connected interfaces (like ip's veth).

This solves a problem I have running a freebsd-based IPv6 stack on top of dpdk and using the tap interface, as both the kernel and freebsd stacks configure the MAC derived IPv6 address on the interface (as can be seen with ifconfig for the kernel), and they both complain about duplicate IPv6 address and the freebsd disables IPv6 as a result.

Signed-off-by Isaac Boukris <[email protected]>

---
 doc/guides/nics/tap.rst       | 18 +++++++++++++++++
 drivers/net/tap/rte_eth_tap.c | 37 ++++++++++++++++++++++++++++++-----
 drivers/net/tap/rte_eth_tap.h |  1 +
 3 files changed, 51 insertions(+), 5 deletions(-)

doc/guides/nics/tap.rst +5 -0

index f01663c700..fd936c6084 100644
@@ -71,6 +71,24 @@ But this behavior can be overridden by the use of the persist flag, example::
 
   --vdev=net_tap0,iface=tap0,persist ...
 
+Normally, the MAC address of the kernel interface is the same as in the
+interface in dpdk, as they represent the same interface. If a MAC address is set
+by the dpdk application using ``rte_eth_dev_default_mac_addr_set()`` then the
+kernel interface changes accordingly (but not vice versa).
+

Quoted 10 lines · Isaac Boukris

Normally, the MAC address of the kernel interface is the same as in the interface in dpdk, as they represent the same interface. It is useful to allow viewing them as separate connected interfaces (like ip's veth).

This solves a problem I have running a freebsd-based IPv6 stack on top of dpdk and using the tap interface, as both the kernel and freebsd stacks configure the MAC derived IPv6 address on the interface (as can be seen with ifconfig for the kernel), and they both complain about duplicate IPv6 address and the freebsd disables IPv6 as a result.

How kernel side knows about the IPv6 address DPDK side uses, what is your tap usecase?

I can see 'macpair' is coming from "veth pair" above, but I don't think it is very explanatory in the context of tap. Do you have any more suggestion?

Another concern is how to test this, tap PMD got multiple arguments now, and as the pmd keeps getting more updates and features, how can we sure this devarg usecase is not broken?

Normally, the MAC address of the kernel interface is the same as in the interface in dpdk, as they represent the same interface. It is useful to allow viewing them as separate connected interfaces (like ip's veth).

This solves a problem I have running a freebsd-based IPv6 stack on top of dpdk and using the tap interface, as both the kernel and freebsd stacks configure the MAC derived IPv6 address on the interface (as can be seen with ifconfig for the kernel), and they both complain about duplicate IPv6 address and the freebsd disables IPv6 as a result.

Signed-off-by Isaac Boukris <[email protected]>

2. Need more error checks. macpair won't work with TUN device. and what happens if mac address is specified?

3. Should the mac address devarg, take two args?