Re: [PATCH] remove BKL from drivers' release functions

[Rick Lindsley]

    Even register_blkdev(), etc hold BKL, without these there will be
    a lot of problems, all these need to be taken care of if BKL is
    ever replaced.

I think it's pretty clear that the BKL has become so, um, multi-functional
over the years that any replacement will not involve a single new variable
(or algorithm).  On the plus side, it also means it can be a stepwise
project. If we carefully identify each function and replace it or obsolete
it, we can gradually deprecate the BKL at a safe and sane pace.

Everybody keep track of these usages; there will come a day where somebody
is asking you about it again :) open/release is the tip of the iceberg,
but it takes a long time for an iceberg to melt ....

Rick