Re: [patch] floppy: suspend/resume fix

[Ingo Molnar <mingo@elte.hu>]

* Russell King <rmk+lkml@arm.linux.org.uk> wrote:

> On Sun, Nov 12, 2006 at 11:40:28PM +0100, Mikael Pettersson wrote:
> > The bug occurs regardless of whether I leave the floppy disc in the drive
> > during suspend or not. 2.6.19-rc5 (vanilla and with Ingo's suspend/resume
> > hooks) fails the following use case as well:
> > 
> > 1. boot
> > 2. insert floppy disc
> > 3. tar tvf /dev/fd0 (works)
> > 4. manually eject floppy disc
> > 5. suspend, later resume 
> > 6. insert floppy disc
> > 7. tar tvf /dev/fd0 (fails with I/O errors)
> > 8. tar tvf /dev/fd0 (works)
> > 
> > Like Ingo said, something happens to the HW during suspend and we
> > need to figure out how to reinitialise the HW and the driver so that
> > things work immediately after resume.
> 
> Now this is interesting - I know there's been a long standing bug with 
> kernels on my Thinkpad which behave in a similar way to your 
> description above.  Basically whenever I change the disk in the drive 
> I tend to need _two_ goes to do anything with it - the first mostly 
> always fails with IO errors.

yeah. But somehow the pre-lockdep-change driver gets this right - purely 
by virtue of unregistering the IRQ line and the DMA channel - neither of 
which should have any material effect on behavior ... [and when i 
restored this in suspend/resume it didnt fix the bug] Weird.

	Ingo