Re: BCM43228

Re: BCM43228

[Larry Finger]

On 11/01/2013 04:55 AM, Ruben De Smet wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi folks,
>
> I'm currently running Broamcoms prorietary wl drivers, which were
> running fine until I found a bug in the bluetooth part of it, which
> panics the kernel at a certain position.
> I don't know if it's a part in the GPL part of the kernel which
> panics, or if it's in the Broadcom part, as the kernel is tainted by
> the wireless driver and cannot be debugged that way.
>
> lspci shows me this:
> 03:00.0 Network controller: Broadcom Corporation BCM43228 802.11a/b/g/n
>
> My question: will b43 ever support this chip? If so, I'd be able to
> give someone access to this computer for debugging purposes.
> I'd give any help possible, though kernels and modules aren't my
> domain yet.

The short answer is "probably not". The reverse engineering process is very time 
consuming and boring. What seems to work best is to find a MIPS driver and 
disassemble it to form the kind of "specs" that are seen at 
http://bcm-v4.sipsolutions.net/. Next a different person needs to take that 
prescription and turn it into code. To preserve the clean-room setting, those 
two parts must be done by separate people. The process worked reasonably well 
for 802.11g devices, but  the code gets very complicated for 802.11n. Further 
complicating the issue is that at least one of two BCM43228 units has an LCNXN 
PHY, and we have done little with that although the web site shows WIP for that 
PHY with a BCM43227. The 2.4 GHz part of the 227 probably matches that of the 
228. Note that b43 has never worked well with the 5 GHz radio in any of the chips.

Of course, if the RE is done perfectly, then you end up with all the bugs of wl. :)

One other possibility is that the device might eventually be supported by 
brcmsmac. That set of authors has the advantage of having access to the Broadcom 
documentation and the wl sources. In case that is a possibility, I added the 
linux-wireless mailing list to this reply.

As shown at http://wireless.kernel.org/en/users/Drivers/b43#Supported_devices, 
there are two different PCI IDs that are called BCM43228. When you ask this kind 
of question, you should include the output of 'lspci -nn' so that the ID is listed.

Larry

Re: BCM43228

[Ruben De Smet]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



On 11/01/2013 04:31 PM, Larry Finger wrote:
>> Hi folks,
>> 
>> I'm currently running Broamcoms prorietary wl drivers, which
>> were running fine until I found a bug in the bluetooth part of
>> it, which panics the kernel at a certain position. I don't know
>> if it's a part in the GPL part of the kernel which panics, or if
>> it's in the Broadcom part, as the kernel is tainted by the
>> wireless driver and cannot be debugged that way.
>> 
>> lspci shows me this: 03:00.0 Network controller: Broadcom
>> Corporation BCM43228 802.11a/b/g/n
>> 
>> My question: will b43 ever support this chip? If so, I'd be able
>> to give someone access to this computer for debugging purposes. 
>> I'd give any help possible, though kernels and modules aren't my 
>> domain yet.
> 
> The short answer is "probably not". The reverse engineering process
> is very time consuming and boring. What seems to work best is to
> find a MIPS driver and disassemble it to form the kind of "specs"
> that are seen at http://bcm-v4.sipsolutions.net/. Next a different
> person needs to take that prescription and turn it into code. To
> preserve the clean-room setting, those two parts must be done by
> separate people. The process worked reasonably well for 802.11g
> devices, but  the code gets very complicated for 802.11n. Further
> complicating the issue is that at least one of two BCM43228 units
> has an LCNXN PHY, and we have done little with that although the
> web site shows WIP for that PHY with a BCM43227. The 2.4 GHz part
> of the 227 probably matches that of the 228. Note that b43 has
> never worked well with the 5 GHz radio in any of the chips.
> 
> Of course, if the RE is done perfectly, then you end up with all
> the bugs of wl. :)
> 
> One other possibility is that the device might eventually be
> supported by brcmsmac. That set of authors has the advantage of
> having access to the Broadcom documentation and the wl sources. In
> case that is a possibility, I added the linux-wireless mailing list
> to this reply.
> 
> As shown at 
> http://wireless.kernel.org/en/users/Drivers/b43#Supported_devices,
> there are two different PCI IDs that are called BCM43228. When you
> ask this kind of question, you should include the output of 'lspci
> -nn' so that the ID is listed.
> 
> Larry
> 

AFAIK, broadcom only offers it's drivers for intel x86 and x64_86; I
guess by MIPS, you mean the architecture? (why wouldn't the intel
driver work? Pretty unknown on the subject ;) )

I read the WIP mail about the 227 chip, I could possibly wait for that
one to be released and test it out on mine?

For the record: the issue I'm having is as follows: when connecting
via the bluetooth part of the chip to my phone and using DUN to
connect to the internet, everything works fine until I shut down the
connection. Then my kernel panics. (Another issue of course is the
fact that wl is the only closed source thing in my kernel running atm
on my notebook...)

On brcmsmac: is linux-wireless the right mailinglist to post to then?
I see that the range of supported devices is more close to my chip
than b43 is ;)
I added brcm80211-dev-list@broadcom.com to this mail too, which looks
like the more appropriate mailinglist to post the issue to.

R

PS. snippet of the output of 'lspci -nn':

03:00.0 Network controller [0280]: Broadcom Corporation BCM43228
802.11a/b/g/n [14e4:4359]

Other information on request, will try to answer asap.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSc9AYAAoJEPgmWogSE9wQIGYIAM34w5YzEW3njtkOX0STZf/m
2nDYgORLDgksjQGf7XbSLIn+M7W0u9HMVQXT/aBBxejnSidD/WYIfNLuJddjrGZm
pCl3vO/Z7Px/PI/TBa58MIlV9zZOR/T2dLJmijm7u466AFE23PXT1mJemYzlDNdE
LLykV+Gp6of8uYLzj2GdqVwHEOztdO9x2q5PGVlDiBfIMkqnt1xn6VqFDZdp75Di
viC03tE4HPkN4IIxJbPmZUzBo4L47eRLLJxXAHsVst/sD+YnHkhvEk0yIvSvF3b4
0kZ1jSsDGjryquZ2f4Bvkw8prKS/LyNmPIn8+SaCZ516+C8uUX1e2Z1em2v5DiI=
=Z3eG
-----END PGP SIGNATURE-----

Re: BCM43228

[Larry Finger]

On 11/01/2013 11:00 AM, Ruben De Smet wrote:
> AFAIK, broadcom only offers it's drivers for intel x86 and x64_86; I
> guess by MIPS, you mean the architecture? (why wouldn't the intel
> driver work? Pretty unknown on the subject ;) )
>
> I read the WIP mail about the 227 chip, I could possibly wait for that
> one to be released and test it out on mine?
>
> For the record: the issue I'm having is as follows: when connecting
> via the bluetooth part of the chip to my phone and using DUN to
> connect to the internet, everything works fine until I shut down the
> connection. Then my kernel panics. (Another issue of course is the
> fact that wl is the only closed source thing in my kernel running atm
> on my notebook...)
>
> On brcmsmac: is linux-wireless the right mailinglist to post to then?
> I see that the range of supported devices is more close to my chip
> than b43 is ;)
> I added brcm80211-dev-list@broadcom.com to this mail too, which looks
> like the more appropriate mailinglist to post the issue to.
>
> R
>
> PS. snippet of the output of 'lspci -nn':
>
> 03:00.0 Network controller [0280]: Broadcom Corporation BCM43228
> 802.11a/b/g/n [14e4:4359]
>
> Other information on request, will try to answer asap.

The MIPS drivers are part of the codes used in various SoC implementations used 
in wireless routers. Disassembling an x86 code stream is much harder than for a 
machine like MIPS where all instructions are the same size.

Larry

Re: BCM43228

[Arend van Spriel]

On 11/01/2013 05:00 PM, Ruben De Smet wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
>
> On 11/01/2013 04:31 PM, Larry Finger wrote:
>>> Hi folks,
>>>
>>> I'm currently running Broamcoms prorietary wl drivers, which
>>> were running fine until I found a bug in the bluetooth part of
>>> it, which panics the kernel at a certain position. I don't know
>>> if it's a part in the GPL part of the kernel which panics, or if
>>> it's in the Broadcom part, as the kernel is tainted by the
>>> wireless driver and cannot be debugged that way.
>>>
>>> lspci shows me this: 03:00.0 Network controller: Broadcom
>>> Corporation BCM43228 802.11a/b/g/n
>>>
>>> My question: will b43 ever support this chip? If so, I'd be able
>>> to give someone access to this computer for debugging purposes.
>>> I'd give any help possible, though kernels and modules aren't my
>>> domain yet.
>>
>> The short answer is "probably not". The reverse engineering process
>> is very time consuming and boring. What seems to work best is to
>> find a MIPS driver and disassemble it to form the kind of "specs"
>> that are seen at http://bcm-v4.sipsolutions.net/. Next a different
>> person needs to take that prescription and turn it into code. To
>> preserve the clean-room setting, those two parts must be done by
>> separate people. The process worked reasonably well for 802.11g
>> devices, but  the code gets very complicated for 802.11n. Further
>> complicating the issue is that at least one of two BCM43228 units
>> has an LCNXN PHY, and we have done little with that although the
>> web site shows WIP for that PHY with a BCM43227. The 2.4 GHz part
>> of the 227 probably matches that of the 228. Note that b43 has
>> never worked well with the 5 GHz radio in any of the chips.
>>
>> Of course, if the RE is done perfectly, then you end up with all
>> the bugs of wl. :)
>>
>> One other possibility is that the device might eventually be
>> supported by brcmsmac. That set of authors has the advantage of
>> having access to the Broadcom documentation and the wl sources. In
>> case that is a possibility, I added the linux-wireless mailing list
>> to this reply.
>>
>> As shown at
>> http://wireless.kernel.org/en/users/Drivers/b43#Supported_devices,
>> there are two different PCI IDs that are called BCM43228. When you
>> ask this kind of question, you should include the output of 'lspci
>> -nn' so that the ID is listed.
>>
>> Larry
>>
>
> AFAIK, broadcom only offers it's drivers for intel x86 and x64_86; I
> guess by MIPS, you mean the architecture? (why wouldn't the intel
> driver work? Pretty unknown on the subject ;) )
>
> I read the WIP mail about the 227 chip, I could possibly wait for that
> one to be released and test it out on mine?
>
> For the record: the issue I'm having is as follows: when connecting
> via the bluetooth part of the chip to my phone and using DUN to
> connect to the internet, everything works fine until I shut down the
> connection. Then my kernel panics. (Another issue of course is the
> fact that wl is the only closed source thing in my kernel running atm
> on my notebook...)
>
> On brcmsmac: is linux-wireless the right mailinglist to post to then?
> I see that the range of supported devices is more close to my chip
> than b43 is ;)
> I added brcm80211-dev-list@broadcom.com to this mail too, which looks
> like the more appropriate mailinglist to post the issue to.

Let me chime in on this. We always keep taps on linux-wireless so you 
are always right to post there. I created the brcm80211-dev-list so 
people can post their issues there regardless changes in the dev team. 
So address brcm80211 issues to brcm80211-dev-list and Cc the 
linux-wireless list. As a bonus that assures they are publicly archived.

Regards,
Arend

> R
>
> PS. snippet of the output of 'lspci -nn':
>
> 03:00.0 Network controller [0280]: Broadcom Corporation BCM43228
> 802.11a/b/g/n [14e4:4359]
>
> Other information on request, will try to answer asap.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2.0.22 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
>
> iQEcBAEBAgAGBQJSc9AYAAoJEPgmWogSE9wQIGYIAM34w5YzEW3njtkOX0STZf/m
> 2nDYgORLDgksjQGf7XbSLIn+M7W0u9HMVQXT/aBBxejnSidD/WYIfNLuJddjrGZm
> pCl3vO/Z7Px/PI/TBa58MIlV9zZOR/T2dLJmijm7u466AFE23PXT1mJemYzlDNdE
> LLykV+Gp6of8uYLzj2GdqVwHEOztdO9x2q5PGVlDiBfIMkqnt1xn6VqFDZdp75Di
> viC03tE4HPkN4IIxJbPmZUzBo4L47eRLLJxXAHsVst/sD+YnHkhvEk0yIvSvF3b4
> 0kZ1jSsDGjryquZ2f4Bvkw8prKS/LyNmPIn8+SaCZ516+C8uUX1e2Z1em2v5DiI=
> =Z3eG
> -----END PGP SIGNATURE-----
>

Re: BCM43228

[Ruben De Smet]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 11/01/2013 06:07 PM, Larry Finger wrote:
> On 11/01/2013 11:00 AM, Ruben De Smet wrote:
>> AFAIK, broadcom only offers it's drivers for intel x86 and
>> x64_86; I guess by MIPS, you mean the architecture? (why wouldn't
>> the intel driver work? Pretty unknown on the subject ;) )
>> 
>> I read the WIP mail about the 227 chip, I could possibly wait for
>> that one to be released and test it out on mine?
>> 
>> For the record: the issue I'm having is as follows: when
>> connecting via the bluetooth part of the chip to my phone and
>> using DUN to connect to the internet, everything works fine until
>> I shut down the connection. Then my kernel panics. (Another issue
>> of course is the fact that wl is the only closed source thing in
>> my kernel running atm on my notebook...)
>> 
>> On brcmsmac: is linux-wireless the right mailinglist to post to
>> then? I see that the range of supported devices is more close to
>> my chip than b43 is ;) I added brcm80211-dev-list@broadcom.com to
>> this mail too, which looks like the more appropriate mailinglist
>> to post the issue to.
>> 
>> R
>> 
>> PS. snippet of the output of 'lspci -nn':
>> 
>> 03:00.0 Network controller [0280]: Broadcom Corporation BCM43228 
>> 802.11a/b/g/n [14e4:4359]
>> 
>> Other information on request, will try to answer asap.
> 
> The MIPS drivers are part of the codes used in various SoC 
> implementations used in wireless routers. Disassembling an x86
> code stream is much harder than for a machine like MIPS where all 
> instructions are the same size.
> 
> Larry
> 
> 

AFAIK this chip won't be incorporated in routers nor access points.
As Arend van Spriel stated in a followup mail, I'll keep
brcm80211-dev-list@broadcom.com in the To field.

Any chance that the brcm80211 team knows something on this topic? For
what I feel here, they are the people who will be closest to the issue.

Thank you for your time.

R
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJSc+5SAAoJEPgmWogSE9wQau4IALRlKUQ3VpSjaX6/wgr2pY7X
d0pyHGCXVQAdKsvJoyIXqGAfpqRNM7jMsaZlP0ZEAwy0klq4Muh288UrF4SJGeHo
ZJx4j3AZhvu4bZrlWUc0tede0V1pe2Cughy6QIpgUwZdNli+FHpYsLXl+ktk2wR4
KQvUYzhB+09/YRSUgaGhbF+uR0XfW6uqK32NE8+hWjZF1VgGj7mZJd7ApYjbGTq/
91sOUM92F6JSFRGx9skCUwf0M/XhzCG+1w/vcw8eN7rbLqIh+GNYFhzhMC35xfjW
cENvCZ9hi0UyQQ+scPwbmQRHhwqsavN6Wk0irJdst6JxMML2jDjHsVAJR/vGTFk=
=drMC
-----END PGP SIGNATURE-----