Problems with USB on x86_64

Problems with USB on x86_64

[Michael Zapf]

Hi all,

I have some trouble using a memory stick of LG in my Athlon64 system. 
When I plug it in, dmesg gives messages like this:

ehci_hcd 0000:00:02.2: port 6 reset error -110
hub 1-0:1.0: hub_port_status failed (err = -32)
ehci_hcd 0000:00:02.2: port 6 reset error -110
hub 1-0:1.0: hub_port_status failed (err = -32)
hub 1-0:1.0: Cannot enable port 6.  Maybe the USB cable is bad?


and the stick cannot be mounted. It took some time of experimenting with 
different kernels to find out that

* using the 32-bit version of 2.6.8 or 2.6.11, the stick works fine, no 
error output!
* using the 64-bit version of 2.6.8 or 2.6.11, the problems occur as 
mentioned.

(2.6.8 as given by SuSE 9.2, 2.6.11 as by SuSE 9.3; both distributions 
offer a 32 and a 64-bit installation)
I also tried a Kubuntu 5.04 with 2.6.10, same result for 64 bit.

Does anybody have a good theory on what is happening here? Could this be 
a hardware problem?

Thanks for any hint,

Michael

Fixed the prob (was: Re: Problems with USB on x86_64)

[Michael Zapf]

Michael Zapf schrieb:

> I have some trouble using a memory stick of LG in my Athlon64 system. 
> When I plug it in, dmesg gives messages like this:
>
> ehci_hcd 0000:00:02.2: port 6 reset error -110
> hub 1-0:1.0: hub_port_status failed (err = -32)
> ehci_hcd 0000:00:02.2: port 6 reset error -110
> hub 1-0:1.0: hub_port_status failed (err = -32)
> hub 1-0:1.0: Cannot enable port 6.  Maybe the USB cable is bad?

just minutes before starting to pack my barebone to send it to service, 
it seems that I solved the issue with the USB stick by patching the 
ehci-hub.c file.

Actually, the messages in the log file

ehci_hcd 0000:00:02.2: port 6 reset error -110

showed that I had a timeout problem (110=ETIMEOUT). I found a call to a 
function "handshake" in ehci-hub.c which set the timeout to 500µs.

Increasing this timeout to 600µs allows the onboard hub to complete the 
reset. The stick is correctly mounted afterwards. (I tried 550, but this 
was still not enough.)

Any chance to have this included in future patches and versions? I guess 
there could be other people around with such a problem. The increase of 
the timeout should not hurt too much, should it?

Michael