On 12/2/22 02:25, Petr Mladek wrote:
On Thu 2022-12-01 16:58:38, Shuah Khan wrote:
On 11/30/22 15:22, Joe Lawrence wrote:
On 7/15/22 10:45 AM, Petr Mladek wrote:
On Fri 2022-07-01 16:13:50, Shuah Khan wrote:
On 7/1/22 1:48 AM, Miroslav Benes wrote:
On Thu, 30 Jun 2022, Shuah Khan wrote:
Sorry Nack on this. Let's not add modules under selftests. Any usage of
module_init()
doesn't belong under selftests.
Yes I did and after reviewing and thinking about it some more, I decided this
is the right direction go down on.
Do you have some particular reason why building modules in selftests
directory might cause problems, please?
My reasons are that with this change module_init() propagates out of
strictly kernel space and now is in selftests which are user-space.
Any changes to this interface will be tied to user-space change.
I am sorry but I do not understand the meaning here. module_init() is
called when module is loaded. It is not called in userspace.
Maybe, you mean that modules under lib/ are clearly in-tree
modules. If we move then under tools/ then they will be build
like out-of-tree modules. Except that they will be maintained in-tree
so that it will be easy to keep them in sync.
Yes. That is what I mean.
And I am sure that they will be actively maintained. The fixes are
there to make sure that livepatching still works as expected.
They must pass when any change is done in the livepatch subsystem.
And they must pass when any kernel is released.
In other words, livepatch and kernel are revision matched.
The only concern might be how build failure is handled. IMHO, we
need to handle it the same way and test failure.
Yes. This is another concern - build failures. Let's experiment with
modules under selftests and see if this becomes a problem.
In general, newer tests offer the best coverage, hence the recommendation
to run newer tests on older kernels assuming that the tests are built
on a newer kernel and backwards should run in a backwards compatible
way on older kernels.
This works for the userspace interface that should always be backward
compatible. But it does not work for kABI.
This is broader than revision matching. Tests should gracefully exit
with skip when a config option they depend on is disabled. The same
gets extended to older kernel versions.
Do you have a requirement that livepatch test has to be revision
matched with the kernel? Even if that is the case, there is no real
reason to move modules under selftests other than keeping them in
one location.
Yes, kABI is not backward compatible. But building the tests
modules out-of-tree way would allow to build test modules with
different kABI from the same sources.
Okay. This is a solid reason for livepatch modules to live under
sefltests. Let's capture this in README and the other updates that
need to be made to it in v3.
thanks,
-- Shuah
