iwlwifi 35% loss rate after v3.16-rc5

iwlwifi 35% loss rate after v3.16-rc5

[Jim Rees]

I've got a Thinkpad x220 with iwlwifi Ultimate-N 6300 AGN. After v3.16-rc5
it became unusable due to 35% packet loss rate. If I revert these commits,
it works again:

dc271ee iwlwifi: mvm: disable CTS to Self
43d826c iwlwifi: dvm: don't enable CTS to self
e48393e iwlwifi: mvm: Fix broadcast filtering
a42c9fc Revert "iwlwifi: remove IWL_UCODE_TLV_FLAGS_UAPSD_SUPPORT flag"
b3c063a iwlwifi: update the 7265 series HW IDs
8987941 iwlwifi: mvm: rework sched scan channel configuration

I can bisect further if that would be useful. Dmesg is here:
http://pastebin.com/wD3UMxUL

It's still broken in 3.17.0-rc2. Any suggestions?

Re: iwlwifi 35% loss rate after v3.16-rc5

[Luca Coelho]

Hi Jim,

On Mon, 2014-08-25 at 22:07 -0400, Jim Rees wrote:
> I've got a Thinkpad x220 with iwlwifi Ultimate-N 6300 AGN. After v3.16-rc5
> it became unusable due to 35% packet loss rate. If I revert these commits,
> it works again:
> 
> dc271ee iwlwifi: mvm: disable CTS to Self
> 43d826c iwlwifi: dvm: don't enable CTS to self
> e48393e iwlwifi: mvm: Fix broadcast filtering
> a42c9fc Revert "iwlwifi: remove IWL_UCODE_TLV_FLAGS_UAPSD_SUPPORT flag"
> b3c063a iwlwifi: update the 7265 series HW IDs
> 8987941 iwlwifi: mvm: rework sched scan channel configuration
> 
> I can bisect further if that would be useful. Dmesg is here:
> http://pastebin.com/wD3UMxUL
> 
> It's still broken in 3.17.0-rc2. Any suggestions?

This is probably the commit that is to blame:

43d826c iwlwifi: dvm: don't enable CTS to self

Could you try again without just that one? If it's confirmed we can
revert that patch.

Also, any chance you can get sniffer logs of the problem?

--
Cheers,
Luca.

Re: iwlwifi 35% loss rate after v3.16-rc5

[Jim Rees]

Luca Coelho wrote:

  This is probably the commit that is to blame:
  
  43d826c iwlwifi: dvm: don't enable CTS to self
  
  Could you try again without just that one? If it's confirmed we can
  revert that patch.

Yes, reverting that one commit fixes it.
  
  Also, any chance you can get sniffer logs of the problem?

I would be happy to if you can give me a hint how to do it. Do you just mean
a pcap, or something else?

Re: iwlwifi 35% loss rate after v3.16-rc5

[Luca Coelho]

On Tue, 2014-08-26 at 07:41 -0400, Jim Rees wrote:
> Luca Coelho wrote:
> 
>   This is probably the commit that is to blame:
>   
>   43d826c iwlwifi: dvm: don't enable CTS to self
>   
>   Could you try again without just that one? If it's confirmed we can
>   revert that patch.
> 
> Yes, reverting that one commit fixes it.

Emmanuel, do you think we should revert this patch? Do you remember
exactly what we tried to address with it?

  
>   Also, any chance you can get sniffer logs of the problem?
> 
> I would be happy to if you can give me a hint how to do it. Do you just mean
> a pcap, or something else?

Jim, I meant an air capture with something like Wireshark.  You need to
have another WiFi NIC that supports monitor mode, though (like Intel's
7260 or any ath9k-based NIC).

--
Cheers,
Luca.

Re: iwlwifi 35% loss rate after v3.16-rc5

[Jim Rees]

Luca Coelho wrote:

  Also, any chance you can get sniffer logs of the problem?

It's here:
http://jim.rees.org/airportSniffyHlDMa.pcap

The Thinkpad is 00:24:d7:11:43:04, the AP (which I'm pinging) is
2C:67:FB:D1:70:56. The first 19 pings work, then it skips 5, then resumes.