Re: [PATCH AUTOSEL 4.18 51/92] tools/testing/nvdimm: Fix support for emulating controller temperature

[Dan Williams]

On Fri, Sep 14, 2018 at 6:40 PM, Sasha Levin
<Alexander.Levin@xxxxxxxxxxxxx> wrote:
> On Fri, Sep 14, 2018 at 06:34:43PM -0700, Dan Williams wrote:
>>On Fri, Sep 14, 2018 at 6:30 PM, Sasha Levin
>><Alexander.Levin@xxxxxxxxxxxxx> wrote:
>>> From: Dan Williams <dan.j.williams@xxxxxxxxx>
>>>
>>> [ Upstream commit e5d772fbe7685aae0dff99f3b54158a0ec32155e ]
>>>
>>> In addition to populating the value the payload also needs to set the
>>> "controller temperature valid" flag.
>>>
>>> Fixes: cdd77d3e1930 ("nfit, libnvdimm: deprecate the generic SMART ioctl")
>>> Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
>>> Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxxxx>
>>> ---
>>>  tools/testing/nvdimm/test/nfit.c | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/tools/testing/nvdimm/test/nfit.c b/tools/testing/nvdimm/test/nfit.c
>>> index e2926f72a821..94c3bdf82ff7 100644
>>> --- a/tools/testing/nvdimm/test/nfit.c
>>> +++ b/tools/testing/nvdimm/test/nfit.c
>>> @@ -1308,7 +1308,8 @@ static void smart_init(struct nfit_test *t)
>>>                         | ND_INTEL_SMART_ALARM_VALID
>>>                         | ND_INTEL_SMART_USED_VALID
>>>                         | ND_INTEL_SMART_SHUTDOWN_VALID
>>> -                       | ND_INTEL_SMART_MTEMP_VALID,
>>> +                       | ND_INTEL_SMART_MTEMP_VALID
>>> +                       | ND_INTEL_SMART_CTEMP_VALID,
>>>                 .health = ND_INTEL_SMART_NON_CRITICAL_HEALTH,
>>>                 .media_temperature = 23 * 16,
>>>                 .ctrl_temperature = 25 * 16,
>>
>>Does no harm, but also does no benefit since this is just unit test
>>infrastructure. You may want to blacklist this directory for stable
>>checking in the future.
>
> We backport selftests back to stable trees. People run selftests on
> stable kernels as well, so it's both important to make sure that a
> backport didn't brake anything, and to locate possible missing stable
> patches.

Sure, I don't mind backporting this one, but I suspect the tests will
still have issues on older kernels. Ideally we would have ndctl, the
project that houses the userspace tests, in the kernel tree.
Unfortunately that's never going to happen because upstream is
allergic to autotools, and distros don't like the idea of more
utilities being built from kernel source.


> It is also the case that backporting selftests can't "break" the kernel.
> They are disconnected from the kernel's build process so we can be less
> conservative about backporting them.

I'm on board with the policy in general, but in the specific case of
tools/testing/nvdimm/ I'm not optimistic that we can reasonably keep
this in sync with an out of tree stimulus.