[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20191002141553.7682-1-mzhivich@akamai.com>
Date: Wed, 2 Oct 2019 10:15:53 -0400
From: Michael Zhivich <mzhivich@...mai.com>
To: <linux-kernel@...r.kernel.org>
CC: <tglx@...utronix.de>, <mingo@...hat.com>, <bp@...en8.de>,
<hpa@...or.com>, <rafael.j.wysocki@...el.com>,
Michael Zhivich <mzhivich@...mai.com>
Subject: [PATCH] clocksource: tsc: respect tsc bootparam for clocksource_tsc_early
Introduction of clocksource_tsc_early broke functionality of "tsc=reliable"
and "tsc=nowatchdog" boot params, since clocksource_tsc_early is *always*
registered with CLOCK_SOURCE_MUST_VERIFY and thus put on the watchdog list.
If CPU is very busy during boot, the watchdog clocksource may not be
read frequently enough, resulting in a large difference that is treated as
"negative" by clocksource_delta() and incorrectly truncated to 0.
This false alarm causes TSC to be declared unstable:
clocksource: timekeeping watchdog on CPU1: Marking clocksource
'tsc-early' as unstable because the skew is too large:
clocksource: 'refined-jiffies' wd_now: fffb7019 wd_last: fffb6e28
mask: ffffffff
clocksource: 'tsc-early' cs_now: 52c3918b89d6 cs_last: 52c31d736d2e
mask: ffffffffffffffff
tsc: Marking TSC unstable due to clocksource watchdog
Work around this issue by respecting "tsc=reliable" or "tsc=nowatchdog"
boot params when registering clocksource_tsc_early.
Signed-off-by: Michael Zhivich <mzhivich@...mai.com>
---
arch/x86/kernel/tsc.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index c59454c382fd..7e322e2daaf5 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -1505,6 +1505,9 @@ void __init tsc_init(void)
return;
}
+ if (tsc_clocksource_reliable || no_tsc_watchdog)
+ clocksource_tsc_early.flags &= ~CLOCK_SOURCE_MUST_VERIFY;
+
clocksource_register_khz(&clocksource_tsc_early, tsc_khz);
detect_art();
}
--
2.17.1
Powered by blists - more mailing lists