RE: [PATCH 08/10] clocksource/drivers/hyper-v: Handle sched_clock differences inline

[Michael Kelley <mikelley@microsoft.com>]

From: Wei Liu <wei.liu@kernel.org> Sent: Monday, February 1, 2021 10:55 AM
> 
> On Wed, Jan 27, 2021 at 12:23:43PM -0800, Michael Kelley wrote:
> [...]
> > +/*
> > + * Reference to pv_ops must be inline so objtool
> > + * detection of noinstr violations can work correctly.
> > + */
> > +static __always_inline void hv_setup_sched_clock(void *sched_clock)
> 
> sched_clock_register is not trivial. Having __always_inline here is
> going to make the compiled object bloated.
> 
> Given this is a static function, I don't think we need to specify any
> inline keyword. The compiler should be able to determine whether this
> function should be inlined all by itself.
> 
> Wei.

There was an explicit request from Peter Zijlstra and Thomas Gleixner
to force it inline.  See https://lore.kernel.org/patchwork/patch/1283635/ and
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/x86/include/asm/mshyperv.h?id=b9d8cf2eb3ceecdee3434b87763492aee9e28845

Michael

> 
> > +{
> > +#ifdef CONFIG_GENERIC_SCHED_CLOCK
> > +	/*
> > +	 * We're on an architecture with generic sched clock (not x86/x64).
> > +	 * The Hyper-V sched clock read function returns nanoseconds, not
> > +	 * the normal 100ns units of the Hyper-V synthetic clock.
> > +	 */
> > +	sched_clock_register(sched_clock, 64, NSEC_PER_SEC);
> > +#else
> > +#ifdef CONFIG_PARAVIRT
> > +	/* We're on x86/x64 *and* using PV ops */
> > +	pv_ops.time.sched_clock = sched_clock;
> > +#endif
> > +#endif
> > +}
> > +
> >  static bool __init hv_init_tsc_clocksource(void)
> >  {
> >  	u64		tsc_msr;
> > --
> > 1.8.3.1
> >