Block all calls to get drivers until after init has completed to avoid race during sorting

Signed-off-by: Neil R. Spruit <neil.r.spruit@intel.com>

Author: nrspruit

CHANGELOG.md

@@ -1,5 +1,7 @@
 # Level zero loader changelog
 
+## v1.22.4
+* Block all calls to get until after init has completed to avoid race during sorting.
 ## v1.22.3
 * Fix sysman-only initialization to block loader context retrieval when versions are incompatible
 * Add ability to register a TeardownCallback to notify release of L0 resources

CMakeLists.txt

@@ -13,7 +13,7 @@ if(MSVC AND (MSVC_VERSION LESS 1900))
 endif()
 
 # This project follows semantic versioning (https://semver.org/)
-project(level-zero VERSION 1.22.3)
+project(level-zero VERSION 1.22.4)
 
 include(GNUInstallDirs)
 

PRODUCT_GUID.txt

@@ -1,2 +1,2 @@
-1.22.3
-b3491b8d-9942-47bf-be47-8741f9614566
+1.22.4
+55e69860-8ae7-4f60-8ad9-728c1ed2a2e7

scripts/templates/ldrddi.cpp.mako

@@ -79,6 +79,8 @@ namespace loader
         %elif re.match(r"\w+DriverGet$", th.make_func_name(n, tags, obj)) or re.match(r"\w+InitDrivers$", th.make_func_name(n, tags, obj)):
         uint32_t total_driver_handle_count = 0;
 
+    {
+        std::lock_guard<std::mutex> lock(loader::context->sortMutex);
         if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
             %if namespace != "zes":
             %if not re.match(r"\w+InitDrivers$", th.make_func_name(n, tags, obj)):
@@ -97,6 +99,7 @@ namespace loader
             %endif
             loader::context->sortingInProgress.store(false);
         }
+    }
 
         %if namespace != "zes":
         for( auto& drv : loader::context->zeDrivers )

source/loader/ze_ldrddi.cpp

@@ -56,11 +56,14 @@ namespace loader
 
         uint32_t total_driver_handle_count = 0;
 
-        if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
-            std::call_once(loader::context->coreDriverSortOnce, []() {
-                loader::context->driverSorting(&loader::context->zeDrivers, nullptr, false);
-            });
-            loader::context->sortingInProgress.store(false);
+        {
+            std::lock_guard<std::mutex> lock(loader::context->sortMutex);
+            if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
+                std::call_once(loader::context->coreDriverSortOnce, []() {
+                    loader::context->driverSorting(&loader::context->zeDrivers, nullptr, false);
+                });
+                loader::context->sortingInProgress.store(false);
+            }
         }
 
         for( auto& drv : loader::context->zeDrivers )
@@ -138,12 +141,14 @@ namespace loader
         ze_result_t result = ZE_RESULT_SUCCESS;
 
         uint32_t total_driver_handle_count = 0;
-
-        if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
-            std::call_once(loader::context->coreDriverSortOnce, [desc]() {
-                loader::context->driverSorting(&loader::context->zeDrivers, desc, false);
-            });
-            loader::context->sortingInProgress.store(false);
+        {
+            std::lock_guard<std::mutex> lock(loader::context->sortMutex);
+            if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
+                std::call_once(loader::context->coreDriverSortOnce, [desc]() {
+                    loader::context->driverSorting(&loader::context->zeDrivers, desc, false);
+                });
+                loader::context->sortingInProgress.store(false);
+            }
         }
 
         for( auto& drv : loader::context->zeDrivers )

source/loader/ze_loader_internal.h

@@ -158,6 +158,7 @@ namespace loader
         std::once_flag coreDriverSortOnce;
         std::once_flag sysmanDriverSortOnce;
         std::atomic<bool> sortingInProgress = {false};
+        std::mutex sortMutex;
         bool instrumentationEnabled = false;
         dditable_t tracing_dditable = {};
         std::shared_ptr<Logger> zel_logger;

source/loader/zes_ldrddi.cpp

@@ -60,11 +60,14 @@ namespace loader
 
         uint32_t total_driver_handle_count = 0;
 
-        if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
-            std::call_once(loader::context->sysmanDriverSortOnce, []() {
-                loader::context->driverSorting(loader::context->sysmanInstanceDrivers, nullptr, true);
-            });
-            loader::context->sortingInProgress.store(false);
+        {
+            std::lock_guard<std::mutex> lock(loader::context->sortMutex);
+            if (!loader::context->sortingInProgress.exchange(true) && !loader::context->instrumentationEnabled) {
+                std::call_once(loader::context->sysmanDriverSortOnce, []() {
+                    loader::context->driverSorting(loader::context->sysmanInstanceDrivers, nullptr, true);
+                });
+                loader::context->sortingInProgress.store(false);
+            }
         }
 
         for( auto& drv : *loader::context->sysmanInstanceDrivers )