Re: swsusp: don't eat ide disks

[Linus Torvalds]

On Sun, 3 Nov 2002, Pavel Machek wrote:
> 
> Is that really the right way to prepare disks for suspend? 

It probably is, although I suspect it should just be a default action, and 
drivers can choose to implement their own "suspend()" functionality if 
they want to.

> I sleep all devices by telling driverfs to sleep them. Should I tell
> all block devices, then tell driverfs? Seems hacky to me. Or should
> idedisk_suspend generate request for itself, then pass it through
> queues?

I would strongly encourage letting the device hierarchy suspend() (now
called sysfs, not driverfs) call be the _only_ call the disk controller
ever gets. Having two different suspend mechanisms is just too confusing
for words, and there's no point.

			Linus