Re: [PATCH v2 03/14] notifiers: Assert that RCU is watching in notify_die

From: Ingo Molnar
Date: Tue Jun 23 2015 - 04:56:39 EST



* Borislav Petkov <bp@xxxxxxxxx> wrote:

> On Mon, Jun 22, 2015 at 09:26:13AM -0700, Andy Lutomirski wrote:
>
> > notify_die is misnamed and has little to do with death. It's really just
> > notifying about an exception, and we might end up oopsing, sending a signal,
> > or neither.
>
> But if we oops and wedge solid afterwards, it might happen that only the first
> splat comes out on the console, no? And that will be the lockdep splat which
> would be useless for debugging the actual problem...

So I think the theory is that crashes do happen, and that any RCU warning only
matters to (usually) small race windows.

So by the time a difficult crash truly happens, exactly in that race window, we'd
have fixed the RCU warning long ago.

I.e. the placement of the RCU warning isn't really relevant in the long run, as it
should not trigger.

In the short run it's probably more important to have it first, because if we have
that RCU race then we don't know whether we can trust anything that happens after
executing the (flawed) notifier chain.

Does that logic make sense to you?

Thanks,

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/