Currently MBSSID parameters in struct ieee80211_bss_conf
are not reset upon connection. This could be problematic
with some drivers in a scenario where the device first
connects to a non-transmit BSS and then connects to a
transmit BSS of a Multi BSS AP. The MBSSID parameters
which are set after connecting to a non-transmit BSS will
not be reset and the same parameters will be passed on to
the driver during the subsequent connection to a transmit
BSS of a Multi BSS AP.
For example, firmware running on the ath11k device uses the
Multi BSS data for tracking the beacon of a non-transmit BSS
and reports the driver when there is a beacon miss. If we do
not reset the MBSSID parameters during the subsequent
connection to a transmit BSS, then the driver would have
wrong MBSSID data and FW would be looking for an incorrect
BSSID in the MBSSID beacon of a Multi BSS AP and reports
beacon loss leading to an unstable connection.
Reset the MBSSID parameters upon every connection to solve this
problem.
Signed-off-by: Manikanta Pubbisetty <[email protected]>
---
net/mac80211/mlme.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/net/mac80211/mlme.c b/net/mac80211/mlme.c
index b857915881e0..07a96f7c5dc3 100644
--- a/net/mac80211/mlme.c
+++ b/net/mac80211/mlme.c
@@ -3657,6 +3657,12 @@ static bool ieee80211_assoc_success(struct ieee80211_sub_if_data *sdata,
cbss->transmitted_bss->bssid);
bss_conf->bssid_indicator = cbss->max_bssid_indicator;
bss_conf->bssid_index = cbss->bssid_index;
+ } else {
+ bss_conf->nontransmitted = false;
+ memset(bss_conf->transmitter_bssid, 0,
+ sizeof(bss_conf->transmitter_bssid));
+ bss_conf->bssid_indicator = 0;
+ bss_conf->bssid_index = 0;
}
/*
--
2.35.1
On Wed, 2022-04-27 at 23:53 +0530, Manikanta Pubbisetty wrote:
> Currently MBSSID parameters in struct ieee80211_bss_conf
> are not reset upon connection. This could be problematic
> with some drivers in a scenario where the device first
> connects to a non-transmit BSS and then connects to a
> transmit BSS of a Multi BSS AP. The MBSSID parameters
> which are set after connecting to a non-transmit BSS will
> not be reset and the same parameters will be passed on to
> the driver during the subsequent connection to a transmit
> BSS of a Multi BSS AP.
>
> For example, firmware running on the ath11k device uses the
> Multi BSS data for tracking the beacon of a non-transmit BSS
> and reports the driver when there is a beacon miss. If we do
> not reset the MBSSID parameters during the subsequent
> connection to a transmit BSS, then the driver would have
> wrong MBSSID data and FW would be looking for an incorrect
> BSSID in the MBSSID beacon of a Multi BSS AP and reports
> beacon loss leading to an unstable connection.
>
> Reset the MBSSID parameters upon every connection to solve this
> problem.
Oops. Please add (or at least provide) a Fixes line though.
johannes
On 4/28/2022 12:41 AM, Johannes Berg wrote:
> On Wed, 2022-04-27 at 23:53 +0530, Manikanta Pubbisetty wrote:
>> Currently MBSSID parameters in struct ieee80211_bss_conf
>> are not reset upon connection. This could be problematic
>> with some drivers in a scenario where the device first
>> connects to a non-transmit BSS and then connects to a
>> transmit BSS of a Multi BSS AP. The MBSSID parameters
>> which are set after connecting to a non-transmit BSS will
>> not be reset and the same parameters will be passed on to
>> the driver during the subsequent connection to a transmit
>> BSS of a Multi BSS AP.
>>
>> For example, firmware running on the ath11k device uses the
>> Multi BSS data for tracking the beacon of a non-transmit BSS
>> and reports the driver when there is a beacon miss. If we do
>> not reset the MBSSID parameters during the subsequent
>> connection to a transmit BSS, then the driver would have
>> wrong MBSSID data and FW would be looking for an incorrect
>> BSSID in the MBSSID beacon of a Multi BSS AP and reports
>> beacon loss leading to an unstable connection.
>>
>> Reset the MBSSID parameters upon every connection to solve this
>> problem.
>
> Oops. Please add (or at least provide) a Fixes line though.
>
Sure Johannes, just sent V2 patch with the Fixes line added.
Thanks,
Manikanta