ACK/cmnt: [X/Z/A][PATCH 0/5] fix memory-hotplug test

Stefan Bader stefan.bader at canonical.com
Fri Aug 18 13:38:10 UTC 2017


On 15.08.2017 13:58, Po-Hsu Lin (Sam) wrote:
> BugLink: https://bugs.launchpad.net/bugs/1710868
> 
> SRU Justification:
> 
> Impact:
> The memory-hotplug test in the kernel tree needs to be fixed:
> 1. The test will pass even the memory blocks are failed to be offlined.
> 2. It's using a random number generator to decide if each memory block
> needs to be offlined, which is not ideal for testing as the block it
> tries to unplug might be different between two runs.
> 3. The "-r" flag, percent-of-memory-to-offline, does not match what is
> says. It's being used as the percentage to get a memory block offlined
> instead.
> 
> Fix:
> Apply the following patch from upstream can fix these issues:
> 593f9278 - selftests: typo correction for memory-hotplug test
> a34b28c9 - selftests: check hot-pluggagble memory for memory-hotplug test
> 72441ea5 - selftests: check percentage range for memory-hotplug test
> 02d8f075 - selftests: add missing test name in memory-hotplug test
> 5ff0c60b - selftests: fix memory-hotplug test
> 
> Testcase:
> Run the script directly, it will fail when all memory blocks are busy.
> And it will try to unplug memory blocks in a fixed order.
> 
> Regression Potential:
> Negligible, it's just affecting this specific test case among those
> testing tools.
> 
> On Thu, Aug 10, 2017 at 2:32 PM, Po-Hsu Lin <po-hsu.lin at canonical.com> wrote:
>> The memory-hotplug test in the kernel selftest directory has a flag which is
>> called "percent-of-memory-to-offline". However it is using a random number
>> to decide if each memory block needs to be offlined (with 10% chance).
>> This might cause all memory blocks being skipped, or throw an error if the
>> block is busy. Also, it does not match with the meaning of the flag.
>>
>> This patch set will fix the test by offlining a certain amount of memory blocks
>> and it will try to offline them one-by-one instead of doing this randomly to
>> avoid the block is busy issue and make it more reproducible.
>>
>> Skipping Trusty as the ubuntu_kernel_selftests test suite was blacklisted on it.
>>
>> Po-Hsu Lin (5):
>>   selftests: typo correction for memory-hotplug test
>>   selftests: check hot-pluggagble memory for memory-hotplug test
>>   selftests: check percentage range for memory-hotplug test
>>   selftests: add missing test name in memory-hotplug test
>>   selftests: fix memory-hotplug test
>>
>>  .../selftests/memory-hotplug/mem-on-off-test.sh    | 86 +++++++++++++++++-----
>>  1 file changed, 67 insertions(+), 19 deletions(-)
>>
>> --
>> 2.7.4
>>
> 
Acked-by: Stefan Bader <stefan.bader at canonical.com>

It sounds reasonable to consolidate the correct testing across the releases and
if that is upstream, better. So yeah, with adding bug references this should be
acceptable. I wonder whether the commits might be marked with "Ignore" like
startnewrelease if the changes have no impact on the delivered kernel(s). OTOH
the same is true for changes to the build scripts and still we did not hide those.

-Stefan


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20170818/a329b627/attachment.sig>


More information about the kernel-team mailing list