Most recent kernel where this bug did not occur: 2.6.11.5 Distribution: Debian unstable Hardware Environment: Toshiba R100 laptop, see http://www.freed.net/opensource/R100.html Software Environment: Vanilla 2.6.13.1 from kernel.org Problem Description: Geting the message "ehci_hcd 0000:00:1d.7: Unlink after no-IRQ? Controller is probably using the wrong IRQ" when trying to get at USB, following some bad PCI stuff above, see dmesg. Steps to reproduce: Boot the machine into 2.6.13.1
Created attachment 6013 [details] The .config file
Created attachment 6014 [details] the dmesg output
Created attachment 6015 [details] lspci -v
Created attachment 6016 [details] lspci
Please report whether (and how!) the IRQ assignments are different between 2.6.11.5 and 2.6.13.1 ... if they're different (like EHCI not being on irq11) that suggests the issue is ACPI IRQ setup again. Also, that "UHCI irq happens too early" thing seems likely to be related to BIOS issues. Try the "usb-handoff" kernel command line parameter, and disabling USB keyboard/mouse support in the BIOS.
Created attachment 6025 [details] 2.6.11.5 messsages
This is my first Kernel Bug, so bear with me --- I am attachning the /var/log/messages and lspci -v from the 2.6.11.5. I will try later to boot with the option you mentioned. and dig into the BIOS to play with the USB keyboard stuff.
Created attachment 6026 [details] 2.6.11.5 lspci -v
"usb-handoff" worked. Oddly, it did not work as an "append=" in lilo.conf. I suppose now it becomes a doc issue.
Ok, closing then.