mce: Panic when a core has reached a timeout
authorBorislav Petkov <bp@suse.de>
Fri, 23 May 2014 09:06:35 +0000 (11:06 +0200)
committerBorislav Petkov <bp@suse.de>
Fri, 30 May 2014 20:05:31 +0000 (22:05 +0200)
There is very little and maybe practically nothing we can do to recover
from a system where at least one core has reached a timeout during the
whole monarch cores gathering. So panic when that happens.

Link: http://lkml.kernel.org/r/20140523091041.GA21332@pd.tnic
Signed-off-by: Borislav Petkov <bp@suse.de>
arch/x86/kernel/cpu/mcheck/mce.c

index 0078761219a2418a691dd96db5780cbe7742ef86..6cc800381d1462867a6ae425d5eb60710570ab7f 100644 (file)
@@ -704,8 +704,7 @@ static int mce_timed_out(u64 *t)
        if (!mca_cfg.monarch_timeout)
                goto out;
        if ((s64)*t < SPINUNIT) {
-               /* CHECKME: Make panic default for 1 too? */
-               if (mca_cfg.tolerant < 1)
+               if (mca_cfg.tolerant <= 1)
                        mce_panic("Timeout synchronizing machine check over CPUs",
                                  NULL, NULL);
                cpu_missing = 1;