Bananymous
9f4271f6d8
This moves locking to the inodes themselves which allows reducing lock times significantly. Main inodes (ext2 and tmpfs) still do contain a single big mutex that gets locked during operations but now we have the architecture to optimize these.