patch: fix 30 second hang while resuming

[Nate Lawson]

Garance A Drosihn wrote:
> At 3:19 PM -0800 2/27/05, Nate Lawson wrote:
> 
>> Poul-Henning Kamp wrote:
>>
>>>
>>> Have you tried sos@ new ATAng mk III patches ?  As far as I know
>>> he plans to commit those shortly.
> 
> One minor point: Søren is on vacation right now.  I *think* he will
> be getting back around March 5th (assuming I understand what he said
> in a recent message).  So, *if* this is something we want to get in
> before the code-freeze, we may have to act without his input.

Augh.  Not the most convenient circumstances.  I'm happy to work with 
someone else if necessary.

>> Not yet.  In any case, I'd prefer these problems be fixed before
>> the import since the patches are minor and data corruption is
>> generally a bad thing for a little while even if a large, new
>> something is coming sometime soon.
> 
> Your original message only mentioned a long-delay when resuming.
> Is this bug also causing data-corruption in some circumstances?
> (I don't have the slightest idea what this area of code does, I
> just want to make sure we understand how serious the issue is...).
> 
> Is this something which only comes up in 6.x?  Or is the same
> thing happening in 5.x too?

There are two bugs I sent in two separate messages.  The first gives 
several overwrites of 512 bytes of memory adjacent to the ATA param 
structure.  The second is just a very long delay on resume.  The first 
is more serious and could occur on different hardware although I think 
only systems with the "phantom slave" problem will trigger it.  It makes 
booting with INVARIANTS panic immediately and causes an indeterminant 
problem without.  The previous time this bug appeared, it was pretty 
nasty -- inserting a PCMCIA card would give a panic or at some point the 
ifnet list would be corrupted.  I don't know the effect this time.

-- 
Nate