Re: [RFC] race in request_module()

[Keith Owens <kaos@ocs.com.au>]

On Mon, 22 Apr 2002 20:49:40 -0400 (EDT), 
Alexander Viro <viro@math.psu.edu> wrote:
>	Looks like request_module() has quite a few problems:
>
>* there is no way to distinguish between failing modprobe and successful
>  one followed by rmmod -a (e.g. called by cron).  For one thing, we
>  don't pass exit value of modprobe to caller of request_module().

There is no such thing as a failing modprobe.  It either works and the
module is loaded or modprobe does not work and the module is not
loaded.  This excludes the case where a module oops during init, but
that is not what you are worried about.

When a module is loaded, it is marked !MOD_USED_ONCE.  An explicit
rmmod will get rid of the module but rmmod -a will not.  rmmod -a will
not remove a module unless __MOD_INC_USE_COUNT has been issued on the
module at least once, or the module is loaded to satisfy unresolved
symbols from another module.

Rusty and I have a completely new design for module loading and
unloading in 2.5, we believe it is race free.  I do not have time to
work on the new design until I have got kbuild 2.5 into the kernel.