Patch claiming to remove enable_irq_nosync() had left it alive but killed
disable_irq_nosync() instead...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>