Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 14887 - Passing kernel options noapic and pci=noacpi does not disable either
Summary: Passing kernel options noapic and pci=noacpi does not disable either
Status: RESOLVED NEEDINFO
Alias: None
Product: Gentoo Release Media
Classification: Unclassified
Component: Everything (show other bugs)
Hardware: x86 Linux
: High major (vote)
Assignee: Bob Johnson (RETIRED)
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2003-01-31 21:05 UTC by anthony
Modified: 2004-01-20 16:55 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments
dmesg output of boot kernel (gentoo.boot.dmesg.out,15.01 KB, text/plain)
2003-01-31 21:06 UTC, anthony
Details

Note You need to log in before you can comment on or make changes to this bug.
Description anthony 2003-01-31 21:05:43 UTC
Booting off the 1.4 rc2 install iso for Athlon XP.  Md5 digest successful.  Have
re-burned and duplicated bug as well.  I am on an SIS-based Desknote and IO-APIC
trashes the built-in ethernet, making the machine unusable and the install
impossible, so I must disable it, yet when I pass the appropriate options the
kernel enables it anyway.

Reproducible: Always
Steps to Reproduce:
1. Boot off a 1.4 rc2 iso
2. type gentoo noapic at the boot prompt
3. Watch the kernel ignore you

Actual Results:  
Kernel starts with apic enabled

Expected Results:  
apic should be disabled

can't run emerge info
Comment 1 anthony 2003-01-31 21:06:24 UTC
Created attachment 7794 [details]
dmesg output of boot kernel
Comment 2 CG 2003-02-15 14:53:35 UTC
the help has an error:
disable acpi with:

boot: gentoo acpi=off


I have to disable this because i have a laptop
http://larve.net/people/hugo/2002/12/evo410
With ACPI, the fan doesn't work and the laptop turns itself down after a while.

Perhaps a kernel upgrade/patch would be good.
Comment 3 Rob Davies 2003-07-10 05:54:00 UTC
On BP6 Dual Celeron board, running with noapic boot parameter is vital, as the 
LKML FAQ notes, the APIC checksum is weak.  I have noticed this with 
gentoo-sources kernel 2.4.20 r2. 
 
The noapic option means CPU0 deals with all hardware interrupts but the LOC 
IPI's needed for SMP synchronisation, behave as normal. 
 
Have had problems booting a kernel, built natively under Gentoo perhaps because 
of this error. 
Comment 4 Brian Jackson (RETIRED) gentoo-dev 2003-09-27 11:47:59 UTC
is this still an issue?
Comment 5 ferret 2004-01-20 03:33:44 UTC
I've recently had BIG problems because of this bug. Because of it I've had to trash my system and bootstrap from stage1 five times before I worked out how to fix it (or more precisely, Livewire and krunk helped me with it in #gentoo).

I have an nforce2 motherboard and so I can't enable APIC. I always enable it in the kernel config and disable it by passing the "noapic" parameter in LILO. This works fine, and has for several months. Then recently I started using GRUB instead, and despite the fact I was still passing "noapic" as a kernel parameter, APIC is still being found and enabled.

Checking dmesg output and /proc/cmdline the "noapic" is still getting passed, and as far as I can see, passed correctly. How having GRUB makes any difference I have no idea... but it does! I didn't change anything else on the system to cause this. I have briefly tested it afterwards to find that this is exactly what is happening.

Of course, it's pretty much impossible for GRUB or LILO to affect whether APIC will be loaded or not if "noapic" is passed to the kernel in each case... but I tell you it really happened. :)

For me this problem is fixed, since I'm disabling APIC support in the kernel. I also don't want to test these results any more, since it makes my system unstable.

For nForce2 users there's supposedly a patch out there somewhere for 2.6 kernels, so I'm going to look for it and pester akpm and/or the maintainer of gentoo-dev-sources to include it.

As far as I'm concerned, this matter is closed. :) Someone (whose computer doesn't require "noapic", perhaps) should try booting with different versions of LILO and GRUB to see if they can find out what's going on. More precisely, someone who has this problem with GRUB needs to try and get it with LILI too! If they can get it with LILO, then just ignore me, maybe I've altered something subtle in my kernel config recently without thinking.

Bye! :)