From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gregory Haskins Subject: Re: [KVM PATCH v2 2/2] kvm: use POLLHUP to close an irqfd instead of an explicit ioctl Date: Thu, 18 Jun 2009 12:29:31 -0400 Message-ID: <4A3A6B6B.3020705@novell.com> References: <20090604124047.10544.38861.stgit@dev.haskins.net> <4A3639B2.7080006@novell.com> <20090615125439.GI6351@redhat.com> <200906181446.31373.rusty@rustcorp.com.au> <20090618064908.GB11155@redhat.com> <4A3A2C67.4030406@novell.com> <20090618122219.GE16838@redhat.com> <4A3A4938.60604@novell.com> <20090618143547.GA19408@redhat.com> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="------------enig76A886F4B34674803E78BC78" Cc: Rusty Russell , kvm@vger.kernel.org, linux-kernel@vger.kernel.org, avi@redhat.com, davidel@xmailserver.org, paulmck@linux.vnet.ibm.com, akpm@linux-foundation.org To: "Michael S. Tsirkin" Return-path: In-Reply-To: <20090618143547.GA19408@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org This is an OpenPGP/MIME signed message (RFC 2440 and 3156) --------------enig76A886F4B34674803E78BC78 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Michael S. Tsirkin wrote: > On Thu, Jun 18, 2009 at 10:03:36AM -0400, Gregory Haskins wrote: > =20 >> Michael S. Tsirkin wrote: >> =20 >>> On Thu, Jun 18, 2009 at 08:00:39AM -0400, Gregory Haskins wrote: >>> =20 >>> =20 >>>> Michael S. Tsirkin wrote: >>>> =20 >>>> >>>> =20 >>>>> BTW, Gregory, this can be used to fix the race in the design: creat= e a >>>>> thread and let it drop the module reference with module_put_and_exi= t. >>>>> =20 >>>>> =20 >>>>> =20 >>>> I had thought of doing something like this initially too, but I thin= k >>>> its racy as well. Ultimately, you need to make sure the eventfd >>>> callback is completely out before its safe to run, and deferring to = a >>>> thread would not change this race. The only sane way I can see to d= o >>>> that is to have the caller infrastructure annotate the event somehow= >>>> (either directly with a module_put(), or indirectly with some kind o= f >>>> state transition that can be tracked with something like >>>> synchronize_sched(). >>>> =20 >>>> =20 >>> Here's what one could do: create a thread for each irqfd, and increme= nt >>> module ref count, put that thread to sleep. When done with >>> irqfd, don't delete it and don't decrement module refcount, wake thre= ad >>> instead. thread kills irqfd and calls module_put_and_exit. >>> >>> I don't think it's racy >>> =20 >> I believe it is. How would you prevent the thread from doing the >> module_put_and_exit() before the eventfd callback thread is known to >> have exited the relevant .text section? >> =20 > > Right. > > =20 >> All this talk does give me an idea, tho. Ill make a patch. >> =20 > > OK, but ask yourself whether this bag of tricks is worth it, and whethe= r > we'll find another hole later. Let's reserve the trickiness for > fast-path, where it's needed, and keep at least the assign/deassign sim= ple. > =20 Understood. OTOH, going back to the model where two steps are needed for close() is ugly too, so I don't want to just give up and revert that fix too easily. At some point we will call it one way or the other, but I am not there quite yet. > =20 >>> =20 >>> =20 >>>>> Which will work, but I guess at this point we should ask ourselves >>>>> whether all the hearburn with srcu, threads and module references i= s >>>>> better than just asking the user to call and ioctl. >>>>> =20 >>>>> =20 >>>>> =20 >>>> I am starting to agree with you, here. :) >>>> >>>> Note one thing: the SRCU stuff is mostly orthogonal from the rest of= the >>>> conversation re: the module_put() races. I only tied it into the >>>> current thread because the eventfd_notifier_register() thread gave m= e a >>>> convenient way to hook some other context to do the module_put(). I= n >>>> the long term, the srcu changes are for the can_sleep() stuff. So o= n >>>> that note, lets see if I can convince Davide that the srcu stuff is = not >>>> so evil before we revert the POLLHUP patches, since the module_put()= fix >>>> is trivial once that is in place. >>>> =20 >>>> =20 >>> Can this help with DEASSIGN as well? We need it for migration. >>> =20 >>> =20 >> No, but afaict you do not need this for migration anyway. Migrate the= >> GSI and re-call kvm_irqfd() on the other side. Would the fd even be >> relevant across a migration anyway? I would think not, but admittedly= I >> know little about how qemu/kvm migration actually works. >> =20 > > Yes but that's not live migration. For live migration, the trick is tha= t > you are running locally but send changes to remote guest. For that, we= > need to put qemu in the middle between the device and the guest, so it > can detect activity and update the remote side. > > And the best way to do that is to take poll eventfd that device assigns= > and push eventfd that kvm polls. To switch between this setup > and the one where kvm polls the ventfd from device directly, > you need deassign. > =20 So its still not clear why the distinction between deassign-the-gsi-but-leave-the-fd-valid is needed over a simple close(). Can you elaborate? -Greg --------------enig76A886F4B34674803E78BC78 Content-Type: application/pgp-signature; name="signature.asc" Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" -----BEGIN PGP SIGNATURE----- Version: GnuPG/MacGPG2 v2.0.11 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iEYEARECAAYFAko6a2sACgkQlOSOBdgZUxmxWACeO6gN6uY7NptLGzhUHeodTYqa FAEAmwVYjz2YhqZEmKUooNZFeoNQoXsE =0ic4 -----END PGP SIGNATURE----- --------------enig76A886F4B34674803E78BC78--