2009-12-10 16:39:56

by Gernot Hillier

[permalink] [raw]
Subject: [PATCH] [REMINDER] x86: fix warning for unsynchronized TSCs (was: avoid modifying global flag tsc_unstable in unsychronized)

From: Gernot Hillier <[email protected]>

This is a small reminder for a patch I sent some weeks ago.
It was already confirmed by Alok (see
http://article.gmane.org/gmane.linux.kernel/901641). I also
decided to give it a better summary. Here's the original
description from last mail:

The current behavior of unsynchronized_tsc() prevents the warning for
unsychronized TSCs to be printed: tsc_init() calls unsynchronized_tsc()
first to check for such problems; if found, mark_tsc_unstable() is
called. However, the latter bails out immediately as unsynchronized_
tsc() did already set the global flag tsc_unstable. Thus, the intended
warning message "Marking TSC unstable due to TSCs unsynchronized" won't
show up.

This patch changes the behavior for unsynchronized_tsc() to only *check*
for the quirk condition, but not to modify any global flags. If I'm not
mistaken, the only place to *set* the flag shall be mark_tsc_unstable().

Signed-off-by: Gernot Hillier <[email protected]>
Reviewed-by: Jan Kiszka <[email protected]>
Reviewed-by: Wolfgang Mauerer <[email protected]>
Reviewed-by: Alok Kataria <[email protected]>

diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index cd982f4..ffbb786 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -836,10 +836,10 @@ __cpuinit int unsynchronized_tsc(void)
if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL) {
/* assume multi socket systems are not synchronized: */
if (num_possible_cpus() > 1)
- tsc_unstable = 1;
+ return 1;
}

- return tsc_unstable;
+ return 0;
}

static void __init init_tsc_clocksource(void)