[libvirt] [PATCH v4 05/20] qemu-cgroup: allow accel rendernode access

Cole Robinson crobinso at redhat.com
Fri Sep 20 20:12:30 UTC 2019


On 9/13/19 8:50 AM, marcandre.lureau at redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau at redhat.com>
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau at redhat.com>
> Reviewed-by: Ján Tomko <jtomko at redhat.com>
> ---
>   src/qemu/qemu_cgroup.c | 24 ++++++++++++++++++++++++
>   1 file changed, 24 insertions(+)
> 
> diff --git a/src/qemu/qemu_cgroup.c b/src/qemu/qemu_cgroup.c
> index ecd96efb0a..eb6f993d8e 100644
> --- a/src/qemu/qemu_cgroup.c
> +++ b/src/qemu/qemu_cgroup.c
> @@ -503,6 +503,25 @@ qemuSetupGraphicsCgroup(virDomainObjPtr vm,
>   }
>   
>   
> +static int
> +qemuSetupVideoAccelCgroup(virDomainObjPtr vm,
> +                          virDomainVideoAccelDefPtr def)
> +{
> +    qemuDomainObjPrivatePtr priv = vm->privateData;
> +    int ret;
> +
> +    if (!def->rendernode ||
> +        !virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_DEVICES))
> +        return 0;
> +
> +    ret = virCgroupAllowDevicePath(priv->cgroup, def->rendernode,
> +                                   VIR_CGROUP_DEVICE_RW, false);
> +    virDomainAuditCgroupPath(vm, priv->cgroup, "allow", def->rendernode,
> +                             "rw", ret);
> +    return ret;
> +}
> +
> +
>   static int
>   qemuSetupBlkioCgroup(virDomainObjPtr vm)
>   {
> @@ -803,6 +822,11 @@ qemuSetupDevicesCgroup(virDomainObjPtr vm)
>               goto cleanup;
>       }
>   
> +    for (i = 0; i < vm->def->nvideos; i++) {
> +        if (qemuSetupVideoAccelCgroup(vm, vm->def->videos[i]->accel) < 0)
> +            goto cleanup;
> +    }
> +
>       for (i = 0; i < vm->def->ninputs; i++) {
>           if (qemuSetupInputCgroup(vm, vm->def->inputs[i]) < 0)
>               goto cleanup;
> 

->accel can be NULL here otherwise we crash like I mentioned yesterday. 
All the other functions here act on a device as a whole so I think this 
should be qemuSetupVideoCgroup while you are here

FWIW after fixing this the VM starts, but I see this in the logs with 
stock Fedora 31 qemu 4.1, not sure if it's indicative of a qemu error or 
something missing on libvirt side. The VM gets to the initrd load stage 
but then graphic output stops updating

2019-09-20T20:06:51.238950Z qemu-system-x86_64: Failed initializing 
vhost-user memory map, consider using -object memory-backend-file share=on
2019-09-20T20:06:51.238971Z qemu-system-x86_64: vhost_set_mem_table 
failed: Resource temporarily unavailable (11)
2019-09-20T20:06:51.238976Z qemu-system-x86_64: Error start vhost dev

- Cole




More information about the libvir-list mailing list