Segmentation fault during cudnnCreate

[iamarcel]

Issue summary

The Caffe tests seem to fail whenever cudnnCreate is called. I compiled the latest master (08a95a4) using Cmake and the following options:

-DBLAS=open 
-DUSE_LMDB=off 
-DUSE_LEVELDB=off 
-DCUDNN_INCLUDE=/cudnn/cudnn5.1/include 
-DCUDNN_LIBRARY=/cudnn/cudnn5.1/lib64/libcudnn.so

Here's the backtrace with a debug build:

$ gdb ./test/test.testbin-d
(gdb) catch signal SIGSEGV
(gdb) r
...
Cuda number of devices: 2
Current device id: 0
Current device name: Tesla K40c
[==========] Running 2073 tests from 279 test cases.
... (tests from PoolingLayerTest/0,1,2,3 that all succeed)

[----------] 8 tests from CuDNNPoolingLayerTest/0, where TypeParam = float
[ RUN      ] CuDNNPoolingLayerTest/0.TestSetupCuDNN

Thread 1 "test.testbin-d" hit Catchpoint 1 (signal SIGSEGV), 0x00007fffaec999ce in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
(gdb) bt
#0  0x00007fffaec999ce in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#1  0x00007fffaef2dc0f in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#2  0x00007fffaef2dc69 in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#3  0x00007fffaeb648fd in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#4  0x00007fffaeb6c53b in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#5  0x00007fffaf20dced in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#6  0x00007fffaeb6f240 in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#7  0x00007fffaeb709e3 in ?? () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#8  0x00007fffaeb6767c in __cuda_CallJitEntryPoint () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#9  0x00007fffaebe3366 in nvPTXCompilerCompile () from /usr/lib/nvidia-384/libnvidia-ptxjitcompiler.so.1
#10 0x00007fffb068d969 in fatBinaryCtl_Compile () from /usr/lib/nvidia-384/libnvidia-fatbinaryloader.so.384.111
#11 0x00007fffb0aad02f in ?? () from /usr/lib/x86_64-linux-gnu/libcuda.so.1
#12 0x00007fffb0aadb33 in ?? () from /usr/lib/x86_64-linux-gnu/libcuda.so.1
#13 0x00007fffb09fb9ed in ?? () from /usr/lib/x86_64-linux-gnu/libcuda.so.1
#14 0x00007fffb09fbd00 in ?? () from /usr/lib/x86_64-linux-gnu/libcuda.so.1
#15 0x00007fffe927024d in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#16 0x00007fffe9265c70 in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#17 0x00007fffe92734c6 in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#18 0x00007fffe9276ad1 in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#19 0x00007fffe926a63c in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#20 0x00007fffe9253cc2 in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#21 0x00007fffe928c38f in ?? () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#22 0x00007fffe8d97c54 in cudnnCreate () from /cudnn/cudnn5.1/lib64/libcudnn.so.5
#23 0x00007ffff77d90f7 in caffe::CuDNNPoolingLayer<float>::LayerSetUp (this=0x7fffffffdb80, bottom=std::vector of length 1, capacity 1 = {...}, top=std::vector of length 1, capacity 1 = {...}) at /home/mlsamyn/caffe/src/caffe/layers/cudnn_pooling_layer.cpp:12
#24 0x0000000000ad1546 in caffe::Layer<float>::SetUp (this=0x7fffffffdb80, bottom=std::vector of length 1, capacity 1 = {...}, top=std::vector of length 1, capacity 1 = {...}) at /home/mlsamyn/caffe/include/caffe/layer.hpp:70
#25 0x0000000000ac09a6 in caffe::CuDNNPoolingLayerTest_TestSetupCuDNN_Test<float>::TestBody (this=0x1742ed0) at /home/mlsamyn/caffe/src/caffe/test/test_pooling_layer.cpp:974
#26 0x0000000000fa881a in testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void> (object=0x1742ed0, method=&virtual testing::Test::TestBody(), location=0x1075473 "the test body") at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3393
#27 0x0000000000fa3927 in testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void> (object=0x1742ed0, method=&virtual testing::Test::TestBody(), location=0x1075473 "the test body") at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3429
#28 0x0000000000f8f262 in testing::Test::Run (this=0x1742ed0) at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3465
#29 0x0000000000f8fa54 in testing::TestInfo::Run (this=0x161bc40) at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3641
#30 0x0000000000f9009f in testing::TestCase::Run (this=0x15f8140) at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3748
#31 0x0000000000f9554d in testing::internal::UnitTestImpl::RunAllTests (this=0x1612f20) at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:5540
#32 0x0000000000fa9a1d in testing::internal::HandleSehExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool> (object=0x1612f20, method=(bool (testing::internal::UnitTestImpl::*)(testing::internal::UnitTestImpl * const)) 0xf9528e <testing::internal::UnitTestImpl::RunAllTests()>, 
    location=0x1075f30 "auxiliary test code (environments or event listeners)") at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3393
#33 0x0000000000fa461a in testing::internal::HandleExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool> (object=0x1612f20, method=(bool (testing::internal::UnitTestImpl::*)(testing::internal::UnitTestImpl * const)) 0xf9528e <testing::internal::UnitTestImpl::RunAllTests()>, 
    location=0x1075f30 "auxiliary test code (environments or event listeners)") at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:3429
#34 0x0000000000f9415c in testing::UnitTest::Run (this=0x1466f00 <testing::UnitTest::GetInstance()::instance>) at /home/mlsamyn/caffe/src/gtest/gtest-all.cpp:5174
#35 0x0000000000e77f28 in main (argc=1, argv=0x7fffffffe3d8) at /home/mlsamyn/caffe/src/caffe/test/test_caffe_main.cpp:37
(gdb) f 23
#23 0x00007ffff77d90f7 in caffe::CuDNNPoolingLayer<float>::LayerSetUp (this=0x7fffffffdb80, bottom=std::vector of length 1, capacity 1 = {...}, top=std::vector of length 1, capacity 1 = {...}) at /home/mlsamyn/caffe/src/caffe/layers/cudnn_pooling_layer.cpp:12
12	  CUDNN_CHECK(cudnnCreate(&handle_));

(gdb) p *this

$1 = {> = {> = {_vptr.Layer = 0x1404f68 +16>, layer_param_ = { = {}, static kNameFieldNumber = 1, static kTypeFieldNumber = 2, static kBottomFieldNumber = 3, 
        static kTopFieldNumber = 4, static kPhaseFieldNumber = 10, static kLossWeightFieldNumber = 5, static kParamFieldNumber = 6, static kBlobsFieldNumber = 7, static kPropagateDownFieldNumber = 11, static kIncludeFieldNumber = 8, static kExcludeFieldNumber = 9, static kTransformParamFieldNumber = 100, 
        static kLossParamFieldNumber = 101, static kAccuracyParamFieldNumber = 102, static kArgmaxParamFieldNumber = 103, static kBatchNormParamFieldNumber = 139, static kBiasParamFieldNumber = 141, static kConcatParamFieldNumber = 104, static kContrastiveLossParamFieldNumber = 105, 
        static kConvolutionParamFieldNumber = 106, static kCropParamFieldNumber = 144, static kDataParamFieldNumber = 107, static kDropoutParamFieldNumber = 108, static kDummyDataParamFieldNumber = 109, static kEltwiseParamFieldNumber = 110, static kEluParamFieldNumber = 140, static kEmbedParamFieldNumber = 137, 
        static kExpParamFieldNumber = 111, static kFlattenParamFieldNumber = 135, static kHdf5DataParamFieldNumber = 112, static kHdf5OutputParamFieldNumber = 113, static kHingeLossParamFieldNumber = 114, static kImageDataParamFieldNumber = 115, static kInfogainLossParamFieldNumber = 116, 
        static kInnerProductParamFieldNumber = 117, static kInputParamFieldNumber = 143, static kLogParamFieldNumber = 134, static kLrnParamFieldNumber = 118, static kMemoryDataParamFieldNumber = 119, static kMvnParamFieldNumber = 120, static kParameterParamFieldNumber = 145, static kPoolingParamFieldNumber = 121, 
        static kPowerParamFieldNumber = 122, static kPreluParamFieldNumber = 131, static kPythonParamFieldNumber = 130, static kRecurrentParamFieldNumber = 146, static kReductionParamFieldNumber = 136, static kReluParamFieldNumber = 123, static kReshapeParamFieldNumber = 133, static kScaleParamFieldNumber = 142, 
        static kSigmoidParamFieldNumber = 124, static kSoftmaxParamFieldNumber = 125, static kSppParamFieldNumber = 132, static kSliceParamFieldNumber = 126, static kTanhParamFieldNumber = 127, static kThresholdParamFieldNumber = 128, static kTileParamFieldNumber = 138, static kWindowDataParamFieldNumber = 129, 
        _unknown_fields_ = {fields_ = 0x0}, _has_bits_ = {0, 512}, name_ = 0x1607c90, type_ = 0x1607c90, bottom_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, allocated_size_ = 0, total_size_ = 0}, }, 
        top_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, allocated_size_ = 0, total_size_ = 0}, }, loss_weight_ = {static kInitialSize = , elements_ = 0x0, current_size_ = 0, total_size_ = 0}, 
        param_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, allocated_size_ = 0, total_size_ = 0}, }, blobs_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, 
            allocated_size_ = 0, total_size_ = 0}, }, propagate_down_ = {static kInitialSize = , elements_ = 0x0, current_size_ = 0, total_size_ = 0}, include_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, 
            allocated_size_ = 0, total_size_ = 0}, }, exclude_ = { = {static kInitialSize = 0, elements_ = 0x0, current_size_ = 0, allocated_size_ = 0, total_size_ = 0}, }, transform_param_ = 0x0, loss_param_ = 0x0, 
        accuracy_param_ = 0x0, argmax_param_ = 0x0, batch_norm_param_ = 0x0, bias_param_ = 0x0, concat_param_ = 0x0, contrastive_loss_param_ = 0x0, convolution_param_ = 0x0, crop_param_ = 0x0, data_param_ = 0x0, dropout_param_ = 0x0, dummy_data_param_ = 0x0, eltwise_param_ = 0x0, elu_param_ = 0x0, 
        embed_param_ = 0x0, exp_param_ = 0x0, flatten_param_ = 0x0, hdf5_data_param_ = 0x0, hdf5_output_param_ = 0x0, hinge_loss_param_ = 0x0, image_data_param_ = 0x0, infogain_loss_param_ = 0x0, inner_product_param_ = 0x0, input_param_ = 0x0, log_param_ = 0x0, lrn_param_ = 0x0, memory_data_param_ = 0x0, 
        mvn_param_ = 0x0, parameter_param_ = 0x0, pooling_param_ = 0x6fb0ff0, power_param_ = 0x0, prelu_param_ = 0x0, python_param_ = 0x0, recurrent_param_ = 0x0, reduction_param_ = 0x0, relu_param_ = 0x0, reshape_param_ = 0x0, scale_param_ = 0x0, sigmoid_param_ = 0x0, softmax_param_ = 0x0, spp_param_ = 0x0, 
        slice_param_ = 0x0, tanh_param_ = 0x0, threshold_param_ = 0x0, tile_param_ = 0x0, window_data_param_ = 0x0, phase_ = 0, _cached_size_ = 0, static default_instance_ = 0x16196e0}, phase_ = caffe::TRAIN, blobs_ = std::vector of length 0, capacity 0, 
      param_propagate_down_ = std::vector of length 0, capacity 0, loss_ = std::vector of length 0, capacity 0}, kernel_h_ = 3, kernel_w_ = 3, stride_h_ = 2, stride_w_ = 2, pad_h_ = 0, pad_w_ = 0, channels_ = 0, height_ = 0, width_ = -8560, pooled_height_ = 32767, pooled_width_ = -141436653, 
    global_pooling_ = false, rand_idx_ = {data_ = {px = 0x0, pn = {pi_ = 0x0}}, diff_ = {px = 0x0, pn = {pi_ = 0x0}}, shape_data_ = {px = 0x0, pn = {pi_ = 0x0}}, shape_ = std::vector of length 0, capacity 0, count_ = 0, capacity_ = 0}, max_idx_ = {data_ = {px = 0x0, pn = {pi_ = 0x0}}, diff_ = {px = 0x0, pn = {
          pi_ = 0x0}}, shape_data_ = {px = 0x0, pn = {pi_ = 0x0}}, shape_ = std::vector of length 0, capacity 0, count_ = 0, capacity_ = 0}}, handles_setup_ = false, handle_ = 0x0, bottom_desc_ = 0x3f800000, top_desc_ = 0xffffffff3f800000, pooling_desc_ = 0x7fffffffdf70, mode_ = (unknown: 4294958944)}

Steps to reproduce

$ git clone https://github.com/BVLC/caffe.git
$ cd caffe
$ mkdir build && cd build
$ cmake .. -DBLAS=open -DUSE_LMDB=off -DUSE_LEVELDB=off -DCUDNN_INCLUDE=/cudnn/cudnn5.1/include -DCUDNN_LIBRARY=/cudnn/cudnn5.1/lib64/libcudnn.so
$ make all
$ make runtest

Your system configuration

Operating system

$ lsb_release -a
No LSB modules are available.
Distributor ID:	Ubuntu
Description:	Ubuntu 16.04.2 LTS
Release:	16.04
Codename:	xenial

Compiler: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
CUDA version (if applicable): 8.0.44
CUDNN version (if applicable): 5.1
BLAS: open
GPU configuration:

$ nvidia-smi
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.111                Driver Version: 384.111                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K40c          Off  | 00000000:02:00.0 Off |                  Off |
| 23%   42C    P8    20W / 235W |     18MiB / 12205MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
|   1  Tesla K40c          Off  | 00000000:03:00.0 Off |                  Off |
| 23%   41C    P8    20W / 235W |      1MiB / 12205MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+