[dpdk-dev] [PATCH 2/2] net/tap: add link status notification

[Pascal Mazon <pascal.mazon@6wind.com>]

As tap is a virtual device, there's no physical way a link can be cut.
However, it has an associated kernel netdevice and possibly a remote
netdevice too. These netdevices link status may change outside of the
DPDK scope, through an external command such as:

  ip link set dev tapX down

This commit implements link status notification through netlink.

Signed-off-by: Pascal Mazon <pascal.mazon@6wind.com>
---
 doc/guides/nics/features/tap.ini |  1 +
 drivers/net/tap/rte_eth_tap.c    | 65 ++++++++++++++++++++++++++++++++++++++--
 drivers/net/tap/rte_eth_tap.h    |  1 +
 drivers/net/tap/tap_netlink.c    | 11 +++++--
 drivers/net/tap/tap_netlink.h    |  2 +-
 drivers/net/tap/tap_tcmsgs.c     |  2 +-
 6 files changed, 76 insertions(+), 6 deletions(-)

diff --git a/doc/guides/nics/features/tap.ini b/doc/guides/nics/features/tap.ini
index 9d73f61cca3b..20cbeee72f52 100644
--- a/doc/guides/nics/features/tap.ini
+++ b/doc/guides/nics/features/tap.ini
@@ -5,6 +5,7 @@
 ;
 [Features]
 Link status          = Y
+Link status event    = Y
 Jumbo frame          = Y
 Promiscuous mode     = Y
 Allmulticast mode    = Y
diff --git a/drivers/net/tap/rte_eth_tap.c b/drivers/net/tap/rte_eth_tap.c
index 54122fd13c66..6567bba75b47 100644
--- a/drivers/net/tap/rte_eth_tap.c
+++ b/drivers/net/tap/rte_eth_tap.c
@@ -61,6 +61,7 @@
 
 #include <rte_eth_tap.h>
 #include <tap_flow.h>
+#include <tap_netlink.h>
 #include <tap_tcmsgs.h>
 
 /* Linux based path to the TUN device */
@@ -111,6 +112,8 @@ static int
 tap_ioctl(struct pmd_internals *pmd, unsigned long request,
 	  struct ifreq *ifr, int set);
 
+static int tap_intr_handle_set(struct rte_eth_dev *dev, int set);
+
 /* Tun/Tap allocation routine
  *
  * name is the number of the interface to use, unless NULL to take the host
@@ -520,6 +523,11 @@ tap_link_set_up(struct rte_eth_dev *dev)
 static int
 tap_dev_start(struct rte_eth_dev *dev)
 {
+	int err;
+
+	err = tap_intr_handle_set(dev, 1);
+	if (err)
+		return err;
 	return tap_link_set_up(dev);
 }
 
@@ -528,6 +536,7 @@ tap_dev_start(struct rte_eth_dev *dev)
 static void
 tap_dev_stop(struct rte_eth_dev *dev)
 {
+	tap_intr_handle_set(dev, 0);
 	tap_link_set_down(dev);
 }
 
@@ -976,6 +985,55 @@ tap_set_mc_addr_list(struct rte_eth_dev *dev __rte_unused,
 	return 0;
 }
 
+static int
+tap_nl_msg_handler(struct nlmsghdr *nh, void *arg)
+{
+	struct rte_eth_dev *dev = arg;
+	struct pmd_internals *pmd = dev->data->dev_private;
+	struct ifinfomsg *info = NLMSG_DATA(nh);
+
+	if (nh->nlmsg_type != RTM_NEWLINK ||
+	    (info->ifi_index != pmd->if_index &&
+	     info->ifi_index != pmd->remote_if_index))
+		return 0;
+	return tap_link_update(dev, 0);
+}
+
+static void
+tap_dev_intr_handler(struct rte_intr_handle *intr_handle __rte_unused,
+		     void *cb_arg)
+{
+	struct rte_eth_dev *dev = cb_arg;
+	struct pmd_internals *pmd = dev->data->dev_private;
+
+	nl_recv(pmd->intr_handle.fd, tap_nl_msg_handler, dev);
+}
+
+static int
+tap_intr_handle_set(struct rte_eth_dev *dev, int set)
+{
+	struct pmd_internals *pmd = dev->data->dev_private;
+
+	/* In any case, disable interrupt if the conf is no longer there. */
+	if (!dev->data->dev_conf.intr_conf.lsc) {
+		if (pmd->intr_handle.fd != -1)
+			nl_final(pmd->intr_handle.fd);
+		rte_intr_callback_unregister(
+			&pmd->intr_handle, tap_dev_intr_handler, dev);
+		return 0;
+	}
+	if (set) {
+		pmd->intr_handle.fd = nl_init(RTMGRP_LINK);
+		if (unlikely(pmd->intr_handle.fd == -1))
+			return -EBADF;
+		return rte_intr_callback_register(
+			&pmd->intr_handle, tap_dev_intr_handler, dev);
+	}
+	nl_final(pmd->intr_handle.fd);
+	return rte_intr_callback_unregister(&pmd->intr_handle,
+					    tap_dev_intr_handler, dev);
+}
+
 static const uint32_t*
 tap_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 {
@@ -1117,7 +1175,7 @@ eth_dev_tap_create(const char *name, char *tap_name, char *remote_iface)
 	data->dev_private = pmd;
 	data->port_id = dev->data->port_id;
 	data->mtu = dev->data->mtu;
-	data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+	data->dev_flags = RTE_ETH_DEV_DETACHABLE | RTE_ETH_DEV_INTR_LSC;
 	data->kdrv = RTE_KDRV_NONE;
 	data->drv_name = pmd_tap_drv.driver.name;
 	data->numa_node = numa_node;
@@ -1133,6 +1191,9 @@ eth_dev_tap_create(const char *name, char *tap_name, char *remote_iface)
 	dev->rx_pkt_burst = pmd_rx_burst;
 	dev->tx_pkt_burst = pmd_tx_burst;
 
+	pmd->intr_handle.type = RTE_INTR_HANDLE_EXT;
+	pmd->intr_handle.fd = -1;
+
 	/* Presetup the fds to -1 as being not valid */
 	for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
 		pmd->rxq[i].fd = -1;
@@ -1147,7 +1208,7 @@ eth_dev_tap_create(const char *name, char *tap_name, char *remote_iface)
 	 * If no netlink socket can be created, then it will fail when
 	 * creating/destroying flow rules.
 	 */
-	pmd->nlsk_fd = nl_init();
+	pmd->nlsk_fd = nl_init(0);
 	if (strlen(remote_iface)) {
 		pmd->remote_if_index = if_nametoindex(remote_iface);
 		snprintf(pmd->remote_iface, RTE_ETH_NAME_MAX_LEN,
diff --git a/drivers/net/tap/rte_eth_tap.h b/drivers/net/tap/rte_eth_tap.h
index a559f6b0a3b0..f1496dcfdb1a 100644
--- a/drivers/net/tap/rte_eth_tap.h
+++ b/drivers/net/tap/rte_eth_tap.h
@@ -90,6 +90,7 @@ struct pmd_internals {
 	LIST_HEAD(tap_implicit_flows, rte_flow) implicit_flows;
 	struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
 	struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
+	struct rte_intr_handle intr_handle;          /* LSC interrupt handle. */
 };
 
 #endif /* _RTE_ETH_TAP_H_ */
diff --git a/drivers/net/tap/tap_netlink.c b/drivers/net/tap/tap_netlink.c
index 9710e41a7801..6de896ab17b6 100644
--- a/drivers/net/tap/tap_netlink.c
+++ b/drivers/net/tap/tap_netlink.c
@@ -55,14 +55,21 @@ struct nested_tail {
 /**
  * Initialize a netlink socket for communicating with the kernel.
  *
+ * @param nl_groups
+ *   Set it to a netlink group value (e.g. RTMGRP_LINK) to receive messages for
+ *   specific netlink multicast groups. Otherwise, no subscription will be made.
+ *
  * @return
  *   netlink socket file descriptor on success, -1 otherwise.
  */
 int
-nl_init(void)
+nl_init(uint32_t nl_groups)
 {
 	int fd, sndbuf_size = SNDBUF_SIZE, rcvbuf_size = RCVBUF_SIZE;
-	struct sockaddr_nl local = { .nl_family = AF_NETLINK };
+	struct sockaddr_nl local = {
+		.nl_family = AF_NETLINK,
+		.nl_groups = nl_groups,
+	};
 
 	fd = socket(AF_NETLINK, SOCK_RAW | SOCK_CLOEXEC, NETLINK_ROUTE);
 	if (fd < 0) {
diff --git a/drivers/net/tap/tap_netlink.h b/drivers/net/tap/tap_netlink.h
index 52ba8c030dcc..98e1390268d8 100644
--- a/drivers/net/tap/tap_netlink.h
+++ b/drivers/net/tap/tap_netlink.h
@@ -53,7 +53,7 @@ struct nlmsg {
 
 #define NLMSG_TAIL(nlh) (void *)((char *)(nlh) + NLMSG_ALIGN((nlh)->nlmsg_len))
 
-int nl_init(void);
+int nl_init(uint32_t nl_groups);
 int nl_final(int nlsk_fd);
 int nl_send(int nlsk_fd, struct nlmsghdr *nh);
 int nl_recv(int nlsk_fd, int (*callback)(struct nlmsghdr *, void *), void *arg);
diff --git a/drivers/net/tap/tap_tcmsgs.c b/drivers/net/tap/tap_tcmsgs.c
index 9a146d165b08..af1c9aec0d22 100644
--- a/drivers/net/tap/tap_tcmsgs.c
+++ b/drivers/net/tap/tap_tcmsgs.c
@@ -107,7 +107,7 @@ qdisc_del(int nlsk_fd, uint16_t ifindex, struct qdisc *qinfo)
 	msg.t.tcm_parent = qinfo->parent;
 	/* if no netlink socket is provided, create one */
 	if (!nlsk_fd) {
-		fd = nl_init();
+		fd = nl_init(0);
 		if (fd < 0) {
 			RTE_LOG(ERR, PMD,
 				"Could not delete QDISC: null netlink socket\n");
-- 
2.12.0.306.g4a9b9b3