On 6/14/19 6:38 AM, Johannes Berg wrote:
On Mon, 2019-04-15 at 10:21 -0700, greearb@xxxxxxxxxxxxxxx wrote:
From: Ben Greear <greearb@xxxxxxxxxxxxxxx>
Report time stamp of when sta became associated.
I guess that makes sense.
Signed-off-by: Ben Greear <greearb@xxxxxxxxxxxxxxx>
---
include/net/cfg80211.h | 2 ++
include/uapi/linux/nl80211.h | 2 ++
net/wireless/nl80211.c | 1 +
3 files changed, 5 insertions(+)
diff --git a/include/net/cfg80211.h b/include/net/cfg80211.h
index f49eb1464b7a..a3ad78b9d9f4 100644
--- a/include/net/cfg80211.h
+++ b/include/net/cfg80211.h
@@ -1268,6 +1268,7 @@ struct cfg80211_tid_stats {
* indicate the relevant values in this struct for them
* @connected_time: time(in secs) since a station is last connected
* @inactive_time: time since last station activity (tx/rx) in milliseconds
+ * @assoc_at_ms: time in ms of the last association
I think the "_at_ms" doesn't really make sense. "time in ms" also
doesn't make sense as documentation. I think you need to say what should
be assigned here.
Also, you're actually using ktime_get_real() for this, which again
doesn't make much sense. For exposing an absolute time, I think we're
better off with CLOCK_BOOTTIME like we use elsewhere:
The point of my patch was to allow 'iw' to return a more precise time
that the station has been associated, so I am not sure that BOOTIME is
a good thing to use for that?
Here are the pertinent parts of my iw patches:
diff --git a/station.c b/station.c
index 25cbbc3..e7738cc 100644
--- a/station.c
+++ b/station.c
@@ -314,6 +314,12 @@ static int print_sta_handler(struct nl_msg *msg, void *arg)
[NL80211_STA_INFO_ACK_SIGNAL_AVG] = { .type = NLA_U8 },
};
char *chain;
+ struct timeval now;
+ unsigned long long now_ms;
+
+ gettimeofday(&now, NULL);
+ now_ms = now.tv_sec * 1000;
+ now_ms += (now.tv_usec / 1000);
nla_parse(tb, NL80211_ATTR_MAX, genlmsg_attrdata(gnlh, 0),
genlmsg_attrlen(gnlh, 0), NULL);
@@ -557,8 +563,11 @@ static int print_sta_handler(struct nl_msg *msg, void *arg)
if (sinfo[NL80211_STA_INFO_CONNECTED_TIME])
printf("\n\tconnected time:\t%u seconds",
nla_get_u32(sinfo[NL80211_STA_INFO_CONNECTED_TIME]));
+ if (sinfo[NL80211_STA_INFO_ASSOC_AT_MS])
+ printf("\n\tassociated at:\t%llu ms",
+ (unsigned long long)nla_get_u64(sinfo[NL80211_STA_INFO_ASSOC_AT_MS]));
- printf("\n");
+ printf("\n\tcurrent time:\t%llu ms\n", now_ms);
return NL_SKIP;
}
Thanks,
Ben
* @boottime_ns: CLOCK_BOOTTIME timestamp the frame was received at, this is
* needed only for beacons and probe responses that update the scan cache.
+ * @NL80211_STA_INFO_ASSOC_AT_MS: Timestamp of last association
_ASSOC_TIMESTAMP or something would make more sense too as the attribute
name, and clearly the documentation needs to spell out the semantics
here too.
johannes
--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc http://www.candelatech.com
