Red Hat Bugzilla – Bug 92417
(IRQ_ROUTING)kernel 2.4.20-13.9smp breaks USB Epson scanner
Last modified: 2007-04-18 12:54:20 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2.1) Gecko/20030225
Description of problem:
My Epson Perfection 1650 USB scanner is not recognized by kernel 2.4.20-13.9smp
or 2.4.20-18.9smp. It was recognized by earlier kernels.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Boot kernel 2.4.20-9smp
2. lsusb - scanner is recognized (see attached lsusb.out-2.4.20-9smp)
3. Reboot kernel 2.4.20-13.9smp
4. lsusb - scanner is not recognized (see attached
Actual Results: Scanner not recognized by kernel 2.4.20-13.9smp
Expected Results: Scanner should be recognized.
Syslog entries when scanner is plugged in:
Jun 5 16:18:40 home kernel: hub.c: new USB device 00:07.2-2, assigned address 3
Jun 5 16:18:43 home /etc/hotplug/usb.agent: Setup scanner for USB product
Jun 5 16:18:44 home devlabel: devlabel service started/restarted
Jun 5 16:44:36 home kernel: hub.c: new USB device 00:07.2-2, assigned address 4
Jun 5 16:44:41 home kernel: usb_control/bulk_msg: timeout
Jun 5 16:44:41 home kernel: usb.c: USB device not accepting new address=4
Jun 5 16:44:41 home kernel: hub.c: new USB device 00:07.2-2, assigned address 5
Jun 5 16:44:46 home kernel: usb_control/bulk_msg: timeout
Jun 5 16:44:46 home kernel: usb.c: USB device not accepting new address=5
Created attachment 92186 [details]
lsusb -vv output from kernel 2.4.20-9smp
Created attachment 92187 [details]
lsusb -vv output from kernel 2.4.20-13.9smp
I would like two things attached:
- dmesg output, taken after scanner was connected
(/var/log/messages is dirty and incomplete).
This also shows if this is the ASUS or some other VIA.
- /proc/interrupts, on both kernels
Created attachment 92212 [details]
dmesg output from kernel 2.4.20-13.9smp
Created attachment 92213 [details]
dmesg output from kernel 2.4.20-9smp
Created attachment 92214 [details]
/proc/bus/usb/devices from kernel 2.4.20-13.9smp
Created attachment 92215 [details]
/proc/bus/usb/devices from kernel 2.4.20-9smp
Created attachment 92216 [details]
/proc/interrupts from kernel 2.4.20-13.9smp
Created attachment 92217 [details]
/proc/interrupts from kernel 2.4.20-9smp
Created attachment 92218 [details]
lspci -vv output (identical across kernels)
Ask and you shall receive!
FYI, this is an Abit VP6 with a VIA "Apollo Pro 133A" chipset. I have attached
the output of 'lspci -vv' (which is identical on either kernel) if you need more
Suspect this may be an issue related to bug 91041, but please don't dup just yet.
I have tried the workaround suggested in bug 91041 (setting the BIOS MPS level
to 1.1, rather than 1.4). The scanner *is* recognized when I do this. I can
also get the scanner recognized when the MPS level is set to 1.4 by booting with
the 'noapic' kernel option.
This bug is also present in kernel 2.6.0-0.test1.1.26smp.
My scanner is recognized when I boot 2.4.21-20.1.2024.2.1.nptlsmp or
2.6.0-0.test2.1.28smp with ACPI turned on. These are the first SMP kernels that
I have been able to boot with ACPI enabled on the VP6.
Booting either kernel with 'acpi=off' reverts to the broken behavior. (The
scanner is not recognized.)
This is a live issue again with Fedora Core 1, since its ACPI
implementation causes a kernel panic on this system. (See bug #109464.)
This bug (along with bug #109464) is still present in
Ian, what about FC2? This is where I realistically can poke our
experts in IRQ routing and ACPI.
To be honest, I've pretty much given up on using the IO-APIC on this
board. I was able to work around the USB IRQ problems with 2.4
kernels by using ACPI IRQ routing (which was an adventure in itself;
see https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109464 and
As of around 2.6.2, however, booting with ACPI on causes my logs to
fill with messages about APIC errors:
APIC error on CPU0: 40(40)
APIC error on CPU1: 40(40)
I just saw these with 2.6.8-1.521smp. Booting without ACPI, but with
the IO-APIC on, I get a bunch of messages about IRQ #145 being
disabled. /proc/interrupts lists this as the IRQ used by one of my
PDC20269 IDE controllers, so that can't be good news.
To net it out, I've been booting with "acpi=off apm=power-off noapic",
and the system is pretty stable (modulo a *very* intermittent hard
lock when switching from a VT running X to a text VT). It's not like
IO-APIC support is really a necessity on a dual-processor system,
Given the vintage of this hardware, I don't see much purpose in
burning a lot of cycles on this (i.e. WONTFIX seems appropriate).
Just promise you'll remember how reasonable I'm being when I finally
get my dual-Opteron system and report my first kernel bug. :-)
Thanks for the bug report. However, Red Hat no longer maintains this version of
the product. Please upgrade to the latest version and open a new bug if the problem
The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases,
and if you believe this bug is interesting to them, please report the problem in
the bug tracker at: http://bugzilla.fedora.us/