Comment 4 for bug 890569

It was one time crash only for me. But upstream doesn't have much changes in that area.

Only patches that I quickly could name even potentially related are in sysfs side:
f6d90b4f9ce018bff429d6e01ee672de712b8641 sysfs: Make sysfs_rename safe with sysfs_dirents in rbtrees
Mikulas Patocka 's rb_tree changes.

in dcache side there is more interesting commits but only one of them talk about d_parent not being NULL. But looking (upstream) code d_parent shouldn't ever be NULL.

If same applies to 3.0 kernel then It starts to look like some race condition that I happened to hit only once.