[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Stephen Gutknecht]

Howdy,

First off, I am not a OpenWRT developer, just a participatory user.
However, I'm hoping someone here is also an openwrt user and we can
trade source files that I can test on the router hardware and pass
results back on to the OpenWRT developers.

I am on the latest development release of OpenWRT; but the included
madwifi nor ath9k driver works on these ar9102 or ar9103 adapters.
not in client mode, not at all.

First off, do newer updates to ath9k possibly support these chips?
Then I can try to update my OpenWRT with the newer source files.

If no drivers exist for the ar9102, can they be adapted?  just a
matter of a new pci hardware ID or is this a major change?

Any help and tips appreciated.  I started a new thread over on the
OpenWRT forum:
http://forum.openwrt.org/viewtopic.php?id=18113

Thank you.

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Luis R. Rodriguez]

On Wed, Dec 17, 2008 at 10:53:46AM -0800, Stephen Gutknecht (hilltx) wrote:
> Howdy,
> 
> First off, I am not a OpenWRT developer, just a participatory user.
> However, I'm hoping someone here is also an openwrt user and we can
> trade source files that I can test on the router hardware and pass
> results back on to the OpenWRT developers.
> 
> I am on the latest development release of OpenWRT; but the included
> madwifi nor ath9k driver works on these ar9102 or ar9103 adapters.
> not in client mode, not at all.
> 
> First off, do newer updates to ath9k possibly support these chips?
> Then I can try to update my OpenWRT with the newer source files.
> 
> If no drivers exist for the ar9102, can they be adapted?  just a
> matter of a new pci hardware ID or is this a major change?
> 
> Any help and tips appreciated.  I started a new thread over on the
> OpenWRT forum:
> http://forum.openwrt.org/viewtopic.php?id=18113

What is an AR9102 or AR9103? Please see:

http://wireless.kernel.org/en/users/Drivers/ath9k#supportedchipsets

  Luis

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Stephen Gutknecht]

An update to my posting.  the ar9102, ar9103, ar9106 used in routers
with the AHB bus.  Although interest has been stirred up on the
OpenWRT side, there so far isn't anyone who has experience with the
AHB bus porting.

The router GPL downloads provided by the various sellers come with
binary-only madwifi drivers for 2.6.15 kernel (MIPS32).  Output
includes:

insmod /lib/modules/2.6.15/net/ath_hal.ko
ath_hal: 0.9.17.1 (AR5416, DEBUG, REGOPS_FUNC, WRITE_EEPROM, 11D)

insmod /lib/modules/2.6.15/net/ath_ahb.ko wireless_led_gpio=7
ath_ahb: 0.9.4.5 (Atheros/multi-bss)
Howl Revision ID 0xb4
No MBSSID aggregation support<6>
wifi0: Atheros AR9100 WiSoC: mem=0xb80c0000, irq=2

As you can see, they use the ath_ahb driver.  Any overview or details
relative to what it will take to adapt the ath9k driver to work with
the AHB bus are appreciated.  There is enough interest in these
drivers that a small bounty and/or hardware donations are on the
table.  OpenWRT has been fully ported to these routers... all except
working wireless drivers.

I can personally put one of these routers on the internet for you
running 2.6.27.10 kernel with OpenWRT running (or the original
firmware 2.6.15 environment); any poking around to to determine what's
needed for drivers would be welcome.  Help please ;)

Thank you.

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Pavel Roskin]

On Sat, 2008-12-27 at 15:45 -0800, Stephen Gutknecht (hilltx) wrote:

> As you can see, they use the ath_ahb driver.  Any overview or details
> relative to what it will take to adapt the ath9k driver to work with
> the AHB bus are appreciated.

The first step would be to take current MadWifi sources, go through the
differences between if_ath_ahb.[ch] and if_ath_pci.[ch] and apply those
differences to the ath9k driver.

When the modified driver is working with the AHB device, the differences
should be restructured to have a PCI module, an AHB module and a common
library.

Perhaps there will be some questions that will need to be communicated
to Atheros.

-- 
Regards,
Pavel Roskin

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Stephen Gutknecht]

A major discovery, good news!  Through exhaustive searching of every
bit I could find about AHB drivers, I stumbled into the source code
for the Atheros Radio Testing (ART) for these routers.  From my AP81
router (Trendnet TEW-652BRP, DLink DIR-615 C1) running OpenWRT trunk:


root at OpenWrt:/tmp# insmod art.ko
CHIP REV ID: b4
dk_dev_init::status after register_chrdev(dk) = 0
 Base Phsycal address :0x180c0000
DK::Add client 0
DK::Reg phy addr = 180c0000 vir Addr = b80c0000
DK::Irq = 2
DK::Allocated 0x200000 bytes memory
DK::memory vir Addr = 80c00000 phy Addr = c00000
Device Status is : 0x11
root at OpenWrt:/tmp# uname -a
Linux OpenWrt 2.6.27.10 #18 Sat Dec 27 13:54:44 CST 2008 mips unknown


The art driver was made for 2.6.15, I updated it to work on newer
kernels.  It has all the juicy details that an experienced ath9k
developer should need to get these chips working.

driver source AP81/AP83:
http://svn.openfoundry.org/jpgdauto/atheros/art_modules_v0_7_b9.tbz

Patch for newer kernels:
http://forum.openwrt.org/viewtopic.php?pid=78718#p78718


Going over the source, I really don't see anything specific to AHB
bus.  Any experienced ath9k developers who can review the source and
comment on this appreciated.  Especially tips on where in the ath9k
driver I need to be hacking.  Bribes of pizza, beer, New Years
sparkling wine are on the table ;)

Thank you.

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Luis R. Rodriguez]

On Sun, Dec 28, 2008 at 02:22:31PM -0800, Stephen Gutknecht (hilltx) wrote:
> A major discovery, good news!  Through exhaustive searching of every
> bit I could find about AHB drivers, I stumbled into the source code
> for the Atheros Radio Testing (ART) for these routers.  From my AP81
> router (Trendnet TEW-652BRP, DLink DIR-615 C1) running OpenWRT trunk:
> 
> 
> root at OpenWrt:/tmp# insmod art.ko
> CHIP REV ID: b4
> dk_dev_init::status after register_chrdev(dk) = 0
>  Base Phsycal address :0x180c0000
> DK::Add client 0
> DK::Reg phy addr = 180c0000 vir Addr = b80c0000
> DK::Irq = 2
> DK::Allocated 0x200000 bytes memory
> DK::memory vir Addr = 80c00000 phy Addr = c00000
> Device Status is : 0x11
> root at OpenWrt:/tmp# uname -a
> Linux OpenWrt 2.6.27.10 #18 Sat Dec 27 13:54:44 CST 2008 mips unknown
> 
> 
> The art driver was made for 2.6.15, I updated it to work on newer
> kernels.  It has all the juicy details that an experienced ath9k
> developer should need to get these chips working.
> 
> driver source AP81/AP83:
> http://svn.openfoundry.org/jpgdauto/atheros/art_modules_v0_7_b9.tbz
> 
> Patch for newer kernels:
> http://forum.openwrt.org/viewtopic.php?pid=78718#p78718

These should *not* be posted as they are not part of the driver, whoever
posted this is violating our license and this will be looked into.

Avoid using this as a source of documentation.

  Luis

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Stephen Gutknecht]

On Mon, Dec 29, 2008 at 8:07 AM, Luis R. Rodriguez
<lrodriguez@atheros.com> wrote:
> On Sun, Dec 28, 2008 at 02:22:31PM -0800, Stephen Gutknecht (hilltx) wrote:
>> A major discovery, good news!  Through exhaustive searching of every
>> bit I could find about AHB drivers, I stumbled into the source code
>> for the Atheros Radio Testing (ART) for these routers.  From my AP81
>> router (Trendnet TEW-652BRP, DLink DIR-615 C1) running OpenWRT trunk:
>>
>>
>> root at OpenWrt:/tmp# insmod art.ko
>> CHIP REV ID: b4
>> dk_dev_init::status after register_chrdev(dk) = 0
>>  Base Phsycal address :0x180c0000
>> DK::Add client 0
>> DK::Reg phy addr = 180c0000 vir Addr = b80c0000
>> DK::Irq = 2
>> DK::Allocated 0x200000 bytes memory
>> DK::memory vir Addr = 80c00000 phy Addr = c00000
>> Device Status is : 0x11
>> root at OpenWrt:/tmp# uname -a
>> Linux OpenWrt 2.6.27.10 #18 Sat Dec 27 13:54:44 CST 2008 mips unknown
>>
>>
>> The art driver was made for 2.6.15, I updated it to work on newer
>> kernels.  It has all the juicy details that an experienced ath9k
>> developer should need to get these chips working.
>>
>> driver source AP81/AP83:
>> http://svn.openfoundry.org/jpgdauto/atheros/art_modules_v0_7_b9.tbz
>>
>> Patch for newer kernels:
>> http://forum.openwrt.org/viewtopic.php?pid=78718#p78718
>
> These should *not* be posted as they are not part of the driver, whoever
> posted this is violating our license and this will be looked into.
>
> Avoid using this as a source of documentation.
>
>  Luis


The module is clearly identified by author and shows GPL license.  The
binary version distributed by companies such as DLink is also marked
in the same way.

root at ipcmedialinux1:/DLink615C1_GPL/AthSDK# strings
./rootfs/target/lib/modules/art_ap81.ko | grep  "="
cfiamd: write %s @ %p (%x != %x)
i=0:fPtr=0x%x:fS=0x%x:ebSize=0x%x
i=1:fPtr=0x%x:fS=0x%x:ebSize=0x%x
i=2:fPtr=0x%x:fS=0x%x:ebSize=0x%x
sysFlashConfigWrite: failed rc=%d
dk_dev_init::status after register_chrdev(dk) = %d
DK::Free reg space phy=%x vir=%x
DK::Free irq = %x
page order=%d
DK::Reg phy addr = %x vir Addr = %x
DK::Irq = %x
DK::memory vir Addr = %x phy Addr = %x
description=Linux MDK driver 1.0
author=muddin at atheros.com
license=GPL
depends=
vermagic=2.6.15- MIPS32_R2 32BIT gcc-3.4



root at ipcmedialinux1:/home/OpenWRT6/trunk/art_modules/modules# grep -R "MOD_" *
include/dk.h:#define MOD_LICENCE "GPL"
include/dk.h:#define MOD_AUTHOR "muddin at atheros.com"
include/dk.h:#define MOD_DESCRIPTION "Linux MDK driver 1.0"
main.c: MODULE_LICENSE(MOD_LICENCE);
main.c:MODULE_AUTHOR(MOD_AUTHOR);
main.c:MODULE_DESCRIPTION(MOD_DESCRIPTION);

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Luis R. Rodriguez]

On Mon, Dec 29, 2008 at 08:24:54AM -0800, Stephen Gutknecht (hilltx) wrote:
> On Mon, Dec 29, 2008 at 8:07 AM, Luis R. Rodriguez
> <lrodriguez@atheros.com> wrote:
> > On Sun, Dec 28, 2008 at 02:22:31PM -0800, Stephen Gutknecht (hilltx) wrote:
> >> A major discovery, good news!  Through exhaustive searching of every
> >> bit I could find about AHB drivers, I stumbled into the source code
> >> for the Atheros Radio Testing (ART) for these routers.  From my AP81
> >> router (Trendnet TEW-652BRP, DLink DIR-615 C1) running OpenWRT trunk:
> >>
> >>
> >> root at OpenWrt:/tmp# insmod art.ko
> >> CHIP REV ID: b4
> >> dk_dev_init::status after register_chrdev(dk) = 0
> >>  Base Phsycal address :0x180c0000
> >> DK::Add client 0
> >> DK::Reg phy addr = 180c0000 vir Addr = b80c0000
> >> DK::Irq = 2
> >> DK::Allocated 0x200000 bytes memory
> >> DK::memory vir Addr = 80c00000 phy Addr = c00000
> >> Device Status is : 0x11
> >> root at OpenWrt:/tmp# uname -a
> >> Linux OpenWrt 2.6.27.10 #18 Sat Dec 27 13:54:44 CST 2008 mips unknown
> >>
> >>
> >> The art driver was made for 2.6.15, I updated it to work on newer
> >> kernels.  It has all the juicy details that an experienced ath9k
> >> developer should need to get these chips working.
> >>
> >> driver source AP81/AP83:
> >> http://svn.openfoundry.org/jpgdauto/atheros/art_modules_v0_7_b9.tbz
> >>
> >> Patch for newer kernels:
> >> http://forum.openwrt.org/viewtopic.php?pid=78718#p78718
> >
> > These should *not* be posted as they are not part of the driver, whoever
> > posted this is violating our license and this will be looked into.
> >
> > Avoid using this as a source of documentation.
> >
> >  Luis
> 
> 
> The module is clearly identified by author and shows GPL license.  The
> binary version distributed by companies such as DLink is also marked
> in the same way.
> 
> root at ipcmedialinux1:/DLink615C1_GPL/AthSDK# strings
> ./rootfs/target/lib/modules/art_ap81.ko | grep  "="
> cfiamd: write %s @ %p (%x != %x)
> i=0:fPtr=0x%x:fS=0x%x:ebSize=0x%x
> i=1:fPtr=0x%x:fS=0x%x:ebSize=0x%x
> i=2:fPtr=0x%x:fS=0x%x:ebSize=0x%x
> sysFlashConfigWrite: failed rc=%d
> dk_dev_init::status after register_chrdev(dk) = %d
> DK::Free reg space phy=%x vir=%x
> DK::Free irq = %x
> page order=%d
> DK::Reg phy addr = %x vir Addr = %x
> DK::Irq = %x
> DK::memory vir Addr = %x phy Addr = %x
> description=Linux MDK driver 1.0
> author=muddin at atheros.com
> license=GPL
> depends=
> vermagic=2.6.15- MIPS32_R2 32BIT gcc-3.4
> 
> 
> 
> root at ipcmedialinux1:/home/OpenWRT6/trunk/art_modules/modules# grep -R "MOD_" *
> include/dk.h:#define MOD_LICENCE "GPL"
> include/dk.h:#define MOD_AUTHOR "muddin at atheros.com"
> include/dk.h:#define MOD_DESCRIPTION "Linux MDK driver 1.0"
> main.c: MODULE_LICENSE(MOD_LICENCE);
> main.c:MODULE_AUTHOR(MOD_AUTHOR);
> main.c:MODULE_DESCRIPTION(MOD_DESCRIPTION);

The MODULE_LICENSE() macro has been incorrectly used before as well, this is
another example. If you see the header of the file though you will so no GPL
copyright notices. Trust me, this is not GPL/ISC licensed material and should be
avoided.

  Luis

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Stephen Gutknecht]

On Mon, Dec 29, 2008 at 10:05 AM, Luis R. Rodriguez
<lrodriguez@atheros.com> wrote:
> The MODULE_LICENSE() macro has been incorrectly used before as well, this is
> another example. If you see the header of the file though you will so no GPL
> copyright notices. Trust me, this is not GPL/ISC licensed material and should be
> avoided.
>
>  Luis


This is very discouraging.  I've done my homework in good faith.  It
has been like pulling teeth to locate the details required to support
these chips.  Plus, I've never worked on a wireless driver before, so
I've been trying to learn the ropes and doing a lot of reading.

On December 17 you sent me a private email saying: "We don't yet
support SoC stuff, but openwrt guys are enthusiastic to get this going
it seems. Please use the openwrt lists and ask there."  Well, I did
ask, and the driver source can't be found, so seems to me I'm the
person to do the work for the community.  So I have been trying to
learn what exactly it means to do AHB programming and if this is
nothing more than a device ID issue (there isn't a whole lot out there
on AHB, PCI is far more common).

I admit I'm new to wireless driver hacking, but it sure seems to me
these chips are very similar to the other Atheros 9000 (9K) chips that
ATH9K driver supports.  I'm trying to collaborate with ATH9K project
to test and contribute patches to get these drivers working on these
chips.

I did my homework before embarking on this venture 2 months ago:
http://www.phoronix.com/scan.php?page=news_item&px=Njc1MA

September 27, 2008 dated story says that the HAL was released for the
"AR9100 WiSoC".  Well, it doesn't work on my build of madwifi on
OpenWRT.  So I go into this with the goal of tracking down technical
details of why.

In November, I went out and purchased a few routers based on the
"AR9100 WiSoC" with the intention of getting them to work with
OpenWRT.   I never signed any kind of NDA with Atheros, my interest in
this project is as a consumer of routers that I have traced back to a
common platform.  The Netgear WNR2000 was publicized as being based on
OpenWRT - and is built on the Atheros platform.   Except one major
problem, the GPL download from Netgear doesn't include any source code
for the drivers.  And it seems to be some version of ath_ahb driver
that was never released back to the madwifi project.  Or maybe I'm
just loading them wrong, as the whole DEVID thing on AHB is still
fuzzy to me (there is no equivalent to lspci on AHB bus that I can
find).   The binary only drivers that are in the GPL downloads work
with 2.6.15 kernel, which is nearly three years old (January 3, 2006).

I can't find the madwifi HAL source code that works on these routers.
So I've tried to make do without bothering you, as your December 17
email seemed to give me the brush off.   I know this is new, and I've
been respectful, but Atheros sure isn't making this job easy. I see
news stories about Atheros releasing a AR9100 WiSoC HAL, but where's
the code?

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[Luis R. Rodriguez]

On Mon, Dec 29, 2008 at 11:17:41AM -0800, Stephen Gutknecht (hilltx) wrote:
> On Mon, Dec 29, 2008 at 10:05 AM, Luis R. Rodriguez
> <lrodriguez@atheros.com> wrote:
> > The MODULE_LICENSE() macro has been incorrectly used before as well, this is
> > another example. If you see the header of the file though you will so no GPL
> > copyright notices. Trust me, this is not GPL/ISC licensed material and should be
> > avoided.
> >
> >  Luis
> 
> 
> This is very discouraging.  I've done my homework in good faith.  It
> has been like pulling teeth to locate the details required to support
> these chips.  Plus, I've never worked on a wireless driver before, so
> I've been trying to learn the ropes and doing a lot of reading.
> 
> On December 17 you sent me a private email saying: "We don't yet
> support SoC stuff, but openwrt guys are enthusiastic to get this going
> it seems. Please use the openwrt lists and ask there."  Well, I did
> ask, and the driver source can't be found, so seems to me I'm the
> person to do the work for the community.  So I have been trying to
> learn what exactly it means to do AHB programming and if this is
> nothing more than a device ID issue (there isn't a whole lot out there
> on AHB, PCI is far more common).
> 
> I admit I'm new to wireless driver hacking, but it sure seems to me
> these chips are very similar to the other Atheros 9000 (9K) chips that
> ATH9K driver supports.  I'm trying to collaborate with ATH9K project
> to test and contribute patches to get these drivers working on these
> chips.
> 
> I did my homework before embarking on this venture 2 months ago:
> http://www.phoronix.com/scan.php?page=news_item&px=Njc1MA
> 
> September 27, 2008 dated story says that the HAL was released for the
> "AR9100 WiSoC".  Well, it doesn't work on my build of madwifi on
> OpenWRT.  So I go into this with the goal of tracking down technical
> details of why.
> 
> In November, I went out and purchased a few routers based on the
> "AR9100 WiSoC" with the intention of getting them to work with
> OpenWRT.   I never signed any kind of NDA with Atheros, my interest in
> this project is as a consumer of routers that I have traced back to a
> common platform.  The Netgear WNR2000 was publicized as being based on
> OpenWRT - and is built on the Atheros platform.   Except one major
> problem, the GPL download from Netgear doesn't include any source code
> for the drivers.  And it seems to be some version of ath_ahb driver
> that was never released back to the madwifi project.  Or maybe I'm
> just loading them wrong, as the whole DEVID thing on AHB is still
> fuzzy to me (there is no equivalent to lspci on AHB bus that I can
> find).   The binary only drivers that are in the GPL downloads work
> with 2.6.15 kernel, which is nearly three years old (January 3, 2006).
> 
> I can't find the madwifi HAL source code that works on these routers.
> So I've tried to make do without bothering you, as your December 17
> email seemed to give me the brush off.   I know this is new, and I've
> been respectful, but Atheros sure isn't making this job easy. I see
> news stories about Atheros releasing a AR9100 WiSoC HAL, but where's
> the code?

You shouldn't be discouraged because someone posted code and you cannot
use it, things like this happen, just move on with what is publicly available
under a license which does allow you contribute. When in doubt you can ask.
We are engaged enough with the community to be replying to you and contributing
to ath9k, and working with the community where possible. What you may have
read about "HAL" code being released is the ath9k driver has the hardware
access routines necessary to operate the cards. Then we also have released
"HAL" code for our legacy chipsets too, you can read more about this here:

http://wireless.kernel.org/en/users/Drivers/ath5k#Documentationavailable

Instead of doing PCI writes you are going to be using an internal bus
for the read/writes. To work on this you need code to work with the MIPS
CPUs on the routers. My suggestion to check with openwrt folks comes
from the fact that they have some code to support the MIPS boards where
these new Atheros 11n card are coming on. What would be left to write is
some ahb sort of glue into ath9k to wrap reads/writes and instead of
doing PCI reads/writes do the specfific platform device reads/writes.

  Luis

[ath9k-devel] ath9k for ar9102 and ar9103, OpenWRT, AP81 router hardware

[RHS Linux User]

Hi All,

   It seems to me that this WHOLE wireless mesh business began when
someone at MIT hacked the Atheros driver....  I assume they either had
sources or a good MIPS reverse compiler?  Maybe both?

   The openwrt.org 'more or less' open Atheros driver doesn't work
reliably on my Meraki Minis :(.

   Maybe someone (at MIT?) should again hack the current family of Atheros
drivers and delete the buggy stuff and provide a simple RadioTap style
interface to send and receive packets. That way at least the Linux stuff 
would work reliably.

   And related to this, I have yet to find anything on the Atheros.com
website that says they support open wireless in any form ??!! 

   So the less Atheros code the better !!

   Maybe we should begin to think in terms of making tools so we can
easily reverse and re-engineer whosever buggy code so it works :). 

   warm regards to all and Happy New Year :).



On Mon, 29 Dec 2008, Luis R. Rodriguez wrote:

> On Mon, Dec 29, 2008 at 11:17:41AM -0800, Stephen Gutknecht (hilltx) wrote:
> > On Mon, Dec 29, 2008 at 10:05 AM, Luis R. Rodriguez
> > <lrodriguez@atheros.com> wrote:
> > > The MODULE_LICENSE() macro has been incorrectly used before as well, this is
> > > another example. If you see the header of the file though you will so no GPL
> > > copyright notices. Trust me, this is not GPL/ISC licensed material and should be
> > > avoided.
> > >
> > >  Luis
> > 
> > 
> > This is very discouraging.  I've done my homework in good faith.  It
> > has been like pulling teeth to locate the details required to support
> > these chips.  Plus, I've never worked on a wireless driver before, so
> > I've been trying to learn the ropes and doing a lot of reading.
> > 
> > On December 17 you sent me a private email saying: "We don't yet
> > support SoC stuff, but openwrt guys are enthusiastic to get this going
> > it seems. Please use the openwrt lists and ask there."  Well, I did
> > ask, and the driver source can't be found, so seems to me I'm the
> > person to do the work for the community.  So I have been trying to
> > learn what exactly it means to do AHB programming and if this is
> > nothing more than a device ID issue (there isn't a whole lot out there
> > on AHB, PCI is far more common).
> > 
> > I admit I'm new to wireless driver hacking, but it sure seems to me
> > these chips are very similar to the other Atheros 9000 (9K) chips that
> > ATH9K driver supports.  I'm trying to collaborate with ATH9K project
> > to test and contribute patches to get these drivers working on these
> > chips.
> > 
> > I did my homework before embarking on this venture 2 months ago:
> > http://www.phoronix.com/scan.php?page=news_item&px=Njc1MA
> > 
> > September 27, 2008 dated story says that the HAL was released for the
> > "AR9100 WiSoC".  Well, it doesn't work on my build of madwifi on
> > OpenWRT.  So I go into this with the goal of tracking down technical
> > details of why.
> > 
> > In November, I went out and purchased a few routers based on the
> > "AR9100 WiSoC" with the intention of getting them to work with
> > OpenWRT.   I never signed any kind of NDA with Atheros, my interest in
> > this project is as a consumer of routers that I have traced back to a
> > common platform.  The Netgear WNR2000 was publicized as being based on
> > OpenWRT - and is built on the Atheros platform.   Except one major
> > problem, the GPL download from Netgear doesn't include any source code
> > for the drivers.  And it seems to be some version of ath_ahb driver
> > that was never released back to the madwifi project.  Or maybe I'm
> > just loading them wrong, as the whole DEVID thing on AHB is still
> > fuzzy to me (there is no equivalent to lspci on AHB bus that I can
> > find).   The binary only drivers that are in the GPL downloads work
> > with 2.6.15 kernel, which is nearly three years old (January 3, 2006).
> > 
> > I can't find the madwifi HAL source code that works on these routers.
> > So I've tried to make do without bothering you, as your December 17
> > email seemed to give me the brush off.   I know this is new, and I've
> > been respectful, but Atheros sure isn't making this job easy. I see
> > news stories about Atheros releasing a AR9100 WiSoC HAL, but where's
> > the code?
> 
> You shouldn't be discouraged because someone posted code and you cannot
> use it, things like this happen, just move on with what is publicly available
> under a license which does allow you contribute. When in doubt you can ask.
> We are engaged enough with the community to be replying to you and contributing
> to ath9k, and working with the community where possible. What you may have
> read about "HAL" code being released is the ath9k driver has the hardware
> access routines necessary to operate the cards. Then we also have released
> "HAL" code for our legacy chipsets too, you can read more about this here:
> 
> http://wireless.kernel.org/en/users/Drivers/ath5k#Documentationavailable
> 
> Instead of doing PCI writes you are going to be using an internal bus
> for the read/writes. To work on this you need code to work with the MIPS
> CPUs on the routers. My suggestion to check with openwrt folks comes
> from the fact that they have some code to support the MIPS boards where
> these new Atheros 11n card are coming on. What would be left to write is
> some ahb sort of glue into ath9k to wrap reads/writes and instead of
> doing PCI reads/writes do the specfific platform device reads/writes.
> 
>   Luis
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel
>