RFC: Kexec Cmdline Support

RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

(sorry for crossposting this,but I think more of the interested parties read
oe-devel then angstrom-devel)

A while back, mwester made some patches to add the ability to modify the
cmdline when using kexec to load a kernel. I did some testing on poodle and
posted the results on the bug tracker
(http://bugs.openembedded.org/show_bug.cgi?id=2969)
I was wondering what the rest of the developers think of these. I would like
to see them incoprorated in one way or another, since this will allow us to
support booting a root-fs directly from the SD card. This is especially
important for flash-challenged devices (collie, poodle, sl-c700) but also
makes testing new images much simpler.

Thank you for your time,

Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Uli Luckas]

On Monday, 24. September 2007, Hans Henry von Tresckow wrote:
> (sorry for crossposting this,but I think more of the interested parties
> read oe-devel then angstrom-devel)
>
> A while back, mwester made some patches to add the ability to modify the
> cmdline when using kexec to load a kernel. I did some testing on poodle and
> posted the results on the bug tracker
> (http://bugs.openembedded.org/show_bug.cgi?id=2969)
> I was wondering what the rest of the developers think of these. 
Hi Henry,
A while back I posted a different approach [1] to this mailing list for 
review. To bad, there have not been any comments.

The differences of my approach in contrast to mwester's approach have been 
discussed on the arm kernel mailing list. If there is any interest in this 
(userspace) approach at all, I will improve on the known problems (like hard 
coded ATAG base address).

I'd really appreciate feedback from the developers.

Best regerds
Uli

1) 
http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/002840.html

-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit

Re: RFC: Kexec Cmdline Support

[Paul Sokolovsky]

Hello Uli,

Tuesday, September 25, 2007, 5:20:52 PM, you wrote:

> On Monday, 24. September 2007, Hans Henry von Tresckow wrote:
>> (sorry for crossposting this,but I think more of the interested parties
>> read oe-devel then angstrom-devel)
>>
>> A while back, mwester made some patches to add the ability to modify the
>> cmdline when using kexec to load a kernel. I did some testing on poodle and
>> posted the results on the bug tracker
>> (http://bugs.openembedded.org/show_bug.cgi?id=2969)
>> I was wondering what the rest of the developers think of these. 
> Hi Henry,
> A while back I posted a different approach [1] to this mailing list for
> review. To bad, there have not been any comments.

> The differences of my approach in contrast to mwester's approach have been
> discussed on the arm kernel mailing list. If there is any interest in this
> (userspace) approach at all, I will improve on the known problems (like hard
> coded ATAG base address).

> I'd really appreciate feedback from the developers.

  Full commandline, etc. support for kexec is much, much needed
feature. Unfortunately, most developers are likely busy with more
mundane stuff and give good review/testing to it. But those who're
interested in this area for sure should apply such patches to the
device kernel they can test, to facilitate wider testing, and the
authors of the patches for sure should post the for review to general
Linux list, and hopefully communicate to find the best solution.

> Best regerds
> Uli

> 1) 
> http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/002840.html



-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 9/25/07, Uli Luckas <u.luckas@road.de> wrote:
>
> On Monday, 24. September 2007, Hans Henry von Tresckow wrote:
> > (sorry for crossposting this,but I think more of the interested parties
> > read oe-devel then angstrom-devel)
> >
> > A while back, mwester made some patches to add the ability to modify the
> > cmdline when using kexec to load a kernel. I did some testing on poodle
> and
> > posted the results on the bug tracker
> > (http://bugs.openembedded.org/show_bug.cgi?id=2969)
> > I was wondering what the rest of the developers think of these.
> Hi Henry,
> A while back I posted a different approach [1] to this mailing list for
> review. To bad, there have not been any comments.
>
> The differences of my approach in contrast to mwester's approach have been
> discussed on the arm kernel mailing list. If there is any interest in this
> (userspace) approach at all, I will improve on the known problems (like
> hard
> coded ATAG base address).
>
> I'd really appreciate feedback from the developers.
>
> Best regerds
> Uli
>
> 1)
>
> http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/002840.html
>
> --
>
> ------- ROAD ...the handyPC Company - - -  ) ) )
>
> Uli Luckas
> Software Development
>
> ROAD GmbH
> Bennigsenstr. 14 | 12159 Berlin | Germany
> fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
> url: www.road.de
>
> Amtsgericht Charlottenburg: HRB 96688 B
> Managing directors: Hans-Peter Constien, Hubertus von Streit



Uli,
Do you have a link to the kernel list discussion? At this point I am more
interested in getting cmdline support than in any particular implementation.
I did try your patch on top of mike's but I had no luck. One of the issues I
have run into is the BROKEN Sharp bootloader that makes things more
interesting. Mike's patches can work around this problem by setting up a
fake ATAG list at a non-standard location.

I would love to see more progress on this, since I would like to improve
support for booting devices from non-standard locations (ie SD card) without
custom kernels.

Thank you for your help,

Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Uli Luckas]

On Tuesday, 25. September 2007, Hans Henry von Tresckow wrote:
> On 9/25/07, Uli Luckas <u.luckas@road.de> wrote:
> > On Monday, 24. September 2007, Hans Henry von Tresckow wrote:
> > > (sorry for crossposting this,but I think more of the interested parties
> > > read oe-devel then angstrom-devel)
> > >
> > > A while back, mwester made some patches to add the ability to modify
> > > the cmdline when using kexec to load a kernel. I did some testing on
> > > poodle
> >
> > and
> >
> > > posted the results on the bug tracker
> > > (http://bugs.openembedded.org/show_bug.cgi?id=2969)
> > > I was wondering what the rest of the developers think of these.
> >
> > Hi Henry,
> > A while back I posted a different approach [1] to this mailing list for
> > review. To bad, there have not been any comments.
> >
> > The differences of my approach in contrast to mwester's approach have
> > been discussed on the arm kernel mailing list. If there is any interest
> > in this (userspace) approach at all, I will improve on the known problems
> > (like hard
> > coded ATAG base address).
> >
> > I'd really appreciate feedback from the developers.
> >
> > Best regerds
> > Uli
> >
> > 1)
> >
> > http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/00284
> >0.html
> >
> > --
> >
> > ------- ROAD ...the handyPC Company - - -  ) ) )
> >
> > Uli Luckas
> > Software Development
> >
> > ROAD GmbH
> > Bennigsenstr. 14 | 12159 Berlin | Germany
> > fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
> > url: www.road.de
> >
> > Amtsgericht Charlottenburg: HRB 96688 B
> > Managing directors: Hans-Peter Constien, Hubertus von Streit
>
> Uli,
> Do you have a link to the kernel list discussion? 
>
kexec & 2 RAM banks [1] 

> At this point I am more 
> interested in getting cmdline support than in any particular
> implementation. I did try your patch on top of mike's but I had no luck.
>
My Patch is a pure user space implementation and is supposed to work with a 
vanilla kernel. 

> One of the issues I have run into is the BROKEN Sharp bootloader that makes
> things more interesting. Mike's patches can work around this problem by
> setting up a fake ATAG list at a non-standard location.
>
The Idea for both approaches is to take the running kenel's ATAG list as a 
template for the new kernel's ATAG list. Mike Westerhof has this already 
implemented. I kind of stalled on that one as no one showed any interest. If 
you feel like testing such an approach, I could put the missing pieces 
together for you.

> I would love to see more progress on this, since I would like to improve
> support for booting devices from non-standard locations (ie SD card)
> without custom kernels.
>
By the way, who is the original author of the arm-kexec-utils patch? I'd like 
to see kexec support pushed upstream which requires a sign-off of the 
original author.

> Thank you for your help,
>
> Henry von Tresckow (hvontres)

1) kexec & 2 RAM banks
http://marc.info/?l=linux-arm-kernel&m=118554158026945&w=4


-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 9/26/07, Uli Luckas <u.luckas@road.de> wrote:
>
> On Tuesday, 25. September 2007, Hans Henry von Tresckow wrote:
> > On 9/25/07, Uli Luckas <u.luckas@road.de> wrote:
> > > On Monday, 24. September 2007, Hans Henry von Tresckow wrote:
> > > > (sorry for crossposting this,but I think more of the interested
> parties
> > > > read oe-devel then angstrom-devel)
> > > >
> > > > A while back, mwester made some patches to add the ability to modify
> > > > the cmdline when using kexec to load a kernel. I did some testing on
> > > > poodle
> > >
> > > and
> > >
> > > > posted the results on the bug tracker
> > > > (http://bugs.openembedded.org/show_bug.cgi?id=2969)
> > > > I was wondering what the rest of the developers think of these.
> > >
> > > Hi Henry,
> > > A while back I posted a different approach [1] to this mailing list
> for
> > > review. To bad, there have not been any comments.
> > >
> > > The differences of my approach in contrast to mwester's approach have
> > > been discussed on the arm kernel mailing list. If there is any
> interest
> > > in this (userspace) approach at all, I will improve on the known
> problems
> > > (like hard
> > > coded ATAG base address).
> > >
> > > I'd really appreciate feedback from the developers.
> > >
> > > Best regerds
> > > Uli
> > >
> > > 1)
> > >
> > >
> http://lists.linuxtogo.org/pipermail/openembedded-devel/2007-August/00284
> > >0.html
> > >
> > > --
> > >
> > > ------- ROAD ...the handyPC Company - - -  ) ) )
> > >
> > > Uli Luckas
> > > Software Development
> > >
> > > ROAD GmbH
> > > Bennigsenstr. 14 | 12159 Berlin | Germany
> > > fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
> > > url: www.road.de
> > >
> > > Amtsgericht Charlottenburg: HRB 96688 B
> > > Managing directors: Hans-Peter Constien, Hubertus von Streit
> >
> > Uli,
> > Do you have a link to the kernel list discussion?
> >
> kexec & 2 RAM banks [1]
>
> > At this point I am more
> > interested in getting cmdline support than in any particular
> > implementation. I did try your patch on top of mike's but I had no luck.
> >
> My Patch is a pure user space implementation and is supposed to work with
> a
> vanilla kernel.
>
> > One of the issues I have run into is the BROKEN Sharp bootloader that
> makes
> > things more interesting. Mike's patches can work around this problem by
> > setting up a fake ATAG list at a non-standard location.
> >
> The Idea for both approaches is to take the running kenel's ATAG list as a
> template for the new kernel's ATAG list. Mike Westerhof has this already
> implemented. I kind of stalled on that one as no one showed any interest.
> If
> you feel like testing such an approach, I could put the missing pieces
> together for you.
>
> > I would love to see more progress on this, since I would like to improve
> > support for booting devices from non-standard locations (ie SD card)
> > without custom kernels.
> >
> By the way, who is the original author of the arm-kexec-utils patch? I'd
> like
> to see kexec support pushed upstream which requires a sign-off of the
> original author.
>
> > Thank you for your help,
> >
> > Henry von Tresckow (hvontres)
>
> 1) kexec & 2 RAM banks
> http://marc.info/?l=linux-arm-kernel&m=118554158026945&w=4
>
>
> --
>
> ------- ROAD ...the handyPC Company - - -  ) ) )
>
> Uli Luckas
> Software Development
>
> ROAD GmbH
> Bennigsenstr. 14 | 12159 Berlin | Germany
> fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
> url: www.road.de
>
> Amtsgericht Charlottenburg: HRB 96688 B
> Managing directors: Hans-Peter Constien, Hubertus von Streit



IIRC, Richard Purdie wrote the original kexec patch for arm. I would love to
try out your solution. Do you know which version of the vanilla kernel now
has the proper patches in place? I would most likely need to do some
tweaking, since the original sharp bootloader passes a cmdline that is
incompatible with 2.6 kernels. But I would love to see some support for
kexec cmdline editing in the near future.

Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Uli Luckas]

On Wednesday, 26. September 2007, you wrote:
> On 9/26/07, Uli Luckas <u.luckas@road.de> wrote:
> > By the way, who is the original author of the arm-kexec-utils patch? I'd
> > like
> > to see kexec support pushed upstream which requires a sign-off of the
> > original author.
> >
>
> IIRC, Richard Purdie wrote the original kexec patch for arm. I would love
> to try out your solution. Do you know which version of the vanilla kernel
> now has the proper patches in place?
I use kexec with 2.6.21.

> I would most likely need to do some tweaking, since the original sharp
> bootloader passes a cmdline that is incompatible with 2.6 kernels. 
For now you might have to modify the hard coded base address.

> But I would love to see some support for kexec cmdline editing in the near
> future. 
>
> Henry von Tresckow (hvontres)

Thanks,
Uli

-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 9/26/07, Uli Luckas <u.luckas@road.de> wrote:
>
> On Wednesday, 26. September 2007, you wrote:
> > On 9/26/07, Uli Luckas <u.luckas@road.de> wrote:
> > > By the way, who is the original author of the arm-kexec-utils patch?
> I'd
> > > like
> > > to see kexec support pushed upstream which requires a sign-off of the
> > > original author.
> > >
> >
> > IIRC, Richard Purdie wrote the original kexec patch for arm. I would
> love
> > to try out your solution. Do you know which version of the vanilla
> kernel
> > now has the proper patches in place?
> I use kexec with 2.6.21.
>
> > I would most likely need to do some tweaking, since the original sharp
> > bootloader passes a cmdline that is incompatible with 2.6 kernels.
> For now you might have to modify the hard coded base address.
>
> > But I would love to see some support for kexec cmdline editing in the
> near
> > future.
> >
> > Henry von Tresckow (hvontres)
>
> Thanks,
> Uli
>
> --
>
> ------- ROAD ...the handyPC Company - - -  ) ) )
>
> Uli Luckas
> Software Development
>
> ROAD GmbH
> Bennigsenstr. 14 | 12159 Berlin | Germany
> fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
> url: www.road.de
>
> Amtsgericht Charlottenburg: HRB 96688 B
> Managing directors: Hans-Peter Constien, Hubertus von Streit
>

Ok, I have tested your patch and it works very well. I think the only real
stumbling block left is figuring out a way to get the base address for the
ATAG list without hard-coding it. I think if we can somehow access the
machine_desc structure in the kernel from user space, this would be easy.
Unfortunately I am not very good at kernel hacking yet, so if you or someone
else on the list has an idea, please let me know.

-- 
Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 10/19/07, Uli Luckas <luckas@musoft.de> wrote:
>
> Am Tuesday, 9. October 2007 schrieb Hans Henry von Tresckow:
> > Ok, I have tested your patch and it works very well. I think the only
> real
> > stumbling block left is figuring out a way to get the base address for
> the
> > ATAG list without hard-coding it. I think if we can somehow access the
> > machine_desc structure in the kernel from user space, this would be
> easy.
> > Unfortunately I am not very good at kernel hacking yet, so if you or
> > someone else on the list has an idea, please let me know.
>
> Hi Henry,
> sorry for letting you wait so long. I am out of office until 10/29. After
> that
> I can provide a kernel patch to export the ATAG list's base address and
> also
> the old ATAG list. For now the patch will be against 2.6.21 though, as I
> don't have a running, more recent kernel available.
>
> Regards
> Uli
>
> --
> Früher oder später emulieren wir euch
> - Linux.de
>

Thanks. That sounds excellent :) 2.6.21 should be OK, since it looks like
that will be the version for Angstrom 2007 iirc. I might look at backporting
it for 2.6.17 as well, since there may be some devices (Tosa due to WiFi
issues and poodle due to the display getting corrupted on resume) that would
initially need to run 2.6.17.

After we beat up on that patch, are you planning on sending it upstream as
well?

-- 
Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hans Henry von Tresckow schreef:
> On 10/19/07, Uli Luckas <luckas@musoft.de> wrote:
>> Am Tuesday, 9. October 2007 schrieb Hans Henry von Tresckow:
>>> Ok, I have tested your patch and it works very well. I think the only
>> real
>>> stumbling block left is figuring out a way to get the base address for
>> the
>>> ATAG list without hard-coding it. I think if we can somehow access the
>>> machine_desc structure in the kernel from user space, this would be
>> easy.
>>> Unfortunately I am not very good at kernel hacking yet, so if you or
>>> someone else on the list has an idea, please let me know.
>> Hi Henry,
>> sorry for letting you wait so long. I am out of office until 10/29. After
>> that
>> I can provide a kernel patch to export the ATAG list's base address and
>> also
>> the old ATAG list. For now the patch will be against 2.6.21 though, as I
>> don't have a running, more recent kernel available.
>>
>> Regards
>> Uli
>>
>> --
>> Früher oder später emulieren wir euch
>> - Linux.de
>>
> 
> Thanks. That sounds excellent :) 2.6.21 should be OK, since it looks like
> that will be the version for Angstrom 2007 iirc.

Only for a few devices, e.g. the zaurus device are probably going to use
.23, the powerpc devices .22, etc

regards,

Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHGbJEMkyGM64RGpERAqwwAJ422t+AkOZC6HgFTbaF5QgVWC2ysACgsnlX
YADqHrw6e3HtqYGI+bS1u34=
=OjIL
-----END PGP SIGNATURE-----

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 10/19/07, Uli Luckas <luckas@musoft.de> wrote:
>
> Am Tuesday, 9. October 2007 schrieb Hans Henry von Tresckow:
> > Ok, I have tested your patch and it works very well. I think the only
> real
> > stumbling block left is figuring out a way to get the base address for
> the
> > ATAG list without hard-coding it. I think if we can somehow access the
> > machine_desc structure in the kernel from user space, this would be
> easy.
> > Unfortunately I am not very good at kernel hacking yet, so if you or
> > someone else on the list has an idea, please let me know.
>
> Hi Henry,
> sorry for letting you wait so long. I am out of office until 10/29. After
> that
> I can provide a kernel patch to export the ATAG list's base address and
> also
> the old ATAG list. For now the patch will be against 2.6.21 though, as I
> don't have a running, more recent kernel available.
>
> Regards
> Uli
>
> --
> Früher oder später emulieren wir euch
> - Linux.de
>

Uli,
Any luck getting that patch put together? I would really love to start
testing this soon :)

-- 
Henry von Tresckow (hvontres)

Re: RFC: Kexec Cmdline Support

[Uli Luckas]

On Friday, 2. November 2007, Hans Henry von Tresckow wrote:
> Uli,
> Any luck getting that patch put together? I would really love to start
> testing this soon :)
Henry,
My schedule is a lot tighter than I thought it would be. I'll try to find a 
time slot during this week. If that does not work out I'll take my spare time 
this weekend.

Uli

-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit

Re: RFC: Kexec Cmdline Support

[Hans Henry von Tresckow]

On 11/6/07, Uli Luckas <u.luckas@road.de> wrote:
>
> On Friday, 2. November 2007, Hans Henry von Tresckow wrote:
> > Uli,
> > Any luck getting that patch put together? I would really love to start
> > testing this soon :)
> Henry,
> My schedule is a lot tighter than I thought it would be. I'll try to find
> a
> time slot during this week. If that does not work out I'll take my spare
> time
> this weekend.
>
> Uli
>
> --
>
> No problem. I have a couple of other things to get done as well. So take
your time and enjoy your weekend...:)


-- 
Henry von Tresckow (hvontres)