[OpenMP] Protect unrecogonized CUDA error code

[ye-luo]

Summary

If an error code can not be recognized by cuGetErrorString, errStr remains null and causes crashing at DP() printing.
Protect this case.

[jhuber6]

I've noticed this happens at the end of the program when Libomptarget is being unregistered. It'd be nice to know what causes this but checking the case is definitely better than crashing.

[ye-luo]

The root cause is a known issue and I put up a bug report to track the status.
https://bugs.llvm.org/show_bug.cgi?id=47595
Anyway, this patch should be sufficient for users at the moment.

[jhuber6]

LGTM

[tianshilei1992]

I also suggest to check nullptr before using errStr.

[jhuber6]

In D87980#2285545, @tianshilei1992 wrote:

I also suggest to check nullptr before using errStr.

I think the errStr is only NULL on when it returns CUDA_FAILURE so it would be redundant.

[tianshilei1992]

In D87980#2285555, @jhuber6 wrote:

In D87980#2285545, @tianshilei1992 wrote:

I also suggest to check nullptr before using errStr.

I think the errStr is only NULL on when it returns CUDA_FAILURE so it would be redundant.

You're right. I double checked the CUDA document and it supports your comment.

[ye-luo]

Hold a second. I'm exploring a bit more in the error message.

[ye-luo]

After a bit more experiment, the return status of cuGetErrorString can be more than CUDA_SUCCESS, CUDA_ERROR_INVALID_VALUE.
In this particular case when the CUDA is deinitialized, the error code cannot be translated by cuGetErrorString any more.
So now only print errStr with CUDA_SUCCESS.
Treat CUDA_ERROR_INVALID_VALUE different from generic !=CUDA_SUCCESS

[ye-luo]

Should be good to go now.

[jhuber6]

Should be a good solution for now.

openmp/libomptarget/plugins/cuda/src/rtl.cpp

 #define TARGET_NAME CUDA
 #define DEBUG_PREFIX "Target " GETNAME(TARGET_NAME) " RTL"
 
 // Utility for retrieving and printing CUDA error string.
 #ifdef OMPTARGET_DEBUG
 #define CUDA_ERR_STRING(err)                                                   \
   do {                                                                         \
     if (getDebugLevel() > 0) {                                                 \
-      const char *errStr;                                                      \
-      cuGetErrorString(err, &errStr);                                          \
-      DP("CUDA error is: %s\n", errStr);                                       \
+      const char *errStr = nullptr;                                            \
+      CUresult errStr_status = cuGetErrorString(err, &errStr);                 \
+      if (errStr_status == CUDA_ERROR_INVALID_VALUE)                           \
+        DP("unrecognized CUDA error code: %d\n", err);                         \
+      else if (errStr_status == CUDA_SUCCESS)                                  \
+        DP("CUDA error is: %s\n", errStr);                                     \
+      else                                                                     \
+        DP("Unresolved cuGetErrorString return status: %d\n", errStr_status);  \
     }                                                                          \
   } while (false)
 #else // OMPTARGET_DEBUG
 #define CUDA_ERR_STRING(err) {}
 #endif // OMPTARGET_DEBUG
 
 #include "../../common/elf_common.c"