Don't crash when backup timeout races with agent completion

There's a narrow window of time in which an agent reporting that its
operation has completed races with timeouts such that we wind up
handling the completion callback just after certain fundamental state
has been reset.  Detect this race and proceed gracefully instead of
crashing.

Bug 19498669

Change-Id: I5a475527db1a55a8e567366ddfb10112e427682e

1 files changed, 16 insertions, 1 deletions

diff --git a/services/backup/java/com/android/server/backup/BackupManagerService.java b/services/backup/java/com/android/server/backup/BackupManagerService.java
index 4d7ebed..acf4d39 100644
--- a/services/backup/java/com/android/server/backup/BackupManagerService.java
+++ b/services/backup/java/com/android/server/backup/BackupManagerService.java
@@ -2778,7 +2778,22 @@ public class BackupManagerService {
 
         @Override
         public void operationComplete() {
-            // Okay, the agent successfully reported back to us!
+            // The agent reported back to us!
+
+            if (mBackupData == null) {
+                // This callback was racing with our timeout, so we've cleaned up the
+                // agent state already and are on to the next thing.  We have nothing
+                // further to do here: agent state having been cleared means that we've
+                // initiated the appropriate next operation.
+                final String pkg = (mCurrentPackage != null)
+                        ? mCurrentPackage.packageName : "[none]";
+                if (DEBUG) {
+                    Slog.i(TAG, "Callback after agent teardown: " + pkg);
+                }
+                addBackupTrace("late opComplete; curPkg = " + pkg);
+                return;
+            }
+
             final String pkgName = mCurrentPackage.packageName;
             final long filepos = mBackupDataName.length();
             FileDescriptor fd = mBackupData.getFileDescriptor();