RDMA/nldev: Add support to get status of all counters
authorAharon Landau <aharonl@nvidia.com>
Fri, 8 Oct 2021 12:24:33 +0000 (15:24 +0300)
committerJason Gunthorpe <jgg@nvidia.com>
Tue, 12 Oct 2021 15:48:05 +0000 (12:48 -0300)
This patch adds the ability to get the name, index and status of all
counters for each link through RDMA netlink. This can be used for
user-space to get the current optional-counter mode.

Examples:
$ rdma statistic mode
link rocep8s0f0/1 optional-counters cc_rx_ce_pkts

$ rdma statistic mode supported
link rocep8s0f0/1 supported optional-counters cc_rx_ce_pkts,cc_rx_cnp_pkts,cc_tx_cnp_pkts
link rocep8s0f1/1 supported optional-counters cc_rx_ce_pkts,cc_rx_cnp_pkts,cc_tx_cnp_pkts

Link: https://lore.kernel.org/r/20211008122439.166063-8-markzhang@nvidia.com
Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Neta Ostrovsky <netao@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
drivers/infiniband/core/nldev.c
include/uapi/rdma/rdma_netlink.h

index 67519730b1acaf76f6b0cfd4727fbfabfdbd442c..210057fef7bdc505b4cd13e1eb3399fbee5f3035 100644 (file)
@@ -154,6 +154,8 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
        [RDMA_NLDEV_NET_NS_FD]                  = { .type = NLA_U32 },
        [RDMA_NLDEV_SYS_ATTR_NETNS_MODE]        = { .type = NLA_U8 },
        [RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK]      = { .type = NLA_U8 },
+       [RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX]  = { .type = NLA_U32 },
+       [RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC] = { .type = NLA_U8 },
 };
 
 static int put_driver_name_print_type(struct sk_buff *msg, const char *name,
@@ -2264,6 +2266,99 @@ static int nldev_stat_get_dumpit(struct sk_buff *skb,
        return ret;
 }
 
+static int nldev_stat_get_counter_status_doit(struct sk_buff *skb,
+                                             struct nlmsghdr *nlh,
+                                             struct netlink_ext_ack *extack)
+{
+       struct nlattr *tb[RDMA_NLDEV_ATTR_MAX], *table, *entry;
+       struct rdma_hw_stats *stats;
+       struct ib_device *device;
+       struct sk_buff *msg;
+       u32 devid, port;
+       int ret, i;
+
+       ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+                         nldev_policy, extack);
+       if (ret || !tb[RDMA_NLDEV_ATTR_DEV_INDEX] ||
+           !tb[RDMA_NLDEV_ATTR_PORT_INDEX])
+               return -EINVAL;
+
+       devid = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+       device = ib_device_get_by_index(sock_net(skb->sk), devid);
+       if (!device)
+               return -EINVAL;
+
+       port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
+       if (!rdma_is_port_valid(device, port)) {
+               ret = -EINVAL;
+               goto err;
+       }
+
+       stats = ib_get_hw_stats_port(device, port);
+       if (!stats) {
+               ret = -EINVAL;
+               goto err;
+       }
+
+       msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+       if (!msg) {
+               ret = -ENOMEM;
+               goto err;
+       }
+
+       nlh = nlmsg_put(
+               msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq,
+               RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_STAT_GET_STATUS),
+               0, 0);
+
+       ret = -EMSGSIZE;
+       if (fill_nldev_handle(msg, device) ||
+           nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port))
+               goto err_msg;
+
+       table = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS);
+       if (!table)
+               goto err_msg;
+
+       mutex_lock(&stats->lock);
+       for (i = 0; i < stats->num_counters; i++) {
+               entry = nla_nest_start(msg,
+                                      RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY);
+               if (!entry)
+                       goto err_msg_table;
+
+               if (nla_put_string(msg,
+                                  RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_NAME,
+                                  stats->descs[i].name) ||
+                   nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX, i))
+                       goto err_msg_entry;
+
+               if ((stats->descs[i].flags & IB_STAT_FLAG_OPTIONAL) &&
+                   (nla_put_u8(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC,
+                               !test_bit(i, stats->is_disabled))))
+                       goto err_msg_entry;
+
+               nla_nest_end(msg, entry);
+       }
+       mutex_unlock(&stats->lock);
+
+       nla_nest_end(msg, table);
+       nlmsg_end(msg, nlh);
+       ib_device_put(device);
+       return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid);
+
+err_msg_entry:
+       nla_nest_cancel(msg, entry);
+err_msg_table:
+       mutex_unlock(&stats->lock);
+       nla_nest_cancel(msg, table);
+err_msg:
+       nlmsg_free(msg);
+err:
+       ib_device_put(device);
+       return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
        [RDMA_NLDEV_CMD_GET] = {
                .doit = nldev_get_doit,
@@ -2353,6 +2448,9 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
                .dump = nldev_res_get_mr_raw_dumpit,
                .flags = RDMA_NL_ADMIN_PERM,
        },
+       [RDMA_NLDEV_CMD_STAT_GET_STATUS] = {
+               .doit = nldev_stat_get_counter_status_doit,
+       },
 };
 
 void __init nldev_init(void)
index 75a1ae2311d86345bbcc096a26fc98319b9c327f..e50c357367db03e5b24719c08e9f0f1999021d52 100644 (file)
@@ -297,6 +297,8 @@ enum rdma_nldev_command {
 
        RDMA_NLDEV_CMD_RES_SRQ_GET, /* can dump */
 
+       RDMA_NLDEV_CMD_STAT_GET_STATUS,
+
        RDMA_NLDEV_NUM_OPS
 };
 
@@ -549,6 +551,9 @@ enum rdma_nldev_attr {
 
        RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK,       /* u8 */
 
+       RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX,   /* u32 */
+       RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC, /* u8 */
+
        /*
         * Always the end
         */