Re: [RFC] race in request_module()

[Alexander Viro]

On Mon, 22 Apr 2002, Matthew Dharm wrote:

> The question then becomes one of how do I distinguish a race condition from
> a legitimate load/unload cycle?

> Isn't the problem here just the misuse of rmmod -a?  Perhaps we should
> attach a warning to the documentation to indicate the possible badness that
> can happen.

Not really.  _Any_ use of rmmod -a (i.e. unload stuff not in use) can
trigger that.

As for legitimate load/unload cycle... in this situation thing should be
considered busy.  It's that simple - what happens here is that we are asking
for module because we want to make it busy as soon as we get it.  Race
window is between the sys_create_module() from modprobe and try_inc_use_count()
in whatever wants it in the kernel or sys_init_module() for module depending
on it.

IOW, unload is _not_ legitimate here.