Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Feng Tang <feng.tang@intel.com>
To:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@intel.com>,
        "H . Peter Anvin" <hpa@zytor.com>,
        Peter Zijlstra <peterz@infradead.org>, x86@kernel.org,
        linux-kernel@vger.kernel.org, paulmck@kernel.org
Cc:     rui.zhang@intel.com, Feng Tang <feng.tang@intel.com>,
        Yu Liao <liaoyu15@huawei.com>
Subject: [PATCH] x86/tsc: Extend watchdog check exemption to 4-Sockets platform
Date:   Wed,  7 Jun 2023 15:54:33 +0800
Message-Id: <20230607075433.387075-1-feng.tang@intel.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

There were reports again that the tsc clocksource on 4 sockets x86
servers was wrongly judged as 'unstable' by 'jiffies' and other
watchdogs, and disabled [1][2].

Commit b50db7095fe0 ("x86/tsc: Disable clocksource watchdog for TSC
on qualified platorms") was introduce to deal with these false
alarms of tsc unstable issues, covering qualified platforms for 2
sockets or smaller ones. And from history of chasing TSC issues,
Thomas and Peter only saw real TSC synchronization issue on 8 socket
machines.

So extend the exemption to 4 sockets to fix the issue.

Rui also proposed another way to disable 'jiffies' as clocksource
watchdog [3], which can also solve problem in [1]. in an architecture
independent way, but can't cure the problem in [2]. whose watchdog
is HPET or PMTIMER, while 'jiffies' is mostly used as watchdog in
boot phase.

'nr_online_nodes' has known inaccurate problem for cases like
platform with cpu-less memory nodes, sub numa cluster enabled,
fakenuma, kernel cmdline parameter 'maxcpus=', etc. The harmful case
is the 'maxcpus' one which could possibly under estimates the package
number, and disable the watchdog, but bright side is it is mostly
for debug usage. All these will be addressed in other patches, as
discussed in thread [4].

[1]. https://lore.kernel.org/all/9d3bf570-3108-0336-9c52-9bee15767d29@huawei.com/
[2]. https://lore.kernel.org/lkml/06df410c-2177-4671-832f-339cff05b1d9@paulmck-laptop/
[3]. https://lore.kernel.org/all/bd5b97f89ab2887543fc262348d1c7cafcaae536.camel@intel.com/
[4]. https://lore.kernel.org/all/20221021062131.1826810-1-feng.tang@intel.com/

Reported-by: Yu Liao <liaoyu15@huawei.com>
Reported-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Feng Tang <feng.tang@intel.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
---
 arch/x86/kernel/tsc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index 344698852146..f15066a1d473 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -1238,7 +1238,7 @@ static void __init check_system_tsc_reliable(void)
 	if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC) &&
 	    boot_cpu_has(X86_FEATURE_NONSTOP_TSC) &&
 	    boot_cpu_has(X86_FEATURE_TSC_ADJUST) &&
-	    nr_online_nodes <= 2)
+	    nr_online_nodes <= 4)
 		tsc_disable_clocksource_watchdog();
 }
 
-- 
2.34.1