> the rest are mostly due to the address space checks. Some of them are
> likely trivial to fix, but the most interesting ones (in the networking
> code) are because the networking code re-uses the same data structures
> for both kernel and user addresses.
By separate mail I sent you half a dozen or perhaps a dozen sparse
fixes, all of the trivial kind. Hope you don't mind.
I did more, but the nfs / aio / sendfile stuff involves somewhat
larger surgery - will send some other time.
> it's not seriously usable yet
I like it, or, rather, the result of having __user annotation
for pointers to user space.
Andries