From 69047b8b02d8f7635fe106daac9e17d72d4a8130 Mon Sep 17 00:00:00 2001

From: Julius Milan <jmilan@redhat.com>

Date: Mon, 10 Apr 2017 16:30:58 +0200

Subject: [PATCH] koops: Improve fatal MCE check when dumping backtrace

Problem was, that regular kernel panic message was treated as fatal MCE,

for which we don't want to have backtrace.

The logic in koops_extract_oopses_from_lines is a bit complex,

so I describe it:

If fatal MCE was recognized, dumping of backtrace was skipped by following

to the next part of cycle (hacky, in the next part of cycle, there is

condition with many allowed strings to continue but not "Call Trace").

The check responsible for this "jump" (kernel.c:396) just checked whether

there is a string beginning with "Kernel panic - not syncing", however

this string appears in all kernel panics, not just fatal MCEs.

Fatal MCE may have form of:

"Kernel panic - not syncing: Fatal Machine check" or

"Kernel panic - not syncing: Machine check" or I found even

"Kernel panic - not syncing: Uncorrected machine check"

so I modified the condition appropriately and added test covering this

case.

Related to rhbz#1374648

---

 src/lib/kernel.c | 2 +-

 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/lib/kernel.c b/src/lib/kernel.c

index 1a9d327..f1a0f2c 100644

--- a/src/lib/kernel.c

+++ b/src/lib/kernel.c

@@ -356,7 +356,7 @@ next_line:

              * In order to capture all these lines, we treat final line

              * as "backtrace" (which is admittedly a hack):

              */

-            if (strstr(curline, "Kernel panic - not syncing"))

+            if (strstr(curline, "Kernel panic - not syncing:") && strcasestr(curline, "Machine check"))

                 inbacktrace = 1;

             else

             if (strnlen(curline, 9) > 8

-- 

1.8.3.1