YAMAMOTO Takashi writes:
> 
> hi,
> 
> the attached diff moves initialization of buffers from rf_CreateDiskQueueData
> to rf_DispatchKernelIO, as it used to be in the original raidframe.
> 
> is it ok?

yes.  I've tested and committed the changes.  (The bug for which the 
"workaround" code was present is gone -- I believe it went at the 
same time I re-worked the memory allocation bits.  In any event, I 
am unable to reproduce the problem even with a physical disk failure.)

Thanks!

Later...

Greg Oster