[v2,1/3] common/mlx5: add Netlink event helpers
Checks
Commit Message
Introduce mlx5_nl_read_events() to read Netlink events
(technically, messages) from a socket that was configured
to listen for them via a new mlx5_nl_init() parameter.
Add mlx5_nl_parse_link_status_update() helper
to extract information from link-related events.
This patch is a shared base for later fixes.
Cc: stable@dpdk.org
Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
Reviewed-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
---
drivers/common/mlx5/linux/mlx5_common_os.c | 2 +-
drivers/common/mlx5/linux/mlx5_nl.c | 102 ++++++++++++++++++++-
drivers/common/mlx5/linux/mlx5_nl.h | 8 +-
drivers/common/mlx5/version.map | 2 +
drivers/net/mlx5/linux/mlx5_os.c | 8 +-
drivers/net/mlx5/linux/mlx5_vlan_os.c | 2 +-
6 files changed, 116 insertions(+), 8 deletions(-)
Comments
On 3/1/2022 12:15 PM, Dmitry Kozlyuk wrote:
> Introduce mlx5_nl_read_events() to read Netlink events
> (technically, messages) from a socket that was configured
> to listen for them via a new mlx5_nl_init() parameter.
> Add mlx5_nl_parse_link_status_update() helper
> to extract information from link-related events.
> This patch is a shared base for later fixes.
>
> Cc: stable@dpdk.org
>
Hi Dmitry,
For clarification, this patch is not fix, but it is requested
to be backported to be able to backport fixes in this patchset,
right?
This looks OK to me, but cc'ed LTS maintainers in case they
have objection.
> Signed-off-by: Dmitry Kozlyuk <dkozlyuk@nvidia.com>
> Reviewed-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
<...>
Hi Ferruh,
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@intel.com>
[...]
> Hi Dmitry,
>
> For clarification, this patch is not fix, but it is requested
> to be backported to be able to backport fixes in this patchset,
> right?
Yes.
On 02/03/2022 15:56, Dmitry Kozlyuk wrote:
> Hi Ferruh,
>
>> -----Original Message-----
>> From: Ferruh Yigit <ferruh.yigit@intel.com>
> [...]
>> Hi Dmitry,
>>
>> For clarification, this patch is not fix, but it is requested
>> to be backported to be able to backport fixes in this patchset,
>> right?
>
> Yes.
The updated API is internal so that should be ok. I'm ok to take this on
21.11 as long as you can confirm it's not breaking any user
compatibility with external sw versions/ABI/API etc from 21.11.0 ?
Assuming that's ok, please send a rebased series for 21.11. I'm not
comfortable rebasing the series with the amount of changes on dpdk main
branch to same functions in mlx5_os.c.
P.S. Better to rebase on patch queue [0] to avoid conflicts with other
backports not pushed to dpdk.org yet.
thanks,
Kevin.
[0] https://github.com/kevintraynor/dpdk-stable.git
Hi Kevin,
> -----Original Message-----
> From: Kevin Traynor <ktraynor@redhat.com>
[...]
> The updated API is internal so that should be ok. I'm ok to take this
> on
> 21.11 as long as you can confirm it's not breaking any user
> compatibility with external sw versions/ABI/API etc from 21.11.0 ?
>
> Assuming that's ok, please send a rebased series for 21.11. I'm not
> comfortable rebasing the series with the amount of changes on dpdk
> main branch to same functions in mlx5_os.c.
Changes are not breaking any external SW.
Backport sent:
http://inbox.dpdk.org/stable/20220308151044.1012413-1-dkozlyuk@nvidia.com
@@ -590,7 +590,7 @@ mlx5_os_get_ibv_device(const struct rte_pci_addr *addr)
static int
mlx5_nl_roce_disable(const char *addr)
{
- int nlsk_fd = mlx5_nl_init(NETLINK_GENERIC);
+ int nlsk_fd = mlx5_nl_init(NETLINK_GENERIC, 0);
int devlink_id;
int enable;
int ret;
@@ -185,19 +185,22 @@ uint32_t atomic_sn;
*
* @param protocol
* Netlink protocol (e.g. NETLINK_ROUTE, NETLINK_RDMA).
+ * @param groups
+ * Groups to listen (e.g. RTMGRP_LINK), can be 0.
*
* @return
* A file descriptor on success, a negative errno value otherwise and
* rte_errno is set.
*/
int
-mlx5_nl_init(int protocol)
+mlx5_nl_init(int protocol, int groups)
{
int fd;
int buf_size;
socklen_t opt_size;
struct sockaddr_nl local = {
.nl_family = AF_NETLINK,
+ .nl_groups = groups,
};
int ret;
@@ -1862,3 +1865,100 @@ mlx5_nl_enable_roce_set(int nlsk_fd, int family_id, const char *pci_addr,
/* Now, need to reload the driver. */
return mlx5_nl_driver_reload(nlsk_fd, family_id, pci_addr);
}
+
+/**
+ * Try to parse a Netlink message as a link status update.
+ *
+ * @param hdr
+ * Netlink message header.
+ * @param[out] ifindex
+ * Index of the updated interface.
+ *
+ * @return
+ * 0 on success, negative on failure.
+ */
+int
+mlx5_nl_parse_link_status_update(struct nlmsghdr *hdr, uint32_t *ifindex)
+{
+ struct ifinfomsg *info;
+
+ switch (hdr->nlmsg_type) {
+ case RTM_NEWLINK:
+ case RTM_DELLINK:
+ case RTM_GETLINK:
+ case RTM_SETLINK:
+ info = NLMSG_DATA(hdr);
+ *ifindex = info->ifi_index;
+ return 0;
+ }
+ return -1;
+}
+
+/**
+ * Read pending events from a Netlink socket.
+ *
+ * @param nlsk_fd
+ * Netlink socket.
+ * @param cb
+ * Callback invoked for each of the events.
+ * @param cb_arg
+ * User data for the callback.
+ *
+ * @return
+ * 0 on success, including the case when there are no events.
+ * Negative on failure and rte_errno is set.
+ */
+int
+mlx5_nl_read_events(int nlsk_fd, mlx5_nl_event_cb *cb, void *cb_arg)
+{
+ char buf[8192];
+ struct sockaddr_nl addr;
+ struct iovec iov = {
+ .iov_base = buf,
+ .iov_len = sizeof(buf),
+ };
+ struct msghdr msg = {
+ .msg_name = &addr,
+ .msg_namelen = sizeof(addr),
+ .msg_iov = &iov,
+ .msg_iovlen = 1,
+ };
+ struct nlmsghdr *hdr;
+ ssize_t size;
+
+ while (1) {
+ size = recvmsg(nlsk_fd, &msg, MSG_DONTWAIT);
+ if (size < 0) {
+ if (errno == EAGAIN)
+ return 0;
+ if (errno == EINTR)
+ continue;
+ DRV_LOG(DEBUG, "Failed to receive netlink message: %s",
+ strerror(errno));
+ rte_errno = errno;
+ return -rte_errno;
+ }
+ hdr = (struct nlmsghdr *)buf;
+ while (size >= (ssize_t)sizeof(*hdr)) {
+ ssize_t msg_len = hdr->nlmsg_len;
+ ssize_t data_len = msg_len - sizeof(*hdr);
+ ssize_t aligned_len;
+
+ if (data_len < 0) {
+ DRV_LOG(DEBUG, "Netlink message too short");
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ aligned_len = NLMSG_ALIGN(msg_len);
+ if (aligned_len > size) {
+ DRV_LOG(DEBUG, "Netlink message too long");
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ cb(hdr, cb_arg);
+ hdr = RTE_PTR_ADD(hdr, aligned_len);
+ size -= aligned_len;
+ }
+ }
+ return 0;
+}
@@ -11,6 +11,7 @@
#include "mlx5_common.h"
+typedef void (mlx5_nl_event_cb)(struct nlmsghdr *hdr, void *user_data);
/* VLAN netdev for VLAN workaround. */
struct mlx5_nl_vlan_dev {
@@ -30,7 +31,7 @@ struct mlx5_nl_vlan_vmwa_context {
};
__rte_internal
-int mlx5_nl_init(int protocol);
+int mlx5_nl_init(int protocol, int groups);
__rte_internal
int mlx5_nl_mac_addr_add(int nlsk_fd, unsigned int iface_idx, uint64_t *mac_own,
struct rte_ether_addr *mac, uint32_t index);
@@ -75,4 +76,9 @@ int mlx5_nl_enable_roce_get(int nlsk_fd, int family_id, const char *pci_addr,
int mlx5_nl_enable_roce_set(int nlsk_fd, int family_id, const char *pci_addr,
int enable);
+__rte_internal
+int mlx5_nl_read_events(int nlsk_fd, mlx5_nl_event_cb *cb, void *cb_arg);
+__rte_internal
+int mlx5_nl_parse_link_status_update(struct nlmsghdr *hdr, uint32_t *ifindex);
+
#endif /* RTE_PMD_MLX5_NL_H_ */
@@ -127,9 +127,11 @@ INTERNAL {
mlx5_nl_mac_addr_flush; # WINDOWS_NO_EXPORT
mlx5_nl_mac_addr_remove; # WINDOWS_NO_EXPORT
mlx5_nl_mac_addr_sync; # WINDOWS_NO_EXPORT
+ mlx5_nl_parse_link_status_update; # WINDOWS_NO_EXPORT
mlx5_nl_port_state; # WINDOWS_NO_EXPORT
mlx5_nl_portnum; # WINDOWS_NO_EXPORT
mlx5_nl_promisc; # WINDOWS_NO_EXPORT
+ mlx5_nl_read_events; # WINDOWS_NO_EXPORT
mlx5_nl_switch_info; # WINDOWS_NO_EXPORT
mlx5_nl_vf_mac_addr_modify; # WINDOWS_NO_EXPORT
mlx5_nl_vlan_vmwa_create; # WINDOWS_NO_EXPORT
@@ -1117,7 +1117,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
sh = mlx5_alloc_shared_dev_ctx(spawn, mkvlist);
if (!sh)
return NULL;
- nl_rdma = mlx5_nl_init(NETLINK_RDMA);
+ nl_rdma = mlx5_nl_init(NETLINK_RDMA, 0);
/* Check port status. */
if (spawn->phys_port <= UINT8_MAX) {
/* Legacy Verbs api only support u8 port number. */
@@ -1180,7 +1180,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
priv->mtu = RTE_ETHER_MTU;
/* 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);
+ priv->nl_socket_route = mlx5_nl_init(NETLINK_ROUTE, 0);
priv->representor = !!switch_info->representor;
priv->master = !!switch_info->master;
priv->domain_id = RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID;
@@ -1927,8 +1927,8 @@ mlx5_os_pci_probe_pf(struct mlx5_common_device *cdev,
* matching ones, gathering into the list.
*/
struct ibv_device *ibv_match[ret + 1];
- int nl_route = mlx5_nl_init(NETLINK_ROUTE);
- int nl_rdma = mlx5_nl_init(NETLINK_RDMA);
+ int nl_route = mlx5_nl_init(NETLINK_ROUTE, 0);
+ int nl_rdma = mlx5_nl_init(NETLINK_RDMA, 0);
unsigned int i;
while (ret-- > 0) {
@@ -135,7 +135,7 @@ mlx5_vlan_vmwa_init(struct rte_eth_dev *dev, uint32_t ifindex)
return NULL;
}
rte_spinlock_init(&vmwa->sl);
- vmwa->nl_socket = mlx5_nl_init(NETLINK_ROUTE);
+ vmwa->nl_socket = mlx5_nl_init(NETLINK_ROUTE, 0);
if (vmwa->nl_socket < 0) {
DRV_LOG(WARNING,
"Can not create Netlink socket"