2.6.19-rc2-mm2: D-Link DUB-E100 Rev. B broken

2.6.19-rc2-mm2: D-Link DUB-E100 Rev. B broken

[Bernhard Rosenkraenzer]

Hi,
D-Link DUB-E100 Rev. B USB Ethernet adapters have worked ok in some recent 
kernels - in rc2-mm2, they're broken again; they're detected correctly and 
the asix module is autoloaded, but after "ifconfig eth0 192.168.0.15 netmask 
255.255.255.0", the box becomes unresponsive and keeps repeating

eth0: Failed to enable software MII access
eth0: Failed to enable hardware MII access
eth0: Failed to enable software MII access
eth0: Failed to enable hardware MII access
eth0: Failed to enable software MII access
eth0: Failed to enable hardware MII access
eth0: Failed to enable software MII access
eth0: Failed to enable hardware MII access
eth0: Failed to enable software MII access
eth0: Failed to enable hardware MII access
eth0: Failed to write Medium Mode to 0x0334: ffffff92

This loops infinitely.

Last known working is 2.6.17 with a patch to bring the asix driver up to the 
14-Jun-2006 revision, first known broken is 2.6.19-rc1-mm1. (I'll try to 
figure out what exact change broke it, but this will take forever on this 
hw).

Re: 2.6.19-rc2-mm2: D-Link DUB-E100 Rev. B broken

[David Hollis]

On Mon, 2006-10-23 at 20:41 +0200, Bernhard Rosenkraenzer wrote:
> Hi,
> D-Link DUB-E100 Rev. B USB Ethernet adapters have worked ok in some recent 
> kernels - in rc2-mm2, they're broken again; they're detected correctly and 
> the asix module is autoloaded, but after "ifconfig eth0 192.168.0.15 netmask 
> 255.255.255.0", the box becomes unresponsive and keeps repeating
> 
> eth0: Failed to enable software MII access
> eth0: Failed to enable hardware MII access
> eth0: Failed to enable software MII access
> eth0: Failed to enable hardware MII access
> eth0: Failed to enable software MII access
> eth0: Failed to enable hardware MII access
> eth0: Failed to enable software MII access
> eth0: Failed to enable hardware MII access
> eth0: Failed to enable software MII access
> eth0: Failed to enable hardware MII access
> eth0: Failed to write Medium Mode to 0x0334: ffffff92

Hmm, the only change that you may not have is to wrap the MII calls with
a Mutex, though that isn't likely the culprit here (even though the
message is talking about MII access).  Could you enable the DEBUG define
at the top of the driver and see what kind of output you get in dmesg?

-- 
David Hollis <dhollis@davehollis.com>

Re: 2.6.19-rc2-mm2: D-Link DUB-E100 Rev. B broken

[Bernhard Rosenkraenzer]

On Monday, 23. October 2006 21:05, David Hollis wrote:
> > eth0: Failed to enable software MII access
> > eth0: Failed to enable hardware MII access
> > eth0: Failed to write Medium Mode to 0x0334: ffffff92
>
> Hmm, the only change that you may not have is to wrap the MII calls with
> a Mutex, though that isn't likely the culprit here (even though the
> message is talking about MII access).  Could you enable the DEBUG define
> at the top of the driver and see what kind of output you get in dmesg?

It isn't the culprit -- I've looked at it some more and noticed it works 
perfectly on another box with the same kernel (but it works as well on the 
broken box with an older kernel).

Looks like the USB port is acting up (only with the new kernel -- so this is 
probably triggered by a USB driver or possibly APIC change [will try with 
pci=noapic next])

The box where it breaks is an Asus Pundit R-350 barebone with an ATI USB 
controller:
00.13.0 0c03: 1002:4347 (rev 01) (prog-if 10)
	Subsystem: 1043:8108
	Flags: bus master, medium devsel, latency 32, IRQ 17
	Memory at fe800000 (32-bit, non-prefetchable) [size=4K]

The working box has a VIA USB 2.0 controller
00:10.3 0c03: 1106:3104 (rev 82) (prog-if 20)
        Subsystem: 1025:0046
        Flags: bus master, medium devsel, latency 64, IRQ 18
        Memory at d0002800 (32-bit, non-prefetchable) [size=256]
        Capabilities: [80] Power Management version 2

Re: 2.6.19-rc2-mm2: D-Link DUB-E100 Rev. B broken

[Bernhard Rosenkraenzer]

On Monday, 23. October 2006 21:45, Bernhard Rosenkraenzer wrote:
> Looks like the USB port is acting up (only with the new kernel -- so this
> is probably triggered by a USB driver or possibly APIC change [will try
> with pci=noapic next])

pci=noapic doesn't change anything, but I remember having weird IRQ problems 
on identical boxes before (in particular, a normal 8139too network card 
doesn't work if ACPI 2.0 support is disabled in the BIOS), so this is a 
likely candidate for "severely broken BIOS needs workarounds and happened to 
work earlier by coincidence"