[Android] Add 64-bit support to memdump (and memory_inspector parser)

tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py

Index: tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py
diff --git a/tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py b/tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py
index 0bb7004a54ac321a298287ce89620a871531377b..5c5e0a1360b35b492f5ac197192d52af9cff8d79 100644
--- a/tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py
+++ b/tools/memory_inspector/memory_inspector/backends/android/memdump_parser.py
@@ -42,7 +42,8 @@ def Parse(lines):
       An instance of |memory_map.Map|.
   """
   RE = (r'^([0-9a-f]+)-([0-9a-f]+)\s+'
-        r'([rwxps-]{4})\s*.*?'
+        r'([rwxps-]{4})\s+'
+        r'([0-9a-f]+)\s+'
         r'private_unevictable=(\d+) private=(\d+) '
         r'shared_app=(.*?) '
         r'shared_other_unevictable=(\d+) shared_other=(\d+) '
@@ -64,20 +65,21 @@ def Parse(lines):
       logging.warning('Skipping unrecognized memdump line "%s"' % line)
       continue
 
-    # TODO(primiano): proper offset handling requires a change in memdump. In
-    # the meanwhile, it should pretty safe assuming zero-offset for libs (for
-    # symbolization). Also, offsets for other mappings don't really matter.
+    start = int(m.group(1), 16)
+    end = int(m.group(2), 16) - 1 # end addr is inclusive in memdump output.
+    if (start > end):
+      continue  # Sadly, this happened. Probably a kernel bug, see b/17402069.

[Sami, 2014/09/09 10:44:26]

Should we log a warning?

[Primiano Tucci (use gerrit), 2014/09/09 11:11:38]

Yeah, probably it's a good idea.

     entry = memory_map.MapEntry(
-        start=int(m.group(1), 16),
-        end=int(m.group(2), 16) - 1, # end addr is inclusive in memdump output.
+        start=start,
+        end=end,
         prot_flags=m.group(3),
-        mapped_file=m.group(9),
-        mapped_offset=0)
-    entry.priv_dirty_bytes = int(m.group(4))
-    entry.priv_clean_bytes = int(m.group(5)) - entry.priv_dirty_bytes
-    entry.shared_dirty_bytes = int(m.group(7))
-    entry.shared_clean_bytes = int(m.group(8)) - entry.shared_dirty_bytes
-    entry.resident_pages = [ord(c) for c in base64.b64decode(m.group(10))]
+        mapped_file=m.group(10),
+        mapped_offset=int(m.group(4), 16))
+    entry.priv_dirty_bytes = int(m.group(5))
+    entry.priv_clean_bytes = int(m.group(6)) - entry.priv_dirty_bytes
+    entry.shared_dirty_bytes = int(m.group(8))
+    entry.shared_clean_bytes = int(m.group(9)) - entry.shared_dirty_bytes
+    entry.resident_pages = [ord(c) for c in base64.b64decode(m.group(11))]

[Sami, 2014/09/09 10:44:26]

Having 11 groups in a regex is probably an argument for using named groups or
using something other than a regex in the first place :) No need to block this
patch on that though.

[Primiano Tucci (use gerrit), 2014/09/09 11:11:39]

Hmm suggestions for alternatives? I can't simply tokenize on spaces because the
mapped file name (the part in quotes) can contain spaces, e.g.,
1234-5678 .... "/dev/ashmem (deleted)" [AAAAA=].

[Sami, 2014/09/09 11:16:20]

Maybe shlex?

     maps.Add(entry)
 
   return maps