@@ -6318,7 +6318,6 @@
as the stability checks done at bootup. Used to enable
high-resolution timer mode on older hardware, and in
virtualized environment.
- [x86] unreliable: mark tsc clocksource as unreliable.
[x86] noirqtime: Do not use TSC to do irq accounting.
Used to run time disable IRQ_TIME_ACCOUNTING on any
platforms where RDTSC is slow and this accounting
@@ -46,7 +46,7 @@ static unsigned int __initdata tsc_early_khz;
static DEFINE_STATIC_KEY_FALSE(__use_tsc);
-int tsc_clocksource_reliable = 1;
+int tsc_clocksource_reliable;
static u32 art_to_tsc_numerator;
static u32 art_to_tsc_denominator;
@@ -297,8 +297,6 @@ static int __init tsc_setup(char *str)
{
if (!strcmp(str, "reliable"))
tsc_clocksource_reliable = 1;
- if (!strcmp(str, "unreliable"))
- tsc_clocksource_reliable = 0;
if (!strncmp(str, "noirqtime", 9))
no_sched_irq_time = 1;
if (!strcmp(str, "unstable"))
This reverts commit da9eb97ad0c58eaee0358f17d5617ee4cbdf8f32. This one is slightly tricky, for most systems this has no affect as TSC_CONSTANT feature is available and thus cores are not assumed to be out of sync at startup. The watchdog remains in place, and the cores are adjusted periodically, which helps after the host is resumed from suspend. Reverting this patch does not show any averse systemd-analyse boot speed impact, probably due to availability of TSC, CONSTANT_TSC TSC_ADJUST features. If we really want to keep this behaviour, it is probably best to simply setup_force_cpu_cap(X86_FEATURE_TSC_RELIABLE) the way VMWare and HyperV do, in the kvm.c guest cpu setup. Signed-off-by: Dimitri John Ledkov <dimitri.ledkov@canonical.com> --- Documentation/admin-guide/kernel-parameters.txt | 1 - arch/x86/kernel/tsc.c | 4 +--- 2 files changed, 1 insertion(+), 4 deletions(-)