Get the proper serial number for device enrollment.

chrome/browser/policy/device_policy_identity_strategy.cc

Index: chrome/browser/policy/device_policy_identity_strategy.cc
===================================================================
--- chrome/browser/policy/device_policy_identity_strategy.cc	(revision 94533)
+++ chrome/browser/policy/device_policy_identity_strategy.cc	(working copy)
@@ -20,8 +20,21 @@
 
 // MachineInfo key names.
 static const char kMachineInfoSystemHwqual[] = "hardware_class";
-static const char kMachineInfoSerialNumber[] = "serial_number";
 
+// These are the machine serial number keys that we check in order until we
+// find a non-empty serial number. The VPD spec says the serial number should be
+// in the "serial_number" key for v2+ VPDs. However, we cannot check this first,
+// since we'd get the "serial_number" value from the SMBIOS (yes, there's a name
+// clash here!), which is different from the serial number we want and not
+// actually per-device. So, we check the the legacy keys first. If we find a
+// serial number for these, we use it, otherwise we must be on a newer device
+// that provides the correct data in "serial_number".
+static const char* kMachineInfoSerialNumberKeys[] = {
+  "sn",            // ZGB
+  "Product_S/N",   // Alex
+  "serial_number"  // VPD v2+ devices

[Louis, 2011/07/29 02:52:05]

I'd like to suggest to re-order the list. Try "serial_number" first (because it
is the official name in the future), then "sn" and "Product_S/N".

By the way, we are going to show VPD 1.0 structure in the vpd_2.0.txt too
(http://gerrit.chromium.org/gerrit/4552
) so that you may add "Product_SN" for CR-48.

[Mattias Nissler (ping if slow), 2011/07/29 09:19:52]

That's unfortunately not possible due to the name clash mentioned in the comment
above. The information we access through GetMachineStatistic not only comes from
VPD, but also other sources. One of them is the mosys output, which also
includes a "serial_number" field. Making "serial_number" the first choice would
result in the value from mosys being used in case the VPD doesn't give us one,
which is not what we want.
OK, I guess this is only relevant once that support has landed? Let's make that
change on ToT then.

[Louis, 2011/08/01 08:18:11]

What argument do you use for mosys? Does its serial_number mean the same thing?
FYI. I've committed in that code.

[Mattias Nissler (ping if slow), 2011/08/01 08:31:17]

/sbin/chromeos_startup does this:

mosys -k smbios info system

Which dumps a serial_number that is different from the one we get from VPD, and,
what's worse, not actually machine-specific (all the Alex devices I tested on
returned the same serial number string).
Thanks, I'll add "Product_SN" to the list then.

+};
+
 namespace policy {
 
 DevicePolicyIdentityStrategy::DevicePolicyIdentityStrategy() {
@@ -31,10 +44,16 @@
                                     &machine_model_)) {
     LOG(ERROR) << "Failed to get machine model.";
   }
-  if (!sys_lib->GetMachineStatistic(kMachineInfoSerialNumber,
-                                    &machine_id_)) {
+  for (unsigned int i = 0; i < arraysize(kMachineInfoSerialNumberKeys); i++) {
+    if (sys_lib->GetMachineStatistic(kMachineInfoSerialNumberKeys[i],
+                                     &machine_id_) &&
+        !machine_id_.empty()) {
+      break;
+    }
+  }
+
+  if (machine_id_.empty())
     LOG(ERROR) << "Failed to get machine serial number.";
-  }
 }
 
 DevicePolicyIdentityStrategy::~DevicePolicyIdentityStrategy() {