kern/60526: Post-PAE stable SMP machine freezes

[S?ren Schmidt]

Eugene Grosbein wrote:
> Doug White wrote:
> 
> 
>>Someone referred me to a EETimes article, and I have hundreds of Tyan
>>S2510/S2518 boards I can reproduce the problem on under multiple operating
>>systems.
> 
> 
> I googled a little for ROSB4.
> 
> Here http://www.ussg.iu.edu/hypermail/linux/kernel/0206.1/1212.html
> I've found the next statement:
> 
> 
>>The system will lock solid when both IDE channels are accessed,
>>and either one is using DMA. Since I want DMA, I simply abandon the
>>secondary channel.
> 
> 
> Is it possible do such a thing using FreeBSD?
> 
> Next, Linux seem to have a workarond(?) since 2002:
> 
> http://www.ussg.iu.edu/hypermail/linux/kernel/0206.1/0996.html
> 
> Soren, could you please look at this?

That fix newer worked AFAIK, today linux just dont allow UDMA on the 
ROSB4 at all.

There is no good SW fix, and its also not established that a HW fix 
exists as well, however I cannot reproduce the problem on my ASUS 
CUR-DLS engineering sample, but ASUS wont tell anything about it...

-- 
-S?ren