Re: wd(4) not waking up on resume

["Jonathan A. Kollasch" <jakllsch%kollasch.net@localhost>]

On Sat, Jun 04, 2016 at 05:02:46PM +0000, coypu%SDF.ORG@localhost wrote:
> Hi,
> 
> I'm interested in working suspend on my Dell Inspiron n4030.
> I've been looking into the various problems.
> 
> Currently, one of the big problems is that my wd(4) disk (which has root
> mounted on it, too) rarely wakes up on resume.
> 
~~~~
> Any suggestions?
> 
~~~~

I'd be looking at debugging in the ATA subsystem.  Having it turned on
just as the suspend is starting could show what commands are being
executed, and if/when they complete.

I'm suspicious that an ACPI interrupt knocks it loose; does a keyboard
(particularly pckbc(4)) event have a similar effect?  On the other hand,
the DSDT could be doing who knows what to the hardware behind our backs
when the power button is pressed.

My first suspicion is that the commands we issue to flush or suspend the
disk is still in flight when the machine transitions to S3.

Or maybe we need to be explicitly reestablishing the SATA link.

I will note that I've seen some serious issues with wd(4) and suspend.
I've observed that UDMA modes are not reestablished on resume.  This is
especially important for PATA disks, and probably also PATA disks on
SATA bridges.  For native SATA disks it is not particularly relevant,
but wouldn't hurt to do either.

I'll also note that wd(4) discovers wedges *before* figuring out if
MDMA/UDMA is possible... but I'm getting a little off the original path
here...

	Jonathan Kollasch