Re: [PATCH] PMCR_EL0.N is RAZ/WI. At least a build failes in Ubuntu 22.04 LTS. Remove the set function.

[Marc Zyngier <maz@kernel.org>]

On Mon, 15 Sep 2025 22:31:31 +0100,
Itaru Kitayama <itaru.kitayama@linux.dev> wrote:
> 
> 
> 
> > On Sep 12, 2025, at 21:11, Marc Zyngier <maz@kernel.org> wrote:
> > 
> > On Fri, 12 Sep 2025 12:33:39 +0100,
> > Itaru Kitayama <itaru.kitayama@gmail.com> wrote:
> >> 
> >> 
> >> 
> >>> On Sep 12, 2025, at 20:01, Marc Zyngier <maz@kernel.org> wrote:
> >>> 
> >>> On Fri, 12 Sep 2025 09:27:40 +0100,
> >>> Itaru Kitayama <itaru.kitayama@linux.dev> wrote:
> >>>> 
> >>>> Signed-off-by: Itaru Kitayama <itaru.kitayama@fujitsu.com>
> >>> 
> >>> This isn't an acceptable commit message.
> >>> 
> >>>> ---
> >>>> Seen a build failure with old Ubuntu 22.04 LTS, while the latest release
> >>>> has no build issue, a write to the bit fields is RAZ/WI, remove the
> >>>> function.
> >>>> ---
> >>>> tools/testing/selftests/kvm/arm64/vpmu_counter_access.c | 6 ------
> >>>> 1 file changed, 6 deletions(-)
> >>>> 
> >>>> diff --git a/tools/testing/selftests/kvm/arm64/vpmu_counter_access.c b/tools/testing/selftests/kvm/arm64/vpmu_counter_access.c
> >>>> index f16b3b27e32ed7ca57481f27d689d47783aa0345..56214a4430be90b3e1d840f2719b22dd44f0b49b 100644
> >>>> --- a/tools/testing/selftests/kvm/arm64/vpmu_counter_access.c
> >>>> +++ b/tools/testing/selftests/kvm/arm64/vpmu_counter_access.c
> >>>> @@ -45,11 +45,6 @@ static uint64_t get_pmcr_n(uint64_t pmcr)
> >>>>   return FIELD_GET(ARMV8_PMU_PMCR_N, pmcr);
> >>>> }
> >>>> 
> >>>> -static void set_pmcr_n(uint64_t *pmcr, uint64_t pmcr_n)
> >>>> -{
> >>>> -    u64p_replace_bits((__u64 *) pmcr, pmcr_n, ARMV8_PMU_PMCR_N);
> >>>> -}
> >>>> -
> >>>> static uint64_t get_counters_mask(uint64_t n)
> >>>> {
> >>>>   uint64_t mask = BIT(ARMV8_PMU_CYCLE_IDX);
> >>>> @@ -490,7 +485,6 @@ static void test_create_vpmu_vm_with_pmcr_n(uint64_t pmcr_n, bool expect_fail)
> >>>>    * Setting a larger value of PMCR.N should not modify the field, and
> >>>>    * return a success.
> >>>>    */
> >>>> -    set_pmcr_n(&pmcr, pmcr_n);
> >>>>   vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_PMCR_EL0), pmcr);
> >>>>   pmcr = vcpu_get_reg(vcpu, KVM_ARM64_SYS_REG(SYS_PMCR_EL0));
> >>>> 
> >>>> 
> >>> 
> >>> So what are you fixing here? A build failure? A semantic defect?
> >>> Something else? What makes this a valid change?
> >>> 
> >>> Frankly, I have no idea.
> >>> 
> >>> But KVM definitely allows PMCR_EL0.N to be written from userspace, and
> >>> that's not going to change.
> >>> 
> >> 
> >> Then I’ll drop this patch.
> > 
> > I'm not asking you to drop it, I'm asking you to explain. If you found
> > a problem, let's discuss it and fix it. But as it stands, you're not
> > giving me much to go on.
> > 
> 
> You are right, while the bit fields are write ignored, to be
> consistent with the handling of other bit fields of the register,
> I’m fully convinced that checking the write operation in the
> vpmu_counter_access.c file should be kept.

The bit field is *not* ignored when written from userspace. That's how
we configure the PMU if the guest runs at EL1.

> The build error I’ve seen with Ubuntu 22.04 LTS is below:

[snip]

Can you please detail what compiler version this is? I'm unlikely to
install an ancient version of Ubuntu, but I can pick the corresponding
compiler version.

	M.

-- 
Without deviation from the norm, progress is not possible.