On Tue, Dec 17, 2024 at 10:00:44AM -0800, Oliver Upton wrote:
On Tue, Dec 17, 2024 at 03:10:28PM +0000, Mark Brown wrote:
No, this isn't a new use - a Fixes: tag indicates that the referenced commit introduced the problem being fixed and that is exactly what's going on here. Like I say the selftests are not a completely separate project, they are a part of the same source release as the rest of the kernel and it is helpful to track information like this.
A Fixes tag suggests a bug in the referenced commit, which isn't the case here.
I agree that having some relation between the two is useful for determining the scope of a backport, but conveniently in this case the test failure was introduced in 6.13.
While the patch introducing the test failure was introduced in -rc3 it was tagged as a fix for d5a32b60dc18 ("KVM: arm64: Allow userspace to change ID_AA64MMFR{0-2}_EL1") which was in v6.7 so I'm expecting to see it being backported to relevant stable kernels, which will in turn cause testsuite failures in those stable kernels if this change doesn't go back with it. Hopefully they'll find it from the commit message.
I would say there's a case that leaving the tests broken is a bug, it's certainly some kind of problem. Obviously we're sometimes a bit relaxed on that within a series, though it's fortunately relatively rare.