From e30c24a5572c33f9ca5157bfb4e504897b1bb7c9 Mon Sep 17 00:00:00 2001

From: Denys Vlasenko <dvlasenk@redhat.com>

Date: Mon, 6 Jan 2014 16:04:37 +0100

Subject: [ABRT PATCH 26/27] MCE: cover cases where kernel version isn't

 detected on Fedora 20.

With this change, both fata and non-fatal MCEs are caught on default

Fedora 20 installation.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>

Related to rhbz#1032077

Signed-off-by: Jakub Filak <jfilak@redhat.com>

---

 doc/MCE_readme.txt            |  9 ++++++++-

 src/lib/kernel.c              |  2 +-

 src/plugins/abrt-dump-oops.c  |  3 ++-

 src/plugins/koops_event.conf  | 11 +++++++++++

 src/plugins/vmcore_event.conf | 14 ++++++++++++--

 5 files changed, 34 insertions(+), 5 deletions(-)

diff --git a/doc/MCE_readme.txt b/doc/MCE_readme.txt

index ed5b627..5dff636 100644

--- a/doc/MCE_readme.txt

+++ b/doc/MCE_readme.txt

@@ -70,7 +70,7 @@ echo "Exitcode:$?"

 It requires files which describe MCE to simulate. I grabbed a few examples

 from mce-test.tar.gz (source tarball of mce-test project).

-I used this this file to cause a non-fatal MCE:

+I used this file to cause a non-fatal MCE:

 CPU 0 BANK 2

 STATUS VAL OVER EN

@@ -84,3 +84,10 @@ RIP 12343434

 MISC 11

 (Not sure what failures exactly they imitate, maybe there are better examples).

+

+

+For testing fatal MCEs you need to set up kdump. Mini-recipe:

+(1) yum install --enablerepo='*debuginfo*' kexec-tools crash kernel-debuginfo

+(2) add "crashkernel=128M" to the kernel's command line, reboot

+(3) before injecting fatal MCE, start kdump service:

+    systemctl start kdump.service

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

index 340ec39..ad20c65 100644

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

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

@@ -66,7 +66,7 @@ static void record_oops(GList **oops_list, struct line_info* lines_info, int oop

         {

             *oops_list = g_list_append(

                         *oops_list,

-                        xasprintf("%s\n%s", (version ? version : "undefined"), oops)

+                        xasprintf("%s\n%s", (version ? version : ""), oops)

             );

         }

         else

diff --git a/src/plugins/abrt-dump-oops.c b/src/plugins/abrt-dump-oops.c

index 5e33f0a..12291be 100644

--- a/src/plugins/abrt-dump-oops.c

+++ b/src/plugins/abrt-dump-oops.c

@@ -115,7 +115,8 @@ static void save_oops_data_in_dump_dir(struct dump_dir *dd, char *oops, const ch

     char *second_line = (char*)strchr(first_line, '\n'); /* never NULL */

     *second_line++ = '\0';

-    dd_save_text(dd, FILENAME_KERNEL, first_line);

+    if (first_line[0])

+        dd_save_text(dd, FILENAME_KERNEL, first_line);

     dd_save_text(dd, FILENAME_BACKTRACE, second_line);

     /* check if trace doesn't have line: 'Your BIOS is broken' */

diff --git a/src/plugins/koops_event.conf b/src/plugins/koops_event.conf

index 37a79a9..b1472ce 100644

--- a/src/plugins/koops_event.conf

+++ b/src/plugins/koops_event.conf

@@ -3,6 +3,17 @@ EVENT=post-create analyzer=Kerneloops

         # >> instead of > is due to bugzilla.redhat.com/show_bug.cgi?id=854266

         abrt-action-analyze-oops &&

         dmesg >>dmesg &&

+        {

+        # action-analyze-oops tries to save kernel version,

+        # but for some oopses it can't do that (e.g. MCEs).

+        # If it failed, try to extract version from dmesg:

+        test -f kernel ||

+            {

+            k=`sed -n '/Linux version/ s/.*Linux version \([^ ]*\) .*/\1/p' dmesg | tail -n1`

+            test "$k" != "" && printf "%s" "$k" >kernel

+            true   # ignore possible failures in previous command

+            }

+        } &&

         abrt-action-save-kernel-data &&

         # Do not fail the event (->do not delete problem dir)

         # if check-oops-for-hw-error exits nonzero:

diff --git a/src/plugins/vmcore_event.conf b/src/plugins/vmcore_event.conf

index 655d842..a525ec7 100644

--- a/src/plugins/vmcore_event.conf

+++ b/src/plugins/vmcore_event.conf

@@ -1,6 +1,7 @@

 # analyze

 EVENT=analyze_VMcore analyzer=vmcore

         # If kdump machinery already extracted dmesg...

+        (

         if test -f vmcore-dmesg.txt; then

             # ...use that

             abrt-dump-oops -o vmcore-dmesg.txt >backtrace || exit $?

@@ -15,8 +16,17 @@ EVENT=analyze_VMcore analyzer=vmcore

             test "$k" != "" && printf "%s" "$k" >kernel

         else

             # No vmcore-dmesg.txt, do it the hard way:

-            abrt-action-analyze-vmcore

-        fi &&

+            abrt-action-analyze-vmcore || exit $?

+            #

+            # Does "kernel" element exist?

+            test -f kernel && exit 0

+            #

+            # Try creating it from dmesg_log (created by abrt-action-analyze-vmcore):

+            test -f dmesg_log || exit 0

+            k=`sed -n '/Linux version/ s/.*Linux version \([^ ]*\) .*/\1/p' dmesg_log | tail -n1`

+            test "$k" != "" && printf "%s" "$k" >kernel

+        fi

+        ) &&

         abrt-action-analyze-oops &&

         abrt-action-save-kernel-data

-- 

1.8.3.1