Re: [RFC] cfg80211: add PM hooks

[Johannes Berg <johannes@sipsolutions.net>]

[-- Attachment #1: Type: text/plain, Size: 1211 bytes --]

On Mon, 2008-08-04 at 18:27 +0200, Marcel Holtmann wrote:

> >>> +	.suspend = wiphy_suspend,
> >>> +	.resume = wiphy_resume,
> >>> };
> >>
> >> this needs to be wrapped into CONFIG_PM, because otherwise the build
> >> will fail when CONFIG_PM is not enabled.
> >
> > I thought so too, but then I looked at include/linux/device.h
> 
> when did this got changed? 

No idea.

> And more to the point, why nobody bothered  
> to remove all the CONFIG_PM and CONFIG_HOTPLUG ifdefs splattered  
> around the kernel. 

Most of them are for struct device_driver and not struct class though,
so I haven't checked.

> It must be part of Greg's changes to remove class  
> devices. I am not sure this is all the right approach since  
> potentially it blots the kernel.

I suppose drivers still have a choice. I just decided that this was so
little code it didn't matter.

The more interesting question I guess is: are we now required to have
these callbacks even if CONFIG_PM(_SLEEP) is not defined? I'm not all
that interested in trying to find answers right now though.

> This would also means that the CONFIG_HOTPLUG for dev_uevent can go  
> away.

Indeed, it could.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]