[powerstation-owners] New experimental Firmware

[Stefan Nürnberger]

Adrian Reber wrote:
> On Thu, Nov 27, 2008 at 10:44:19AM +0100, Stefan Nürnberger wrote:
>>> What happens if you boot to the firmware prompt and turn then the devices
>>> on? Does it still crash in Linux? It could be that the firmware does not
>>> correctly disable the controller before booting Linux and there is still
>>> some DMA transaction active.
>> I tried that. I booted to firmware prompt and switched the usb harddisk  
>> on. The firmware did not detect it, since the "Scan USB" part was  
>> already done. The effect in Linux is still the same:
>>
>> With kernel 2.6.27.x:
>> - system will hang during boot when device is connected/switched on.
>> - system will boot fine if device is switched off during boot.
>> - system hangs as soon as the device is switched on.
>> - no output in dmesg, not even "device detected"
>>
>> With kernel 2.6.24-gentoo-r3:
>> - system will boot fine whether device is switched on or off
>> - system hangs seconds after the first attempt to do something with the  
>> device, e.g. trying to mount the device or even a simple 'lsusb'
>> - normal "device detected" output in dmesg
> 
> If you get the error if the USB devices are not turned on during SLOF's USB
> scan, then I would say it is a Linux bug.

Looks like the firmware is really innocent here. At least I managed to 
get some error output with decent console logging ('dmesg -n8', silly 
me). I verified the problem with another usb 2.0 card. This time it's a 
PCI card with ALI chipset, not PCIe. Behaviour is the same.

Error log: This is kernel 2.6.27.7. I booted the machine, logged in, set 
the log level and switched on the usb disk drive.
--- snip ---
bimini ~ # dmesg -n8
bimini ~ #
usb 5-6: new high speed USB device using ehci_hcd and address 2
Machine check in kernel mode.
Machine check in kernel mode.
Caused by (from SRR1=9000000000009032): Transfer error ack signal
Machine check in kernel mode.
Caused by (from SRR1=9000000000009032): Transfer error ack signal
Machine check in kernel mode.
Caused by (from SRR1=9000000000009032): Transfer error ack signal
Caused by (from SRR1=9000000000009032): Transfer error ack signal
Oops: Machine check, sig: 7 [#1]
SMP NR_CPUS=4 Maple
Modules linked in: snd_usb_audio snd_usb_lib snd_rawmidi snd_seq_device 
snd_hwdep snd_pcm snd_page_alloc and_timer snd evdev soundcore tg3 ehci_hcd


Stefan