Bug 514787 - b43 Fails to load on iMac G5 (PPC)
Summary: b43 Fails to load on iMac G5 (PPC)
Keywords:
Status: CLOSED ERRATA
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: 10
Hardware: powerpc
OS: Linux
low
medium
Target Milestone: ---
Assignee: David Woodhouse
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2009-07-30 18:50 UTC by Jason Rivers
Modified: 2009-10-06 12:28 UTC (History)
5 users (show)

Fixed In Version: 2.6.27.35-170.2.94
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2009-10-06 12:28:33 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
potential fix (737 bytes, patch)
2009-07-31 12:04 UTC, David Woodhouse
no flags Details | Diff

Description Jason Rivers 2009-07-30 18:50:18 UTC
Description of problem:
b43 Firmware fails to load with the following:

b43-pci-bridge 0001:01:01.0: enabling device (0004 -> 0006)
b43-phy0: Broadcom 4306 WLAN found
b43-phy0 debug: Found PHY: Analog 2, Type 2, Revision 2
b43-phy0 debug: Found Radio: Manuf 0x17F, Version 0x2050, Revision 2
wmaster0 (b43): not using net_device_ops yet
wlan0 (b43): not using net_device_ops yet
input: b43-phy0 as /devices/virtual/input/input4
b43 ssb0:0: firmware: requesting b43/ucode5.fw
b43 ssb0:0: firmware: requesting b43/pcm5.fw
b43 ssb0:0: firmware: requesting b43/b0g0initvals5.fw
b43 ssb0:0: firmware: requesting b43/b0g0bsinitvals5.fw
b43-phy0: Loading firmware version 410.2160 (2007-05-26 15:32:10)
b43-phy0 debug: Chip initialized
b43-phy0 ERROR: The machine/kernel does not support the required 30-bit DMA mask
input: b43-phy0 as /devices/virtual/input/input5
b43-phy0 ERROR: Microcode not responding
b43-phy0 ERROR: You must go to http://linuxwireless.org/en/users/Drivers/b43#devicefirmware and download the latest firmware (version 4).


Version-Release number of selected component (if applicable):

Kernel release:2.6.29.6-93.fc10.ppc64
Memory: 512MB DDR
Wifi:Broadcom Corporation BCM4306 802.11b/g Wireless LAN Controller (rev 03)

Also tried with earlier kernels to no avail. this is as far as I got in updating.

How reproducible:

install on G5 iMac with Airport card.

Steps to Reproduce:

1. Install Fedora 10 on Mac G5 with Apple Airport Wireless
2. install b43 firmware
3. try loading b43 module.
  

Additional info:

with a hack in b43/dma.c to return 0 even if it gets the error of "The machine/kernel does not support the required 30-bit DMA mask" and changed from Error to Info. Though this isn't good for a real fix, it does sort my issue out on the PPC with only 512MB Memory.

Comment 1 David Woodhouse 2009-07-31 12:04:48 UTC
Created attachment 355802 [details]
potential fix

Hm, what's the point in whining about a 30-bit DMA mask when the machine only has 512MiB of RAM anyway?

Comment 2 Jason Rivers 2009-07-31 12:33:26 UTC
will test the fix as soon as I'm home.

Thanks for your help David.

Comment 3 Jason Rivers 2009-08-21 09:48:31 UTC
Thanks David. fix seemed to work. Mac now has Wifi on Fedora. (sorry for the delay)

Comment 4 David Woodhouse 2009-08-21 09:53:19 UTC
This is committed to the 2.6.29.6 branch in Fedora 11 CVS, and should be pushed when the next NULL pointer panic hits us.

Comment 5 Fedora Update System 2009-08-26 05:12:24 UTC
kernel-2.6.29.6-217.2.16.fc11 has been submitted as an update for Fedora 11.
http://admin.fedoraproject.org/updates/kernel-2.6.29.6-217.2.16.fc11

Comment 6 Fedora Update System 2009-08-27 02:19:11 UTC
kernel-2.6.29.6-217.2.16.fc11 has been pushed to the Fedora 11 stable repository.  If problems still persist, please make note of it in this bug report.

Comment 7 Josh Boyer 2009-09-04 16:46:47 UTC
This patch actually seems to cause problems rather than fix them on my 17" iMac G5.

The kernel from comment #5 produces a rather nasty oops on boot, which appears to be something scribbling over kernel text pages.  The 2.6.29.6-217-2.9.fc11 kernel doesn't have those problems.  I built a tip-of-branch (the private-fedora-f11-2_6_29_6 CVS branch) F11 kernel today with the patch for this reverted and that does not oops at all.  The only change was to not apply this patch.

I'll try and get a screenshot of the oops soon.

Comment 8 Josh Boyer 2009-09-04 17:18:28 UTC
Hm.  Actually, I think the patch itself is probably fine.  What isn't fine is that it's calling lmb_end_of_DRAM, which is marked as __init in the F11 kernels.  We need the fixes that went into .31 to remove the __init from the function and the header file

Comment 9 Josh Boyer 2009-09-04 17:21:08 UTC
I tested the rawhide kernel from today which is based on 2.6.31-rc8.git2 and contains the __init fixes.  The oops is gone.

Comment 10 Chuck Ebbert 2009-09-07 23:23:38 UTC
Backported fix from the f-11 2.6.29 kernel to 2.6.27.32-170.2.89.rc1.fc10


Note You need to log in before you can comment on or make changes to this bug.