mbox series

[SRU,I/raspi,F/raspi,B/raspi-5.4,0/1] Kernel fails to boot in ScalingStack (LP: #1959102)

Message ID 20220210103806.19659-1-juergh@canonical.com
Headers show
Series [SRU,I/raspi,1/1] UBUNTU: [Config] raspi: Set VIRTIO_PCI=m | expand

Message

Juerg Haefliger Feb. 10, 2022, 10:38 a.m. UTC
[ Impact ]

The raspi kernel fails to boot in ScalingStack which is our OpenStack cluster for running ADT/autopkgtest tests.

[ Test Case ]

Boot a VM in ScalingStack, install raspi kernel, reboot. VM won't come back.

[ Fix ]

Enable VIRTIO PCI driver. ScalingStack uses virtio so the guest kernel needs to support that.

[ Where Problems Could Occur ]

New module means new code which could trigger kernel problems when loaded.

Juerg Haefliger (1):
  UBUNTU: [Config] raspi: Set VIRTIO_PCI=m

 debian.raspi/config/annotations          | 2 ++
 debian.raspi/config/config.common.ubuntu | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)

Comments

Kleber Sacilotto de Souza Feb. 10, 2022, 11:10 a.m. UTC | #1
On 2/10/22 11:38, Juerg Haefliger wrote:
> [ Impact ]
> 
> The raspi kernel fails to boot in ScalingStack which is our OpenStack cluster for running ADT/autopkgtest tests.
> 
> [ Test Case ]
> 
> Boot a VM in ScalingStack, install raspi kernel, reboot. VM won't come back.
> 
> [ Fix ]
> 
> Enable VIRTIO PCI driver. ScalingStack uses virtio so the guest kernel needs to support that.
> 
> [ Where Problems Could Occur ]
> 
> New module means new code which could trigger kernel problems when loaded.
> 
> Juerg Haefliger (1):
>    UBUNTU: [Config] raspi: Set VIRTIO_PCI=m
> 
>   debian.raspi/config/annotations          | 2 ++
>   debian.raspi/config/config.common.ubuntu | 3 ++-
>   2 files changed, 4 insertions(+), 1 deletion(-)
> 

I believe for b/raspi-5.4 the config and annotation files are copied from
f/raspi during open, so the change wouldn't need to be explicitly
applied there.


Acked-by: Kleber Sacilotto de Souza <kleber.souza@canonical.com>

Thanks
Tim Gardner Feb. 10, 2022, 12:55 p.m. UTC | #2
Acked-by: Tim Gardner <tim.gardner@canonical.com>

I'm curious why the original policy passed the config test when 
CONFIG_VIRTIO_PCI was not set, e.g.,

CONFIG_VIRTIO_PCI policy<{'amd64': 'y', 'arm64': 'y', 'armhf': 'y', 
'i386': 'y', 'ppc64el': 'y', 's390x': 'y'}>

rtg

On 2/10/22 3:38 AM, Juerg Haefliger wrote:
> [ Impact ]
> 
> The raspi kernel fails to boot in ScalingStack which is our OpenStack cluster for running ADT/autopkgtest tests.
> 
> [ Test Case ]
> 
> Boot a VM in ScalingStack, install raspi kernel, reboot. VM won't come back.
> 
> [ Fix ]
> 
> Enable VIRTIO PCI driver. ScalingStack uses virtio so the guest kernel needs to support that.
> 
> [ Where Problems Could Occur ]
> 
> New module means new code which could trigger kernel problems when loaded.
> 
> Juerg Haefliger (1):
>    UBUNTU: [Config] raspi: Set VIRTIO_PCI=m
> 
>   debian.raspi/config/annotations          | 2 ++
>   debian.raspi/config/config.common.ubuntu | 3 ++-
>   2 files changed, 4 insertions(+), 1 deletion(-)
>
Juerg Haefliger Feb. 10, 2022, 2:53 p.m. UTC | #3
On Thu, 10 Feb 2022 05:55:36 -0700
Tim Gardner <tim.gardner@canonical.com> wrote:

> Acked-by: Tim Gardner <tim.gardner@canonical.com>
> 
> I'm curious why the original policy passed the config test when 
> CONFIG_VIRTIO_PCI was not set, e.g.,
> 
> CONFIG_VIRTIO_PCI policy<{'amd64': 'y', 'arm64': 'y', 'armhf': 'y', 
> 'i386': 'y', 'ppc64el': 'y', 's390x': 'y'}>

It wasn't enforced.

...Juerg

 
> rtg
> 
> On 2/10/22 3:38 AM, Juerg Haefliger wrote:
> > [ Impact ]
> > 
> > The raspi kernel fails to boot in ScalingStack which is our OpenStack cluster for running ADT/autopkgtest tests.
> > 
> > [ Test Case ]
> > 
> > Boot a VM in ScalingStack, install raspi kernel, reboot. VM won't come back.
> > 
> > [ Fix ]
> > 
> > Enable VIRTIO PCI driver. ScalingStack uses virtio so the guest kernel needs to support that.
> > 
> > [ Where Problems Could Occur ]
> > 
> > New module means new code which could trigger kernel problems when loaded.
> > 
> > Juerg Haefliger (1):
> >    UBUNTU: [Config] raspi: Set VIRTIO_PCI=m
> > 
> >   debian.raspi/config/annotations          | 2 ++
> >   debian.raspi/config/config.common.ubuntu | 3 ++-
> >   2 files changed, 4 insertions(+), 1 deletion(-)
> >   
>
Tim Gardner Feb. 10, 2022, 3:01 p.m. UTC | #4
On 2/10/22 7:53 AM, Juerg Haefliger wrote:
> On Thu, 10 Feb 2022 05:55:36 -0700
> Tim Gardner <tim.gardner@canonical.com> wrote:
> 
>> Acked-by: Tim Gardner <tim.gardner@canonical.com>
>>
>> I'm curious why the original policy passed the config test when
>> CONFIG_VIRTIO_PCI was not set, e.g.,
>>
>> CONFIG_VIRTIO_PCI policy<{'amd64': 'y', 'arm64': 'y', 'armhf': 'y',
>> 'i386': 'y', 'ppc64el': 'y', 's390x': 'y'}>
> 
> It wasn't enforced.
> 
> ...Juerg
> 

Ah, "# FORMAT: 2". Seems to me that policy settings don't do much good 
unless they are enforced by default. I guess I'd better check all of the 
cloud kernels.

>   
>> rtg
>>
>> On 2/10/22 3:38 AM, Juerg Haefliger wrote:
>>> [ Impact ]
>>>
>>> The raspi kernel fails to boot in ScalingStack which is our OpenStack cluster for running ADT/autopkgtest tests.
>>>
>>> [ Test Case ]
>>>
>>> Boot a VM in ScalingStack, install raspi kernel, reboot. VM won't come back.
>>>
>>> [ Fix ]
>>>
>>> Enable VIRTIO PCI driver. ScalingStack uses virtio so the guest kernel needs to support that.
>>>
>>> [ Where Problems Could Occur ]
>>>
>>> New module means new code which could trigger kernel problems when loaded.
>>>
>>> Juerg Haefliger (1):
>>>     UBUNTU: [Config] raspi: Set VIRTIO_PCI=m
>>>
>>>    debian.raspi/config/annotations          | 2 ++
>>>    debian.raspi/config/config.common.ubuntu | 3 ++-
>>>    2 files changed, 4 insertions(+), 1 deletion(-)
>>>    
>>
>