On Tue, 11 Mar 2003, Greywolf wrote:

> Thus spake Bill Studenmund ("BS> ") sometime Today...
>
> BS> That I got. But he wanted unmount to send the signal. I think the kernel
> BS> (specifically the fs code in the kernel) to send it.
> BS>
> BS> I'm envisioning the fs and the helper keeping a control socket open during
> BS> the lifetime of the mount, if for no other reason than to receive this
> BS> signal. Note also that the "signal" would be a socket write, not a SIG.
>
> Okay, now what happens if the helper actually gets kill()ed somehow?
> Will the filesystem automagically unmount?  If there are processes
> with something attached to said filesystem, do they go *poof*?
> Does the kernel panic?  Or is this "implementation-defined behaviour"?

It would have to be implementation-defined. Maybe another helper will get
re-spawned. Maybe the kernel (fs) can spawn one. On the other hand, maybe
the file system should go poof.

The kernel shouldn't panic.

Take care,

Bill