[makedumpfile] config file -- OSRELEASE

[makedumpfile] config file -- OSRELEASE

[Bernhard Walle]

Hi,

is there any reason why the kernel version saved in the configuration
file is the version of the running kernel and not the version for
which the config file is for?

It's a big problem because my plan was to ship the configuration files
with the kernel RPMs in SLES. However, we don't build the kernels on
systems which run the kernel that is just built -- so that's not
possible.

I think from implementation point of view it's not a problem, that's
why I ask and didn't create a patch.


Thanks,
   Bernhard


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Jay Lan]

Bernhard Walle wrote:
> Hi,
> 
> is there any reason why the kernel version saved in the configuration
> file is the version of the running kernel and not the version for
> which the config file is for?

Yep, i figured that was my problem.
I copied my vmlinux, which is a 2.6.18-8.el5, to my disk server,
which is a 2.6.16.43-0.5-default.

I would think the OSRELEASE of the CONFIGFILE generated from
'makedumpfile -g CONFIGFILE -x VMLINUX' would be the version
of VMLINUX. Instead, it said: OSRELEASE=2.6.16.43-0.5-default,
which caused 'makedumpfile -i' to complain about:
   CONFIGFILE and VMCORE don't match

I think it is a bug? You would want to ensure the version of
the VMLINUX and VMCORE match.

> 
> It's a big problem because my plan was to ship the configuration files
> with the kernel RPMs in SLES. 

I love this idea!

Thanks,
 - jay


> However, we don't build the kernels on
> systems which run the kernel that is just built -- so that's not
> possible.
> 
> I think from implementation point of view it's not a problem, that's
> why I ask and didn't create a patch.
> 
> 
> Thanks,
>    Bernhard
> 


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Ken'ichi Ohmichi]

Hi Bernhard,

2007/05/29 21:30:58 +0200, Bernhard Walle <bwalle@suse.de> wrote:
>is there any reason why the kernel version saved in the configuration
>file is the version of the running kernel and not the version for
>which the config file is for?

The reason is that makedumpfile gets a running kernel's page_size as
a 1st-kernel's page_size. If they are different, makedumpfile cannot
analyze /proc/vmcore and it fails. So I think a configuration file
should be created while 1st-kernel is running.


>It's a big problem because my plan was to ship the configuration files
>with the kernel RPMs in SLES. However, we don't build the kernels on
>systems which run the kernel that is just built -- so that's not
>possible.

It is a good idea to ship the configuration files with the kernel RPMs. 
I think you can do it with current implementation, because you need
only one configuration file for one kernel image.

I think you can ship configuration files as follows:

1. Build a kernel file and a debuginfo file.
2. Boot the system with the above kernel file.
3. Generate a configration file from a debuginfo file.
4. Ship the configration file with the kernel RPM.

Any problem to follow the method above ?


Thanks
Ken'ichi Ohmichi

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Bernhard Walle]

Hello,

* Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-05-30 13:13]:
> 2007/05/29 21:30:58 +0200, Bernhard Walle <bwalle@suse.de> wrote:
> >is there any reason why the kernel version saved in the configuration
> >file is the version of the running kernel and not the version for
> >which the config file is for?
> 
> The reason is that makedumpfile gets a running kernel's page_size as
> a 1st-kernel's page_size. If they are different, makedumpfile cannot
> analyze /proc/vmcore and it fails. So I think a configuration file
> should be created while 1st-kernel is running.

I also didn't find a possibility to get the page size from a kernel
image without running the executable.

> >It's a big problem because my plan was to ship the configuration files
> >with the kernel RPMs in SLES. However, we don't build the kernels on
> >systems which run the kernel that is just built -- so that's not
> >possible.
> 
> It is a good idea to ship the configuration files with the kernel RPMs. 
> I think you can do it with current implementation, because you need
> only one configuration file for one kernel image.
> 
> I think you can ship configuration files as follows:
> 
> 1. Build a kernel file and a debuginfo file.
> 2. Boot the system with the above kernel file.
> 3. Generate a configration file from a debuginfo file.
> 4. Ship the configration file with the kernel RPM.
> 
> Any problem to follow the method above ?

No, that's not possible because it would change our whole build
process. However, I just replace the OSRELEASE now after generating
the config file with the kernel that's built and change the page size
according to the .config on IA64. That works for now. I know it's not
the best solution, but the only practicable for now.


Thanks,
   Bernhard
-- 
SUSE LINUX Products GmbH          Tel. +49 (911) 74053-0
Maxfeldstr. 5                     GF: Markus Rex
90409 Nürnberg, Germany           HRB 16746 (AG Nürnberg)
OpenPGP DDAF6454: F61F 34CC 09CA FB82 C9F6  BA4B 8865 3696 DDAF 6454

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Eric W. Biederman]

Bernhard Walle <bwalle@suse.de> writes:

> Hello,
>
> * Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-05-30 13:13]:
>> 2007/05/29 21:30:58 +0200, Bernhard Walle <bwalle@suse.de> wrote:
>> >is there any reason why the kernel version saved in the configuration
>> >file is the version of the running kernel and not the version for
>> >which the config file is for?
>> 
>> The reason is that makedumpfile gets a running kernel's page_size as
>> a 1st-kernel's page_size. If they are different, makedumpfile cannot
>> analyze /proc/vmcore and it fails. So I think a configuration file
>> should be created while 1st-kernel is running.
>
> I also didn't find a possibility to get the page size from a kernel
> image without running the executable.
>
>> >It's a big problem because my plan was to ship the configuration files
>> >with the kernel RPMs in SLES. However, we don't build the kernels on
>> >systems which run the kernel that is just built -- so that's not
>> >possible.
>> 
>> It is a good idea to ship the configuration files with the kernel RPMs. 
>> I think you can do it with current implementation, because you need
>> only one configuration file for one kernel image.
>> 
>> I think you can ship configuration files as follows:
>> 
>> 1. Build a kernel file and a debuginfo file.
>> 2. Boot the system with the above kernel file.
>> 3. Generate a configration file from a debuginfo file.
>> 4. Ship the configration file with the kernel RPM.
>> 
>> Any problem to follow the method above ?
>
> No, that's not possible because it would change our whole build
> process. However, I just replace the OSRELEASE now after generating
> the config file with the kernel that's built and change the page size
> according to the .config on IA64. That works for now. I know it's not
> the best solution, but the only practicable for now.

Do we need more information in vmlinux to make it more amenable to
automatic processing?  version and page size for example.


Eric

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Ken'ichi Ohmichi]

Hi Eric,

2007/05/31 13:14:20 -0600, ebiederm@xmission.com (Eric W. Biederman) wrote:
>> * Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-05-30 13:13]:
>>> 2007/05/29 21:30:58 +0200, Bernhard Walle <bwalle@suse.de> wrote:
>>> >is there any reason why the kernel version saved in the configuration
>>> >file is the version of the running kernel and not the version for
>>> >which the config file is for?
>>> 
>>> The reason is that makedumpfile gets a running kernel's page_size as
>>> a 1st-kernel's page_size. If they are different, makedumpfile cannot
>>> analyze /proc/vmcore and it fails. So I think a configuration file
>>> should be created while 1st-kernel is running.
>>
>> I also didn't find a possibility to get the page size from a kernel
>> image without running the executable.
>>
>>> >It's a big problem because my plan was to ship the configuration files
>>> >with the kernel RPMs in SLES. However, we don't build the kernels on
>>> >systems which run the kernel that is just built -- so that's not
>>> >possible.
>>> 
>>> It is a good idea to ship the configuration files with the kernel RPMs. 
>>> I think you can do it with current implementation, because you need
>>> only one configuration file for one kernel image.
>>> 
>>> I think you can ship configuration files as follows:
>>> 
>>> 1. Build a kernel file and a debuginfo file.
>>> 2. Boot the system with the above kernel file.
>>> 3. Generate a configration file from a debuginfo file.
>>> 4. Ship the configration file with the kernel RPM.
>>> 
>>> Any problem to follow the method above ?
>>
>> No, that's not possible because it would change our whole build
>> process. However, I just replace the OSRELEASE now after generating
>> the config file with the kernel that's built and change the page size
>> according to the .config on IA64. That works for now. I know it's not
>> the best solution, but the only practicable for now.
>
>Do we need more information in vmlinux to make it more amenable to
>automatic processing?  version and page size for example.

Now, the items makedumpfile cannot get from vmlinux statically are
only version and page size. Do you have any idea for getting them
from vmlinux ?


Thanks
Ken'ichi Ohmichi

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Bernhard Walle]

* Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-06-01 11:07]:
> 
> Now, the items makedumpfile cannot get from vmlinux statically are
> only version and page size. Do you have any idea for getting them
> from vmlinux ?

I already tried to address a similar issue (version + command line
length for kexec-tools) by adding a ELF note. But it seems that kernel
hackers didn't like it.

http://article.gmane.org/gmane.comp.boot-loaders.fastboot.general/3996/.



Thanks,
   Bernhard

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Eric W. Biederman]

Bernhard Walle <bwalle@suse.de> writes:

> * Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-06-01 11:07]:
>> 
>> Now, the items makedumpfile cannot get from vmlinux statically are
>> only version and page size. Do you have any idea for getting them
>> from vmlinux ?
>
> I already tried to address a similar issue (version + command line
> length for kexec-tools) by adding a ELF note. But it seems that kernel
> hackers didn't like it.
>
> http://article.gmane.org/gmane.comp.boot-loaders.fastboot.general/3996/.

I'm not seeing the discussion there.

Looking at the patch I'm not quite certain of the composition of your
notes. 

However as a general direction of using ELF notes I'm more or less
in favor of it.

Eric

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [makedumpfile] config file -- OSRELEASE

[Bernhard Walle]

* Eric W. Biederman <ebiederm@xmission.com> [2007-06-01 15:40]:
> Bernhard Walle <bwalle@suse.de> writes:
> 
> > * Ken'ichi Ohmichi <oomichi@mxs.nes.nec.co.jp> [2007-06-01 11:07]:
> >> 
> >> Now, the items makedumpfile cannot get from vmlinux statically are
> >> only version and page size. Do you have any idea for getting them
> >> from vmlinux ?
> >
> > I already tried to address a similar issue (version + command line
> > length for kexec-tools) by adding a ELF note. But it seems that kernel
> > hackers didn't like it.
> >
> > http://article.gmane.org/gmane.comp.boot-loaders.fastboot.general/3996/.
> 
> I'm not seeing the discussion there.

Because there was no real discussion about ELF notes here. I just
created this first patch as attempt to add ELF notes. The idea was
from Vivek Goyal.

> Looking at the patch I'm not quite certain of the composition of your
> notes. 

Well, it was a first attempt. I'm open to any suggestions about that
topic.


Thanks,
   Bernhard

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec