iwlwifi: mvm: check time event is over before disconnecting

On the 7260 time event was often ended before end_time and connections failed
with "No association and the time event is over already...".
This checks that the time event is actually over before disconnecting.

Signed-off-by: Filip Ayazi <filipayazi@gmail.com>
---
 drivers/net/wireless/iwlwifi/mvm/time-event.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

On Sat, Feb 28, 2015 at 1:10 AM, Filip Ayazi <filipayazi@gmail.com> wrote:
> On the 7260 time event was often ended before end_time and connections failed
> with "No association and the time event is over already...".
> This checks that the time event is actually over before disconnecting.
>
> Signed-off-by: Filip Ayazi <filipayazi@gmail.com>

While this patch is wrong I'd like to know if it helps.
The patch is wrong because even if the driver thinks the time event
should still be running, if the firmware indicates it has already
ended, we should still disconnect. The reason for the disconnection is
that we can't be sure that the firmware will have the proper timing
for the beacon and hence for the DTIM. Both are critical to get a
decent behavior while saving power.
In any case, I doubt this patch does actually something because the
firmware and the driver should really be close in their timings. If
they aren't, it is fundamental bug in the firmware. While the firmware
does have bugs just like any piece of software, I doubt it has such
big ones :)

> ---
>  drivers/net/wireless/iwlwifi/mvm/time-event.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/iwlwifi/mvm/time-event.c b/drivers/net/wireless/iwlwifi/mvm/time-event.c
> index 54fafbf..b0aa892 100644
> --- a/drivers/net/wireless/iwlwifi/mvm/time-event.c
> +++ b/drivers/net/wireless/iwlwifi/mvm/time-event.c
> @@ -256,7 +256,8 @@ static void iwl_mvm_te_handle_notif(struct iwl_mvm *mvm,
>                 }
>         }
>
> -       if (le32_to_cpu(notif->action) & TE_V2_NOTIF_HOST_EVENT_END) {
> +       if (le32_to_cpu(notif->action) & TE_V2_NOTIF_HOST_EVENT_END &&
> +           time_after(jiffies, te_data->end_jiffies)) {
>                 IWL_DEBUG_TE(mvm,
>                              "TE ended - current time %lu, estimated end %lu\n",
>                              jiffies, te_data->end_jiffies);
> --
> 1.9.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On 02/28/2015 08:55 PM, Emmanuel Grumbach wrote:
> On Sat, Feb 28, 2015 at 1:10 AM, Filip Ayazi <filipayazi@gmail.com> wrote:
>> On the 7260 time event was often ended before end_time and connections failed
>> with "No association and the time event is over already...".
>> This checks that the time event is actually over before disconnecting.
>>
>> Signed-off-by: Filip Ayazi <filipayazi@gmail.com>
> While this patch is wrong I'd like to know if it helps.
It helped here (7260 on Lenovo E440).
I get a TE with unknown action message and successful connection now,
used to be TE ended with current time < end time and a disconnect,
even on networks with excellent signal strength. It used to happen
about 30% of the time with no apparent pattern, sometimes restarting
the interface fixed the issue.
> The patch is wrong because even if the driver thinks the time event
> should still be running, if the firmware indicates it has already
> ended, we should still disconnect. The reason for the disconnection is
> that we can't be sure that the firmware will have the proper timing
> for the beacon and hence for the DTIM. Both are critical to get a
> decent behavior while saving power.
You are right, few hours ago, after waking from suspend, I could not switch
the interface on (ifconfig said SIOCSIFFLAGS: Input/output error), reloading
iwlmvm module helped, so this was likely the cause.
> In any case, I doubt this patch does actually something because the
> firmware and the driver should really be close in their timings. If
> they aren't, it is fundamental bug in the firmware. While the firmware
> does have bugs just like any piece of software, I doubt it has such
> big ones :)
I do not see any other explanation apart from faulty hardware,
which doesn't seem likely :)
>
>> ---
>>   drivers/net/wireless/iwlwifi/mvm/time-event.c | 3 ++-
>>   1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/wireless/iwlwifi/mvm/time-event.c b/drivers/net/wireless/iwlwifi/mvm/time-event.c
>> index 54fafbf..b0aa892 100644
>> --- a/drivers/net/wireless/iwlwifi/mvm/time-event.c
>> +++ b/drivers/net/wireless/iwlwifi/mvm/time-event.c
>> @@ -256,7 +256,8 @@ static void iwl_mvm_te_handle_notif(struct iwl_mvm *mvm,
>>                  }
>>          }
>>
>> -       if (le32_to_cpu(notif->action) & TE_V2_NOTIF_HOST_EVENT_END) {
>> +       if (le32_to_cpu(notif->action) & TE_V2_NOTIF_HOST_EVENT_END &&
>> +           time_after(jiffies, te_data->end_jiffies)) {
>>                  IWL_DEBUG_TE(mvm,
>>                               "TE ended - current time %lu, estimated end %lu\n",
>>                               jiffies, te_data->end_jiffies);
>> --
>> 1.9.1
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html