SMP ATA woes on 6.0-RELEASE

[Lowell Gilbert]

"Steve Coles" <stcoles at tripos.com> writes:

> I have a strange deterministic boot problem with ATA devices on 2-way dell
> precision machines which boot from their SCSI disks. I have 4 of these
> boxes, and the only difference is the add-in ATA controller in the affected
> box.
> 
> A good summary would be "which kernels work":
>  
> a) GENERIC from the distribution boots fine (uniprocessor)
> b) GENERIC built from 6.0-RELEASE boots fine
> c) SMP built from 6.0-RELEASE hangs during the ATA probe of my disks
> d) <custom> SMP kernel built from 6.0-RELEASE with no ATA options/devices
> boots fine
> e) <custom> SMP kernel built from 6.0-RELEASE with ATA options/devices hangs
> as in c) above
> 
> .. It gets weirder
> 
> F) <custom> SMP kernel built from 6.0-RELEASE with no ATA options/devices
> boots fine (same as option d) above, then I load the ata modules from my
> normal login shell:
> 
> $> kldload atapci (loads ata.ko etc )
> $> Kldload atadisk
> 
> .. And it produces a stable, fast, ATA disk subsystem
> 
> G) <custom> SMP kernel built from 6.0-RELEASE with no ATA options/devices
> boots fine (option d) above with a
> 
> atapci_load="YES"
> atadisk_load="YES"
> 
> In loader.conf produces the same hang as c)
> 
> So...I presume the problem is related to the fact that the second CPU must
> be launched else the ATA code gets confused with my ATA subsystem (Naieve
> explanation of facts)
> 
> Q: Is this a known bug ?
> Q: If it is, is it fixed in -STABLE ?
> Q: If so - when is a good cvsup date that gives me a fixed ata driver with a
> "stable stable"

I'm not sure about exactly the situation you are seeing, but there do
seem to have been a number of changes related to ATA interactions with
SMP, added since 6.0.  If this system isn't in production yet, try
just updating to the latest -STABLE.