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 > Debian > Debian Kernel

 
 
LinkBack Thread Tools
 
Old 12-22-2009, 11:57 PM
Frans Pop
 
Default Bits from the kernel team

Kurt Roeckx wrote:
> Now that we have a 2.6.32 kernel in unstable, can you updates us
> on the various things mentioned in this mail?

I have another question. The naming policy for linux-image packages seems
to have changed: instead of an ABI we now have "trunk". First I thought
this was a bug, but now that meta packages have been updated to use trunk
too that seems unlikely.

I've not seen any announcement for this, nor any discussion on how this may
affect other packages (such as packages building out of tree modules and
Debian Installer).

I've always considered the fact that a kernel update with a different ABI
did not replace the current kernel an important feature (reducing the need
for an immediate reboot).

Are we no longer interested in the ABI?
What will happen during/after upgrades if the ABI does change?

Would someone care to explain to the rest of the project?

Cheers,
FJP


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-23-2009, 01:56 AM
Ben Hutchings
 
Default Bits from the kernel team

On Wed, 2009-12-23 at 01:51 +0100, Frans Pop wrote:
> Ben Hutchings wrote:
> > Currently you can install kernel images from unstable or backports
> > without any extra dependencies. I'm not aware of any significant
> > breakage though some packages may rely on deprecated and removed stuff
> > in procfs or sysfs.
>
> I've been running upstream kernels without any problems on Lenny.
>
> The only issue I'm aware of is that the init script of nfs-kernel-server is
> not compatible with 2.6.32.
> The following change in /etc/init.d/nfs-kernel-server fixes this:
> # See if our running kernel supports the NFS kernel server
> - if [ -f /proc/kallsyms ] && ! grep -qE 'init_nf(sd| )' /proc/kallsyms; then
> + if ! [ -d /sys/module/nfsd ]; then

I made that change myself, how could I forget it?!

Ben.

--
Ben Hutchings
Experience is what causes a person to make new mistakes instead of old ones.
 
Old 12-23-2009, 02:06 AM
Ben Hutchings
 
Default Bits from the kernel team

On Wed, 2009-12-23 at 01:57 +0100, Frans Pop wrote:
> Kurt Roeckx wrote:
> > Now that we have a 2.6.32 kernel in unstable, can you updates us
> > on the various things mentioned in this mail?
>
> I have another question. The naming policy for linux-image packages seems
> to have changed: instead of an ABI we now have "trunk". First I thought
> this was a bug, but now that meta packages have been updated to use trunk
> too that seems unlikely.

The error was in uploading 2.6.32-rc8 to unstable rather than
experimental as was intended. Following that, we could only move
forward to 2.6.32 even though we have not stabilised its configurations
and the resulting ABIs.

> I've not seen any announcement for this, nor any discussion on how this may
> affect other packages (such as packages building out of tree modules and
> Debian Installer).
>
> I've always considered the fact that a kernel update with a different ABI
> did not replace the current kernel an important feature (reducing the need
> for an immediate reboot).
>
> Are we no longer interested in the ABI?
> What will happen during/after upgrades if the ABI does change?
>
> Would someone care to explain to the rest of the project?

Once we are finished with major changes to 2.6.32 (such as the libata
transition) the ABI version will change from 'trunk' to '1' and we will
then try to avoid unnecessary ABI changes.

Ben.

--
Ben Hutchings
Experience is what causes a person to make new mistakes instead of old ones.
 
Old 12-23-2009, 02:06 AM
Ben Hutchings
 
Default Bits from the kernel team

On Wed, 2009-12-23 at 01:57 +0100, Frans Pop wrote:
> Kurt Roeckx wrote:
> > Now that we have a 2.6.32 kernel in unstable, can you updates us
> > on the various things mentioned in this mail?
>
> I have another question. The naming policy for linux-image packages seems
> to have changed: instead of an ABI we now have "trunk". First I thought
> this was a bug, but now that meta packages have been updated to use trunk
> too that seems unlikely.

The error was in uploading 2.6.32-rc8 to unstable rather than
experimental as was intended. Following that, we could only move
forward to 2.6.32 even though we have not stabilised its configurations
and the resulting ABIs.

> I've not seen any announcement for this, nor any discussion on how this may
> affect other packages (such as packages building out of tree modules and
> Debian Installer).
>
> I've always considered the fact that a kernel update with a different ABI
> did not replace the current kernel an important feature (reducing the need
> for an immediate reboot).
>
> Are we no longer interested in the ABI?
> What will happen during/after upgrades if the ABI does change?
>
> Would someone care to explain to the rest of the project?

Once we are finished with major changes to 2.6.32 (such as the libata
transition) the ABI version will change from 'trunk' to '1' and we will
then try to avoid unnecessary ABI changes.

Ben.

--
Ben Hutchings
Experience is what causes a person to make new mistakes instead of old ones.
 
Old 12-23-2009, 07:12 AM
Michael Gernoth
 
Default Bits from the kernel team

On Wed, Dec 23, 2009 at 01:51:01AM +0100, Frans Pop wrote:
> The following change in /etc/init.d/nfs-kernel-server fixes this:
> # See if our running kernel supports the NFS kernel server
> - if [ -f /proc/kallsyms ] && ! grep -qE 'init_nf(sd| )' /proc/kallsyms; then
> + if ! [ -d /sys/module/nfsd ]; then

Only when nfsd is a module, not when it is compiled in:
$ grep NFSD /boot/config-2.6.32.2
CONFIG_NFSD=y
$ ls -dl /sys/module/nfs*
drwxr-xr-x 3 root root 0 Dec 23 09:05 /sys/module/nfs
$

Testing for /proc/fs/nfs/exports could probably work everywhere.

Regards,
Michael

--
Michael Gernoth Department of Computer Science IV
Martensstrasse 1 D-91058 Erlangen Germany University of Erlangen-Nuremberg
http://www4.informatik.uni-erlangen.de/~gernoth/


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-23-2009, 08:24 AM
Frans Pop
 
Default Bits from the kernel team

Michael Gernoth wrote:
> On Wed, Dec 23, 2009 at 01:51:01AM +0100, Frans Pop wrote:
>> The following change in /etc/init.d/nfs-kernel-server fixes this:
>> # See if our running kernel supports the NFS kernel server
>> - if [ -f /proc/kallsyms ] && ! grep -qE 'init_nf(sd| )'
>> /proc/kallsyms; then
>> + if ! [ -d /sys/module/nfsd ]; then
>
> Only when nfsd is a module, not when it is compiled in:
> $ grep NFSD /boot/config-2.6.32.2
> CONFIG_NFSD=y
> $ ls -dl /sys/module/nfs*
> drwxr-xr-x 3 root root 0 Dec 23 09:05 /sys/module/nfs
> $

Right. Looks like that's already being discussed in #550153.

> Testing for /proc/fs/nfs/exports could probably work everywhere.

That's mentioned in the BR as well.

Cheers,
FJP


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 
Old 12-23-2009, 08:29 AM
Michael Tokarev
 
Default Bits from the kernel team

Ben Hutchings wrote:
> On Wed, 2009-12-23 at 01:51 +0100, Frans Pop wrote:
>> Ben Hutchings wrote:
>>> Currently you can install kernel images from unstable or backports
>>> without any extra dependencies. I'm not aware of any significant
>>> breakage though some packages may rely on deprecated and removed stuff
>>> in procfs or sysfs.
>> I've been running upstream kernels without any problems on Lenny.
>>
>> The only issue I'm aware of is that the init script of nfs-kernel-server is
>> not compatible with 2.6.32.
>> The following change in /etc/init.d/nfs-kernel-server fixes this:
>> # See if our running kernel supports the NFS kernel server
>> - if [ -f /proc/kallsyms ] && ! grep -qE 'init_nf(sd| )' /proc/kallsyms; then
>> + if ! [ -d /sys/module/nfsd ]; then
>
> I made that change myself, how could I forget it?!

That apparently breaks non-modular nfsd -- see #561674.

/mjt


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
 

Thread Tools




All times are GMT. The time now is 06:50 AM.

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