[fedora-arm] smsc95xx performance bug: eth vs usb

[Niels de Vos]

On Tue, May 31, 2011 at 7:55 PM, DJ Delorie <dj at redhat.com> wrote:
>
>> I was interested to find out if this USB-device is conflicting with
>> the storage. Therefore you might be able to disable smsc95xx and see
>> if your storage performs much better.
>
> The storage is *also* connected via the smsc9514 chip.  It's a USB
> docking port for sata disks.
>
> So, ethernet and storage are going through the same physical USB hub
> (the smsc9514).

This is confusing me now... I can only find
linux-2.6/drivers/net/usb/smsc95xx.c, which definitely is not
providing a usb-hub (like linux-2.6/drivers/usb/core/hub.c) and also
not a USB-storage interface.

The different USB-devices can be on the same chip, but they will still
be different devices in the USB-stack. If you can attach or copy&paste
your /proc/bus/usb/devices, I should be able to tell you what bus/port
the smsc95xx is connected to, and how you can 'disconnect' it without
unloading the smsc95xx-module. Using unbind/bind over sysfs you can
logically disconnect the driver from the device, which would allow you
to do some tests without side-effects caused by the smsc95xx driver.