Bug#533357: linux-2.6: please add b43: Add fw capabilities, to support OpenFWWF 5.1

2009-09-06 Thread Moritz Muehlenhoff
On Tue, Jun 16, 2009 at 09:33:46PM +0200, Stefan Lippers-Hollmann wrote:
 Package: linux-2.6
 Version: 2.6.30-1
 Severity: wishlist
 Tags: patch
 
 Hi
 
 I am currently packaging [1], [2] the OpenFWWF (GPL2) [3] firmware for 
 core revision 5 Broadcom AirForce wlan chipsets [4], while the most 
 important patches to support OpenFWWF have already been merged for 2.6.30,
 b43: Add fw capabilities 403a3a136122457165321e90b7569a321cc9ac12
 has only been merged mainline yesterday [5].

Sorry, nooone has found the time to merge this for 2.6.30 :-/

2.6.31 packages will appear soon.

Cheers,
Moritz



-- 
To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#533357: linux-2.6: please add b43: Add fw capabilities, to support OpenFWWF 5.1

2009-06-16 Thread Stefan Lippers-Hollmann
Package: linux-2.6
Version: 2.6.30-1
Severity: wishlist
Tags: patch

Hi

I am currently packaging [1], [2] the OpenFWWF (GPL2) [3] firmware for 
core revision 5 Broadcom AirForce wlan chipsets [4], while the most 
important patches to support OpenFWWF have already been merged for 2.6.30,
b43: Add fw capabilities 403a3a136122457165321e90b7569a321cc9ac12
has only been merged mainline yesterday [5].

I would really appreciate if you could consider adding that patch to the 
next 2.6.30 kernel upload for Debian (I am using it successfully since 
kernel 2.6.29), as it would make packaging openfwwf significantly easier 
and avoid ugly upgrade handling. 

To explain the reasoning for my request (an overview is presented at [6]),
as of now OpenFWWF doesn't support hardware encryption and QoS yet, which 
are available in the proprietary Broadcom firmware (b43-fwcutter/ 
non-free), while b43 (until 2.6.30-git9) enables both on default, making
b43 to failing to authentificate and resulting numerous PHY resets 
and firmware crashes/ restarts (potential log spam). While it is possible
to force b43 into disabling both features through a modprobe.d override
[7], this would only be required until kernel 2.6.31 enters the archive and
present some problems with the upgrade handling (introducing 
/etc/modprobe.d/openfwwf.conf as a Debian conffile for ~3 months and 
subsequently removing it) and additionally hard-disables QoS even in case
OpenFWWF and the proprietary firmware (b43-fwcutter/ non-free) are 
coinstalled (since kernel 2.6.30, b43 first probes ${FIRMWARE_DIRS}/b43/, 
before falling back to ${FIRMWARE_DIRS}/b43-open/, provided by OpenFWWF).
Assuming this patch would be merged into 2.6.30-2, this would not affect 
compatibility with previous kernels, because ${FIRMWARE_DIRS}/b43-open/ 
is ignored until kernel 2.6.30 and therefore isn't eligible to the 
described problems (only 2.6.30-1 would be).

openfwwf and b43-asm (its build-dependency), as a dfsg-free firmware for, 
many devices covered by b43 are basically ready and only pending your 
decision, as this patch has already been merged upstream for 2.6.31, it
only applies to kernel 2.6.30. Backports to lenny's 2.6.26 are not 
reasonable (as it would require installing openfwwf under
${FIRMWARE_DIRS}/b43/ in conflict with b43-fwcutter/ non-free and the 
modprobe overrides at [7]), while it would be compatible lenny-and-a-half,
assuming b43: Add fw capabilities is present.

Regarding DebianKernelPatchAcceptanceGuidelines:
- Security Fixes:   FAIL
- Driver Fixes: depending on the interpretation, it was meant 
to go into 
2.6.30, but fell through the 
cracks and basically makes a new feature
of 2.6.30 (transparent OpenFWWF 
compatibility) work for the first 
time and would allow to operate 
these devices solely with packages 
from main, once openfwwf enters 
the archive.
- Stability Fixes:  FAIL

- New features: YES, strictly speaking
- Out-of Tree Drivers:  NO
- My favourite patch-set:   NO, single patch already merged upstream

I am aware that this patch is borderline following these patch acceptance 
guidelines, but hope that you might consider it nevertheless.

A short word regarding the current state of OpenFWWF, for core revision 5
devices [4], it is mostly stable, but as a reverse engineering efforts 
there may be still unknown corner cases and behaviour under (articifically)
high package pressure is said to be still slightly erratic (I haven't been 
able to observe such issues myself) and a decent replacement for the 
proprietary firmware. core revision 6-10 devices are likely to work, but
might expose several problems in practice and are not claimed to be 
supported upstream yet. core revision =11 devices do not work yet. I am
personally using this patch successfully since early 2.6.29 kernels on 
BCM4306 (rev 5 core) in STA and AP modi, BCM4318 (rev 9 core) works, but 
not perfectly and I have positive feedback from BCM4311/1 (rev 5 core) and 
BCM4318 (rev 5 core) users. 

Potential real world compatibility scenarios:
- for users who have the proprietary firmware installed already, this patch
  is basically a no-op, because ${FIRMWARE_DIRS}/b43/ is always preferred 
  by b43.
- users with kernel = 2.6.29 and only openfwwf installed won't experience
  any change, their b43 module ignores ${FIRMWARE_DIRS}/b43-open/ 
  (openfwwf) - recent compat-wireless snapshots contain this patch already.
- users with vanilla 2.6.30 or linux-2.6 2.6.30-1 and _only_ openfwwf 
  installed, are not able to get a connection with b43 and openfwwf and do 
  experience numerous PHY resets and firmware restarts (triggered by the 
  netdev watchdog), manifesting itself in syslog/ kern.log (not rate 
  limited).