[libvirt PATCH v2 5/5] qemu_migrate: Do not forbif vDPA devices if can query blockers

Eugenio Perez Martin eperezma at redhat.com
Wed Jul 20 11:11:57 UTC 2022


On Wed, Jul 20, 2022 at 12:36 PM Jiri Denemark <jdenemar at redhat.com> wrote:
>
> On Wed, Jul 20, 2022 at 11:11:54 +0200, Eugenio Pérez wrote:
> > vDPA devices will be migratable soon. Since they are not migratable
> > before qemu 6.0, and qemu pre-6.0 didn't have the capability of asking
> > for migration blockers, let it hardcoded in that case.
> >
> > Otherwise, ask qemu about the explicit blocker.
> >
> > Signed-off-by: Eugenio Pérez <eperezma at redhat.com>
> > ---
> >  src/qemu/qemu_migration.c | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c
> > index 4224339f39..2f5c1d8121 100644
> > --- a/src/qemu/qemu_migration.c
> > +++ b/src/qemu/qemu_migration.c
> > @@ -1454,9 +1454,11 @@ qemuMigrationSrcIsAllowed(virQEMUDriver *driver,
> >      int pauseReason;
> >      size_t i;
> >      int r;
> > +    bool blockedReasonsCap = virQEMUCapsGet(priv->qemuCaps,
> > +                                          QEMU_CAPS_MIGRATION_BLOCKED_REASONS);
>
> Wrong indentation of the second line (QEMU... should be aligned with
> priv).
>

I'm fine with that indentation, but that will be a line with more than 80 chars.

> >
> >      /* Ask qemu if it have a migration blocker */
> > -    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_BLOCKED_REASONS)) {
> > +    if (blockedReasonsCap) {
> >          g_auto(GStrv) blockers = NULL;
> >          r = qemuDomainGetMigrationBlockers(driver, vm, &blockers);
> >          if (r != 0) {
> > @@ -1579,7 +1581,8 @@ qemuMigrationSrcIsAllowed(virQEMUDriver *driver,
> >              virDomainNetDef *net = vm->def->nets[i];
> >              qemuSlirp *slirp;
> >
> > -            if (net->type == VIR_DOMAIN_NET_TYPE_VDPA) {
> > +
>
> Extra empty line.
>

I'll delete it in the next version.

> > +            if (!blockedReasonsCap && net->type == VIR_DOMAIN_NET_TYPE_VDPA) {
>
> So possibly we could skip more checks in the function if migration
> blockers are supported by QEMU, but this is a good conservative
> approach. The other checks (if any) can be taken care of separately.
>

I did a fast search but I didn't see something super obvious to me.

Thanks!

> >                  virReportError(VIR_ERR_OPERATION_INVALID, "%s",
> >                                 _("vDPA devices cannot be migrated"));
> >                  return false;
>
> Jirka
>



More information about the libvir-list mailing list