Jason Thorpe wrote:
>
> So, it seems like it would be possible to fix this for the i386 case 
> by completely revoking the mapping, yes?
If you wanted to fix it, it would be by changing kcopy() (as it is used 
in uiomove()) to have the same workaround as copyout().

That's probably easier and more straightforward, since revoking the 
mapping would also hit read access, and probably would require some more 
bookkeeping.

- Frank