Re: [PATCH] clocksource/drivers/hyper-v: Re-enable VDSO_CLOCKMODE_HVCLOCK on X86

From: Vitaly Kuznetsov <vkuznets@redhat.com>
To: Thomas Gleixner <tglx@linutronix.de>, linux-hyperv@vger.kernel.org
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>,
	Michael Kelley <mikelley@microsoft.com>,
	Mohammed Gamal <mgamal@redhat.com>, Wei Liu <wei.liu@kernel.org>,
	"K. Y. Srinivasan" <kys@microsoft.com>,
	Haiyang Zhang <haiyangz@microsoft.com>,
	Dexuan Cui <decui@microsoft.com>,
	Stephen Hemminger <sthemmin@microsoft.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] clocksource/drivers/hyper-v: Re-enable VDSO_CLOCKMODE_HVCLOCK on X86
Date: Thu, 13 May 2021 09:18:37 +0200	[thread overview]
Message-ID: <8735urw182.fsf@vitty.brq.redhat.com> (raw)
In-Reply-To: <87tun766kv.ffs@nanos.tec.linutronix.de>

Thomas Gleixner <tglx@linutronix.de> writes:

> On Wed, May 12 2021 at 10:46, Vitaly Kuznetsov wrote:
>> Mohammed reports (https://bugzilla.kernel.org/show_bug.cgi?id=213029)
>> the commit e4ab4658f1cf ("clocksource/drivers/hyper-v: Handle vDSO
>> differences inline") broke vDSO on x86. The problem appears to be that
>> VDSO_CLOCKMODE_HVCLOCK is an enum value in 'enum vdso_clock_mode' and
>> '#ifdef VDSO_CLOCKMODE_HVCLOCK' branch evaluates to false (it is not
>> a define). Replace it with CONFIG_X86 as it is the only arch which
>> has this mode currently.
>>
>> Reported-by: Mohammed Gamal <mgamal@redhat.com>
>> Fixes: e4ab4658f1cf ("clocksource/drivers/hyper-v: Handle vDSO differences inline")
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>> ---
>>  drivers/clocksource/hyperv_timer.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/clocksource/hyperv_timer.c b/drivers/clocksource/hyperv_timer.c
>> index 977fd05ac35f..e17421f5e47d 100644
>> --- a/drivers/clocksource/hyperv_timer.c
>> +++ b/drivers/clocksource/hyperv_timer.c
>> @@ -419,7 +419,7 @@ static void resume_hv_clock_tsc(struct clocksource *arg)
>>  	hv_set_register(HV_REGISTER_REFERENCE_TSC, tsc_msr);
>>  }
>>  
>> -#ifdef VDSO_CLOCKMODE_HVCLOCK
>> +#ifdef CONFIG_X86
>>  static int hv_cs_enable(struct clocksource *cs)
>>  {
>>  	vclocks_set_used(VDSO_CLOCKMODE_HVCLOCK);
>> @@ -435,7 +435,7 @@ static struct clocksource hyperv_cs_tsc = {
>>  	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
>>  	.suspend= suspend_hv_clock_tsc,
>>  	.resume	= resume_hv_clock_tsc,
>> -#ifdef VDSO_CLOCKMODE_HVCLOCK
>> +#ifdef CONFIG_X86
>>  	.enable = hv_cs_enable,
>>  	.vdso_clock_mode = VDSO_CLOCKMODE_HVCLOCK,
>>  #else
>
> That's lame as it needs to be patched differently once ARM64 gains
> support. What about the below?
>

The solution I liked the most was Michael's: no need to do anything
except for adding VDSO_CLOCKMODE_HVCLOCK to the enum. Too bad it didn't
work)

You proposal seems to be slightly better than mine: when adding
VDSO_CLOCKMODE_HVCLOCK to ARM the change will be limited to
drivers/clocksource/hyperv_timer.c will stay intact.

I'll send v2 shortly, thanks!

> Thanks,
>
>         tglx
> ---
> --- a/arch/x86/include/asm/vdso/clocksource.h
> +++ b/arch/x86/include/asm/vdso/clocksource.h
> @@ -7,4 +7,6 @@
>  	VDSO_CLOCKMODE_PVCLOCK,	\
>  	VDSO_CLOCKMODE_HVCLOCK
>  
> +#define HAVE_VDSO_CLOCKMODE_HVCLOCK
> +
>  #endif /* __ASM_VDSO_CLOCKSOURCE_H */
> --- a/drivers/clocksource/hyperv_timer.c
> +++ b/drivers/clocksource/hyperv_timer.c
> @@ -419,7 +419,7 @@ static void resume_hv_clock_tsc(struct c
>  	hv_set_register(HV_REGISTER_REFERENCE_TSC, tsc_msr);
>  }
>  
> -#ifdef VDSO_CLOCKMODE_HVCLOCK
> +#ifdef HAVE_VDSO_CLOCKMODE_HVCLOCK
>  static int hv_cs_enable(struct clocksource *cs)
>  {
>  	vclocks_set_used(VDSO_CLOCKMODE_HVCLOCK);
> @@ -435,7 +435,7 @@ static struct clocksource hyperv_cs_tsc
>  	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
>  	.suspend= suspend_hv_clock_tsc,
>  	.resume	= resume_hv_clock_tsc,
> -#ifdef VDSO_CLOCKMODE_HVCLOCK
> +#ifdef HAVE_VDSO_CLOCKMODE_HVCLOCK
>  	.enable = hv_cs_enable,
>  	.vdso_clock_mode = VDSO_CLOCKMODE_HVCLOCK,
>  #else
>

-- 
Vitaly