Re: [PATCH 0/2] Initial 11ac support using ath10k

Re: [PATCH 0/2] Initial 11ac support using ath10k

[Kalle Valo]

Hi Sven,

adding ath10k@lists.infradead.org.

Sven Eckelmann <sven@open-mesh.com> writes:

> Hi,
>
> support for ath10k cards was added a while ago to OpenWrt. Unfortunately, the 
> actual 802.11ac support was never added (patches on the mailing lists were 
> just ignored). This is a rather unpleasant situation when trying to get some 
> 11ac test hardware working.

Yeah, it would be great to get proper ath10k support to OpenWrt. Thanks
for working on this!

> I've picked up the patches and rebased them on top of the current master 
> branch/trunk.

Where did you take the ath10k patches? ath10k is under heavy
development, so I strongly recommend taking patches directly from
ath-next branch of my ath.git. That way you would be using the latest
and greatest.

> Just as general information to the state of ath10k, my current test showed 
> some problems with it:
>
>  * Samsung GT-I9300 only got horrible slow connections to an QCA9880 but
>    an Intel N6205 (using iwlwifi from v3.12) or Samsung S4 worked
>    "fine"

Could you report the I9300 issue to the ath10k list with more details,
please? We can try to investigate it.

>  * Adhoc mode doesn't work at all

That was with 10.1 firmware? AFAIK it doesn't support adhoc at all. I
think we need to add a firmware feature flag to disable adhoc with that
firmware.

>  * TX bitrate information is bogus (iw XXX station dump)

Yeah, this is annoying. But firmware doesn't report this to the host at
all so there is not much ath10k can do right now. But I'll send a query
to the firmware team if they could provide that information to ath10k.

>  * Crashed relative often on reconfiguration:
>    Data bus error, epc 87b7a864, ra 87b7a864

On some QCA988x boards cold reset seems to cause problems on the PCI
bus. We are investigating this.

>  * Sometimes adding of station seems to fail:
>    ath10k: Failed to add peer 5c:0a:5b:4e:6a:c4 for vdev 1 when adding a new
>    sta: -145

This is new to me. Can you report this to the ath10k list, please?

>  * Adhoc interface +  "iw dev wlan0 scan trigger" just crashes the firmware:
>    [ 296.160000] ath10k: firmware crashed!
>    [ 296.170000] ath10k: hardware name qca988x hw2.0 version 0x4100016c
>    [ 296.170000] ath10k: firmware version: 65.467.0.0

What firmware version was this? The version printed here seems to be
corrupted.

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: [PATCH 0/2] Initial 11ac support using ath10k

[Kalle Valo]

(resending my mail due to a stupid rule in ath10k-devel, I need to be
subcribed before posting to that list)

Hi Sven,

adding ath10k@lists.infradead.org.

Sven Eckelmann <sven@open-mesh.com> writes:

> Hi,
>
> support for ath10k cards was added a while ago to OpenWrt. Unfortunately, the 
> actual 802.11ac support was never added (patches on the mailing lists were 
> just ignored). This is a rather unpleasant situation when trying to get some 
> 11ac test hardware working.

Yeah, it would be great to get proper ath10k support to OpenWrt. Thanks
for working on this!

> I've picked up the patches and rebased them on top of the current master 
> branch/trunk.

Where did you take the ath10k patches? ath10k is under heavy
development, so I strongly recommend taking patches directly from
ath-next branch of my ath.git. That way you would be using the latest
and greatest.

> Just as general information to the state of ath10k, my current test showed 
> some problems with it:
>
>  * Samsung GT-I9300 only got horrible slow connections to an QCA9880 but
>    an Intel N6205 (using iwlwifi from v3.12) or Samsung S4 worked
>    "fine"

Could you report the I9300 issue to the ath10k list with more details,
please? We can try to investigate it.

>  * Adhoc mode doesn't work at all

That was with 10.1 firmware? AFAIK it doesn't support adhoc at all. I
think we need to add a firmware feature flag to disable adhoc with that
firmware.

>  * TX bitrate information is bogus (iw XXX station dump)

Yeah, this is annoying. But firmware doesn't report this to the host at
all so there is not much ath10k can do right now. But I'll send a query
to the firmware team if they could provide that information to ath10k.

>  * Crashed relative often on reconfiguration:
>    Data bus error, epc 87b7a864, ra 87b7a864

On some QCA988x boards cold reset seems to cause problems on the PCI
bus. We are investigating this.

>  * Sometimes adding of station seems to fail:
>    ath10k: Failed to add peer 5c:0a:5b:4e:6a:c4 for vdev 1 when adding a new
>    sta: -145

This is new to me. Can you report this to the ath10k list, please?

>  * Adhoc interface +  "iw dev wlan0 scan trigger" just crashes the firmware:
>    [ 296.160000] ath10k: firmware crashed!
>    [ 296.170000] ath10k: hardware name qca988x hw2.0 version 0x4100016c
>    [ 296.170000] ath10k: firmware version: 65.467.0.0

What firmware version was this? The version printed here seems to be
corrupted.

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: [PATCH 0/2] Initial 11ac support using ath10k

[Yeoh Chun-Yeow]

>>  * Adhoc mode doesn't work at all
>
> That was with 10.1 firmware? AFAIK it doesn't support adhoc at all. I
> think we need to add a firmware feature flag to disable adhoc with that
> firmware.

I managed to try out the firmware version 999.999.0.636 on adhoc mode.
It works only after I fool the drv_get_tsf (currently not implemented
in ath10k) to return 0.

Also, the firmware seems only allow the legacy rate and not HT/VHT
rate in adhoc mode. FYI, drv_sta_rc_update also not implemented in
ath10k.

Let me know if you have any updates. Thanks

----
Chun-Yeow

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: [PATCH 0/2] Initial 11ac support using ath10k

[Sven Eckelmann]

[-- Attachment #1.1: Type: text/plain, Size: 3194 bytes --]

On Saturday 18 January 2014 01:14:28 Kalle Valo wrote:
> > I've picked up the patches and rebased them on top of the current master
> > branch/trunk.
> 
> Where did you take the ath10k patches? ath10k is under heavy
> development, so I strongly recommend taking patches directly from
> ath-next branch of my ath.git. That way you would be using the latest
> and greatest.

They were taken from the wireless-testing repo. But it seems that (for obvious 
reasons) the OpenWrt maintainers doesn't want so apply so many patches. But it 
is a good source [1] for future tests.

> > Just as general information to the state of ath10k, my current test showed
> > 
> > some problems with it:
> >  * Samsung GT-I9300 only got horrible slow connections to an QCA9880 but
> >  
> >    an Intel N6205 (using iwlwifi from v3.12) or Samsung S4 worked
> >    "fine"
> 
> Could you report the I9300 issue to the ath10k list with more details,
> please? We can try to investigate it.

This is a little bit problematic. I could reproduce this problem quite well on 
a specific date and when I tried some other day it worked fine. So it is hard 
to report anything. The only thing I could report would be the "Failed to add 
peer" problem which seemed to happen at the same time.

Maybe I find a way to reproduce it reliable but I don't know how at the 
moment.

> >  * Adhoc mode doesn't work at all
> 
> That was with 10.1 firmware? AFAIK it doesn't support adhoc at all. I
> think we need to add a firmware feature flag to disable adhoc with that
> firmware.

It used the latest and greatest, stable(?) version of the firmware when the 
patchset was generated: ap/firmware-2.bin_10.1.467-1

The statement "doesn't support adhoc at all" makes me a little bit nervous. 
Would you say that it is better to completely ignore the "ap" directory of the 
firmware repo [2]? Which benefits does the ap firmware has over the ath10k 
firmware?

At the end I would most likely run something like this:

 * 1 AP dev with WPA (1/2, TKIP and/or CCMP, PSK or Enterprise)
 * 1 AP dev without WPA
 * 1 adhoc dev with IBSS RSN (running something like batman-adv over
   infrastructure mode is a little bit lame ;) )

But it also has to be able to use a dev in managed mode.

This worked quite well in the past with ath9k and thus I was checking ath10k 
to decide if it is an interesting platform/driver for the future.

> >  * Sometimes adding of station seems to fail:
> >    ath10k: Failed to add peer 5c:0a:5b:4e:6a:c4 for vdev 1 when adding a
> >    new
> >    sta: -145
> 
> This is new to me. Can you report this to the ath10k list, please?

I will try to reproduce it and report it in a separate thread when I have more 
useful information.

> >  * Adhoc interface +  "iw dev wlan0 scan trigger" just crashes the 
firmware:
> >    [ 296.160000] ath10k: firmware crashed!
> >    [ 296.170000] ath10k: hardware name qca988x hw2.0 version 0x4100016c
> >    [ 296.170000] ath10k: firmware version: 65.467.0.0
> 
> What firmware version was this? The version printed here seems to be
> corrupted.

ap/firmware-2.bin_10.1.467-1

Kind regards,
	Sven

[1] https://github.com/kvalo/ath
[2] https://github.com/kvalo/ath10k-firmware

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

[-- Attachment #2: Type: text/plain, Size: 146 bytes --]

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: [PATCH 0/2] Initial 11ac support using ath10k

[Kalle Valo]

Sven Eckelmann <sven@open-mesh.com> writes:

> On Saturday 18 January 2014 01:14:28 Kalle Valo wrote:
>
>> > Just as general information to the state of ath10k, my current test showed
>> > 
>> > some problems with it:
>> >  * Samsung GT-I9300 only got horrible slow connections to an QCA9880 but
>> >  
>> >    an Intel N6205 (using iwlwifi from v3.12) or Samsung S4 worked
>> >    "fine"
>> 
>> Could you report the I9300 issue to the ath10k list with more details,
>> please? We can try to investigate it.
>
> This is a little bit problematic. I could reproduce this problem quite well on 
> a specific date and when I tried some other day it worked fine. So it is hard 
> to report anything. The only thing I could report would be the "Failed to add 
> peer" problem which seemed to happen at the same time.
>
> Maybe I find a way to reproduce it reliable but I don't know how at the 
> moment.

The bug report doesn't need to be perfect, not all bugs are easily
reproducable, but just knowing about a bug would even help. That's why
ask people to send bug reports to ath10k@lists.infradead.org.

>> >  * Adhoc mode doesn't work at all
>> 
>> That was with 10.1 firmware? AFAIK it doesn't support adhoc at all. I
>> think we need to add a firmware feature flag to disable adhoc with that
>> firmware.
>
> It used the latest and greatest, stable(?) version of the firmware when the 
> patchset was generated: ap/firmware-2.bin_10.1.467-1
>
> The statement "doesn't support adhoc at all" makes me a little bit nervous. 
> Would you say that it is better to completely ignore the "ap" directory of the 
> firmware repo [2]? Which benefits does the ap firmware has over the ath10k 
> firmware?

So we have two firmware branches right now, "main" and "10.1". 10.1 is
solely focused on AP mode, nothing else. The main branch again has more
features, like P2P, but doesn't support AP mode as well as 10.1 does.

-- 
Kalle Valo

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k