[ath9k-devel] [PATCH] ath9k: Support 4.9Ghz channels on AR9580 adapter.

Discussion:

g***@candelatech.com

2016-05-12 22:40:57 UTC

Permalink

--
1.9.3

Kalle Valo

2016-06-20 20:34:31 UTC

Permalink

Post by g***@candelatech.com
NOTE: These channels must not be used in most regulatory
domains unless you have a license from the FCC or similar!
A proper regulatory database is also required to actually use
these channels.

I feel that the license is a problem and this doesn't belong to the kernel.
The patch has been dropped.

--
Sent by pwcli
https://patchwork.kernel.org/patch/9086491/

Ben Greear

2016-06-20 20:41:09 UTC

Permalink

Post by Kalle Valo

I feel that the license is a problem and this doesn't belong to the kernel.
The patch has been dropped.

It's not a lot different from having to be in normal FCC regulations w/regard to
DFS channels, tx power, and such, but maybe having driver support is considered to make it
too easy for users to hack around restrictions?

Another part of this is that Cisco APs, at least, use fractional center
frequencies (4942.5, for instance) when using 5Mhz bandwidths on 4.9Ghz
channels, and that requires more patches that I don't think I even bothered
to post.

I'll keep this in my trees, someone needing it can just clone it and/or borrow
patches as needed.

Thanks,
Ben

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

Julian Calaby

2016-06-20 23:53:56 UTC

Permalink

Hi Ben,

Post by Ben Greear

Post by Kalle Valo

I feel that the license is a problem and this doesn't belong to the kernel.
The patch has been dropped.

Maybe hide this behind CFG80211_CERTIFICATION_ONUS or equivalent?
(Maybe a new symbol CFG80211_LICENSE_REQUIRED with big scary
feds-will-come-knocking warnings in the help text?)

Thanks,

--
Julian Calaby

Email: ***@gmail.com
Profile: http://www.google.com/profiles/julian.calaby/

Ben Greear

2016-06-21 00:00:13 UTC

Permalink

Post by Julian Calaby
Hi Ben,

Post by Ben Greear

Post by Kalle Valo

I feel that the license is a problem and this doesn't belong to the kernel.
The patch has been dropped.

Maybe hide this behind CFG80211_CERTIFICATION_ONUS or equivalent?
(Maybe a new symbol CFG80211_LICENSE_REQUIRED with big scary
feds-will-come-knocking warnings in the help text?)

That would be fine with me. As I said though, there are more patches
than this one needed for full features and interoperability.
The person that sent me the mac80211/wireless patches,
and hostapd for that matter, may be able to put something together for upstream,
but I am not sure if they will have time.

I don't currently feel motivated enough to try to clean it all up and
try to push it upstream myself.

Thanks,
Ben

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

Julian Calaby

2016-06-21 01:02:20 UTC

Permalink

From: Ben Greear <***@candelatech.com>

NOTE: These channels must not be used in most regulatory
domains unless you have a license from the FCC or similar!

A proper regulatory database is also required to actually use
these channels.

Signed-off-by: Ben Greear <***@candelatech.com>
[Hide this support behind a Kconfig option]
Signed-off-by: Julian Calaby <***@gmail.com>
---
drivers/net/wireless/ath/ath9k/Kconfig | 19 +++++++++++
drivers/net/wireless/ath/ath9k/ath9k.h | 2 +-
drivers/net/wireless/ath/ath9k/common-init.c | 49 +++++++++++++++++++++++-----
drivers/net/wireless/ath/ath9k/hw.h | 4 +--
4 files changed, 63 insertions(+), 11 deletions(-)

Hi Kalle,

I've only done this work as I hate to see people's efforts go to
waste and I feel that there's enough roadblocks in the way of
actually using this functionality that casual idiots won't be able
to.

I've tried to keep the code hidden by the Kconfig option as minimal
as possible, however the #ifdef in ath9k_49ghz_capable() is arguably
not required as returning true from that function still results in no
action if the option is not enabled.

This is compile tested only as I cannot test this for real as I lack
both the hardware and license required.

Thanks,

Julian Calaby

diff --git a/drivers/net/wireless/ath/ath9k/Kconfig b/drivers/net/wireless/ath/ath9k/Kconfig
index f68cb00..57b0a41 100644
--- a/drivers/net/wireless/ath/ath9k/Kconfig
+++ b/drivers/net/wireless/ath/ath9k/Kconfig
@@ -114,6 +114,25 @@ config ATH9K_DFS_CERTIFIED
developed. At this point enabling this option won't do anything
except increase code size.

+config ATH9K_49_GHZ_CHAN
+ bool "Support 4.9Ghz public safety channels on some devices"
+ depends on ATH9K && CFG80211_CERTIFICATION_ONUS
+ default n
+ ---help---
+ This option enables support for 4.9 GHz channels on AR9580.
+
+ These are PUBLIC SAFETY CHANNELS and MUST NOT BE USED in most
+ regulatory domains UNLESS YOU HAVE A FULL LICENSE for their use from
+ your local radio regulator, e.g. the FCC or equivalent. Using these
+ channels without proper authorisation may result in serious legal
+ consequences.
+
+ You will also have to build a regulatory database with these channels
+ enabled to actually use them.
+
+ If you are a distro kernel builder or have any doubt whatsoever about
+ your legal ability to use these channels, say N.
+
config ATH9K_DYNACK
bool "Atheros ath9k ACK timeout estimation algorithm (EXPERIMENTAL)"
depends on ATH9K
diff --git a/drivers/net/wireless/ath/ath9k/ath9k.h b/drivers/net/wireless/ath/ath9k/ath9k.h
index 93b3793..d8744b5 100644
--- a/drivers/net/wireless/ath/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath/ath9k/ath9k.h
@@ -954,7 +954,7 @@ struct ath_softc {
struct device *dev;

struct survey_info *cur_survey;
- struct survey_info survey[ATH9K_NUM_CHANNELS];
+ struct survey_info survey[ATH9K_MAX_NUM_CHANNELS];

struct tasklet_struct intr_tq;
struct tasklet_struct bcon_tasklet;
diff --git a/drivers/net/wireless/ath/ath9k/common-init.c b/drivers/net/wireless/ath/ath9k/common-init.c
index 8b4f7fd..f6a495b 100644
--- a/drivers/net/wireless/ath/ath9k/common-init.c
+++ b/drivers/net/wireless/ath/ath9k/common-init.c
@@ -86,6 +86,24 @@ static const struct ieee80211_channel ath9k_5ghz_chantable[] = {
CHAN5G(5785, 35), /* Channel 157 */
CHAN5G(5805, 36), /* Channel 161 */
CHAN5G(5825, 37), /* Channel 165 */
+
+#ifdef ATH9K_49_GHZ_CHAN
+ /* 4.9Ghz channels, public safety channels, license is required in US
+ * and most other regulatory domains!
+ */
+ CHAN5G(4915, 38), /* Channel 183 */
+ CHAN5G(4920, 39), /* Channel 184 */
+ CHAN5G(4925, 40), /* Channel 185 */
+ CHAN5G(4935, 41), /* Channel 187 */
+ CHAN5G(4940, 42), /* Channel 188 */
+ CHAN5G(4945, 43), /* Channel 189 */
+ CHAN5G(4960, 44), /* Channel 192 */
+ CHAN5G(4970, 45), /* Channel 194 */
+ CHAN5G(4980, 46), /* Channel 196 */
+#define ATH9K_NUM_49GHZ_CHANNELS 9
+#else
+#define ATH9K_NUM_49GHZ_CHANNELS 0
+#endif
};

/* Atheros hardware rate code addition for short premble */
@@ -122,14 +140,31 @@ static struct ieee80211_rate ath9k_legacy_rates[] = {
IEEE80211_RATE_SUPPORTS_10MHZ)),
};

+static bool ath9k_49ghz_capable(struct ath_hw *ah)
+{
+#ifdef ATH9K_49_GHZ_CHAN
+ /* Seems AR9580 supports 4.9ghz, at least. */
+ switch (ah->hw_version.devid) {
+ case AR9300_DEVID_AR9580:
+ return true;
+ }
+#endif
+
+ return false;
+}
+
int ath9k_cmn_init_channels_rates(struct ath_common *common)
{
struct ath_hw *ah = (struct ath_hw *)common->ah;
void *channels;
+ int num_5ghz_chan = ARRAY_SIZE(ath9k_5ghz_chantable);
+
+ if (!ath9k_49ghz_capable(ah))
+ num_5ghz_chan -= ATH9K_NUM_49GHZ_CHANNELS;

BUILD_BUG_ON(ARRAY_SIZE(ath9k_2ghz_chantable) +
- ARRAY_SIZE(ath9k_5ghz_chantable) !=
- ATH9K_NUM_CHANNELS);
+ ARRAY_SIZE(ath9k_5ghz_chantable) >
+ ATH9K_MAX_NUM_CHANNELS);

if (ah->caps.hw_caps & ATH9K_HW_CAP_2GHZ) {
channels = devm_kzalloc(ah->dev,
@@ -149,17 +184,16 @@ int ath9k_cmn_init_channels_rates(struct ath_common *common)
}

if (ah->caps.hw_caps & ATH9K_HW_CAP_5GHZ) {
- channels = devm_kzalloc(ah->dev,
- sizeof(ath9k_5ghz_chantable), GFP_KERNEL);
+ int ch_sz = num_5ghz_chan * sizeof(ath9k_5ghz_chantable[0]);
+
+ channels = devm_kzalloc(ah->dev, ch_sz, GFP_KERNEL);
if (!channels)
return -ENOMEM;

- memcpy(channels, ath9k_5ghz_chantable,
- sizeof(ath9k_5ghz_chantable));
+ memcpy(channels, ath9k_5ghz_chantable, ch_sz);
common->sbands[NL80211_BAND_5GHZ].channels = channels;
common->sbands[NL80211_BAND_5GHZ].band = NL80211_BAND_5GHZ;
- common->sbands[NL80211_BAND_5GHZ].n_channels =
- ARRAY_SIZE(ath9k_5ghz_chantable);
+ common->sbands[NL80211_BAND_5GHZ].n_channels = num_5ghz_chan;
common->sbands[NL80211_BAND_5GHZ].bitrates =
ath9k_legacy_rates + 4;
common->sbands[NL80211_BAND_5GHZ].n_bitrates =
diff --git a/drivers/net/wireless/ath/ath9k/hw.h b/drivers/net/wireless/ath/ath9k/hw.h
index 9cbca12..8115d86 100644
--- a/drivers/net/wireless/ath/ath9k/hw.h
+++ b/drivers/net/wireless/ath/ath9k/hw.h
@@ -73,7 +73,7 @@

#define ATH9K_RSSI_BAD -128

-#define ATH9K_NUM_CHANNELS 38
+#define ATH9K_MAX_NUM_CHANNELS 47

/* Register read/write primitives */
#define REG_WRITE(_ah, _reg, _val) \
@@ -777,7 +777,7 @@ struct ath_hw {
struct ath9k_hw_version hw_version;
struct ath9k_ops_config config;
struct ath9k_hw_capabilities caps;
- struct ath9k_channel channels[ATH9K_NUM_CHANNELS];
+ struct ath9k_channel channels[ATH9K_MAX_NUM_CHANNELS];
struct ath9k_channel *curchan;

union {

--
2.8.1

Jouni Malinen

2016-06-21 09:41:46 UTC

Permalink

Post by Julian Calaby
I've only done this work as I hate to see people's efforts go to
waste and I feel that there's enough roadblocks in the way of
actually using this functionality that casual idiots won't be able
to.

Are these really ready to go to the upstream kernel in this state and
without the other changes that would be needed to operate correctly?
What is the use case for these and how have these been tested?

Post by Julian Calaby
This is compile tested only as I cannot test this for real as I lack
both the hardware and license required.

I don't think this is sufficient when touching this type of area. I
would not apply these without proper testing and full set of
functionality being available. I see no point in ath9k defining
additional channels if all those new channels can cause is harm and not
correct functionality. This channel list addition looks like the easiest
part to handle compared to the other patches needed for 4.9 GHz and this
would be the last patch on my list to get accepted..

Post by Julian Calaby
diff --git a/drivers/net/wireless/ath/ath9k/common-init.c b/drivers/net/wireless/ath/ath9k/common-init.c
+#ifdef ATH9K_49_GHZ_CHAN
+ /* 4.9Ghz channels, public safety channels, license is required in US
+ * and most other regulatory domains!
+ */
+ CHAN5G(4915, 38), /* Channel 183 */
+ CHAN5G(4920, 39), /* Channel 184 */
+ CHAN5G(4925, 40), /* Channel 185 */
+ CHAN5G(4935, 41), /* Channel 187 */
+ CHAN5G(4940, 42), /* Channel 188 */
+ CHAN5G(4945, 43), /* Channel 189 */
+ CHAN5G(4960, 44), /* Channel 192 */
+ CHAN5G(4970, 45), /* Channel 194 */
+ CHAN5G(4980, 46), /* Channel 196 */

Where are these channels defined and are these really correct
frequencies for them? Please note that many of the 4.9 GHz channels have
channel starting frequencies like 4.9375 GHz and 4.0025 GHz, i.e.,
fractional MHz.. While US public safety may not have all those cases,
even there are some 0.5 MHz cases. In addition, those channel numbers
sound more like some of the channels defined in Japan rather than US
public safety operating class. In addition, some of these channels seem
to be outside the US public safety range.

Is this trying to add 4.9 GHz channels in general for multiple different
use cases? And if so, what are those use cases? Or is this only for some
public safety cases? And if so, for which regulatory domains?

To be frank, I really don't see how this would be even close to a state
that should be accepted into the upstream tree.

--
Jouni Malinen PGP id EFC895FA

Julian Calaby

2016-06-21 10:16:00 UTC

Permalink

Hi Jouni,

Post by Jouni Malinen

This patch is a edited version (adding the Kconfig option) of Ben's
patch. I picked it up as he seemed to have given up on it and I could
see nothing directly wrong with the patch.

Post by Jouni Malinen

Post by Julian Calaby
This is compile tested only as I cannot test this for real as I lack
both the hardware and license required.

It isn't, Ben himself has said that fractional MHz need to be
supported for this to be 100% correct, however I believe that defining
the channels is minimally sufficient for his purposes. I do know he's
done more work than this, however I can't pick that up as getting it
ready for upstream requires testing I can't do.

Post by Jouni Malinen

From my perspective, this is enough to reject the patch outright. I
assumed the frequencies and numbers were correct, which is why I
picked it up.

I believe he has a client that requires this support in a Linux kernel.

Post by Jouni Malinen
To be frank, I really don't see how this would be even close to a state
that should be accepted into the upstream tree.

Fair enough I'm dropping this.

Kalle, I've marked this as rejected and archived on Patchwork.

Thanks,

--
Julian Calaby

Email: ***@gmail.com
Profile: http://www.google.com/profiles/julian.calaby/

Ben Greear

2016-06-21 14:17:27 UTC

Permalink

Post by Jouni Malinen

More patches are needed to make this work properly. My full patchset, with
modifications to supplicant, *has* been tested, including against a Cisco
AP running 5Mhz channels.

Post by Jouni Malinen

Post by Julian Calaby
This is compile tested only as I cannot test this for real as I lack
both the hardware and license required.

I got these from WiKi pedia. There is a later patch that adds the 5Mhz
channels that Cisco uses. I figured that since this is all restricted anyway,
that users of it would know how to make it work (and they would have to hack
regdb in order to use any of them anyway).

Post by Jouni Malinen
Is this trying to add 4.9 GHz channels in general for multiple different
use cases? And if so, what are those use cases? Or is this only for some
public safety cases? And if so, for which regulatory domains?
To be frank, I really don't see how this would be even close to a state
that should be accepted into the upstream tree.

It was a starting point, and posted mainly to let folks know how to get started.

As I said, I don't have interest in trying to push all the changes upstream,
but they are in my kernel tree on dmz2.candelatech.com and hostap tree on
github, so maybe someone who has interest in dealing with upstreaming them
can take them over.

Thanks,
Ben

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

Dave Taht

2016-06-21 14:25:22 UTC

Permalink

Post by Jouni Malinen

So far as I know the use case for these is to make it possible to build
open source wifi systems that enable emergency services. This
strikes me as a worthy goal.