Re: Patch "selftests: vDSO: skip getrandom test if architecture is unsupported" has been added to the 6.11-stable tree

[Greg KH <greg@kroah.com>]

On Wed, Oct 02, 2024 at 06:13:45AM +0200, Jason A. Donenfeld wrote:
> On Tue, Oct 01, 2024 at 09:29:45AM -0600, Shuah Khan wrote:
> > On 10/1/24 09:03, Jason A. Donenfeld wrote:
> > > On Tue, Oct 01, 2024 at 08:56:43AM -0600, Shuah Khan wrote:
> > >> On 10/1/24 08:45, Jason A. Donenfeld wrote:
> > >>> On Tue, Oct 01, 2024 at 08:43:05AM -0600, Shuah Khan wrote:
> > >>>> On 9/30/24 21:56, Jason A. Donenfeld wrote:
> > >>>>> This is not stable material and I didn't mark it as such. Do not backport.
> > >>>>
> > >>>> The way selftest work is they just skip if a feature isn't supported.
> > >>>> As such this test should run gracefully on stable releases.
> > >>>>
> > >>>> I would say backport unless and skip if the feature isn't supported.
> > >>>
> > >>> Nonsense. 6.11 never returns ENOSYS from vDSO. This doesn't make sense.
> > >>
> > >> Not sure what you mean by Nonsense. ENOSYS can be used to skip??
> > > 
> > > The branch that this patch adds will never be reached in 6.11 because
> > > the kernel does not have the corresponding code.
> > 
> > What should/would happen if this test is run on a kernel that doesn't
> > support the feature?
> 
> The build system doesn't compile it for kernels without the feature.
> 

That's not how the kselftests should be working.  They can run on any
kernel image (build is separate from running on many test systems), and
so they should just fail with whatever the "feature not present" error
is if the feature isn't present in the system-that-is-being-tested.

thanks,

greg k-h