Hi list,
I am using compat-wireless-20132-04-16 with kernel 3.3.8.
I experienced a kernel oops after applying the "get_expected_throughput"
patches series from Antonio Quartulli.
It happens when a station_dump is issued from userspace while an association
has begun but does not complete immediately due to bad channel.
After some debugging I made this patch which solves the issue:
===========================================================================
--- a/net/mac80211/rc80211_minstrel.c
+++ b/net/mac80211/rc80211_minstrel.c
@@@ -708,6 +708,15 @@ static u32 minstrel_get_expected_throughput(
struct minstrel_sta_info *mi = priv_sta;
int idx = mi->max_tp_rate[0];
+ if (!mi->r) {
+ /*
+ * Happens if get_expected_throughput() is called by
+ * ieee80211_dump_station during association from HT
+ * client to AP, before getting the HT IE from the AP
+ */
+ printk(KERN_ERR "minstrel: uninitialized station rates");
+ return 0; /* value suggested by J. Berg 25-apr-2014 */
+ }
/* convert pkt per sec in kbps (1200 is the average pkt size used
for
* computing cur_tp
*/
===========================================================================
With this patch I get this trace (the timeout is expected, since the SNR is
very low):
[ 359.798796] wlan0: authenticate with 92:a4:de:21:4f:85
[ 359.813180] wlan0: direct probe to 92:a4:de:21:4f:85 (try 1/3)
[ 359.922306] wlan0: direct probe to 92:a4:de:21:4f:85 (try 2/3)
[ 360.005424] minstrel: uninitialized station rates
[ 360.030318] wlan0: direct probe to 92:a4:de:21:4f:85 (try 3/3)
[ 360.138294] wlan0: authentication with 92:a4:de:21:4f:85 timed out
Maybe station_dump should not dump the stations just created by an
yet-unanswered authentication request?
Or is my patch the right thing to do?
Regards
Jean-Pierre Tosoni
Great, it works! No more oops, and my debug printk never appears anymore.
But it looks like one line is missing in your patch, 12/15 lines announced
but only 11/14 lines present?
Thanks you so much!
-----Message d'origine-----
De?: Felix Fietkau [mailto:[email protected]]
Envoy??: mardi 22 juillet 2014 15:47
??: Jean-Pierre Tosoni; [email protected]
Objet?: Re: get_expected_throughput introduces a kernel oops on
sta_set_sinfo calls
On 2014-07-22 11:11, Jean-Pierre Tosoni wrote:
> Hi list,
>
> I am using compat-wireless-20132-04-16 with kernel 3.3.8.
> I experienced a kernel oops after applying the "get_expected_throughput"
> patches series from Antonio Quartulli.
> It happens when a station_dump is issued from userspace while an
> association has begun but does not complete immediately due to bad
channel.
> After some debugging I made this patch which solves the issue:
Please try this patch:
---
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -472,12 +472,15 @@ static void sta_set_sinfo(struct sta_info *sta, struct
station_info *sinfo) {
struct ieee80211_sub_if_data *sdata = sta->sdata;
struct ieee80211_local *local = sdata->local;
- struct rate_control_ref *ref = local->rate_ctrl;
+ struct rate_control_ref *ref = NULL;
struct timespec uptime;
u64 packets = 0;
u32 thr = 0;
int i, ac;
+ if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
+ ref = local->rate_ctrl;
+
sinfo->generation = sdata->local->sta_generation;
sinfo->filled = STATION_INFO_INACTIVE_TIME |
On 2014-07-22 11:11, Jean-Pierre Tosoni wrote:
> Hi list,
>
> I am using compat-wireless-20132-04-16 with kernel 3.3.8.
> I experienced a kernel oops after applying the "get_expected_throughput"
> patches series from Antonio Quartulli.
> It happens when a station_dump is issued from userspace while an association
> has begun but does not complete immediately due to bad channel.
> After some debugging I made this patch which solves the issue:
Please try this patch:
---
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -472,12 +472,15 @@ static void sta_set_sinfo(struct sta_info *sta, struct station_info *sinfo)
{
struct ieee80211_sub_if_data *sdata = sta->sdata;
struct ieee80211_local *local = sdata->local;
- struct rate_control_ref *ref = local->rate_ctrl;
+ struct rate_control_ref *ref = NULL;
struct timespec uptime;
u64 packets = 0;
u32 thr = 0;
int i, ac;
+ if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
+ ref = local->rate_ctrl;
+
sinfo->generation = sdata->local->sta_generation;
sinfo->filled = STATION_INFO_INACTIVE_TIME |
In your line-wrapped reply there are 11/14 lines. In my original mail,
I'm counting 12/15 ;) The difference is the '{' part.
I will submit this change as a proper patch now, thanks for testing.
- Felix
On 2014-07-22 16:55, Jean-Pierre Tosoni wrote:
> Great, it works! No more oops, and my debug printk never appears anymore.
>
> But it looks like one line is missing in your patch, 12/15 lines announced
> but only 11/14 lines present?
>
> Thanks you so much!
>
> -----Message d'origine-----
> De : Felix Fietkau [mailto:[email protected]]
> Envoy? : mardi 22 juillet 2014 15:47
> ? : Jean-Pierre Tosoni; [email protected]
> Objet : Re: get_expected_throughput introduces a kernel oops on
> sta_set_sinfo calls
>
> On 2014-07-22 11:11, Jean-Pierre Tosoni wrote:
>> Hi list,
>>
>> I am using compat-wireless-20132-04-16 with kernel 3.3.8.
>> I experienced a kernel oops after applying the "get_expected_throughput"
>> patches series from Antonio Quartulli.
>> It happens when a station_dump is issued from userspace while an
>> association has begun but does not complete immediately due to bad
> channel.
>> After some debugging I made this patch which solves the issue:
> Please try this patch:
> ---
> --- a/net/mac80211/cfg.c
> +++ b/net/mac80211/cfg.c
> @@ -472,12 +472,15 @@ static void sta_set_sinfo(struct sta_info *sta, struct
> station_info *sinfo) {
> struct ieee80211_sub_if_data *sdata = sta->sdata;
> struct ieee80211_local *local = sdata->local;
> - struct rate_control_ref *ref = local->rate_ctrl;
> + struct rate_control_ref *ref = NULL;
> struct timespec uptime;
> u64 packets = 0;
> u32 thr = 0;
> int i, ac;
>
> + if (test_sta_flag(sta, WLAN_STA_RATE_CONTROL))
> + ref = local->rate_ctrl;
> +
> sinfo->generation = sdata->local->sta_generation;
>
> sinfo->filled = STATION_INFO_INACTIVE_TIME |
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>