Precisely. Note that the actual value of BKL-covered code shrinks every time
we push the BKL out of commonly used code.
So even if there's still usage sites around, there will be a point where it
makes sense to just desupport it or do a final, crude conversion. If there are
no active users who are willing to help us debug potential bugs then there's
frankly no value in us being just as careful with those drivers as we are with
other, more commonly used code.
For the rest, lockdep & softlockup detector will be plenty enough to debug any
bugs that may slip through - just like you did it with ReiserFS.
Ingo
--
