Exposing enum ieee80211_channel_flags to user

Exposing enum ieee80211_channel_flags to user

[Rafał Miłecki]

Hi,

Recently I spent quite some time trying to figure out why hostapd
can't start 80 MHz AP for me with following error:
radio1 (1617): nl80211: Failed to set channel (freq=5180): -22
(Invalid argument)

My config included:
hw_mode=a
channel=36
vht_oper_chwidth=1
vht_oper_centr_freq_seg0_idx=42

I added some debugging to hostapd but it didn't help much:
radio1 (1617): Allowed channel: mode=2 chan=36 freq=5180 MHz max_tx_power=20 dBm
radio1 (1617): Allowed channel: mode=2 chan=40 freq=5200 MHz max_tx_power=20 dBm
radio1 (1617): Allowed channel: mode=2 chan=44 freq=5220 MHz max_tx_power=20 dBm
radio1 (1617): Allowed channel: mode=2 chan=48 freq=5240 MHz max_tx_power=20 dBm
radio1 (1617): hw vht capab: 0x20, conf vht capab: 0x20
radio1 (1617): Completing interface initialization
radio1 (1617): Mode: IEEE 802.11a  Channel: 36  Frequency: 5180 MHz
radio1 (1617): nl80211: Set freq 5180 (ht_enabled=1, vht_enabled=1,
bandwidth=80 MHz, cf1=5210 MHz, cf2=0 MHz)
radio1 (1617):   * freq=5180
radio1 (1617):   * vht_enabled=1
radio1 (1617):   * ht_enabled=1
radio1 (1617):   * bandwidth=80
radio1 (1617):   * channel_width=3
radio1 (1617):   * center_freq1=5210
radio1 (1617):   * center_freq2=0
radio1 (1617): nl80211: Failed to set channel (freq=5180): -22
(Invalid argument)

iw also wasn't helpful:
Frequencies:
        * 5180 MHz [36] (30.0 dBm)
        * 5200 MHz [40] (30.0 dBm)
        * 5220 MHz [44] (30.0 dBm)
        * 5240 MHz [48] (30.0 dBm)

After some in-kernel debugging (nl80211_set_wiphy,
cfg80211_chandef_usable, cfg80211_secondary_chans_ok) I finally
realized it was because of:
IEEE80211_CHAN_NO_80MHZ

Do you have any idea how we could handle such cases nicely? To let
user find out what's going on (wrong)?

Should "iw" be extended to print flag names? Or should hostapd check
for channels in some smarter way? Any other ideas?

-- 
Rafał

Re: Exposing enum ieee80211_channel_flags to user

[Johannes Berg]

On Thu, 2015-06-11 at 11:35 +0200, Rafał Miłecki wrote:

> iw also wasn't helpful:
> Frequencies:
>         * 5180 MHz [36] (30.0 dBm)
>         * 5200 MHz [40] (30.0 dBm)
>         * 5220 MHz [44] (30.0 dBm)
>         * 5240 MHz [48] (30.0 dBm)
> 
> After some in-kernel debugging (nl80211_set_wiphy,
> cfg80211_chandef_usable, cfg80211_secondary_chans_ok) I finally
> realized it was because of:
> IEEE80211_CHAN_NO_80MHZ
> 
> Do you have any idea how we could handle such cases nicely? To let
> user find out what's going on (wrong)?
> 
> Should "iw" be extended to print flag names? Or should hostapd check
> for channels in some smarter way? Any other ideas?

We already have the flags exposed to iw, it's just not printing them.
There was a patch to print them, but it wasn't complete, and the
submitter kinda went away. You can see the discussion here:

http://mid.gmane.org/1425452686-28196-1-git-send-email-arnagara@qti.qualcomm.com

johannes

Re: Exposing enum ieee80211_channel_flags to user

[Johannes Berg]

On Thu, 2015-06-11 at 11:45 +0200, Johannes Berg wrote:
> On Thu, 2015-06-11 at 11:35 +0200, Rafał Miłecki wrote:
> 
> > iw also wasn't helpful:
> > Frequencies:
> >         * 5180 MHz [36] (30.0 dBm)
> >         * 5200 MHz [40] (30.0 dBm)
> >         * 5220 MHz [44] (30.0 dBm)
> >         * 5240 MHz [48] (30.0 dBm)
> > 
> > After some in-kernel debugging (nl80211_set_wiphy,
> > cfg80211_chandef_usable, cfg80211_secondary_chans_ok) I finally
> > realized it was because of:
> > IEEE80211_CHAN_NO_80MHZ
> > 
> > Do you have any idea how we could handle such cases nicely? To let
> > user find out what's going on (wrong)?
> > 
> > Should "iw" be extended to print flag names? Or should hostapd check
> > for channels in some smarter way? Any other ideas?
> 
> We already have the flags exposed to iw, it's just not printing them.
> There was a patch to print them, but it wasn't complete, and the
> submitter kinda went away. You can see the discussion here:
> 
> http://mid.gmane.org/1425452686-28196-1-git-send-email-arnagara@qti.qualcomm.com

Upon rereading, I think my preference would be to remove much of the
channel list information and add a separate iw command.

johannes

Re: Exposing enum ieee80211_channel_flags to user

[Rafał Miłecki]

On 11 June 2015 at 11:46, Johannes Berg <johannes@sipsolutions.net> wrote:
> On Thu, 2015-06-11 at 11:45 +0200, Johannes Berg wrote:
>> On Thu, 2015-06-11 at 11:35 +0200, Rafał Miłecki wrote:
>>
>> > iw also wasn't helpful:
>> > Frequencies:
>> >         * 5180 MHz [36] (30.0 dBm)
>> >         * 5200 MHz [40] (30.0 dBm)
>> >         * 5220 MHz [44] (30.0 dBm)
>> >         * 5240 MHz [48] (30.0 dBm)
>> >
>> > After some in-kernel debugging (nl80211_set_wiphy,
>> > cfg80211_chandef_usable, cfg80211_secondary_chans_ok) I finally
>> > realized it was because of:
>> > IEEE80211_CHAN_NO_80MHZ
>> >
>> > Do you have any idea how we could handle such cases nicely? To let
>> > user find out what's going on (wrong)?
>> >
>> > Should "iw" be extended to print flag names? Or should hostapd check
>> > for channels in some smarter way? Any other ideas?
>>
>> We already have the flags exposed to iw, it's just not printing them.
>> There was a patch to print them, but it wasn't complete, and the
>> submitter kinda went away. You can see the discussion here:
>>
>> http://mid.gmane.org/1425452686-28196-1-git-send-email-arnagara@qti.qualcomm.com
>
> Upon rereading, I think my preference would be to remove much of the
> channel list information and add a separate iw command.

Do you mean removing channel list from "iw phy X info"? Can you
suggest the new command name?

-- 
Rafał

Re: Exposing enum ieee80211_channel_flags to user

[Johannes Berg]

On Thu, 2015-06-11 at 12:00 +0200, Rafał Miłecki wrote:

> > Upon rereading, I think my preference would be to remove much of the
> > channel list information and add a separate iw command.
> 
> Do you mean removing channel list from "iw phy X info"? 

I'm not sure I'd remove it completely - but perhaps remove all the
detail information like DFS time etc. that's just making it really
large? I.e. go back to just having a single line of information.

> Can you suggest the new command name?

"iw phy X channels" would work?

johannes

Re: Exposing enum ieee80211_channel_flags to user

[Ben Greear]

On 06/11/2015 03:01 AM, Johannes Berg wrote:
> On Thu, 2015-06-11 at 12:00 +0200, Rafał Miłecki wrote:
>
>>> Upon rereading, I think my preference would be to remove much of the
>>> channel list information and add a separate iw command.
>>
>> Do you mean removing channel list from "iw phy X info"?
>
> I'm not sure I'd remove it completely - but perhaps remove all the
> detail information like DFS time etc. that's just making it really
> large? I.e. go back to just having a single line of information.
>
>> Can you suggest the new command name?
>
> "iw phy X channels" would work?

Maybe also allow the full dump with an optional --verbose flag:  'iw --verbose phy X info' ?

Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com

Re: Exposing enum ieee80211_channel_flags to user

[Johannes Berg]

On Thu, 2015-06-11 at 06:16 -0700, Ben Greear wrote:

> Maybe also allow the full dump with an optional --verbose flag:  'iw --verbose phy X info' ?

Well, as I said, it's not so simple to do this in the "info" handler
since you need to collect all the info first and then print it.

johannes