[gentoo-user] What's the story with 2.6.35 kernel?

[gentoo-user] What's the story with 2.6.35 kernel?

[Ajai Khattri]

I upgraded several machines and some failed to boot 2.6.35, they just hang 
after grub starts loading the kernel. Some of these I managed to fix by 
comparing kernel configs with working machines, others dont work at all.

The worst case is one where Ive upgraded udev to the latest which only 
works with kernels 2.6.25 or higher, and the last working kernel on that 
machine is 2.6.24 (grr!). Wondering if I can download a binary package of 
udev-149 from somewhere (or can I build it on another machine in a sandbox 
and package it with quickpkg?). Im kind of under pressure to fix this box 
but dont have udev-149 on any other machine available :-(

That'll teach me to upgrade to the latest and greatest too fast...



-- 
A

[gentoo-user] Re: What's the story with 2.6.35 kernel?

[Nikos Chantziaras]

On 09/11/2010 01:18 AM, Ajai Khattri wrote:
>
> I upgraded several machines and some failed to boot 2.6.35, they just
> hang after grub starts loading the kernel.

If you updated from 2.6.34 or even 33, then there shouldn't be a 
problem; "make oldconfig" should have worked OK and presented 
new/changed config options in a sane way.

However, if you updated from a very old kernel, then essentially you 
should have configured the new kernel from scratch, with all the work 
this entails.

So even though this advise isn't gonna help in your current situation, 
doing updates ASAP so that they're smaller is much better than waiting 
for too long and doing huge ones; this often breaks too much stuff.

In your situation, configuring the kernel from scratch might be the best 
option.

[gentoo-user] Re: What's the story with 2.6.35 kernel?

[walt]

On 09/10/2010 03:18 PM, Ajai Khattri wrote:
>
> I upgraded several machines and some failed to boot 2.6.35, they just hang after grub starts loading the kernel. Some of these I managed to fix by comparing kernel configs with working machines...

What kernel config items did you change to fix those broken machines?

Re: [gentoo-user] Re: What's the story with 2.6.35 kernel?

[Ajai Khattri]

On Fri, 10 Sep 2010, walt wrote:

> On 09/10/2010 03:18 PM, Ajai Khattri wrote:
>> 
>> I upgraded several machines and some failed to boot 2.6.35, they just hang 
>> after grub starts loading the kernel. Some of these I managed to fix by 
>> comparing kernel configs with working machines...
>
> What kernel config items did you change to fix those broken machines?

The two main ones seemed to be disabling the deprecated ATA drivers and 
another one was disabling some sysfs options. But these changes didn't 
help in this case. At this point, I think Im better off sticking to the 
older kernel until I figure this out properly so Ive focussed my efforts 
on getting that working.

I have since booted off a recent boot CD, mounted the RAID partitions and 
tried downgrading udev so I can maybe boot the older kernel. This didn't 
work since the LiveCD is newer than the original CD I used to build this 
box, and the naming conventions for RAID devices seems to have changed 
(and building udev from a LiveCD means Im using newer /proc and /dev!).

The older kernel loads but then says my root device doens't exist.

Under the old scheme I specified /dev/md1 as my root in grub.conf but 
since this doesn't exist the boot fails. In trying to fix this I see that 
the LiveCD uses the following device files for the RAID partitions: 
/dev/md123 /dev/md124 /dev/md125 /dev/md126. Should I be using those in 
grub.conf now?



-- 
A

Re: [gentoo-user] Re: What's the story with 2.6.35 kernel?

[Ajai Khattri]

On Sat, 11 Sep 2010, Nikos Chantziaras wrote:

> On 09/11/2010 01:18 AM, Ajai Khattri wrote:
>> 
>> I upgraded several machines and some failed to boot 2.6.35, they just
>> hang after grub starts loading the kernel.
>
> If you updated from 2.6.34 or even 33, then there shouldn't be a problem; 
> "make oldconfig" should have worked OK and presented new/changed config 
> options in a sane way.

On this particular box I was updating from 2.6.24 to 2.6.35 and I always 
use 'make silentoldconfig' when upgrading.


-- 
A

Re: [gentoo-user] Re: What's the story with 2.6.35 kernel?

[Ajai Khattri]

On Fri, 10 Sep 2010, Ajai Khattri wrote:

> Under the old scheme I specified /dev/md1 as my root in grub.conf but since 
> this doesn't exist the boot fails. In trying to fix this I see that the 
> LiveCD uses the following device files for the RAID partitions: /dev/md123 
> /dev/md124 /dev/md125 /dev/md126. Should I be using those in grub.conf now?

Long story short: I used the LiveCD device names in grub.conf and my fstab 
and I can now successfully boot the older 2.6.24 kernel. Everything seems 
to be working. I will have to read up on the naming scheme changes - 
advice/tips? (/dev/md1 is easier to remember than /dev/md124 :-)



-- 
A

Re: [gentoo-user] Re: What's the story with 2.6.35 kernel?

[John Campbell]

On 09/10/2010 05:30 PM, Ajai Khattri wrote:
> On Fri, 10 Sep 2010, walt wrote:

> The two main ones seemed to be disabling the deprecated ATA drivers and
> another one was disabling some sysfs options. But these changes didn't
> help in this case. At this point, I think Im better off sticking to the
> older kernel until I figure this out properly so Ive focussed my efforts
> on getting that working.

Don't forget your old ata drives are now named /dev/sd[a-z] so rewrite
your grub entries.

Do the systems have a mix of pata and sata drives?  I had this problem
when I updated from ata->pata.  My /dev/hda drive was being mounted as
/dev/sda and thus moving all the sata /dev/sd[a-z] drives up one slot.
Which caused partitions to be mounted, or fail from the wrong drive.

The other problem I had was that my sd[a-z] drives are numbered
depending on their order in the grub device.map file, which had been
shuffled around somehow.

I ended up going into grub menu edit and rewriting the kernel line to
reflect the new naming scheme and then once I'd booted into single user
mode, with no other partitions mounted, editing fstab to reflect the new
drive scheme.

[gentoo-user] Re: What's the story with 2.6.35 kernel?

[Nikos Chantziaras]

On 09/11/2010 03:31 AM, Ajai Khattri wrote:
> On Sat, 11 Sep 2010, Nikos Chantziaras wrote:
>
>> On 09/11/2010 01:18 AM, Ajai Khattri wrote:
>>>
>>> I upgraded several machines and some failed to boot 2.6.35, they just
>>> hang after grub starts loading the kernel.
>>
>> If you updated from 2.6.34 or even 33, then there shouldn't be a
>> problem; "make oldconfig" should have worked OK and presented
>> new/changed config options in a sane way.
>
> On this particular box I was updating from 2.6.24 to 2.6.35 and I always
> use 'make silentoldconfig' when upgrading.

Extremely bad idea.  It's no wonder something did break.  You can't just 
"silentoldconfig" between kernel versions and expect it to always work. 
  At least "oldconfig" would catch new options (which might be options 
replacing old ones.)

Re: [gentoo-user] What's the story with 2.6.35 kernel?

[Graham Murray]

Ajai Khattri <ajai@bway.net> writes:

> I upgraded several machines and some failed to boot 2.6.35, they just
> hang after grub starts loading the kernel. Some of these I managed to
> fix by comparing kernel configs with working machines, others dont
> work at all.
>
> The worst case is one where Ive upgraded udev to the latest which only
> works with kernels 2.6.25 or higher, and the last working kernel on
> that machine is 2.6.24 (grr!). Wondering if I can download a binary
> package of udev-149 from somewhere (or can I build it on another
> machine in a sandbox and package it with quickpkg?). Im kind of under
> pressure to fix this box but dont have udev-149 on any other machine
> available :-(
>
> That'll teach me to upgrade to the latest and greatest too fast...

I think that you may be suffering from bug
https://bugs.gentoo.org/show_bug.cgi?id=334269

Re: [gentoo-user] Re: What's the story with 2.6.35 kernel?

[Ajai Khattri]

On Sat, 11 Sep 2010, Nikos Chantziaras wrote:

> Extremely bad idea.  It's no wonder something did break.  You can't just 
> "silentoldconfig" between kernel versions and expect it to always work.  At 
> least "oldconfig" would catch new options (which might be options replacing 
> old ones.)

silentoldconfig shows all the new options too. I think the main problem 
was the changes in udev.


-- 
A