Re: DWL-520 Troubles

From: Mike Tyson (
Date: 2002-08-15 06:40:25 UTC

Hiya Journi =)

On Tuesday, August 13, 2002, at 03:30 AM, Jouni Malinen wrote:

> Which PCI BIOS revision are you using? I do not have D-Link DWL-520
> cards, but I have heard various reports about problems with PCI BIOS
> revisions older than 2.20.

This is a rev. 2.1 PCI BIOS
PCI: PCI BIOS revision 2.10 entry at 0xfb360, last bus=1

>> I'm not sure if it could be a cause, but the wireless card seems to be
>> sharing an IRQ with the firewire controller (pci info below). This is
>> after i moved the card to the second PCI slot in the box - previously,
>> it was sharing an irq with the ethernet controller (eth0).
> At least Host AP driver should be able to share IRQs. Anyway, the
> failing operation (TX fid alloc) is performed before any IRQs are
> needed, so I don't think this is the issue. Of course you could test
> the
> card without other PCI cards to make sure that something else is not
> causing the problem.

Hmm, unfortunately I've done all the PCI card shifting I can... It's a mini-system, and everything is on-board. There are 2 PCI slots, and I'm using just the one PCI card (the wlan card)... In one slot, it shares the IRQ with eth1, in the other, it's sharing with the firewire controller.

>> prism2_hw_config: initialized in 245 iterations
> This looks quite fast.. Prism2.5 PCI card I have used in tests
> initialize in about 18000 iterations..

Interesting... This is a fairly new machine, but I doubt it'll be going that fast and operating properly.

>> wlan0: CMD=0x000a => res=0x7f
>> wlan0: cannot allocate fid, len=2364
>> wlan0: CMD=0x000a => res=0x7f
>> wlan0: cannot allocate fid, len=1600
> These are more or less the first things done after card init. If the
> BIOS revision does not explain the problem, you could try increasing
> the
> COR sreset delay in hostap_pci.c by defining

I'll give that a try.. Thanks

When I compiled the kernel for this machine, I noticed I couldn't boot into it at first.. I had to append 'pci=bios' as a kernel param... Could this have something to do with the problem?



This archive was generated by hypermail 2.1.4.