From: "Marc-André Lureau" <marcandre.lureau@gmail.com>
To: Eduardo Habkost <ehabkost@redhat.com>
Cc: Laurent Vivier <lvivier@redhat.com>,
"Michael S. Tsirkin" <mst@redhat.com>,
qemu-stable <qemu-stable@nongnu.org>,
QEMU <qemu-devel@nongnu.org>
Subject: Re: [Qemu-devel] [PATCH] virtio-pci: fix missing device properties
Date: Wed, 26 Jun 2019 11:48:20 +0200 [thread overview]
Message-ID: <CAJ+F1CKWRvuDxNuidWPE6+A7gKmbJ2h9G3DcDDVbd24y1XMYKw@mail.gmail.com> (raw)
In-Reply-To: <20190626015503.GX1862@habkost.net>
Hi
On Wed, Jun 26, 2019 at 3:56 AM Eduardo Habkost <ehabkost@redhat.com> wrote:
>
> On Wed, Jun 26, 2019 at 01:23:33AM +0200, Marc-André Lureau wrote:
> > Since commit a4ee4c8baa37154 ("virtio: Helper for registering virtio
> > device types"), virtio-gpu-pci, virtio-vga, and virtio-crypto-pci lost
> > some properties: "ioeventfd" and "vectors". This may cause various
> > issues, such as failing migration or invalid properties.
> >
> > Since those VirtioPCI devices do not have a base name, their class are
> > initialized with virtio_pci_generic_base_class_init(). However, if the
> > VirtioPCIDeviceTypeInfo provided a class_init which sets dc->props,
> > the properties were overwritten by virtio_pci_generic_class_init().
> >
> > Instead, introduce an intermediary base-type to register the generic
> > properties.
> >
> > Fixes: a4ee4c8baa37154f42b4dc6a13fee79268d15238
> > Cc: qemu-stable@nongnu.org
> > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > ---
> > hw/virtio/virtio-pci.c | 28 ++++++++++++++--------------
> > 1 file changed, 14 insertions(+), 14 deletions(-)
> >
> > diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> > index e6d5467e54..62c4977332 100644
> > --- a/hw/virtio/virtio-pci.c
> > +++ b/hw/virtio/virtio-pci.c
> > @@ -1913,13 +1913,6 @@ static void virtio_pci_generic_class_init(ObjectClass *klass, void *data)
> > dc->props = virtio_pci_generic_properties;
> > }
> >
> > -/* Used when the generic type and the base type is the same */
> > -static void virtio_pci_generic_base_class_init(ObjectClass *klass, void *data)
> > -{
> > - virtio_pci_base_class_init(klass, data);
> > - virtio_pci_generic_class_init(klass, NULL);
> > -}
> > -
> > static void virtio_pci_transitional_instance_init(Object *obj)
> > {
> > VirtIOPCIProxy *proxy = VIRTIO_PCI(obj);
> > @@ -1938,14 +1931,13 @@ static void virtio_pci_non_transitional_instance_init(Object *obj)
> >
> > void virtio_pci_types_register(const VirtioPCIDeviceTypeInfo *t)
> > {
> > + char *base_name = NULL;
> > TypeInfo base_type_info = {
> > .name = t->base_name,
> > .parent = t->parent ? t->parent : TYPE_VIRTIO_PCI,
> > .instance_size = t->instance_size,
> > .instance_init = t->instance_init,
> > .class_size = t->class_size,
> > - .class_init = virtio_pci_base_class_init,
> > - .class_data = (void *)t,
> > .abstract = true,
> > };
> > TypeInfo generic_type_info = {
> > @@ -1961,13 +1953,20 @@ void virtio_pci_types_register(const VirtioPCIDeviceTypeInfo *t)
> >
> > if (!base_type_info.name) {
> > /* No base type -> register a single generic device type */
> > - base_type_info.name = t->generic_name;
> > - base_type_info.class_init = virtio_pci_generic_base_class_init;
> > - base_type_info.interfaces = generic_type_info.interfaces;
> > - base_type_info.abstract = false;
> > - generic_type_info.name = NULL;
> > + /* use intermediate %s-base-type to add generic device props */
> > + base_name = g_strdup_printf("%s-base-type", t->generic_name);
> > + base_type_info.name = base_name;
> > + base_type_info.class_init = virtio_pci_generic_class_init;
> > +
> > + generic_type_info.parent = base_name;
> > + generic_type_info.class_init = virtio_pci_base_class_init;
> > + generic_type_info.class_data = (void *)t;
>
> Why are you using virtio_pci_generic_class_init for the base
> class, and virtio_pci_base_class_init for the subclass, but doing
> exactly the opposite when t->base_name is set?
>
> Isn't it simpler to just initialize base_type_info.name and leave
> all the rest alone? Patch below.
That was my initial approach. But then I tested the backport on v4.0.0, you get:
hw/display/virtio-vga.c:200:virtio_vga_class_init: Object
0x56247edbc0e0 is not an instance of type virtio-vga
The problem is that the introduced base class calls the t->class_init,
which expects the final class (virtio-vga, not virtio-vga-base-type).
This seems to be limited to virtio-vga, and gone in upstream since the
registration changed. But I would rather have the same patch
reviewed/applied.
>
> Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
> ---
> diff --git a/hw/virtio/virtio-pci.c b/hw/virtio/virtio-pci.c
> index e6d5467e54..3ee50a0783 100644
> --- a/hw/virtio/virtio-pci.c
> +++ b/hw/virtio/virtio-pci.c
> @@ -1913,13 +1913,6 @@ static void virtio_pci_generic_class_init(ObjectClass *klass, void *data)
> dc->props = virtio_pci_generic_properties;
> }
>
> -/* Used when the generic type and the base type is the same */
> -static void virtio_pci_generic_base_class_init(ObjectClass *klass, void *data)
> -{
> - virtio_pci_base_class_init(klass, data);
> - virtio_pci_generic_class_init(klass, NULL);
> -}
> -
> static void virtio_pci_transitional_instance_init(Object *obj)
> {
> VirtIOPCIProxy *proxy = VIRTIO_PCI(obj);
> @@ -1938,8 +1931,11 @@ static void virtio_pci_non_transitional_instance_init(Object *obj)
>
> void virtio_pci_types_register(const VirtioPCIDeviceTypeInfo *t)
> {
> + char *base_name = t->base_name ?
> + NULL :
> + g_strdup_printf("%s-base-type", t->generic_name);
> TypeInfo base_type_info = {
> - .name = t->base_name,
> + .name = t->base_name ?: base_name,
> .parent = t->parent ? t->parent : TYPE_VIRTIO_PCI,
> .instance_size = t->instance_size,
> .instance_init = t->instance_init,
> @@ -1959,21 +1955,8 @@ void virtio_pci_types_register(const VirtioPCIDeviceTypeInfo *t)
> },
> };
>
> - if (!base_type_info.name) {
> - /* No base type -> register a single generic device type */
> - base_type_info.name = t->generic_name;
> - base_type_info.class_init = virtio_pci_generic_base_class_init;
> - base_type_info.interfaces = generic_type_info.interfaces;
> - base_type_info.abstract = false;
> - generic_type_info.name = NULL;
> - assert(!t->non_transitional_name);
> - assert(!t->transitional_name);
> - }
> -
> type_register(&base_type_info);
> - if (generic_type_info.name) {
> - type_register(&generic_type_info);
> - }
> + type_register(&generic_type_info);
t->generic_name can't be NULL anymore? I thought the condition was
done for a good reason.
>
> if (t->non_transitional_name) {
> const TypeInfo non_transitional_type_info = {
> @@ -2005,6 +1988,7 @@ void virtio_pci_types_register(const VirtioPCIDeviceTypeInfo *t)
> };
> type_register(&transitional_type_info);
> }
> + g_free(base_name);
> }
>
> /* virtio-pci-bus */
>
>
> --
> Eduardo
>
--
Marc-André Lureau
next prev parent reply other threads:[~2019-06-26 9:49 UTC|newest]
Thread overview: 95+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-02 15:06 [Qemu-devel] [PULL 00/22] virtio, pc, pci: features, fixes, cleanups Michael S. Tsirkin
2019-06-25 23:23 ` [Qemu-devel] [PATCH] virtio-pci: fix missing device properties Marc-André Lureau
2019-06-26 1:55 ` Eduardo Habkost
2019-06-26 9:48 ` Marc-André Lureau [this message]
2019-06-26 12:39 ` Eduardo Habkost
2019-07-02 15:07 ` [Qemu-devel] [PULL 12/22] " Michael S. Tsirkin
2019-06-28 20:02 ` [Qemu-devel] [PATCH] pc: Move compat_apic_id_mode variable to PCMachineClass Eduardo Habkost
2019-06-29 10:46 ` Philippe Mathieu-Daudé
2019-06-30 21:30 ` Michael S. Tsirkin
2019-07-02 15:08 ` [Qemu-devel] [PULL 18/22] " Michael S. Tsirkin
2019-07-02 15:06 ` [Qemu-devel] [PULL 01/22] pcie: don't skip multi-mask events Michael S. Tsirkin
2019-07-02 15:06 ` [Qemu-devel] [PULL 02/22] pcie: check that slt ctrl changed before deleting Michael S. Tsirkin
2019-07-11 12:31 ` Peter Maydell
2019-07-02 15:06 ` [Qemu-devel] [PULL 03/22] pcie: work around for racy guest init Michael S. Tsirkin
2019-07-02 15:06 ` [Qemu-devel] [PULL 04/22] pcie: minor cleanups for slot control/status Michael S. Tsirkin
2019-07-02 15:56 ` [Qemu-devel] [PULL 00/22] virtio, pc, pci: features, fixes, cleanups Peter Maydell
2019-07-02 17:00 ` Michael S. Tsirkin
2019-07-02 17:20 ` Peter Maydell
2019-07-02 18:22 ` Michael S. Tsirkin
2019-07-02 18:27 ` Peter Maydell
2019-07-02 19:00 ` Michael S. Tsirkin
2019-07-26 12:39 ` Peter Maydell
2019-07-26 13:43 ` Michael S. Tsirkin
2019-07-04 9:25 ` Peter Maydell
2019-07-04 11:03 ` Pankaj Gupta
2019-07-04 21:24 ` Michael S. Tsirkin
2019-07-05 9:37 ` Pankaj Gupta
2019-07-04 21:29 ` Michael S. Tsirkin
2019-07-05 9:47 ` Peter Maydell
-- strict thread matches above, loose matches on Subject: below --
2019-06-26 7:48 [Qemu-devel] [PATCH v2 0/4] libvhost-user: VHOST_USER_PROTOCOL_F_MQ support Stefan Hajnoczi
2019-06-26 7:48 ` [Qemu-devel] [PATCH v2 1/4] libvhost-user: add vmsg_set_reply_u64() helper Stefan Hajnoczi
2019-07-02 15:08 ` [Qemu-devel] [PULL 19/22] " Michael S. Tsirkin
2019-06-26 7:48 ` [Qemu-devel] [PATCH v2 2/4] libvhost-user: support many virtqueues Stefan Hajnoczi
2019-07-02 15:08 ` [Qemu-devel] [PULL 20/22] " Michael S. Tsirkin
2019-06-26 7:48 ` [Qemu-devel] [PATCH v2 3/4] libvhost-user: implement VHOST_USER_PROTOCOL_F_MQ Stefan Hajnoczi
2019-07-02 15:08 ` [Qemu-devel] [PULL 21/22] " Michael S. Tsirkin
2019-06-26 7:48 ` [Qemu-devel] [PATCH v2 4/4] docs: avoid vhost-user-net specifics in multiqueue section Stefan Hajnoczi
2019-07-02 15:08 ` [Qemu-devel] [PULL 22/22] " Michael S. Tsirkin
2019-06-26 2:31 [Qemu-devel] [PATCH v4 0/5] virtio: fix some issues of "started" and "start_on_kick" flag elohimes
2019-06-26 2:31 ` [Qemu-devel] [PATCH v4 1/5] virtio: add "use-started" property elohimes
2019-06-26 10:17 ` Greg Kurz
2019-06-27 2:20 ` Yongji Xie
2019-07-02 15:07 ` [Qemu-devel] [PULL 13/22] " Michael S. Tsirkin
2019-06-26 2:31 ` [Qemu-devel] [PATCH v4 2/5] virtio: Set "start_on_kick" for legacy devices elohimes
2019-07-02 15:07 ` [Qemu-devel] [PULL 14/22] " Michael S. Tsirkin
2019-06-26 2:31 ` [Qemu-devel] [PATCH v4 3/5] virtio: Set "start_on_kick" on virtio_set_features() elohimes
2019-07-02 15:08 ` [Qemu-devel] [PULL 15/22] " Michael S. Tsirkin
2019-06-26 2:31 ` [Qemu-devel] [PATCH v4 4/5] virtio: Make sure we get correct state of device on handle_aio_output() elohimes
2019-07-02 15:08 ` [Qemu-devel] [PULL 16/22] " Michael S. Tsirkin
2019-06-26 2:31 ` [Qemu-devel] [PATCH v4 5/5] virtio: Don't change "started" flag on virtio_vmstate_change() elohimes
2019-07-02 15:08 ` [Qemu-devel] [PULL 17/22] " Michael S. Tsirkin
2019-06-26 10:43 ` [Qemu-devel] [PATCH v4 0/5] virtio: fix some issues of "started" and "start_on_kick" flag Laurent Vivier
2019-06-27 2:19 ` Yongji Xie
2019-06-19 9:49 [Qemu-devel] [PATCH v2 0/7] Qemu virtio pmem device Pankaj Gupta
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 1/7] virtio-pmem: add virtio device Pankaj Gupta
2019-07-02 11:46 ` Cornelia Huck
2019-07-02 15:07 ` [Qemu-devel] [PULL 05/22] " Michael S. Tsirkin
2019-07-11 12:57 ` Peter Maydell
2019-07-11 14:05 ` Pankaj Gupta
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 2/7] virtio-pci: Allow to specify additional interfaces for the base type Pankaj Gupta
2019-07-02 15:07 ` [Qemu-devel] [PULL 06/22] " Michael S. Tsirkin
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 3/7] virtio-pmem: sync linux headers Pankaj Gupta
2019-07-02 11:50 ` Cornelia Huck
2019-07-02 11:59 ` Pankaj Gupta
2019-07-02 16:58 ` Michael S. Tsirkin
2019-07-02 17:09 ` Pankaj Gupta
2019-07-02 17:11 ` Michael S. Tsirkin
2019-07-02 17:21 ` Pankaj Gupta
2019-07-02 15:10 ` Michael S. Tsirkin
2019-07-02 15:07 ` [Qemu-devel] [PULL 07/22] " Michael S. Tsirkin
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 4/7] virtio-pci: Proxy for virtio-pmem Pankaj Gupta
2019-07-02 11:55 ` Cornelia Huck
2019-07-02 12:00 ` Pankaj Gupta
2019-07-02 17:09 ` Michael S. Tsirkin
2019-07-02 17:14 ` Pankaj Gupta
2019-07-02 15:07 ` [Qemu-devel] [PULL 08/22] " Michael S. Tsirkin
2019-07-11 12:59 ` Peter Maydell
2019-07-11 13:27 ` Pankaj Gupta
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 5/7] hmp: Handle virtio-pmem when printing memory device infos Pankaj Gupta
2019-07-02 8:50 ` Wei Yang
2019-07-02 10:17 ` Pankaj Gupta
2019-07-02 15:07 ` [Qemu-devel] [PULL 09/22] " Michael S. Tsirkin
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 6/7] numa: Handle virtio-pmem in NUMA stats Pankaj Gupta
2019-07-02 15:07 ` [Qemu-devel] [PULL 10/22] " Michael S. Tsirkin
2019-06-19 9:49 ` [Qemu-devel] [PATCH v2 7/7] pc: Support for virtio-pmem-pci Pankaj Gupta
2019-07-02 15:07 ` [Qemu-devel] [PULL 11/22] " Michael S. Tsirkin
2019-07-01 3:53 ` [Qemu-devel] [PATCH v2 0/7] Qemu virtio pmem device Pankaj Gupta
2019-07-02 8:49 ` Wei Yang
2019-07-02 10:07 ` Pankaj Gupta
2019-07-03 0:58 ` Wei Yang
2019-07-03 1:31 ` Pankaj Gupta
2019-07-03 1:57 ` Wei Yang
2019-07-03 2:31 ` Pankaj Gupta
2019-07-03 2:42 ` Wei Yang
2019-07-03 3:21 ` Pankaj Gupta
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAJ+F1CKWRvuDxNuidWPE6+A7gKmbJ2h9G3DcDDVbd24y1XMYKw@mail.gmail.com \
--to=marcandre.lureau@gmail.com \
--cc=ehabkost@redhat.com \
--cc=lvivier@redhat.com \
--cc=mst@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-stable@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).