Re: [PATCH, 2nd try] make disable_console_suspend runtime configurable

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Thursday, 14 June 2007 00:23, Stefan Seyfried wrote:
> On Thu, Jun 14, 2007 at 12:08:00AM +0200, Pavel Machek wrote:
> > Hi!
> > 
> > > I hate having to recompile the kernel, just to be able to debug suspend.
> > > Remove CONFIG_DISABLE_CONSOLE_SUSPEND, replace it by a tunable in
> > > /sys/power/disable_console_suspend.
> > 
> > > Signed-off-by: Stefan Seyfried <seife@suse.de>
> > > Signed-off-by: Frank Seidel <fseidel@suse.de>
> > 
> > I wonder if there's a better name?
> 
> Suggest one.
> 
> > Or maybe this should not be /sys configurable, but just have value for
> > each console "this console can work while suspended"?
> > 
> > (serial can, vesafb can, netconsole can't)?
> 
> Go ahead, submit a patch. It won't be that trivial. And i wonder
> if it is actually worth the hassle. This is a debugging facility.
> 
> > Exporting "crash-me" option to user does not seem that cool to me.
> 
> We have "echo c > /proc/sysrq-trigger" also.
> This is a debugging option, and forcing users to recompile the kernel just
> to debug suspend problems (not resume problems, the "it does not even go to
> sleep" stuff is where this matters most) is IMO a bad idea.
> 
> We can also make this a boot parameter, i don't care, but i want to disable
> console suspend without recompiling the kernel.

It's a bit similar to the pm_trace thing.

As long as there is a sane default (ie. the consoles are disabled), I don't see
a big problem with that.

Moreover, distributions will probably want to have a switch so that they can
tell the user "please do 'echo 1 > ...' and retest" and if they start to add
such things by themselves that wouldn't be nice.

Greetings,
Rafael


-- 
"Premature optimization is the root of all evil." - Donald Knuth