Enable SysInfo::AmountOfPhysicalMemory to be called from within the Linux sandbox.

base/sys_info_linux.cc

 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/sys_info.h"
 
 #include <limits>
 
 #include "base/file_util.h"
 #include "base/logging.h"
@@ skipping 10 matching lines @@
   }
   return static_cast<int64>(pages) * page_size;
 }
 
 }  // namespace
 
 namespace base {
 
 // static
 int64 SysInfo::AmountOfPhysicalMemory() {
-  return AmountOfMemory(_SC_PHYS_PAGES);
+  static int64 physical_memory;
+  static bool physical_memory_valid = false;
+  if (!physical_memory_valid) {

[jar (doing other things), 2013/10/18 19:24:14]

I'm hopeful this is done on only one thread... as this would not be thread safe.

Can you verify that?  DCHECK it?

[jln (very slow on Chromium), 2013/10/18 21:08:42]

Should we just use LazyInstance these things? It's a bit expensive, but none of
these are hot.

[jar (doing other things), 2013/10/18 22:19:30]

+1.  If in doubt about multi-threading, I'd rather see lazy instance.  I don't
think the cost is actually that high, and probably not large considering it is
caching the answer, and replacing code that probably used to take longer to find
the answer again and again.

+    physical_memory = AmountOfMemory(_SC_PHYS_PAGES);
+    physical_memory_valid = true;
+  }
+  return physical_memory;
 }
 
 // static
 int64 SysInfo::AmountOfAvailablePhysicalMemory() {
   return AmountOfMemory(_SC_AVPHYS_PAGES);
 }
 
 // static
 size_t SysInfo::MaxSharedMemorySize() {
   static int64 limit;
@@ skipping 34 matching lines @@
       if (line.compare(0, strlen(kCpuModelPrefix), kCpuModelPrefix) == 0) {
         size_t pos = line.find(": ");
         return line.substr(pos + 2);
       }
     }
   }
   return std::string();
 }
 
 }  // namespace base