Hi Cedric, > Here's one I have been keeping for a while. Nothing really fancy : > basic replacement of kernel_thread and removal of the start > completion which is now covered in kthread. the signal needs to be unblocked before the start completion, since otherwise the signal might be sent while signals are blocked (though this is extremely unlikely). So you can't rely on kthread's start completion I'm afraid, because that happens before the unblocking. Ciao, Duncan.