FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Gentoo > Gentoo Desktop

 
 
LinkBack Thread Tools
 
Old 03-22-2011, 06:15 PM
Duncan
 
Default System problems

Lindsay Haisley posted on Tue, 22 Mar 2011 10:41:24 -0500 as excerpted:

>> > Let's just bring this back to technical discussion of the problem,
>> > shall we? I think that's what's needed here.
>>
>> Let me try one more time, then. On my original reply you only
>> responded to the top half. In the second part I asked a question about
>> whether you're running an initrd/initramfs
>
> Duncan, I'm very sorry to have overlooked your question (twice)! Please
> accept my apology. The answer is no, on the box is question, I'm not
> running an initrd/initramfs, and unless it's necessary I would rather
> not do so. If I were, this would be one explanation for the problem.

I'm absolutely with you on that. If you don't need an initr*, it
definitely makes the whole thing simpler to avoid it (as I too have done).

(FWIW, my style is to make my opinion known, but go ahead and deal with
the question too, to the extent that I can. You saw the opinion part and
skipped the rest, which is understandable, I suppose... In any case it's
straight now.)

>> I'm still waiting on an answer to that, but as I spent quite some time
>> on the details the first AND second time, I'll simply refer you back to
>> them this time.
>
> Thanks for following up. I'll re-read your posts and get back on this.
> I haven't been able to follow up on the problem. It's planting season
> here and beyond doing email I've been kinda tired and not up for late
> night hacking after a day spent building a greenhouse or whatever. At
> this point, with a cooler head and a verified back out path, I'm 99%
> sure I can solve the problem myself. I was kind of freaked out when
> this problem originally happened since I depend so heavily on the
> desktop system for company billing, website development, customer
> support, etc.

Panicking is certainly understandable, particularly coming at an already
busy time of year. But that's behind us now. And simply having no time
to deal with it for a few days is something I'm sure we all deal with,
especially when there ends up a social crisis to deal with too. Sometimes
it's all just too much and something has to give. When that's finally
realized and done... it's like venting steam from an overheated nuclear
reactor (apropos comparison ATM, if I do say so =:^), it might mean a
slightly elevated immediate issue, but tends to deescalate the entire
situation.

Meanwhile, now we know the fork in the path to take. Without an initr*,
the fact that you get even a limited userspace (not just a kernel panic)
means that the kernel has the necessary drivers to get to and do the
initial load of the rootfs. The problem must be beyond that, in the
userspace config, initscripts or binaries.

At this point I'd guess something like the udev/kernel-config-deprecated-
sysfs issue someone else mentioned.

--
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman
 
Old 03-22-2011, 06:48 PM
"Nicholas E. Andrade"
 
Default System problems

Quoting Paul Hartman <paul.hartman+gentoo@gmail.com>:

On Tue, Mar 22, 2011 at 11:35 AM, Lindsay Haisley
<fmouse-gentoo@fmp.com> wrote:

And it bails if _any_ ebuild fails, so I
have to go to the office, figure out the problem, and restart the
process.


With a new-enough portage you can use the --keep-going switch to cause
it to emerge everything it possibly can emerge which is not dependent
on the failed package.

If your system is new enough to support bootable CD/DVD/USB, I'd try
to find a linux live CD which runs the same kernel version that you're
hoping to switch to, and (assuming it works) look at its kernel
config, dmesg, lspci -k, that sort of thing, and try to use it as a
hint for configuring your own kernel on your hardware. You can chroot
into gentoo from the liveCD and do all the kernel compilation stuff
from there so it might be relatively painless.



Also if you do choose to go to a kernel newer than 2.6.32, consider
trying the "make localyesconfig" which takes a loaded, working kernel
(e.g. from a live CD) and creates a .config based on it. For
additional info:
http://kernelnewbies.org/Linux_2_6_32#head-11f54cdac41ad6150ef817fd68597554d9d05a5f





----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
 
Old 03-22-2011, 06:51 PM
 
Default System problems

Portage now has a very useful switch that you can use "emerge --keep-going" that will not stop when a package will not compile and will continue on to the next leaving output at the end of the emerge what programs did not compile. Cheers
Sent from my Verizon Wireless BlackBerry

-----Original Message-----
From: Lindsay Haisley <fmouse-gentoo@fmp.com>
Date: Tue, 22 Mar 2011 11:35:47
To: <gentoo-desktop@lists.gentoo.org>
Reply-to: gentoo-desktop@lists.gentoo.org
Subject: Re: [gentoo-desktop] System problems

Thanks, eamjr56!

On Tue, 2011-03-22 at 07:43 +0000, eamjr56@gmail.com wrote:
> Now that you explained more fully the situation and why and what led
> to your problems I understand some of your frustrations. Sometimes
> wanting to fix one thing leads to other complex issues. Maybe (being
> Gentoo-specific) would be first to build your toolchain
> (linux-headers, glibc, binutils,and gcc) to today's versions then a
> "emerge -e world" so all of your stuff will be new and
> compatable-(your original problem).

The box is scheduled for replacement, as I've noted. The hardware is
old, relatively, and has limitations. Additionally, Gentoo has its
limitations as a desktop environment, although it's much easier to
maintain on servers. As Linux has to some extent penetrated the
marketplace, manufacturers are distributing drivers for things like
printers as .deb or .rpm files - no joy for Gentoo users :-( My sweetie
bought such a printer a while back, and I can't use it for this reason.
Pulling packages apart and installing the contents manually is possible,
but not simple, and I have other things to do.

I'm reluctant to "emerge -e world" on the box because X is seriously out
of date, and I've been trying to avoid having to deal with the issue.
revdep-rebuild takes days! And it bails if _any_ ebuild fails, so I
have to go to the office, figure out the problem, and restart the
process. When I get the box so it'll boot properly with a more recent
kernel I can at least move in that direction.


> All of this can be done in the background while you keep working.
> This is only one option. Feel free to use it or not. Sometimes I cuss
> the Gentoo devs for things they do that screws up my installs too.

Gentoo is substantially more complex as a distribution than are others.
Subtle screw-ups in ebuilds are more than just an annoyance on my
servers, where things break in the background after an emerge and I
don't find out until one or more customers calls me and says "**** isn't
working!"

> As to your kernel question as to "default" switches in Gentoo
> kernel I really don't think there are. Gen-kernel is one answer, but
> it never worked for me, I always had to get dirty with "make
> menuconfig".

I tried genkernel for a while and didn't like it. For one thing, I
recall that it installed an initrd, which introduced a level of
complexity that I didn't feel was necessary. I use an initrd on our
SOHO gateway box since the root filesystem is managed by EVMS, but I had
to manually re-write the initrd filesystem at one point pursuant to a
glibc upgrade! Too much work.

I assume that there must some mods to or initial setting of the default
kernel switches, somewhere, to avoid things such as enabling CONFIG_IDE
in a kernel build where it conflicts with recent releases of udev.

When I get a few moments to work on it again I'm starting with a new,
much more recent kernel and _not_ massaging my .config from a previous
kernel for it. I assume the default settings will at very least produce
a bootable system and I need only add drivers/modules for my hardware -
NIC, sound cards, etc. to get it up and running. I can go in and add
stuff later as needed.

> Hope you figure it out. Cheers

Thanks, really. I mean it! With a cooler head and a verified back down
plan for any changes I may make (plus the few points of light I received
on this list in spite of the misunderstandings), I'm 99% sure I can jump
in an fix the problem myself. If I post here with it again, I'll have
my presentation of the issue MUCH better organized and thought out.

--
Lindsay Haisley |"Windows .....
FMP Computer Services | life's too short!"
512-259-1190 |
http://www.fmp.com | - Brad Johnston
 
Old 03-22-2011, 08:11 PM
Lindsay Haisley
 
Default System problems

On Tue, 2011-03-22 at 12:48 -0700, Nicholas E. Andrade wrote:
> Also if you do choose to go to a kernel newer than 2.6.32, consider
> trying the "make localyesconfig" which takes a loaded, working kernel
> (e.g. from a live CD) and creates a .config based on it. For
> additional info:
> http://kernelnewbies.org/Linux_2_6_32#head-11f54cdac41ad6150ef817fd68597554d9d05a5f

Thanks, that's an interesting possibility. Live CD kernels do tend to
be somewhat overloaded with features/modules since they're designed to
cover all possible systems on which they might be used. Nonetheless,
that's good to know. It might be useful.

--
Lindsay Haisley <fmouse-gentoo@fmp.com>
FMP Computer Services
 
Old 03-22-2011, 08:35 PM
Donnie Berkholz
 
Default System problems

On 02:38 Mon 21 Mar , Lindsay Haisley wrote:
> On Sun, 2011-03-20 at 21:13 -0500, Donnie Berkholz wrote:
> > I also suspect What Jean-Marc said is the problem. I'd recommend
> > completely disabling everything in the old ATA section to ensure it
> > doesn't attempt to control any devices, while building the PATA
> > driver into the kernel and using root=/dev/sdXN in the grub
> > parameters.
>
> If I use the stock configuration from the 2.6.36-gentoo-r5 kernel,
> won't this have the correct basic kernel facilities built in, at least
> as far as the deprecated IDE capabilities are concerned and the libata
> replacement? I assume the Gentoo devs modify kernels so that the
> default config settings are more appropriate than those which come
> with the vanilla kernel from the kernel devs, yes?

Things are pretty vanilla, as per Gentoo philosophy, unless you run
genkernel to build your kernel. I wouldn't rely on anything being built
in for a manually configured kernel.

Take a look at the kernel config guide if you want some pointers:

http://www.gentoo.org/doc/en/handbook/handbook-amd64.xml?part=1&chap=7

> Putting "root=/dev/sda4" on the kernel cmd line in grub actually
> worked, and got me a bit further in the boot process. The kernel
> obviously understood it. However later in the boot process, I got
> "Checking the root filesystem", following by an error message that the
> root filesystem spec of /dev/sda4 wasn't understood. This is a
> complaint about the root fs spec is in /etc/fstab, since I had been
> using a UUID spec there, and got an error at the same point in the
> boot-up about the UUID instead.

That is symptomatic of a missing driver for an ATA controller or root
filesystem.

> > It might be a worthwhile step to boot from a LiveCD and run `lspci
> > -k` to identify the kernel modules.
>
> lsmod will probably give the same useful information.

The useful thing about `lspci -k` is it only shows modules actually used
by hardware on your system, rather than the huge superset of modules
that are loaded.

--
Thanks,
Donnie

Donnie Berkholz
Desktop project lead
Gentoo Linux
Blog: http://dberkholz.com
 
Old 03-22-2011, 08:42 PM
Lindsay Haisley
 
Default System problems

On Tue, 2011-03-22 at 19:15 +0000, Duncan wrote:
> I'm absolutely with you on that. If you don't need an initr*, it
> definitely makes the whole thing simpler to avoid it (as I too have done).

I have an initrd on our firewall/fileserver here at the SOHO since I'm
I'm loading the root fs from a RAID-supported, EVMS drive. I did this
as a kind of proof of concept to learn how to do it, but at one point a
kernel upgrade forced me to rebuild the initrd manually with a newer
glibc. It was a strange problem, and a lot of work to fix. If I could
diagnose and fix _that_ I reckon I can quit whining and fix the
(probably relatively easy) one with my desktop boot.

It's a pity that EVMS is an orphaned project, or was the last time I
checked. IBM dropped it. It was and is a pretty slick system.

Ubuntu uses an initrd in all their stuff. I believe it's involved in
the display of their boot graphic and such. They've achieved _very_
fast boot times, though, and the loaded system is disk-based. I know a
lot of Gentoo people look down on Ubuntu, but every distribution I've
ever worked with has its advantages and disadvantages. Open Source is
about choice :-)

> Panicking is certainly understandable, particularly coming at an already
> busy time of year. But that's behind us now. And simply having no time
> to deal with it for a few days is something I'm sure we all deal with,
> especially when there ends up a social crisis to deal with too.

I'm going to nominate Jorge for sainthood! Everybody's moved on. Kudos
to everybody :-)

> Meanwhile, now we know the fork in the path to take. Without an initr*,
> the fact that you get even a limited userspace (not just a kernel panic)
> means that the kernel has the necessary drivers to get to and do the
> initial load of the rootfs.

I was getting a kernel panic for a while until I added a root spec to
the kernel invocation in menu.lst.

> The problem must be beyond that, in the
> userspace config, initscripts or binaries.

This is probably close to the point. Still working on our greenhouse,
but soon I'll get the time to get back on the case with this problem.

> At this point I'd guess something like the udev/kernel-config-deprecated-
> sysfs issue someone else mentioned.

Excellent shot, Duncan. I'll check it. I won't be using the 2.6.29
kernel again, as per Roman's suggestion, but 2.6.36 instead.

--
Lindsay Haisley | "We are all broken toasters, but we still
FMP Computer Services | manage to make toast"
512-259-1190 |
http://www.fmp.com | - Cheryl Dehut
|
 
Old 03-22-2011, 08:49 PM
Lindsay Haisley
 
Default System problems

On Tue, 2011-03-22 at 19:15 +0000, Duncan wrote:
> I'm absolutely with you on that. If you don't need an initr*, it
> definitely makes the whole thing simpler to avoid it (as I too have done).

I have an initrd on our firewall/fileserver here at the SOHO since I'm
I'm loading the root fs from a RAID-supported, EVMS drive. I did this
as a kind of proof of concept to learn how to do it, but at one point a
kernel upgrade forced me to rebuild the initrd manually with a newer
glibc. It was a strange problem, and a lot of work to fix. If I could
diagnose and fix _that_ I reckon I can quit whining and fix the
(probably relatively easy) one with my desktop boot.

It's a pity that EVMS is an orphaned project, or was the last time I
checked. IBM dropped it. It was and is a pretty slick system.

Ubuntu uses an initrd in all their stuff. I believe it's involved in
the display of their boot graphic and such. They've achieved _very_
fast boot times, though, and the loaded system is disk-based. I know a
lot of Gentoo people look down on Ubuntu, but every distribution I've
ever worked with has its advantages and disadvantages. Open Source is
about choice :-)

> Panicking is certainly understandable, particularly coming at an already
> busy time of year. But that's behind us now. And simply having no time
> to deal with it for a few days is something I'm sure we all deal with,
> especially when there ends up a social crisis to deal with too. Sometimes
> it's all just too much and something has to give. When that's finally
> realized and done... it's like venting steam from an overheated nuclear
> reactor (apropos comparison ATM, if I do say so =:^), it might mean a
> slightly elevated immediate issue, but tends to deescalate the entire
> situation.

I'm going to nominate Jorge for sainthood!

> Meanwhile, now we know the fork in the path to take. Without an initr*,
> the fact that you get even a limited userspace (not just a kernel panic)
> means that the kernel has the necessary drivers to get to and do the
> initial load of the rootfs. The problem must be beyond that, in the
> userspace config, initscripts or binaries.

This is probably close to the point. Still working on our greenhouse,
but soon I'll get the time to get back on the case with this problem.

> At this point I'd guess something like the udev/kernel-config-deprecated-
> sysfs issue someone else mentioned.

Excellent shot, Duncan. I'll check it. I won't be using the 2.6.29
kernel again, as per Roman's suggestion, but 2.6.36 instead.

--
Lindsay Haisley | "Everything works if you let it"
FMP Computer Services | (The Roadie)
512-259-1190 |
http://www.fmp.com |
 
Old 03-23-2011, 01:02 AM
Lindsay Haisley
 
Default System problems

>From the config for the 2.6.29 kernel on my desktop box.

CONFIG_SYSFS_DEPRECATED_V2=y

Oops!

Very possibly the nucleus of the problem. Thanks, Jorge.

--
Lindsay Haisley | "Everything works if you let it"
FMP Computer Services | (The Roadie)
512-259-1190 |
http://www.fmp.com |
 
Old 03-23-2011, 11:39 AM
James Cloos
 
Default System problems

>>>>> "LH" == Lindsay Haisley <fmouse-gentoo@fmp.com> writes:

LH> It seems that all
LH> my /dev/hda? drives have been renamed /dev/sda? so I set gave
LH> "root=/dev/sda4" as a kernel parameter and got a little further. After
LH> "Checking root filesystem" in the boot sequence, I got a message that
LH> the UUID for the root filesystem wasn't understood in /etc/fstab.

LH> So I set the root filesystem in /etc/fstab to /dev/sda4, and got the
LH> same error - that "/dev/sda4" was not understood either, although the
LH> kernel seemed to understand this just fine as a boot parameter, and once
LH> again, I'm dumped into a very limited single-user mode.

It sounds like you do not have the required block special files in /dev.

With current udev, I use devtmpfs for /dev. It looks like having
devtmpfs support compiled into the kernel is enough.

If you don't have that, or if there is something in /etc/fstab competing
against /etc/init.d/udev-mount, that might explain the problem.

-JimC
--
James Cloos <cloos@jhcloos.com> OpenPGP: 1024D/ED7DAEA6
 
Old 03-23-2011, 04:29 PM
Lindsay Haisley
 
Default System problems

On Wed, 2011-03-23 at 08:39 -0400, James Cloos wrote:
> With current udev, I use devtmpfs for /dev. It looks like having
> devtmpfs support compiled into the kernel is enough.
>
> If you don't have that, or if there is something in /etc/fstab
> competing
> against /etc/init.d/udev-mount, that might explain the problem.

Thanks. Jorge Vicetto noted that having CONFIG_SYSFS_DEPRECATED_V2 set
in the kernel config is a major no-no with recent versions of udev, and
will screw up device creation. This kernel config setting was indeed
set, which may well have accounted for this particular incarnation of
the problem. Roman Zilka noted that the 2.6.29 kernel I was using when
this happens has known issues and isn't even in the Gentoo portage tree
anymore. So thanks to Roman and Jorge I have a 2.6.36 kernel built
(without the problem setting) which I'll try when I get a moment. Being
a day late and a dollar short at the moment I'm a bit short on round
tuits :-)

--
Lindsay Haisley | "The difference between a duck is because
FMP Computer Services | one leg is both the same"
512-259-1190 | - Anonymous
http://www.fmp.com |
 

Thread Tools




All times are GMT. The time now is 04:43 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org