[BUG] QWLE900V5-18 - cannot achievie 3 spatial streams link

Wed Dec 11 14:43:37 EST 2013

On 11 December 2013 13:19, Tomasz Tajmajer <tt at elnovel.com> wrote:
>
> On 09.12.2013 13:10, Tomasz Tajmajer wrote:
>>
>>
>> On 09.12.2013 12:29, Tomasz Tajmajer wrote:
>>>
>>>
>>> On 09.12.2013 11:13, Michal Kazior wrote:
>>>>
>>>> On 9 December 2013 11:03, Tomasz Tajmajer <tt at elnovel.com> wrote:
>>>>>
>>>>> On 09.12.2013 10:04, Kalle Valo wrote:
>>>>>>
>>>>>> Janusz Dziedzic <janusz.dziedzic at tieto.com> writes:
>>>>>>
>>>>>>> On 6 December 2013 17:33, Tomasz Tajmajer <tt at elnovel.com> wrote:
>>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Recently, I've been making quite a lot of tests with two
>>>>>>>> QWLE900V5-18
>>>>>>>> cards.
>>>>>>>> I had some issues with the drivers (e.g. regarding the regulatory
>>>>>>>> domain),
>>>>>>>> but I've managed to fix them on my own. However, there is one issue
>>>>>>>> that
>>>>>>>> I
>>>>>>>> cannot deal with.
>>>>>>>> I'm trying to get the highest throughput possible using 3x3 MIMO.
>>>>>>>> The
>>>>>>>> problem is that the wifi card does not set higher rates than
>>>>>>>> 2x433.3Mbps
>>>>>>>> (MCS 9, VHT-NSS 2). The specification says that it should be
>>>>>>>> possible to
>>>>>>>> have 3 spatial streams.
>>>>>>>> First, I thought it is a problem with some options of hostapd, but
>>>>>>>> hostapd
>>>>>>>> even does not allows me to set the [RX-STBC-12] or [RX-STBC-123]
>>>>>>>> options.
>>>>>>>> Only [RX-STBC-1] may be set in the configuration (what in fact, does
>>>>>>>> not
>>>>>>>> disallows the 2x433.3Mbps link, which is another issue).
>>>>>>>> Driver info shows that the driver should support 3 spatial streams.
>>>>>>>> I am not sure where is the problem, is it some bug in hostapd, which
>>>>>>>> forces
>>>>>>>> the card to use 2 spatial streams only or is it the driver, which
>>>>>>>> does
>>>>>>>> not
>>>>>>>> enable 3 streams when it is possible. The physical connectivity
>>>>>>>> (antennas
>>>>>>>> etc.) is not the problem AFAIK.
>>>>>>>>
>>>>>>> When sta assoc to ap you will get dbg mesg (you can enable this eg.
>>>>>>> when loading ath10k_core module with param - debug_mask).
>>>>>>> "mac ht peer %pM mcs cnt %d nss %d\n" - nss is interesting here
>>>>>>
>>>>>> This debug message is enabled with ATH10K_DBG_MAC. More info here:
>>>>>>
>>>>>> http://wireless.kernel.org/en/users/Drivers/ath10k/debug
>>>>>>
>>>>>>> Also rx_mcs_set for STA from ath10k_peer_assoc_h_vht() could be
>>>>>>> interesting - unfortunetly we don't print this. But you can print
>>>>>>> this.
>>>>>>
>>>>>> Indeed, we should include that to the message as well.
>>>>>>
>>>>> I've enabled dbg messages and added rx_mcs_set and tx_mcs_set as you
>>>>> suggested.
>>>>> Here is the output for STA:
>>>>>
>>>>> [ 2278.320312] ath10k: htt target version 2.1
>>>>> [ 2278.320455] ath10k: mac vdev create 0 (add interface) type 2 subtype
>>>>> 0
>>>>> [ 2278.320741] ath10k: mac vdev 0 cts_prot 0
>>>>> [ 2278.320765] ath10k: mac vdev 0 slot_time 1
>>>>> [ 2278.320801] ath10k: mac vdev 0 preamble 1n
>>>>> [ 2278.321320] ath10k: mac config channel 5180 mhz
>>>>> [ 2278.321353] ath10k: mac vdev 0 psmode disable
>>>>> [ 2278.321448] IPv6: ADDRCONF(NETDEV_UP): wlan1: link is not ready
>>>>> [ 2278.321458] ath10k: mac vdev 0 psmode disable
>>>>> [ 2282.499925] wlan1: authenticate with 04:f0:21:0c:a2:3a
>>>>> [ 2282.499946] ath10k: mac config channel 5180 mhz
>>>>> [ 2282.499962] ath10k: mac vdev 0 beacon_interval 100
>>>>> [ 2282.499968] ath10k: mac vdev 0 create peer 04:f0:21:0c:a2:3a
>>>>> [ 2282.500109] ath10k: mac vdev 0 start 04:f0:21:0c:a2:3a
>>>>> [ 2282.500115] ath10k: mac vdev 0 start center_freq 5180 phymode
>>>>> 11ac-vht80
>>>>> [ 2282.506711] wlan1: direct probe to 04:f0:21:0c:a2:3a (try 1/3)
>>>>> [ 2282.708052] wlan1: send auth to 04:f0:21:0c:a2:3a (try 2/3)
>>>>> [ 2282.709096] wlan1: authenticated
>>>>> [ 2282.712122] wlan1: associate with 04:f0:21:0c:a2:3a (try 1/3)
>>>>> [ 2282.713907] wlan1: RX AssocResp from 04:f0:21:0c:a2:3a (capab=0x1
>>>>> status=0 aid=1)
>>>>> [ 2282.714331] ath10k: mac vdev 0 dtim_period 2
>>>>> [ 2282.714367] ath10k: mac vdev 0 slot_time 2
>>>>> [ 2282.714415] ath10k: mac ht peer 04:f0:21:0c:a2:3a mcs cnt 24 nss 3
>>>>> [ 2282.714421] ath10k: mac vht peer 04:f0:21:0c:a2:3a max_mpdu 65535
>>>>> flags
>>>>> 0x601b001 rx_mcs_set 0000ffea tx_mcs_set 0000ffea
>>>>> [ 2282.714427] ath10k: mac peer 04:f0:21:0c:a2:3a phymode 11ac-vht80
>>>>> [ 2282.714435] ath10k: mac vdev 0 up (associated) bssid
>>>>> 04:f0:21:0c:a2:3a
>>>>> aid 1
>>>>> [ 2282.714822] ath10k: mac vdev 0 psmode enable
>>>>> [ 2282.714869] wlan1: associated
>>>>>
>>>>> And for AP:
>>>>> [ 9460.458700] ath10k: htt target version 2.1
>>>>> [ 9460.458874] ath10k: mac vdev create 0 (add interface) type 1 subtype
>>>>> 0
>>>>> [ 9460.459120] ath10k: mac vdev 0 cts_prot 0
>>>>> [ 9460.459155] ath10k: mac vdev 0 slot_time 1
>>>>> [ 9460.459156] ath10k: mac vdev 0 preamble 1n
>>>>> [ 9460.459385] ath10k: mac config channel 5180 mhz
>>>>> [ 9460.459434] IPv6: ADDRCONF(NETDEV_UP): wlan2: link is not ready
>>>>> [ 9461.562262] ath10k: mac vdev 0 rts threshold -1
>>>>> [ 9461.562292] ath10k: mac vdev 0 fragmentation threshold -1
>>>>> [ 9461.564900] ath10k: mac config channel 5180 mhz
>>>>> [ 9461.564919] ath10k: mac vdev 0 beacon_interval 100
>>>>> [ 9461.564923] ath10k: vdev 0 set beacon tx mode to staggered
>>>>> [ 9461.564931] ath10k: mac vdev 0 start center_freq 5180 phymode
>>>>> 11ac-vht80
>>>>> [ 9461.571970] ath10k: mac vdev 0 up
>>>>> [ 9461.572108] ath10k: mac vdev 0 cts_prot 0
>>>>> [ 9461.572117] ath10k: mac vdev 0 slot_time 2
>>>>> [ 9461.572241] IPv6: ADDRCONF(NETDEV_CHANGE): wlan2: link becomes ready
>>>>> [ 9461.572192] ath10k: mac vdev 0 preamble 1n
>>>>> [ 9549.331556] ath10k: vdev 0 set beacon tx mode to staggered
>>>>> [ 9549.331637] ath10k: mac vdev 0 cts_prot 0
>>>>> [ 9549.331644] ath10k: mac vdev 0 preamble 1n
>>>>> [ 9549.332425] ath10k: mac vdev 0 peer create 04:f0:21:0c:a2:3b (new
>>>>> sta)
>>>>> [ 9549.332628] ath10k: mac sta 04:f0:21:0c:a2:3b associated
>>>>> [ 9549.332637] ath10k: mac ht peer 04:f0:21:0c:a2:3b mcs cnt 24 nss 3
>>>>> [ 9549.332643] ath10k: mac vht peer 04:f0:21:0c:a2:3b max_mpdu 1048575
>>>>> flags
>>>>> 0x601b001 rx_mcs_set 0000ffea tx_mcs_set 0000ffea
>>>>> [ 9549.332648] ath10k: mac peer 04:f0:21:0c:a2:3b phymode 11ac-vht80
>>>>>
>>>>>
>>>>> As you can see, the "nss" value is set to 3, but I've never observed
>>>>> VHT-NSS
>>>>> 3 reported by "iw DEV station dump". Also the maximum throughput for
>>>>> UDP
>>>>> measured by iperf was ~630Mbps, which seems to be the netto value for
>>>>> 2x433
>>>>> Mbps PHY throughput.
>>>>
>>>> I wouldn't rely on station dump Tx/Rx stats (especially Tx stats which
>>>> are broken due to FW limitation).
>>>>
>>>> You can create a monitor interface on the receiver side and run
>>>> wireshark in non-promiscuous mode and check radiotap headers to
>>>> confirm whether 3 spatial streams are used or not.
>>>>
>>>>
>>>> Michał
>>>>
>>> I've attached the output from wireshark.
>>>
>>>  Bandwidth: 80 MHz (4)
>>>         User 0: MCS 9
>>>             1001 .... = MCS index 0: 9 (256-QAM 5/6)
>>>             .... 0010 = Spatial streams 0: 2
>>>             Space-time streams 0: 2
>>>             Coding 0: BCC (0)
>>>             [Data Rate: 866,6 Mb/s]
>>>
>>> Unfortunately, it confirms that only 2 spatial streams are used.
>>>
>>>
>>> TomaszT
>>>
>> I thought that maybe one of the card is physically damaged and the third
>> spatial link cannot be established. I've checked that using a 802.11ac
>> router, which I also have (and it supports 3 streams). Both cards connected
>> with it using only 2 spatial streams. Its unlikely that both cards have a
>> malfunction, so its rather a software problem.
>>
>> TomaszT
>
> Hi,
>
> Are there any chances, that this issue would be investigated and hopefully
> fixed before the end of the year ?
>
Seems from driver we configure everything correctly. Rate Control is
in FW - so FW will decide what NSS/MCS to use. Here we see our devices
works fine with 3x3.

BTW.
You can try to use fixed_rate setting. You will have to apply some
patches from mac80211-next/group:

Kernel:
1) nl80211: don't clear bitrate_mask twice
2) nl80211: back to default bitrate_mask correctly
3) nl80211/cfg80211: bitrate_mask, rename mcs to ht_mcs
4) nl80211: add VHT support for set_bitrate_mask
5) ath10k: add set_bitrate_mask callback - this I will send in few minutes

iw:
1) iw: add VHT MCS/NSS set support to set bitrates

After that just run ap and: iw wlanX set bitrates legacy-5 ht-mcs-5
vht-mcs-5 3:0

You should send frames from AP with NSS=3, MCS=0

BR
Janusz