Re: perf usage of arch/arm64/include/asm/cputype.h

[Leo Yan <leo.yan@arm.com>]

On Wed, Jun 18, 2025 at 09:52:53AM +0100, Mark Rutland wrote:

[...]

> > > Other than that, I think that userspace should just maintain its own
> > > infrastructure, and only pull in things from kernel sources when there's
> > > a specific reason to. Otherwise we're just creating busywork.
> > 
> > I agree with the methodology.
> > 
> > Since Arnaldo is facing build failure when sync headers between kernel
> > and perf tool, to avoid long latency, let us split the refactoriing
> > into separate steps.
> > 
> > As a first step, I think my previous suggestion is valid, we can create a
> > header tools/perf/arch/arm64/include/cputype.h with below code:
> > 
> >   #include "../../../../arch/arm64/include/asm/cputype.h"
> 
> Directly including the kernel header introduces the very fragility that
> having a copy was intended to avoid. NAK to that.

My suggestion is not to include the kernel header, nor to modify the
copy header. :)

Instead, I suggested creating a new header within the perf tool (under
perf's arm64 folder) and then include the copy header in tools:

  tools/arch/arm64/include/asm/cputype.h

> I've replied to the same effect Yicong's patch [1,2].
> 
> If we want to share headers between userspace and kernel, we should
> refactor those headers such that this is safe by construction.
> 
> There is no need to update the userspace headers just because the kernel
> headers have changed, so the simple solution in the short term is to
> suppress the warning from check-headers.sh.

Sure, makes sense for me.

@Arnaldo, as Mark suggested, do you want me to send a patch to remove
cputype.h checking in check-headers.sh or it is fine to keep the warning
until finish the header refactoring?

@Yicong, could you confirm if you proceed to refactor the MIDR? thanks!

Just note, I searched tools folder and found kselftest also uses the
cputype.h header. The refactoring should not break the files below.

$ git grep cputype.h
perf/check-headers.sh:check arch/arm64/include/asm/cputype.h '-I "^#include [<\"]\(asm/\)*sysreg.h"'
perf/util/arm-spe.c:#include "../../arch/arm64/include/asm/cputype.h"
testing/selftests/kvm/arm64/psci_test.c:#include <asm/cputype.h>
testing/selftests/kvm/lib/arm64/vgic.c:#include <asm/cputype.h>

Thanks,
Leo

> [1] https://lore.kernel.org/linux-arm-kernel/dc5afc5c-060c-8bcb-c3a7-0de49a7455fb@huawei.com/T/#m23dfbea6af559f3765d89b9d8427213588871ffd
> [2] https://lore.kernel.org/linux-arm-kernel/dc5afc5c-060c-8bcb-c3a7-0de49a7455fb@huawei.com/T/#m6acbfa00002af8ee791266ea86a58f8f994ed710
> 
> Mark.
> 
> > 
> >   static bool is_perf_midr_in_range_list(u32 midr,
> >                                          struct midr_range const *ranges)
> >   {
> >           while (ranges->model) {
> >                   if (midr_is_cpu_model_range(midr, ranges->model,
> >                                   ranges->rv_min, ranges->rv_max))
> >                           return true;
> >                   ranges++;
> >           }
> > 
> >           return false;
> >   }
> > 
> > Then, once we can generate a dynamic MIDR header file, we can use that
> > header and define the midr_range structure specifically in the perf.
> > In the end, perf can avoid to include kernel's cputype.h.
> > 
> > If no objection, Yicong, do you mind preparing the patch mentioned
> > above? Thanks!
> > 
> > Leo