> Quoting Steve Wise <[EMAIL PROTECTED]>:
> Subject: Re: [PATCH 00/12] ofed_1_2 - Neighbour update support
>
> On Fri, 2007-02-02 at 08:03 +0200, Michael S. Tsirkin wrote:
> > > We could use a global refcnt to count the number of pending destructions
> > > and use a completion object to block unload until all the destructors
> > > fire and the refcnt goes to zero.
> >
> > It has the same race as module refcnt. So just use that.
> >
>
> I don't understand the race. Can you explain please? This should be
> able to be done without a race with a refcnt, a spinlock, a bit saying
> we're unloading, and a completion object.
>
> But maybe I'm confused ;-)
In short, the rule is that you can't pass a pointer to your function
to another module, and the unload module safely without synchronizing with that
other module.
Simplified example:
destructor
{
complete(&foo);
A:
return;
}
module_cleanup:
{
wait(foo)
return;
}
Now, assume destructor runs up to point A, then your module unloads,
and the memory its text occupied is overwritten by something else.
An attempt to execute code from point A will now crash.
So completion is not better than just module refcount here.
That said, I think the race is unlikely and just using module
refcount should be sufficient, and it's certainly simple.
--
MST
_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
