From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0; H010818) Description of problem: On a workstation using the VIA VT6202 USB controller, the USB attached keyboard is dead when running BOOT controller. The keyboard works well when the system is running the UP or SMP kernel. This prevents an install on this system when only a USB keyboard is conected. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Boot system from CD1. 2. USB keyboard will not respond after the boot: prompt during the first install screen to choose language. Actual Results: Keyboard is non-responsive during the install or when running the BOOT kernel. Expected Results: Expect USB keyboard to work during install as it does when system is finally installed. Additional info:
This issue was found in Milan.
Please tell me if the problem can be reproduced by booting installed userland with BOOT kernel. The -BOOT kernel RPM can be installed on a plain system and added to /boot/grub/grub.conf just any other kernel. If the problem is reproduced, please use a network login to capture dmesg by running "dmesg >/tmp/dmesg.out", then attach the dmesg.out file to the bug. PLEASE DO NOT DROP dmesg.out INTO THE COMMENTS BOX.
Yes, it does. If I install the -BOOT kernel rpm and select it from grub I lose USB keyboard support. I will get a copy of dmesg and attach it.
Created attachment 77544 [details] dmesg output for -BOOT kernel / USB keyboard fails
Thanks for the dmesg. It indicates a dead interrupt line problem (it normally causes -110 errors). Perhaps -BOOT does parsing of some tables wrong, or has components missing. I would like to have 3 things done here. 1. Try to boot with "noapic", to see if the workaround works. If it does, it will carry us over for now. 2. Get output of "cat /proc/interrupts" with the same technique as dmesg.out, to confirm that it's actually interrupt routing problem. 3. Collect dmesg and /proc/interrupts on a working SMP kernel, so that we can compare those and see what is different. That would be helpful.
I should mention that the keyboard works with the UP kernel. (So that is already noapic). I will include the dmesg from the UP kernel.
Created attachment 77545 [details] dmesg output from -UP kernel / keyboard works.
Booting the -BOOT kernel with noapic WORKS! That is a shocker. I thought the BOOT kernel was noapic by default.
Yes, the BIOS is broken on the box. Windows cuts over to ACPI, so BIOS writers let normal tables to bitrot all across the board. If would be helpful if you downloaded an RPM from my web page and ran the program in it, called "mptable". Please add "-pirq" flag too. URL: http://people.redhat.com/zaitcev/linux/mptable-2.0.15a-1.i386.rpm The result can be used to discuss the problem with BIOS people. The working SMP kernel dmesg that you mentioned in the report won't hurt.
Thanks very much for your prompt attention. I didn't realize that the -BOOT kernel also used the mptables. The problem makes sense, now.