[Yakkety] [PATCH] UBUNTU: [Config] CONFIG_ARM64_ERRATUM_845719=y

[Colin Ian King]

On 06/12/16 21:58, Seth Forshee wrote:
> On Wed, Dec 07, 2016 at 01:04:44AM +0800, Ming Lei wrote:
>> On Wed, Dec 7, 2016 at 12:47 AM, Seth Forshee
>> <seth.forshee at canonical.com> wrote:
>>> On Tue, Dec 06, 2016 at 04:42:48PM +0000, Colin Ian King wrote:
>>>> On 06/12/16 16:37, Ming Lei wrote:
>>>>> Looks this option is missed in Yakkety, and should be
>>>>> enabled always.
>>>>>
>>>>> Signed-off-by: Ming Lei <ming.lei at canonical.com>
>>>>> ---
>>>>>  debian.master/config/config.common.ubuntu | 2 +-
>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/debian.master/config/config.common.ubuntu b/debian.master/config/config.common.ubuntu
>>>>> index aaba225..26ddbf7 100644
>>>>> --- a/debian.master/config/config.common.ubuntu
>>>>> +++ b/debian.master/config/config.common.ubuntu
>>>>> @@ -505,7 +505,7 @@ CONFIG_ARM64_ERRATUM_826319=y
>>>>>  CONFIG_ARM64_ERRATUM_827319=y
>>>>>  CONFIG_ARM64_ERRATUM_832075=y
>>>>>  CONFIG_ARM64_ERRATUM_834220=y
>>>>> -# CONFIG_ARM64_ERRATUM_843419 is not set
>>>>> +CONFIG_ARM64_ERRATUM_843419=y
>>>>>  CONFIG_ARM64_ERRATUM_845719=y
>>>>>  CONFIG_ARM64_HW_AFDBM=y
>>>>>  # CONFIG_ARM64_LSE_ATOMICS is not set
>>>>>
>>>> This is enabled for Xenial and not Yakkety (I wonder why the config
>>>> enforcer is not picking this up...)
>>>>
>>>> There is no bug number in this patch, can that be included?  Also
>>>> reference to "Cortex-A53: 845719: a load might read incorrect data"
>>>> maybe of use in the patch as it informs us what CPU issue it works around.
>>>
>>> Some config options got mistakenly changed in yakkety, especially for
>>> non-x86 architectures. Only options marked as ENFORCED in the
>>> annotations file are enforced.
>>>
>>> So the patch probably should also mark this option as enforced. It will
>>> also be needed in zesty.
>>
>> But all CONFIG_ARM64_ERRATUM_* aren't marked yet as ENFORCED
>> in debian.master/config/annotations for yakkety.
> 
> Undoubtedly not everything that ought to be enforced is marked as
> enforced today. If it's critical that all of those options remain
> enbled, feel free to submit a patch which marks all of them as enforced.
> 
> It's less important for the stable releases, but for the development
> release it will help us make sure we don't turn off options which need
> to be enabled. It would have prevented this problem in yakkety, for
> instance :-)

My gut feeling is this fits into our remit to sort out rather than those
submitting these kind of config changes as it's our kernel team tool and
method for catching these and we shouldn't expect community to be
concerned with the config enforcer details.

Colin

> 
> Seth
>