Re: Kernel Freeze, Atmel at79c505 (RFMD 2958), Fiberline FL-WL240U, usb-uhci.o
Joerg Albert <[email protected]>
| Newsgroups | gmane.linux.drivers.at76c503a.user |
|---|---|
| Message-ID | <[email protected]> |
Hi, On Mon, 12 Jan 2004, Klaus Franken wrote: > Am Samstag, 10. Januar 2004 21:30 schrieb Joerg Albert: > > > > Does anyone have better experiences? Any other ideas? > > > > got my WL240U today. Your crash is caused by a driver bug, which is fixed > > in 0.12beta5 (just checked in, avail. in the snapshot tomorrow). > > > > Unfortunately it just hid another problem: usb-uhci calls the driver's > > disconnect procedure when the device got reset during startup leading to > > _two_ wlanX netdev and a lot of trouble afterwards. uhci doesn't do that, > > nor does the unified uhci-hcd of 2.6.x. I compared uhci and usb-uhci and > > found a small patch to usb-uhci, which helped here (attached in case > > anyone feels like patching usb-uhci and compiling a new kernel). > > > > The problem doesn't show up if you use an external hub. > > > > So I'd recommend to either try and use 0.11, if you must use usb-uhci > > and cannot use an external USB hub or use >= 0.12beta5 with a patched > > usb-uhci. This applies to 2.4.x kernels only. > > Hmm, I'm using 0.12beta5 with patched and unpatched usb-uhci.o without succes > (kernel freezes and Opps ... ). > > But if I'm using an USB hub everything seems fine! yes, I also still have problems (althrough no crashes anymore): with the usb-uhci.o patch and 0.12beta5 wlan0 stays after the device is unplugged. The kernel oopses on stopping the USB subsys. No cure yet :-/ IMHO the problem is that the software implementation of the virtual root hub (i.e. the hub in the host controller) in usb-uhci.c differs to the one of ordinary hubs, but I have no idea how a correct one looks like. /Jörg