How does one get paid to work on the kernel?

How does one get paid to work on the kernel?

[Nigel Cunningham]

Hi all.

I'm about to finish my present employment, and have wondered about the
possibility of getting paid to finish off the 2.4 version of Software
Suspend and do the port to 2.6. How does one go about seeing if anyone
might be interested in funding such a project for a few months?

Regards,

Nigel
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: How does one get paid to work on the kernel?

[Måns Rullgård]

Nigel Cunningham <ncunningham@clear.net.nz> writes:

> I'm about to finish my present employment, and have wondered about the
> possibility of getting paid to finish off the 2.4 version of Software
> Suspend and do the port to 2.6. How does one go about seeing if anyone
> might be interested in funding such a project for a few months?

Hasn't that already been done?

-- 
Måns Rullgård
mru@users.sf.net

Re: How does one get paid to work on the kernel?

[Nigel Cunningham]

There is support in the current kernel for Software Suspend, but the 2.4
version contains a lot of extra functionality that isn't present in 2.6
at the moment. (Support for HighMem, swap files, asynchronous I/O, a
nicer user interface, compression...).

On Fri, 2003-09-19 at 19:51, Måns Rullgård wrote:
> Nigel Cunningham <ncunningham@clear.net.nz> writes:
> 
> > I'm about to finish my present employment, and have wondered about the
> > possibility of getting paid to finish off the 2.4 version of Software
> > Suspend and do the port to 2.6. How does one go about seeing if anyone
> > might be interested in funding such a project for a few months?
> 
> Hasn't that already been done?
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: How does one get paid to work on the kernel?

[Måns Rullgård]

Nigel Cunningham <ncunningham@clear.net.nz> writes:

> There is support in the current kernel for Software Suspend, but the 2.4
> version contains a lot of extra functionality that isn't present in 2.6
> at the moment. (Support for HighMem, swap files, asynchronous I/O, a
> nicer user interface, compression...).

I see.  BTW, is it possible to boot normally, and later resume from
the saved state, provided you don't touch any filesystems or swap
areas involved in the suspend?  I seem to recall reading somewhere
that it would be possible, but I can't find any information on how to
do it.

-- 
Måns Rullgård
mru@users.sf.net

Re: How does one get paid to work on the kernel?

[Nigel Cunningham]

Yes, provided as you say that you don't mount the file systems involved;
mounting them will make journalling filesystems run their recoveries,
which will in turn make the suspend image inconsistent. It's only really
viable if the filesystems were mounted read only to start with... I've
just added functionality to the 2.4 version for such a case.

Regards,

Nigel

On Fri, 2003-09-19 at 20:10, Måns Rullgård wrote:
> Nigel Cunningham <ncunningham@clear.net.nz> writes:
> 
> > There is support in the current kernel for Software Suspend, but the 2.4
> > version contains a lot of extra functionality that isn't present in 2.6
> > at the moment. (Support for HighMem, swap files, asynchronous I/O, a
> > nicer user interface, compression...).
> 
> I see.  BTW, is it possible to boot normally, and later resume from
> the saved state, provided you don't touch any filesystems or swap
> areas involved in the suspend?  I seem to recall reading somewhere
> that it would be possible, but I can't find any information on how to
> do it.
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Resuming from software suspend [was: Re: How does one get paid to work on the kernel?]

[Måns Rullgård]

Nigel Cunningham <ncunningham@clear.net.nz> writes:

> Yes, provided as you say that you don't mount the file systems involved;
> mounting them will make journalling filesystems run their recoveries,
> which will in turn make the suspend image inconsistent. It's only really
> viable if the filesystems were mounted read only to start with... I've
> just added functionality to the 2.4 version for such a case.

OK, is it possible with 2.6?  The Kconfig help says you can.  How is
it done?

-- 
Måns Rullgård
mru@users.sf.net

Re: Resuming from software suspend [was: Re: How does one get paid to work on the kernel?]

[Nigel Cunningham]

If your filesystems were mounted readonly and the boot won't mount them
writable, you should be fine with no special precautions. Last time I
looked at 2.6 code, it didn't fix the suspend header when you use
noresume. If that's still true, you should be able to boot with the
noresume option, and then later normally.

(The 2.4 functionality I spoke of works differently, partly because it
does fix the suspend header if you use the noresume option.)

Regards,

Nigel

On Fri, 2003-09-19 at 21:45, Måns Rullgård wrote:
> Nigel Cunningham <ncunningham@clear.net.nz> writes:
> 
> > Yes, provided as you say that you don't mount the file systems involved;
> > mounting them will make journalling filesystems run their recoveries,
> > which will in turn make the suspend image inconsistent. It's only really
> > viable if the filesystems were mounted read only to start with... I've
> > just added functionality to the 2.4 version for such a case.
> 
> OK, is it possible with 2.6?  The Kconfig help says you can.  How is
> it done?
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: Resuming from software suspend

[Måns Rullgård]

Nigel Cunningham <ncunningham@clear.net.nz> writes:

> If your filesystems were mounted readonly and the boot won't mount them
> writable, you should be fine with no special precautions. Last time I
> looked at 2.6 code, it didn't fix the suspend header when you use
> noresume. If that's still true, you should be able to boot with the
> noresume option, and then later normally.

What I want to do is boot, do some things, and then resume the
suspended state without rebooting between.  Is that possible?  I don't
see any reason why it should be impossible to do, even if it's not
currently supported.

-- 
Måns Rullgård
mru@users.sf.net

Re: Resuming from software suspend

[Nigel Cunningham]

Provided you're not making the data of the filesystem inconsistent with
the state that the suspended image is expecting to see, you won't get
any corruption. As to beginning a resume without rebooting, whether it
would work would depend upon the size of the image, the amount of memory
used when you start the resume and the degree of overlap between the two
sets of memory.

Regards,

Nigel

On Fri, 2003-09-19 at 22:15, Måns Rullgård wrote:
> Nigel Cunningham <ncunningham@clear.net.nz> writes:
> 
> > If your filesystems were mounted readonly and the boot won't mount them
> > writable, you should be fine with no special precautions. Last time I
> > looked at 2.6 code, it didn't fix the suspend header when you use
> > noresume. If that's still true, you should be able to boot with the
> > noresume option, and then later normally.
> 
> What I want to do is boot, do some things, and then resume the
> suspended state without rebooting between.  Is that possible?  I don't
> see any reason why it should be impossible to do, even if it's not
> currently supported.
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: Resuming from software suspend

[Måns Rullgård]

Nigel Cunningham <ncunningham@clear.net.nz> writes:

> Provided you're not making the data of the filesystem inconsistent with
> the state that the suspended image is expecting to see, you won't get
> any corruption. As to beginning a resume without rebooting, whether it
> would work would depend upon the size of the image, the amount of memory
> used when you start the resume and the degree of overlap between the two
> sets of memory.

If all processes are killed and all filesystem unmounted, there's not
much left to use memory.  Once you've decided to resume, there's not
much point in keeping old processes running, is there?

-- 
Måns Rullgård
mru@users.sf.net

Re: Resuming from software suspend

[Pavel Machek]

Hi!

> What I want to do is boot, do some things, and then resume the
> suspended state without rebooting between.  Is that possible?  I don't
> see any reason why it should be impossible to do, even if it's not
> currently supported.

Its not impossible, its just pretty tricky. You'd have to kill all
userland and bring devices back to sane state.
(It is also going to be very tricky to *test*.)
				Pavel
-- 
				Pavel
Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...

Re: Resuming from software suspend

[Bas Mevissen]

Måns Rullgård wrote:

> What I want to do is boot, do some things, and then resume the
> suspended state without rebooting between.  Is that possible?  I don't
> see any reason why it should be impossible to do, even if it's not
> currently supported.
> 

Just after booting, you know the state of the hardware (just initialised 
for most things that are not used to /boot/start resume/ from). You need 
to get the hardware in a sort of just-booted state before revering to 
the swsusp image you saved earlier because the drivers (might) expect 
the hardware to be in a certain state when they are started up by swsusp.

So you have to be carefull about the hardware (and not only the 
filesystem) state just before reverting to the swsusp image.

Bas.

Re: How does one get paid to work on the kernel?

[Bas Mevissen]

Måns Rullgård wrote:

> 
> I see.  BTW, is it possible to boot normally, and later resume from
> the saved state, provided you don't touch any filesystems or swap
> areas involved in the suspend?  I seem to recall reading somewhere
> that it would be possible, but I can't find any information on how to
> do it.
> 

Just wondering: what kind of use do you see for that?

ctually, I'm more thinking of a sort of freezing the state of processes 
rather then the kernel state. It would be nice to generalise this to be 
able to quick-(re)start applications (as long as their config file 
aren't changed).

Bas.

Re: How does one get paid to work on the kernel?

[Måns Rullgård]

Bas Mevissen <ml@basmevissen.nl> writes:

>> I see.  BTW, is it possible to boot normally, and later resume from
>> the saved state, provided you don't touch any filesystems or swap
>> areas involved in the suspend?  I seem to recall reading somewhere
>> that it would be possible, but I can't find any information on how to
>> do it.
>>
>
> Just wondering: what kind of use do you see for that?

For instance, to load a module required to access the suspended image.

-- 
Måns Rullgård
mru@users.sf.net

Re: How does one get paid to work on the kernel?

[Bas Mevissen]

Måns Rullgård wrote:

> 
>>Just wondering: what kind of use do you see for that?
> 
> For instance, to load a module required to access the suspended image.
> 

Why not loading it in the initrd image? Actually, the swsusp itself 
could (partly) be made a module if you load it in initrd :-)

Bas.

Re: How does one get paid to work on the kernel?

[Måns Rullgård]

Bas Mevissen <ml@basmevissen.nl> writes:

>>>Just wondering: what kind of use do you see for that?
>> For instance, to load a module required to access the suspended
>> image.
>>
>
> Why not loading it in the initrd image? Actually, the swsusp itself
> could (partly) be made a module if you load it in initrd :-)

You might want to do more than will fit in an initrd.

-- 
Måns Rullgård
mru@users.sf.net

Re: How does one get paid to work on the kernel?

[Jan Rychter]

[-- Attachment #1: Type: text/plain, Size: 491 bytes --]

>>>>> "Nigel" == Nigel Cunningham <ncunningham@clear.net.nz> writes:
 Nigel> There is support in the current kernel for Software Suspend, but
 Nigel> the 2.4 version contains a lot of extra functionality that isn't
 Nigel> present in 2.6 at the moment. (Support for HighMem, swap files,
 Nigel> asynchronous I/O, a nicer user interface, compression...).

Nigel is being modest and doesn't mention that the 2.4 version actually
works, which is possibly its biggest advantage.

--J.

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: How does one get paid to work on the kernel?

[Pavel Machek]

Hi!

>  Nigel> There is support in the current kernel for Software Suspend, but
>  Nigel> the 2.4 version contains a lot of extra functionality that isn't
>  Nigel> present in 2.6 at the moment. (Support for HighMem, swap files,
>  Nigel> asynchronous I/O, a nicer user interface, compression...).
> 
> Nigel is being modest and doesn't mention that the 2.4 version actually
> works, which is possibly its biggest advantage.

2.6.0-test3 swsusp should work, too, unless you have driver problem.
With ext2, ide and vesafb you should be able to suspend/resume
correctly.
That might not be practical for you, but should be
good enough for fixing drivers.
				Pavel
-- 
				Pavel
Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...