.. and one more comment on that patch: it would probably be a good
idea to make sure that the __I_xyz constants that are used for this
are in the range 0-3.
It doesn't really *matter*, in the sense that it will all just be a
cookie with a random address, but if anybody else ever uses the same
trick (or just uses bit_waitqueue) for another field in the inode, the
two cookies might end up being the same if you are very unlucky.
So from a future-proofing standpoint it would be good if the cookies
that are used are always "within" the address range of i_state.
I don't think any of the bits in i_state have any actual meaning, so
moving the bits around shouldn't be a problem.
Linus
