diff mbox series

[8/9] DFS: switch to background radar channel if available

Message ID ec706e090d7e44bcc9afd2f9789802f8c07e3c3f.1640014128.git.lorenzo@kernel.org
State New
Headers show
Series introduce background radar detection support | expand

Commit Message

Lorenzo Bianconi Dec. 20, 2021, 3:48 p.m. UTC 
On radar detection on the main chain switch to the channel monitored
by the background chain if we have already performed the CAC there.

Tested-by: Owen Peng <owen.peng@mediatek.com>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
---
 src/ap/dfs.c | 60 ++++++++++++++++++++++++++++++++++++++++++++++------
 1 file changed, 53 insertions(+), 7 deletions(-)

Comments

Jouni Malinen March 3, 2022, 10:20 p.m. UTC | #1 
On Mon, Dec 20, 2021 at 04:48:23PM +0100, Lorenzo Bianconi wrote:
> On radar detection on the main chain switch to the channel monitored
> by the background chain if we have already performed the CAC there.

Is it clear that the channel on the background chain, if one is
available with CAC performed, comply with the ETSI uniform spreading
requirements? The comment below feels a bit misleading, but the way
patch 5/9 picks the channel for background CAC might be sufficient for
this. If so, it would be good to note that in the comment here:

> diff --git a/src/ap/dfs.c b/src/ap/dfs.c
> +hostapd_dfs_background_start_channel_switch(struct hostapd_iface *iface,

> +	/*
> +	 * If background radar detection is supported and radar channel
> +	 * monitored by background chain is available switch to it without
> +	 * waiting for the CAC otherwise let's keep a random channel.

That first part leaves it a bit open whether the random channel part from
the end of the sentence applies in practice for the new operating
channel. It is important to make this clearly describe the functionality
needed to meet uniform spreading requirement.
Lorenzo Bianconi March 4, 2022, 1:46 p.m. UTC | #2 
> On Mon, Dec 20, 2021 at 04:48:23PM +0100, Lorenzo Bianconi wrote:
> > On radar detection on the main chain switch to the channel monitored
> > by the background chain if we have already performed the CAC there.
> 
> Is it clear that the channel on the background chain, if one is
> available with CAC performed, comply with the ETSI uniform spreading
> requirements? The comment below feels a bit misleading, but the way
> patch 5/9 picks the channel for background CAC might be sufficient for
> this. If so, it would be good to note that in the comment here:

ack, got your point here. We should force channel_type to DFS_ANY_CHANNEL
for ETSI regdmain in hostpad_dfs_update_background_chain() and not just
DFS_NO_CAC_YET. I will fix it in next revision. Thanks.

Regards,
Lorenzo

> 
> > diff --git a/src/ap/dfs.c b/src/ap/dfs.c
> > +hostapd_dfs_background_start_channel_switch(struct hostapd_iface *iface,
> 
> > +	/*
> > +	 * If background radar detection is supported and radar channel
> > +	 * monitored by background chain is available switch to it without
> > +	 * waiting for the CAC otherwise let's keep a random channel.
> 
> That first part leaves it a bit open whether the random channel part from
> the end of the sentence applies in practice for the new operating
> channel. It is important to make this clearly describe the functionality
> needed to meet uniform spreading requirement.
> 
> -- 
> Jouni Malinen                                            PGP id EFC895FA
>
diff mbox series

Patch

diff --git a/src/ap/dfs.c b/src/ap/dfs.c
index 30be7e349..35d26e725 100644
--- a/src/ap/dfs.c
+++ b/src/ap/dfs.c
@@ -1259,6 +1259,48 @@  static int hostapd_dfs_start_channel_switch_cac(struct hostapd_iface *iface)
 	return err;
 }
 
+static int
+hostapd_dfs_background_start_channel_switch(struct hostapd_iface *iface,
+					    int freq)
+{
+	if (!(iface->drv_flags2 & WPA_DRIVER_RADAR_BACKGROUND))
+		return -1; /* Background radar chain not supported */
+
+	wpa_printf(MSG_DEBUG,
+		   "%s called (background CAC active: %s, CSA active: %s)",
+		   __func__, iface->radar_background.cac_started ? "yes" : "no",
+		   hostapd_csa_in_progress(iface) ? "yes" : "no");
+
+	/* Check if CSA in progress */
+	if (hostapd_csa_in_progress(iface))
+		return 0;
+
+	/*
+	 * If background radar detection is supported and radar channel
+	 * monitored by background chain is available switch to it without
+	 * waiting for the CAC otherwise let's keep a random channel.
+	 * If radar pattern is reported on the background chain, just switch
+	 * monitoring another radar channel.
+	 */
+	if (hostapd_dfs_is_background_event(iface, freq)) {
+		hostpad_dfs_update_background_chain(iface);
+		return 0;
+	}
+
+	/* Background channel not available yet. Perform CAC on the
+	 * main chain.
+	 */
+	if (iface->radar_background.cac_started) {
+		/* We want to switch to monitored channel as soon as
+		 * CAC is completed.
+		 */
+		iface->radar_background.temp_ch = 1;
+		return -1;
+	}
+
+	return hostapd_dfs_start_channel_switch_background(iface);
+}
+
 static int hostapd_dfs_start_channel_switch(struct hostapd_iface *iface)
 {
 	struct hostapd_channel_data *channel;
@@ -1359,15 +1401,19 @@  int hostapd_dfs_radar_detected(struct hostapd_iface *iface, int freq,
 	if (!res)
 		return 0;
 
-	/* Skip if reported radar event not overlapped our channels */
-	res = dfs_are_channels_overlapped(iface, freq, chan_width, cf1, cf2);
-	if (!res)
-		return 0;
+	if (!hostapd_dfs_is_background_event(iface, freq)) {
+		/* Skip if reported radar event not overlapped our channels */
+		res = dfs_are_channels_overlapped(iface, freq, chan_width,
+						  cf1, cf2);
+		if (!res)
+			return 0;
+	}
 
-	/* radar detected while operating, switch the channel. */
-	res = hostapd_dfs_start_channel_switch(iface);
+	if (hostapd_dfs_background_start_channel_switch(iface, freq))
+		/* radar detected while operating, switch the channel. */
+		return hostapd_dfs_start_channel_switch(iface);
 
-	return res;
+	return 0;
 }