ct-bugcheck: Add tools to poll for and report ath10k firmware crashes.

This tool can periodically check for ath10k firmware crashes.
If it finds a crash, it will package up the binary crash dump,
some OS level things like dmesg, lspci, etc into a tar file.

It then notifies the user about the crash and asks them to report
the bug to the appropriate email address.

This is most useful when used with ath10k-ct driver and
CT ath10k firmware, but it should also report issues with stock
ath10k driver and firmware in case one has appropriate contacts
to debug them.

This tool could be extended later for other modules/bugs/etc.

Signed-off-by: Ben Greear <greearb@candelatech.com>

3 files changed, 160 insertions, 0 deletions

diff --git a/package/utils/ct-bugcheck/src/bugcheck.initd b/package/utils/ct-bugcheck/src/bugcheck.initd
new file mode 100644
index 0000000000..b97a415bbf
--- /dev/null
+++ b/package/utils/ct-bugcheck/src/bugcheck.initd
@@ -0,0 +1,16 @@
+#!/bin/sh /etc/rc.common
+# Copyright (C) 2016 OpenWrt.org
+
+START=99
+
+USE_PROCD=1
+PROG=/usr/bin/bugchecker.sh
+
+# To actually make bugchecker.sh run, see comments
+# at top of its file.
+
+start_service () {
+        procd_open_instance
+        procd_set_param command "$PROG"
+        procd_close_instance
+}
diff --git a/package/utils/ct-bugcheck/src/bugcheck.sh b/package/utils/ct-bugcheck/src/bugcheck.sh
new file mode 100755
index 0000000000..85f70c5e79
--- /dev/null
+++ b/package/utils/ct-bugcheck/src/bugcheck.sh
@@ -0,0 +1,115 @@
+#!/bin/sh
+
+# Check for ath10k (and maybe other) bugs, package them up,
+# and let user know what to do with them.
+
+TMPLOC=/tmp
+CRASHDIR=$TMPLOC/bugcheck
+FOUND_BUG=0
+
+# set -x
+
+bugcheck_generic()
+{
+    echo "LEDE crashlog report" > $CRASHDIR/info.txt
+    date >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "uname" >> $CRASHDIR/info.txt
+    uname -a >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "os-release" >> $CRASHDIR/info.txt
+    cat /etc/os-release >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "os-release" >> $CRASHDIR/info.txt
+    cat /etc/os-release >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "dmesg output" >> $CRASHDIR/info.txt
+    dmesg >> $CRASHDIR/info.txt
+    if [ -x /usr/bin/lspci ]
+	then
+	echo >> $CRASHDIR/info.txt
+	echo "lspci" >> $CRASHDIR/info.txt
+	lspci >> $CRASHDIR/info.txt
+    fi
+    echo >> $CRASHDIR/info.txt
+    echo "cpuinfo" >> $CRASHDIR/info.txt
+    cat /proc/cpuinfo >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "meminfo" >> $CRASHDIR/info.txt
+    cat /proc/cpuinfo >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "cmdline" >> $CRASHDIR/info.txt
+    cat /proc/cmdline >> $CRASHDIR/info.txt
+    echo >> $CRASHDIR/info.txt
+    echo "lsmod" >> $CRASHDIR/info.txt
+    lsmod >> $CRASHDIR/info.txt
+}
+
+roll_crashes()
+{
+    # Roll any existing crashes
+    if [ -d $CRASHDIR ]
+	then
+	if [ -d $CRASHDIR.1 ]
+	    then
+	    rm -fr $CRASHDIR.1
+	fi
+	mv $CRASHDIR $CRASHDIR.1
+    fi
+
+    # Prepare location
+    mkdir -p $CRASHDIR
+}
+
+# ath10k, check debugfs entries.
+for i in /sys/kernel/debug/ieee80211/*/ath10k/fw_crash_dump
+do
+  #echo "Checking $i"
+  if cat $i > $TMPLOC/ath10k_crash.bin 2>&1
+      then
+      FOUND_BUG=1
+
+      #echo "Found ath10k crash data in $i"
+      roll_crashes
+
+      ADIR=${i/fw_crash_dump/}
+
+      CTFW=0
+      if grep -- -ct- $TMPLOC/ath10k_crash.bin > /dev/null 2>&1
+	  then
+	  CTFW=1
+      fi
+
+      echo "Send bug reports to:" > $CRASHDIR/report_to.txt
+      if [ -f $ADIR/ct_special -o $CTFW == "1" ]
+	  then
+	  # Looks like this is CT firmware or driver...
+	  echo "greearb@candelatech.com" >> $CRASHDIR/report_to.txt
+	  echo "and/or report or check for duplicates here:" >> $CRASHDIR/report_to.txt
+	  echo "https://github.com/greearb/ath10k-ct/issues" >> $CRASHDIR/report_to.txt
+      else
+	  # Not sure who would want these bug reports for upstream...
+	  echo "https://www.lede-project.org/" >> $CRASHDIR/report_to.txt
+      fi
+      echo >> $CRASHDIR/report_to.txt
+      echo "Please attach all files in this directory to bug reports." >> $CRASHDIR/report_to.txt
+
+      mv $TMPLOC/ath10k_crash.bin $CRASHDIR
+
+      # Add any more ath10k specific stuff here.
+
+      # And call generic bug reporting logic
+      bugcheck_generic
+  fi
+done
+
+if [ $FOUND_BUG == "1" ]
+    then
+    # Notify LUCI somehow?
+    echo "bugcheck.sh found an issue to be reported" > /dev/kmsg
+    echo "See $CRASHDIR for details on how to report this" > /dev/kmsg
+    # Let calling code know something was wrong.
+    exit 1
+fi
+
+exit 0
diff --git a/package/utils/ct-bugcheck/src/bugchecker.sh b/package/utils/ct-bugcheck/src/bugchecker.sh
new file mode 100755
index 0000000000..be305af390
--- /dev/null
+++ b/package/utils/ct-bugcheck/src/bugchecker.sh
@@ -0,0 +1,29 @@
+#!/bin/sh
+
+# Periodically call bugcheck.sh script
+
+CHECKER=bugcheck.sh
+SLEEPFOR=60
+
+DO_BUGCHECK=0
+
+# So, to enable this, you create an /etc/config/bugcheck file
+# with contents like:
+#  DO_BUGCHECK=1
+#  export DO_BUGCHECK
+
+if [ -f /etc/config/bugcheck ]
+    then
+    . /etc/config/bugcheck
+fi
+
+if [ $DO_BUGCHECK == 0 ]
+then
+    exit 0
+fi
+
+while true
+  do
+  $CHECKER
+  sleep $SLEEPFOR
+done