[PATCH 0/1] UBUNTU: SAUCE: Increase ATA_TMOUT_FF_WAIT to 5 seconds.

[TJ]

On Thu, 2009-02-26 at 10:47 +0100, Stefan Bader wrote:
> I am not really convinced this will help. I might still miss something but...

The way I had read the code, ATA_TMOUT_FF_WAIT (via nodev_deadline)
comes into play when -ENODEV, where ready is 'artificially' reset to
prevent the loop returning -ENODEV prematurely:

else if ((link->ap->flags & ATA_FLAG_SATA) &&
 !ata_link_offline(link) && 
 time_before(now, nodev_deadline))

But, looking at it again following you comments, it is possible that the
code-path followed is the prior:

if (ata_link_online(link))

- in other words the link is up but the device hasn't yet cleared reset.

As I commented as an addendum to the bug report my feeling is that this
is more to do with a 'false positive' from the slow-path detection and
any real solution will come from parallel handling of the devices and
some intelligence in slow-path (it could do with maybe_wait(delay)
functionality).

This class of issue is going to affect the push to solve the
suspend/resume reliability but I don't see any easy way to deal with the
underlying problem.

I investigated this bug as part of my sweep to determine how many
suspend/resume reports might be caused by bug #331415
"request_firmware() fails on resume from suspend".

It looks like this one should be marked "Won't Fix - would you agree?