2002-10-13 20:38:16

by Adam J. Richter

[permalink] [raw]
Subject: Re: Patch: linux-2.5.42/kernel/sys.c - warm reboot should not suspend device

Eric Blade wrote:
>On Sun, 2002-10-13 at 15:24, Adam J. Richter wrote:
>> [...] I think the new behavior in IDE
>> of spinning down the hard drives on suspend is correct. The problem
>> is that the warm reboot system call is trying to suspend all of the
>> devices before a warm reboot for no reason. [...]

>Adam,
> I'm not sure the proper thing to do is necessarily remove the
>device_shutdown() call.

If, by this, you are saying that you have in mind some reason
why this should not be done, then please explain.

> Please try this patch [...]

Your patch does not apply and I don't see how renaming
a constant in essentially every place that it is referenced would
change the behavior of the code in a way releveant to the problem
that I described.

I don't see a problem with device_shutdown spinning down the
IDE hard disks. What I have a problem with, and what my patch fixes,
is the relatively new behavior of the warm reboot system call calling
device_shutdown. Why was this added? The reboot notifier chain is
already called for devices that need some preparation before it is
safe to reboot or halt.

Adam J. Richter __ ______________ 575 Oroville Road
[email protected] \ / Milpitas, California 95035
+1 408 309-6081 | g g d r a s i l United States of America
"Free Software For The Rest Of Us."


2002-10-13 21:51:56

by Eric W. Biederman

[permalink] [raw]
Subject: Re: Patch: linux-2.5.42/kernel/sys.c - warm reboot should not suspend device

"Adam J. Richter" <[email protected]> writes:

> Eric Blade wrote:
> >On Sun, 2002-10-13 at 15:24, Adam J. Richter wrote:
> >> [...] I think the new behavior in IDE
> >> of spinning down the hard drives on suspend is correct. The problem
> >> is that the warm reboot system call is trying to suspend all of the
> >> devices before a warm reboot for no reason. [...]
>
> >Adam,
> > I'm not sure the proper thing to do is necessarily remove the
> >device_shutdown() call.
>
> If, by this, you are saying that you have in mind some reason
> why this should not be done, then please explain.

We need it. It doesn't make sense for every device driver
to register a reboot notifier. When especially as they
have to run the same code when they are modular and are
removed.

Why would you not want to do that?

> > Please try this patch [...]
>
> Your patch does not apply and I don't see how renaming
> a constant in essentially every place that it is referenced would
> change the behavior of the code in a way releveant to the problem
> that I described.
>
> I don't see a problem with device_shutdown spinning down the
> IDE hard disks. What I have a problem with, and what my patch fixes,
> is the relatively new behavior of the warm reboot system call calling
> device_shutdown. Why was this added?

Because most device drivers don't implement a reboot notifier?
And they almost certainly need it.

> The reboot notifier chain is
> already called for devices that need some preparation before it is
> safe to reboot or halt.

Error please try again. It is just they don't cause major problems
on reboot so no one notices the problems.

Eric