USB completely broken - SMP 1Ghz intel machine

[ewildgoose]

I have built a gentoo-sources 2.4.20 SMP kernel (and older 2.4.19) with uhci AND usb-uhci as modules so that I can try both.

In both cases doing just an insmod on this module (without even any of the HID stuff, etc) as soon as I plug in a USB mouse, or printer, the log starts to fill up with:

Mar 1 08:59:43 [kernel] usb_control/bulk_msg: timeout
Mar 1 08:59:43 [kernel] usb.c: USB device not accepting new address=94 (error=-110)

etc

I am pretty sure that Redhat with a 2.4.18 kernel made USB work on this machine and I have tried fiddling with the obvious settings in the BIOS (MPS 1.4/1.1, PnP, etc), but with no success.

The only help I found via google was a hint that this might be related to WinXP (dual boot) setting Interrupt on all devices to 9, but Linux setting some other number which the device rejects...?

Any thoughts on how to debug this further

For reference this is a dual 1Ghz intel CPU with a nasty Via 133a based motherboard (EP-D3VA)

Thanks

[cyrillic]

If a redhat kernel will work on your hardware, try this :

[Saruman]

If you are using ACPI, than you are likely going to get IRQ conflicts (at least on my smp pentium3, Asus CUV4X-D, VIA chipset). You can test by doing 'cat /proc/interrupts': if usb-uhci interrupts counter are 0 then you have IRQ conflict because of ACPI. You have two options:
1) boot the kernel with acpi=off and you'll miss the automatic power off when shutting down the system;
2) compile a new kernel with only APM (instead of ACPI) power management to maintain the ability to tun off your Pc when halting.

So far, this is my experience (non kernel tree worked right : -ac, -aa, -lolo, -ck). Should try with redhat-sources perhaps ...
_________________
---
Miser, miser! Modo niger et ustus fortiter!

[Saruman]

Ok, tested redhat-sources-2.4.20.2.48-r1. So far:

Pros:
- ACPI _does_ work
- USB is functioning right (scanner, printer & my memory stick work nice)
- SB live works (it gets assigned the same IRQ of acpi)

Cons:
- Nvidia-kernel driver 3.123 _does NOT_ compile so X is not useful at all for me (xfree drivers are bad)
- swap off (that is, "swapoff -a" command) makes the kernel segfault "at will" until fully blocking the system
- software turning off of the PC at shutdown _does NOT_ work

Till now, I stand on APM: old & ugly but working, and working right
_________________
---
Miser, miser! Modo niger et ustus fortiter!

[ewildgoose]

I found that the problem here is (apparently) quite common with via based motherboards. Seems that you have to add a "noapic" boot option to get things working

HOWEVER, the noapic option works with the vanilla kernel, and the -ac kernel (and perhaps more), but does not work with a gentoo-sources 2.4.20 kernel... Presumably one of the patches included is causing problems...?

Also, for the working kernels ACPI was NOT a problem (on or off worked OK), however, with the gentoo-sources kernel the kernel would not load, an error with ACPI was reported and then the boot process hung.

I'm not sure if/how this is worth escalating to a bug report...?

Thanks to all those who helped

[aequitas]

I had the problem using gaming sources.

Turned apic of in the kernel config and it was solved.

[edit]
but now my NIC has probs. Got to tweak this a little bit
[/edit]
_________________
I am not superstitious, that brings bad luck.