From b9b31093de4ea2a89b53daa9ea8f2a957ea089ee Mon Sep 17 00:00:00 2001
From: Justin Gatzen <justin.gatzen@gmail.com>
Date: Thu, 19 Dec 2024 21:53:52 -0500
Subject: [PATCH] [host] app: fix app state race in lgmpTimer thread

While resuming from hibernation lgmpTimer may transition app.state to
APP_STATE_REINIT_LGMP. Sometimes the transition is lost if the app_main
processing loop also changes app.state simultaneously. This seems to
occur frequently with the hibernation use case.

Separate lgmp timer state into its own enum. Move resulting app.state
transitions to the app_main loop so they are serialized.
---
 host/src/app.c | 37 ++++++++++++++++++++++++++-----------
 1 file changed, 26 insertions(+), 11 deletions(-)

diff --git a/host/src/app.c b/host/src/app.c
index b052c1a9..de23e6ad 100644
--- a/host/src/app.c
+++ b/host/src/app.c
@@ -69,10 +69,16 @@ enum AppState
   APP_STATE_RUNNING,
   APP_STATE_IDLE,
   APP_STATE_TRANSITION_TO_IDLE,
-  APP_STATE_REINIT_LGMP,
   APP_STATE_SHUTDOWN
 };
 
+enum LGMPTimerState
+{
+  LGMP_TIMER_STATE_OK,
+  LGMP_TIMER_STATE_CORRUPTED,
+  LGMP_TIMER_STATE_FAULTED
+};
+
 struct app
 {
   int exitcode;
@@ -105,7 +111,8 @@ struct app
   CaptureInterface * iface;
   bool captureStarted;
 
-  enum AppState state, lastState;
+  enum AppState state;
+  enum LGMPTimerState lgmpTimerState;
   LGTimer  * lgmpTimer;
   LGThread * frameThread;
   bool threadsStarted;
@@ -162,8 +169,7 @@ inline static void setAppState(enum AppState state)
 {
   if (app.state == APP_STATE_SHUTDOWN)
     return;
-  app.lastState = app.state;
-  app.state     = state;
+  app.state = state;
 }
 
 static bool lgmpTimer(void * opaque)
@@ -176,12 +182,12 @@ static bool lgmpTimer(void * opaque)
     {
       DEBUG_ERROR("LGMP reported the shared memory has been corrrupted, "
           "attempting to recover");
-      setAppState(APP_STATE_REINIT_LGMP);
+      atomic_store(&app.lgmpTimerState, LGMP_TIMER_STATE_CORRUPTED);
       return false;
     }
 
     DEBUG_ERROR("lgmpHostProcess Failed: %s", lgmpStatusString(status));
-    setAppState(APP_STATE_SHUTDOWN);
+    atomic_store(&app.lgmpTimerState, LGMP_TIMER_STATE_FAULTED);
     return false;
   }
 
@@ -785,6 +791,7 @@ static bool lgmpSetup(struct IVSHMEM * shmDev)
     app.frameBuffer[i] = (FrameBuffer *)(((uint8_t*)app.frame[i]) + alignOffset);
   }
 
+  atomic_store(&app.lgmpTimerState, LGMP_TIMER_STATE_OK);
   if (!lgCreateTimer(10, lgmpTimer, NULL, &app.lgmpTimer))
   {
     DEBUG_ERROR("Failed to create the LGMP timer");
@@ -811,8 +818,7 @@ int app_main(int argc, char * argv[])
   // make sure rng is actually seeded for LGMP
   srand((unsigned)time(NULL));
 
-  app.lastState = APP_STATE_RUNNING;
-  app.state     = APP_STATE_RUNNING;
+  app.state = APP_STATE_RUNNING;
   ivshmemOptionsInit();
 
   // register capture interface options
@@ -954,16 +960,25 @@ int app_main(int argc, char * argv[])
 
   do
   {
-    switch(app.state)
+    switch(atomic_load(&app.lgmpTimerState))
     {
-      case APP_STATE_REINIT_LGMP:
+      case LGMP_TIMER_STATE_OK:
+        break;
+
+      case LGMP_TIMER_STATE_CORRUPTED:
         DEBUG_INFO("Performing LGMP reinitialization");
         lgmpShutdown();
-        setAppState(app.lastState);
         if (!lgmpSetup(&shmDev))
           goto fail_lgmp;
         break;
 
+      case LGMP_TIMER_STATE_FAULTED:
+        setAppState(APP_STATE_SHUTDOWN);
+        break;
+    }
+
+    switch(app.state)
+    {
       case APP_STATE_IDLE:
         // if there are no clients subscribed, just remain idle
         if (!lgmpHostQueueHasSubs(app.pointerQueue) &&