michaelh/zephyr
1
0
Fork 0
Code Releases Activity

net: Collect net_pkt TX timing statistics when passing IP stack

Browse source 
Collect information how long net_pkt has travelled in IP stack
in certain points. See network documentation what these points
are and how to get information about the timings. This initial
commit adds support to TX timing collection.

Signed-off-by: Jukka Rissanen <jukka.rissanen@linux.intel.com>
This commit is contained in:
Jukka Rissanen 2020-05-13 11:09:10 +03:00 • committed by Jukka Rissanen
commit 76398945c3
7 changed files with 208 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

54
subsys/net/ip/net_if.c
View file

@ -183,6 +183,23 @@ static inline void net_context_send_cb(struct net_context *context,
}
}
static void update_txtime_stats_detail(struct net_pkt *pkt,
uint32_t start_time, uint32_t stop_time)
{
uint32_t val, prev = start_time;
int i;
for (i = 0; i < net_pkt_stats_tick_count(pkt); i++) {
if (!net_pkt_stats_tick(pkt)[i]) {
break;
}
val = net_pkt_stats_tick(pkt)[i] - prev;
prev = net_pkt_stats_tick(pkt)[i];
net_pkt_stats_tick(pkt)[i] = val;
}
}
static bool net_if_tx(struct net_if *iface, struct net_pkt *pkt)
{
struct net_linkaddr ll_dst = {
@ -240,6 +257,13 @@ static bool net_if_tx(struct net_if *iface, struct net_pkt *pkt)
memcpy(&start_timestamp, net_pkt_timestamp(pkt),
sizeof(start_timestamp));
pkt_priority = net_pkt_priority(pkt);
if (IS_ENABLED(CONFIG_NET_PKT_TXTIME_STATS_DETAIL)) {
/* Make sure the statistics information is not
* lost by keeping the net_pkt over L2 send.
*/
net_pkt_ref(pkt);
}
}
status = net_if_l2(iface)->send(iface, pkt);
@ -251,11 +275,35 @@ static bool net_if_tx(struct net_if *iface, struct net_pkt *pkt)
}
}
if (IS_ENABLED(CONFIG_NET_PKT_TXTIME_STATS) && status >= 0) {
if (IS_ENABLED(CONFIG_NET_PKT_TXTIME_STATS)) {
uint32_t end_tick = k_cycle_get_32();
net_pkt_set_tx_stats_tick(pkt, end_tick);
net_stats_update_tc_tx_time(iface,
pkt_priority,
start_timestamp.nanosecond,
k_cycle_get_32());
end_tick);
if (IS_ENABLED(CONFIG_NET_PKT_TXTIME_STATS_DETAIL)) {
update_txtime_stats_detail(
pkt,
start_timestamp.nanosecond,
end_tick);
net_stats_update_tc_tx_time_detail(
iface, pkt_priority,
net_pkt_stats_tick(pkt));
/* For TCP connections, we might keep the pkt
* longer so that we can resend it if needed.
* Because of that we need to clear the
* statistics here.
*/
net_pkt_stats_tick_reset(pkt);
net_pkt_unref(pkt);
}
}
} else {
@ -303,6 +351,8 @@ static void process_tx_packet(struct k_work *work)
pkt = CONTAINER_OF(work, struct net_pkt, work);
net_pkt_set_tx_stats_tick(pkt, k_cycle_get_32());
iface = net_pkt_iface(pkt);
net_if_tx(iface, pkt);