On Oct 18, 2006, at 11:26 AM, Chavdar Ivanov wrote:

> On 10/18/06, Chavdar Ivanov <ci4ic4@gmail.com> wrote:
>> This all is being done under fresh -current, although the main  
>> problem
> ...
>> The BroadCom BCM4401-B0 (bce) started to work OK
>> after I included bmtphy in the GENERIC_LAPTOP configuration. The  
>> sound
>> (azalia) also shows, but I still haven't tested any of it.
>
> The gremlin with the BroadCom appeared again - funny enough after I
> booted into Windows XP to check if the sound was on. Now I am getting
>
> bce0: timeout disabling ethernet mac
> bce0: timeout writing pkt filter ctl

If if isn't already off, try disabling PXE/RPL/Netboot agent in the  
BIOS. At least with the 575x chips, I see problems where our bge  
driver doesn't properly tell the chip to disable the PXE/UNDI stack  
when the chip is reset. It almost always happens when warm rebooting  
from windows to NetBSD until I PXE boot or hard power off the system  
(unplugging it), and sometimes not even then.

> when I try to up it again. It is probably an interrupt routing problem
> - azalia0. wpi0 uhci1 and bce0 all use irq11.

Yes, on the GX280, GX620 and Optiplex 745, there are about 6 devices  
on IRQ 11, NetBSD fails miserably with both APIC and non-APIC, ACPI  
and non-ACPI kernels. Linux works fine.

> The sound worked after I increased the volume with
>
> mixerctl -w outputs.hp=200
>
> but only to the external output - couldn't figure out how to switch it
> to the speaker (not that I will actually *use* the speaker, but there
> you are).

I don't know how well these new chips that use UAA bus (Universal  
Audio Architecture) work, I didn't even try it. Most of the time I  
couldn't even type because the USB controllers keep shutting down,  
and the NIC doesn't work so I was debugging over serial line. I gave  
up when I couldn't even get a PCI ID for the new Broadcom chip  
committed.