add an upstart script: vpd-log.conf

vpd-log.conf

Index: vpd-log.conf
diff --git a/vpd-log.conf b/vpd-log.conf
new file mode 100644
index 0000000000000000000000000000000000000000..e58e1fca0464f121a68e0aaccd08c739df6109ed
--- /dev/null
+++ b/vpd-log.conf
@@ -0,0 +1,30 @@
+# Copyright (c) 2011 The Chromium OS Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+# vpd-log
+#
+# when boot reaches login prompt, dumps VPD information from RO and RW
+# partitions.

[jrbarnette, 2011/02/24 17:34:17]

I've got upcoming changes to try and standardize the style of the
comment header here.  This is the style I want to encourage:

    description     "... brief description ..."
    author          "chromium-os-dev@chromium.org"

    # Description of the purpose of the job, and any non-obvious
    # reasons behind the start and stop conditions.  This shouldn't
    # just repeat the 'description' stanza; if that stanza says
    # everything, leave out these comments.
    start on stopped boot-complete

I'd rather not see phrases that repeat the obvious, like "when boot
reaches login prompt, ...":  The 'start on' stanza makes that clear,
so the comment is too much like "a++; /* increment a by 1 */".
Important things to call out regarding the start conditions would
be if there's some important reason for starting after login other
than the conventional performance rationale, like a hidden
dependency.

[Louis, 2011/02/25 08:50:12]

Done.

+start on stopped boot-complete
+
+script
+  VPD_2_0="/var/log/vpd_2.0.txt"
+  TMP="/tmp/vpd.bin"
+
+  # Since we don't change the VPD data too often, only create the log file
+  # if it does not exist.
+  if [ ! -f "$VPD_2_0" ]; then
+    touch $VPD_2_0

[jrbarnette, 2011/02/24 17:34:17]

This is unnecessary; redirection onto the file below will create it.

[Louis, 2011/02/25 08:50:12]

Done.

+
+    if [ -x /usr/sbin/vpd -a -x /usr/sbin/flashrom ]; then

[jrbarnette, 2011/02/24 17:34:17]

I think this test is pointless for various reasons:
  * Given this job needs them, we shouldn't consider vpd
    or flashrom to be optional, so we shouldn't need to test
    for their presence.
  * This test will prevent vpd from running simply because
    of a missing flashrom, and vice versa.  That's not as robust
    as I'd like to see.

[Louis, 2011/02/25 08:50:12]

Done.

+      flashrom -r $TMP > /dev/null || echo "flashrom execute error."
+      vpd -f "$TMP" -i "RO VPD" -l || echo "RO VPD execute error."
+      vpd -f "$TMP" -i "RW VPD" -l || echo "RW VPD execute error."

[jrbarnette, 2011/02/24 17:34:17]

It seems like the logic above could be replaced by something more
like this:
    if flashrom -r $TMP >/dev/null ; then
        vpd -f ...  "RO ..." ...
        vpd -f ...  "RW ..." ...
    fi
    rm -f $TMP

Also, I'd like to see consistent use of quotes around $TMP - either
always used, or always omitted (in this context, I prefer always
omitted, but consistency is much more important than the
particular usage.

[Louis, 2011/02/25 08:50:12]

Done.

+      rm -f $TMP
+    else
+      echo "Unable to dump VPD 2.0 data."
+    fi >> $VPD_2_0
+  fi
+
+end script