Re: Solving suspend-level confusion

[Nigel Cunningham <ncunningham@linuxmail.org>]

Hi.

On Wed, 2004-08-04 at 12:53, Benjamin Herrenschmidt wrote:
> > I've also done partial-tree support for suspend2 by making a new list
> > (along side the active, off and off_irq lists) and simply moving devices
> > I want to keep on (plus their parents) to this list prior to calling
> > device_suspend. Works well for keeping alive the ide devices being used
> > write the image.
> 
> How so ? By not calling suspend for it at all ? That's broken, the
> driver wants suspend to match the resume it will get when the image
> is reloaded, that's the only way the driver can guarantee a sane state
> saved in the suspend image.

Yes, I don't call suspend for it because I can be sure the drivers are
idle (before beginning to write the image, freeze all process, flush all
dirty buffers and suspend all other drivers, I then wait on my own I/O
until it is flushed too). I know it's broken to do so, but it was a good
work around for wearing out the thing by spinning it down and then
immediately spinning it back up, and I wasn't sure what the right state
to try to put it in is (sound familiar?!). If you want to tell me how I
could tell it to quiesce without spin down, I'll happily do that.

The sooner these issues get sorted, the better.

Regards,

Nigel