From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Wed, 27 Sep 2023 16:19:22 -0400 From: Stefan Hajnoczi Message-ID: <20230927201922.GB529043@fedora> References: <20230915102531.55894-1-hreitz@redhat.com> <20230925204852.GG323580@fedora> <07282c72-7a83-70c5-395d-454281663eb1@redhat.com> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="kjMssMK4IWHFEABh" Content-Disposition: inline In-Reply-To: <07282c72-7a83-70c5-395d-454281663eb1@redhat.com> Subject: Re: [Virtio-fs] [PATCH v3 0/5] vhost-user: Back-end state migration List-Id: Development discussions about virtio-fs List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Hanna Czenczek Cc: Stefan Hajnoczi , "open list:virtiofs" , Eugenio =?iso-8859-1?Q?P=E9rez?= , qemu-devel , "Michael S . Tsirkin" --kjMssMK4IWHFEABh Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Wed, Sep 27, 2023 at 10:32:14AM +0200, Hanna Czenczek wrote: > On 26.09.23 22:10, Stefan Hajnoczi wrote: > > Hi Hanna, > > I was thinking about how this could work without SUSPEND/RESUME. What > > do you think of the following? > >=20 > > 1. The front-end sends VHOST_USER_RESET_DEVICE (or > > VHOST_USER_RESET_OWNER, when necessary) when the guest driver resets > > the device but not on vhost_dev_start()/vhost_dev_stop(). >=20 > This is half the work of SUSPEND/RESUME.=C2=A0 It isn=E2=80=99t easy to d= o. I sent a patch series to bring VHOST_USER_RESET_DEVICE to all vhost-user devices: https://lore.kernel.org/qemu-devel/20230927192737.528280-1-stefanha@redhat.= com/T/#t >=20 > > 2. Suspend the device when all virtqueues are stopped via > > VHOST_USER_GET_VRING_BASE. Resume the device after at least one > > virtqueue is started and enabled. > > 3. Ignore VHOST_USER_SET_STATUS. > >=20 > > Reset would work. The device would suspend and resume without losing > > state. Existing vhost-user backends already behave like this in > > practice (they often don't implement RESET_DEVICE). >=20 > I don=E2=80=99t understand the point, though.=C2=A0 Today, reset in pract= ice is a no-op > anyway, precisely because we only send SET_STATUS 0, don=E2=80=99t fall b= ack to > RESET_OWNER/RESET_DEVICE, and no back-end implements SET_STATUS 0 as a > reset.=C2=A0 By sending RESET_* in case of a guest-initiated reset and no= thing in > case of stop/cont, we effectively don=E2=80=99t change anything about the= latter > (which is what SUSPEND/RESUME would be for), but only fix the former case= =2E=C2=A0 > While I agree that it=E2=80=99s wrong that we don=E2=80=99t really reset = the back-end in > case of a guest-initiated reset, this is the first time in this whole > discussion that that part has been presented as a problem that needs fixi= ng > now. >=20 > So the proposal effectively changes nothing for the vhost_dev_stop()/star= t() > case where we=E2=80=99d want to make use of SUSPEND/RESUME, but only for = the case > where we would not use it. We discussed this on a call today. 2 & 3 are additions to the spec that Hanna has agreed to work on. Stefan --kjMssMK4IWHFEABh Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- iQEzBAEBCAAdFiEEhpWov9P5fNqsNXdanKSrs4Grc8gFAmUUjkoACgkQnKSrs4Gr c8j1Rwf+P5HoMSLa9EGAF8upIl91VlYs2rf3f1VzRUdwl7cHM4kNim4Xs9uBnwad DBPJhnzkf6597F91IAaZLPP8JmwKUxt38NtxwSjWM3CCLZd0Ace6NGfTO1qjTU1L 6i5Oib4k1F/4Jjzb868ureKyInfeIFeASRC1gUnlLMT5pKHbDcpO4oVm7HE7x2+i xROTR/wXZ1pc6GZHUF2WlzdcqJ2HrbtMqpsjvi5JIK09+lwNjzopgscbw2Il4Fia /W9wMyq8GyPtIf9elWgMzK3HkPyUyrWcSMC8ptnR318pgQnDnYLqSNbiryQplGo6 Gs6XdgB+CM21imIQTtqNJRovF3HeTA== =VtOc -----END PGP SIGNATURE----- --kjMssMK4IWHFEABh--