From: Owen Smith <owen.smith@citrix.com>
Date: Tue, 15 Dec 2015 11:30:18 +0000 (+0000)
Subject: BSOD if initial balloon thread has not completed within 20 minutes
X-Git-Url: http://xenbits.xensource.com/gitweb?a=commitdiff_plain;h=be94a72a2d2c0b3849cebafe5e919cd4f7a52797;p=people%2Fpauldu%2Fxenbus.git

BSOD if initial balloon thread has not completed within 20 minutes

Since there is no way of reporting balloon failures to the toolstack,
the only way of stopping a VM from attempting to balloon indefinitely
is to BSOD after a large timeout.

Signed-off-by: Owen Smith <owen.smith@citrix.com>

Largely cosmetic changes (comments and #defines).

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
---

diff --git a/src/xenbus/fdo.c b/src/xenbus/fdo.c
index 9c9e380..da5cdcc 100644
--- a/src/xenbus/fdo.c
+++ b/src/xenbus/fdo.c
@@ -1532,9 +1532,9 @@ loop:
     return STATUS_SUCCESS;
 }
 
-#define TIME_US(_us)            ((_us) * 10)
-#define TIME_MS(_ms)            (TIME_US((_ms) * 1000))
-#define TIME_S(_s)              (TIME_MS((_s) * 1000))
+#define TIME_US(_us)            ((_us) * 10ll)
+#define TIME_MS(_ms)            (TIME_US((_ms) * 1000ll))
+#define TIME_S(_s)              (TIME_MS((_s) * 1000ll))
 #define TIME_RELATIVE(_t)       (-(_t))
 
 static FORCEINLINE NTSTATUS
@@ -3251,7 +3251,8 @@ FdoFilterCmPartialResourceList(
     }
 }
 
-#define BALLOON_PAUSE   60
+#define BALLOON_WARN_TIMEOUT        10
+#define BALLOON_BUGCHECK_TIMEOUT    1200
 
 static NTSTATUS
 FdoStartDevice(
@@ -3356,29 +3357,38 @@ not_active:
         goto fail8;
 
     if (Fdo->BalloonInterface.Interface.Context != NULL) {
-        BOOLEAN Warned;
+        LARGE_INTEGER   Timeout;
 
         ASSERT(__FdoIsActive(Fdo));
 
-        Warned = FALSE;
-
-        for (;;) {
-            LARGE_INTEGER   Timeout;
-
-            Timeout.QuadPart = TIME_RELATIVE(TIME_S(BALLOON_PAUSE));
+        //
+        // Balloon inflation should complete within a reasonable
+        // time (otherwise the target is probably unreasonable).
+        //
+        Timeout.QuadPart = TIME_RELATIVE(TIME_S(BALLOON_WARN_TIMEOUT));
+
+        status = KeWaitForSingleObject(&Fdo->BalloonEvent,
+                                        Executive,
+                                        KernelMode,
+                                        FALSE,
+                                        &Timeout);
+        if (status == STATUS_TIMEOUT) {
+            Warning("waiting for balloon\n");
+
+            //
+            // If inflation does not complete after a lengthy timeout
+            // then it is unlikely that it ever will. In this case we
+            // cause a bugcheck.
+            //
+            Timeout.QuadPart = TIME_RELATIVE(TIME_S((BALLOON_BUGCHECK_TIMEOUT - BALLOON_WARN_TIMEOUT)));
 
             status = KeWaitForSingleObject(&Fdo->BalloonEvent,
-                                           Executive,
-                                           KernelMode,
-                                           FALSE,
-                                           &Timeout);
-            if (status != STATUS_TIMEOUT)
-                break;
-
-            if (!Warned) {
-                Warning("waiting for balloon\n");
-                Warned = TRUE;
-            }
+                                            Executive,
+                                            KernelMode,
+                                            FALSE,
+                                            &Timeout);
+            if (status == STATUS_TIMEOUT)
+                BUG("BALLOON INFLATION TIMEOUT\n");
         }
     }