sched/clock: Print a warning recommending 'tsc=unstable'
authorPeter Zijlstra <peterz@infradead.org>
Fri, 21 Apr 2017 10:52:52 +0000 (12:52 +0200)
committerIngo Molnar <mingo@kernel.org>
Mon, 15 May 2017 08:15:21 +0000 (10:15 +0200)
With our switch to stable delayed until late_initcall(), the most
likely cause of hitting mark_tsc_unstable() is the watchdog. The
watchdog typically only triggers when creative BIOS'es fiddle with the
TSC to hide SMI latency.

Since the watchdog can only detect TSC fiddling after the fact all TSC
clocks (including userspace GTOD) can already have reported funny
values.

The only way to fully avoid this, is manually marking the TSC unstable
at boot. Suggest people do this on their broken systems.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/clock.c

index a2f847c6ada8d65e46b56cbdeeb8cc5b2c47cd4b..1a0d389d2f2b94bc3d221156e2c8de06cef1c436 100644 (file)
@@ -175,6 +175,7 @@ static void __sched_clock_work(struct work_struct *work)
        for_each_possible_cpu(cpu)
                per_cpu(sched_clock_data, cpu) = *scd;
 
+       printk(KERN_WARNING "TSC found unstable after boot, most likely due to broken BIOS. Use 'tsc=unstable'.\n");
        printk(KERN_INFO "sched_clock: Marking unstable (%lld, %lld)<-(%lld, %lld)\n",
                        scd->tick_gtod, __gtod_offset,
                        scd->tick_raw,  __sched_clock_offset);