Description of problem: When loading the bcm43xx module with more then 1 GB memory on x86_64 results in freezing the system. Version-Release number of selected component (if applicable): FC6 kernel 2.6.18-1.2798.fc6 How reproducible: 100 % Steps to Reproduce: 1. Boot a x86_64 system with more then 1 GB memory 2. Install the bcm43xx firmware 3. modprobe bcm43xx Actual results: Limiting the memory by adding mem=1023 to the kernel boot parameters everything works fine. Expected results: It should even work with more then 1GB memory. Additional info: The problem is discussed on the following URL. It seems it is a problem getting DMA capable memory below 1 GB. Most bcm43xx Hardware has only 30/31-Bit addressing and needs DMA below 1GB. http://www.mail-archive.com/bcm43xx-dev@lists.berlios.de/msg02071.html
Test kernels w/ back-ported DMA fixes here: http://people.redhat.com/linville/kernels/fc6/ Please give them a try and post the results here...thanks
Well, i gave it a try and the module loaded smoothly, but bringing up the interface with "# ifconfig eth1 up" panics the kernel. Call Trace: [<ffffffff8026929b>]show_trace+0x34/0x47 [<ffffffff802692c0>]dump_stack+0x12/0x17 [<ffffffff8029dc68>]down_read+0x15/0x23 [<ffffffff80296254>]blocking_notifier_call_chain+0x13/0x36 [<ffffffff80214e4b>]do_exit+0x1f/0x8c2 [<ffffffff802648c0>]do_page_fault+0x794/0x80f [<ffffffff8025ccdd>]error_exit+0x0/0x84 DWARF2 unwinder stuck at error_exit+0x0/0x84 Leftover inexact backtrace: <IRQ> [<ffffffff8035396>] vgacon_cursor+0x0/0x1a5 [<ffffffff882ecb86>] :bcm43xx:bcm43xx_dma_handle_xmitstatus+0x122/0x388 [<ffffffff882db96f>] :bcm43xx:bcm43xx_interrupt_tasklet+0x79f/0x889 [<ffffffff8028edbf>] tasklet_action+0x62/0xac [<ffffffff80211cd8>] __do_softirq+0x5e/0xd5 [<ffffffff802756ff>] end_level_ioapic_vector+0x9/0x16 [<ffffffff8025d1f0>] call_softirq+0x1c/0x28 [<ffffffff8026a454>] do_softirq+0xc2/0x85 [<ffffffff8026a2e2>] do_IRQ+0xec/0xf5 [<ffffffff8025c509>] ret_from_intr+0x0/0xa <EOI> [<ffffffff8025bfaa>] system_call+0x1a/0x83 [<ffffffff8025c00e>]system_call+0x7e/0x83 Kernel pamic - not syncing: Aiee, killing interrupt handler! With best regards Heiko
Just a data point to add to this one: my PowerBook G4 with 1.5GB of RAM in it has zero problems loading up the bcm43xx module.
Jarod, is that with the stock FC6 kernel? Or my test kernel?
Current kernels at the location from comment 1 now have the new d80211 infrastructure available. Please give those kernels a try, and be sure to change references in /etc/modprobe.conf to bcm43xx-d80211 instead of bcm43xx. You probably need to use v4 firmware as well. Does that work any better for you?
John I just tested kernel-2.6.18-1.2861.2.1.fc6.jwltest.16 x86_64 kernel and while the kernel no longer crashes, I could not get my wireless card to work. I'm not sure which version of firmware I used, I just used what was already installed nine months ago the last time I tried this.
I do get this message: Dec 23 10:07:42 rivendell kernel: bcm43xx_d80211: ASSERTION FAILED (radio_attenuation < 10) at: drivers/net/wireless/d80211/bcm43xx/bcm43xx_phy.c:1482:bcm43xx_find_lopair()
This bug seems to be fixed in the recent kernels from kernels.org. Because of this I use a self compiled kernel to run both the wireless LAN and my 2 gigs of memory. Chris` is another well known bug/problem, but in my opinion it has nothing to do with this one...
I wish someone would package a vanilla kernel rpm for extras.
Kernel 2.6.19 fixed this bug for me, and now 2.6.19-1.2895.fc6 is out I've tested the RH kernel which seems to work fine -- at least I'm submitting this comment using the bcm43xx driver with 2GB RAM on a x86_64 machine.
Closing based on comment 10. Please re-open (or open a new bug) if problems continue with kernels later than 2.6.19-1.2895.fc6.