P4 to Git Change 2047424 by lmoriche@lmoriche_opencl_dev2 on 2019/12/17 20:13:01

SWDEV-216705 - [hipclang-vdi-rocm][FBA-80]Test crash when all GPUs are hidden by ROCR_VISIBLE_DEVICES

	Return an error instead of dereferencing a null pointer. This should address the issue described
	in the ticket, but more places need fixing in the runtime to avoid crashes for corner cases.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/hip/hip_context.cpp#26 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_device_runtime.cpp#23 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#92 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_platform.cpp#58 edit


[ROCm/hip commit: f6d38a725c]

projects/hip/api/hip/hip_context.cpp

@@ -106,7 +106,8 @@ amd::HostQueue* getNullStream(amd::Context& context) {
 }
 
 amd::HostQueue* getNullStream() {
-  return getNullStream(*getCurrentContext());
+  amd::Context* context = getCurrentContext();
+  return context ? getNullStream(*context) : nullptr;
 }
 
 };

projects/hip/api/hip/hip_device_runtime.cpp

@@ -437,7 +437,6 @@ int ihipGetDevice() {
       return i;
     }
   }
-  assert(0 && "Current device not found?!");
   return -1;
 }
 
@@ -446,7 +445,9 @@ hipError_t hipGetDevice ( int* deviceId ) {
 
   if (deviceId != nullptr) {
     int dev = ihipGetDevice();
-    assert(dev != -1);
+    if (dev == -1) {
+      HIP_RETURN(hipErrorNoDevice);
+    }
     *deviceId = dev;
     HIP_RETURN(hipSuccess);
   } else {

projects/hip/api/hip/hip_memory.cpp

@@ -69,6 +69,10 @@ hipError_t ihipMalloc(void** ptr, size_t sizeBytes, unsigned int flags)
   amd::Context* amdContext = ((flags & CL_MEM_SVM_FINE_GRAIN_BUFFER) != 0)?
     hip::host_context : hip::getCurrentContext();
 
+  if (amdContext == nullptr) {
+    return hipErrorMemoryAllocation;
+  }
+
   if (amdContext->devices()[0]->info().maxMemAllocSize_ < sizeBytes) {
     return hipErrorMemoryAllocation;
   }

projects/hip/api/hip/hip_platform.cpp

@@ -469,7 +469,9 @@ extern "C" void __hipUnregisterFatBinary(std::vector< std::pair<hipModule_t, boo
       as_amd(reinterpret_cast<cl_program>(module.first))->release();
     }
   });
-  PlatformState::instance().unregisterVar((*modules)[0].first);
+  if (modules->size() > 0) {
+    PlatformState::instance().unregisterVar((*modules)[0].first);
+  }
   PlatformState::instance().removeFatBinary(modules);
 }
 
@@ -532,6 +534,9 @@ extern "C" hipError_t hipLaunchByPtr(const void *hostFunction)
   HIP_INIT_API(NONE, hostFunction);
 
   int deviceId = ihipGetDevice();
+  if (deviceId == -1) {
+    HIP_RETURN(hipErrorNoDevice);
+  }
   hipFunction_t func = PlatformState::instance().getFunc(hostFunction, deviceId);
   if (func == nullptr) {
     HIP_RETURN(hipErrorInvalidDeviceFunction);