DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org CC8BE208D9
MIME-Version: 1.0
References: <20180925081933.28553-1-enrique.giraldo@galgus.net>
 <57143da1-1569-bd98-70e9-91cdcd89d1b8@broadcom.com> <CAEbaMQ+N=hUwXta9NbrBpF3YcoCyUNWB=s3cynp4pzFLuFrmDA@mail.gmail.com>
 <1ff1f335-6962-4ace-6b25-562dab939893@broadcom.com>
In-Reply-To: <1ff1f335-6962-4ace-6b25-562dab939893@broadcom.com>
From:   Enrique Giraldo <hgiraldos@gmail.com>
Date:   Tue, 25 Sep 2018 12:56:19 +0200
Message-ID: <CAEbaMQJgOqpvj-mvfL+s836YSii6htodU2YivkFUN1hQ=c8O=w@mail.gmail.com>
Subject: Re: [PATCH] New functionality for aborting ongoing CAC.
To:     arend.vanspriel@broadcom.com
Cc:     linux-wireless@vger.kernel.org,
        Enrique Giraldo <enrique.giraldo@galgus.net>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk

What you commented was the first thing that I tried, but doing it, the
radio stays in an inconsistent state and the hostapd is not able to
raise an instance again.
Attached is the log of what is happening:

Tue Sep 25 09:58:10 2018 daemon.notice hostapd: wlan0: DFS-CAC-START
freq=3D5500 chan=3D100 sec_chan=3D1, width=3D1, seg0=3D106, seg1=3D0, cac_t=
ime=3D60s
Tue Sep 25 09:58:29 2018 daemon.notice hostapd: wlan0:
DFS-CAC-COMPLETED freq=3D5500 success=3D0 ht_enabled=3D0 chan_offset=3D0
chan_width=3D3 cf1=3D5530 cf2=3D0
Tue Sep 25 09:58:29 2018 daemon.notice hostapd: wlan0: INTERFACE-DISABLED
Tue Sep 25 09:58:29 2018 kern.info kernel: [   70.650485] br-lan: port
3(wlan0) entered disabled state
Tue Sep 25 09:58:29 2018 kern.info kernel: [   70.665228] device wlan0
left promiscuous mode
Tue Sep 25 09:58:29 2018 kern.info kernel: [   70.669794] br-lan: port
3(wlan0) entered disabled state
Tue Sep 25 09:58:30 2018 daemon.err hostapd: Configuration file:
/var/run/hostapd-phy0.conf
Tue Sep 25 09:58:31 2018 kern.warn kernel: [   72.295668] ath10k_pci
0000:00:00.0: 10.1 wmi init: vdevs: 16  peers: 127  tid: 256
Tue Sep 25 09:58:31 2018 kern.info kernel: [   72.312713] ath10k_pci
0000:00:00.0: wmi print 'P 128 V 8 T 410'
Tue Sep 25 09:58:31 2018 kern.info kernel: [   72.318911] ath10k_pci
0000:00:00.0: wmi print 'msdu-desc: 912  sw-crypt: 0'
Tue Sep 25 09:58:31 2018 kern.info kernel: [   72.326063] ath10k_pci
0000:00:00.0: wmi print 'alloc rem: 38688 iram: 27140'
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: wlan0: INTERFACE-ENABLED
Tue Sep 25 09:58:31 2018 kern.info kernel: [   72.402994] IPv6:
ADDRCONF(NETDEV_UP): wlan0: link is not ready
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: wlan0: INTERFACE-DISABLED
Tue Sep 25 09:58:31 2018 daemon.err hostapd: nl80211: Could not
configure driver mode
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: nl80211: deinit
ifname=3Dwlan0 disabled_11b_rates=3D0
Tue Sep 25 09:58:31 2018 daemon.err hostapd: nl80211 driver
initialization failed.
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: wlan0: interface state
UNINITIALIZED->DISABLED
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: wlan0: AP-DISABLED
Tue Sep 25 09:58:31 2018 daemon.notice hostapd: wlan0:
CTRL-EVENT-TERMINATING
Tue Sep 25 09:58:31 2018 daemon.err hostapd: hostapd_free_hapd_data:
Interface wlan0 wasn't started
Tue Sep 25 09:58:31 2018 daemon.notice netifd: radio0 (1809): WARNING
(wireless_add_process): executable path /usr/sbin/hostapd does not
match process 1640 path ()
Tue Sep 25 09:58:31 2018 daemon.notice netifd: radio0 (1809): Device
setup failed: HOSTAPD_START_FAILED

Regards,
EG

El mar., 25 sept. 2018 a las 11:53, Arend van Spriel
(<arend.vanspriel@broadcom.com>) escribi=C3=B3:
>
> On 9/25/2018 11:28 AM, Enrique Giraldo wrote:
> > The main reason is to be able to stop the CAC when you want to make a
> > channel switch and the CAC is ongoing. It's true that the radio would
> > not pass to the next phase, the behavior is the same as when during
> > the CAC a radar event is detected. In the case of aborting, a later
> > action is expected, for example, switch to the desired new channel.
>
> So for mac80211 the function ieee80211_dfs_cancel_cac() indeed seems to
> just abort. So what is the scenario here? If user-space sends
> NL80211_CMD_START_AP with DFS channel, a CAC period is started. So you
> are saying you want to do a channel switch? Does it not make sense to
> issue a NL80211_CMD_STOP_AP and then NL80211_CMD_START_AP for the
> desired channel? Maybe the use-case you are looking at is not AP mode,
> but there is no context whatsoever in the commit message so I can only
> guess.
>
> Regards,
> Arend
>