Michael S. Tsirkin wrote: > On Thu, Jun 25, 2009 at 09:02:28AM -0400, Gregory Haskins wrote: > >>> Here's what I have in mind: >>> kvm does >>> lock >>> dev = find >>> unlock >>> >>> <---------- at this point group device is removed >>> >>> write access to device that has been removed >>> >>> >>> >>> >> Hmm...you are right. This looks like it was introduced with that recent >> locking patch you cited. Well, I can still fix it now easily by putting >> an rcu-read-lock around the access. Longer term we should move to >> srcu. Thoughts? >> >> -Greg >> >> > > > Some callbacks take kvm mutex lock. So it seems rcu won't work, > we need srcu. > > Indeed. I will put something together. We should consider merging the io_bus cleanup I have for deregister support ASAP as well. -Greg
Attachment:
signature.asc
Description: OpenPGP digital signature
