Re: [Xen-devel] [PATCH v2 3/6] x86/time: implement tsc as clocksource

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 29.03.16 at 15:44, <joao.m.martins@xxxxxxxxxx> wrote:
> Introduce support for using TSC as platform time which is the highest
> resolution time and most performant to get (~20 nsecs).  Though there
> are also several problems associated with its usage, and there isn't a
> complete (and architecturally defined) guarantee that all machines
> will provide reliable and monotonic TSC across all CPUs, on different
> sockets and different P/C states.  I believe Intel to be the only that
> can guarantee that. For this reason it's only set when adminstrator
> changes "clocksource" boot option to "tsc". Initializing TSC
> clocksource requires all CPUs to have the tsc reliability checks
> performed. init_xen_time is called before all CPUs are up, and so we
> start with HPET at boot time, and switch later to TSC.

Please don't make possibly wrong statements: There's no guarantee
we'd start with HPET - might as well be PMTMR or PIT.

> The switch then
> happens on verify_tsc_reliability initcall that is invoked when all
> CPUs are up. When attempting to initializing TSC we also check for
> time warps and appropriate CPU features i.e.  TSC_RELIABLE,
> CONSTANT_TSC and NONSTOP_TSC. And in case none of these conditions are
> met, we keep the clocksource that was previously initialized on
> init_xen_time.

DYM "And in case any of these conditions is not met, ..."? Or,
looking at the code, something more complex?

> --- a/xen/arch/x86/time.c
> +++ b/xen/arch/x86/time.c
> @@ -432,6 +432,63 @@ uint64_t ns_to_acpi_pm_tick(uint64_t ns)
>  }
>  
>  /************************************************************
> + * PLATFORM TIMER 4: TSC
> + */
> +static u64 tsc_freq;
> +static unsigned long tsc_max_warp;
> +static void tsc_check_reliability(void);
> +
> +static int __init init_tsctimer(struct platform_timesource *pts)
> +{
> +    bool_t tsc_reliable = 0;
> +
> +    tsc_check_reliability();
> +
> +    if ( tsc_max_warp > 0 )
> +    {
> +        tsc_reliable = 0;

Redundant assignment.

> +static void resume_tsctimer(struct platform_timesource *pts)
> +{
> +}

No need for an empty function - other clock sources don't have
such empty stubs either (i.e. the caller checks for NULL).

> @@ -541,6 +613,10 @@ static int __init try_platform_timer(struct 
> platform_timesource *pts)
>      if ( rc <= 0 )
>          return rc;
>  
> +    /* We have a platform timesource already so reset it */
> +    if ( plt_src.counter_bits != 0 )
> +        reset_platform_timer();

What if any of the time functions get used between here and the
setting of the new clock source? For example, what will happen to
log messages when "console_timestamps=..." is in effect?

> @@ -566,7 +642,9 @@ static void __init init_platform_timer(void)
>      struct platform_timesource *pts = NULL;
>      int i, rc = -1;
>  
> -    if ( opt_clocksource[0] != '\0' )
> +    /* clocksource=tsc is initialized later when all CPUS are up */
> +    if ( (opt_clocksource[0] != '\0') &&
> +         (strcmp(opt_clocksource, "tsc") != 0) )

In line with the inverse check done below (using ! instead of == 0)
please omit the != 0 here.

> @@ -1437,6 +1515,20 @@ static int __init verify_tsc_reliability(void)
>          }
>      }
>  
> +    if ( !strcmp(opt_clocksource, "tsc") )
> +    {
> +        if ( try_platform_timer(&plt_tsc) > 0 )
> +        {
> +            printk(XENLOG_INFO "Switched to Platform timer %s TSC\n",
> +                   freq_string(plt_src.frequency));
> +
> +            init_percpu_time();

So you re-do this for the BP, but not for any of the APs?

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel