Crash on cpp application exit with "corrupted size vs. prev_size while consolidating" followed by Segmentation fault

[gorilux]

🐛 Describe the bug

TLDR; Crash on cpp application exit with "corrupted size vs. prev_size while consolidating" followed by Segmentation fault

Details: I've written a multithreaded c++ application that loads a collection pytorch modules for object detection. All the models are created in the main thread and inference is launched from a different thread into the cuda GPU. The snippet below is just an example that resembles the actual implementation of loading the models.

struct pytorch_model {
  torch::jit::script::Module module;
  torch::Device  device     = torch::kCPU;
};
std::vector<pytorch_model> model_collection {};
for(auto& cfg: configurations)
{
  auto& pm = model_collection.emplace_back( pytorch_model{});
   if (cfg.use_cuda && torch::cuda::is_available()) {
      pm.device = torch::kCUDA;
    } else {
      pm.device = torch::kCPU;
    }

  pm.module =  torch::jit::load(cfg.model_path);
  pm.module.to(pm.device);
}

Everything works perfectly until the application ends and there is a crash on the destructor cpp torch::jit::script::Module::~Module()

Here is a screen capture of the stack trace:

Versions

Collecting environment information... PyTorch version: 2.1.0 Is debug build: False CUDA used to build PyTorch: 12.3 ROCM used to build PyTorch: N/A

OS: Arch Linux (x86_64) GCC version: (GCC) 13.2.1 20230801 Clang version: 16.0.6 CMake version: version 3.27.8 Libc version: glibc-2.38

Python version: 3.11.5 (main, Sep 2 2023, 14:16:33) [GCC 13.2.1 20230801] (64-bit runtime) Python platform: Linux-6.6.1-arch1-1-x86_64-with-glibc2.38 Is CUDA available: True CUDA runtime version: 12.3.52 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA GeForce RTX 3090 Nvidia driver version: 545.29.02 cuDNN version: Probably one of the following: /usr/lib/libcudnn.so.8.9.6 /usr/lib/libcudnn_adv_infer.so.8.9.6 /usr/lib/libcudnn_adv_train.so.8.9.6 /usr/lib/libcudnn_cnn_infer.so.8.9.6 /usr/lib/libcudnn_cnn_train.so.8.9.6 /usr/lib/libcudnn_ops_infer.so.8.9.6 /usr/lib/libcudnn_ops_train.so.8.9.6 HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 43 bits physical, 48 bits virtual Byte Order: Little Endian CPU(s): 24 On-line CPU(s) list: 0-23 Vendor ID: AuthenticAMD Model name: AMD Ryzen 9 3900XT 12-Core Processor CPU family: 23 Model: 113 Thread(s) per core: 2 Core(s) per socket: 12 Socket(s): 1 Stepping: 0 Frequency boost: disabled CPU(s) scaling MHz: 80% CPU max MHz: 5278.7100 CPU min MHz: 2200.0000 BogoMIPS: 8403.36 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip rdpid overflow_recov succor smca sev sev_es Virtualization: AMD-V L1d cache: 384 KiB (12 instances) L1i cache: 384 KiB (12 instances) L2 cache: 6 MiB (12 instances) L3 cache: 64 MiB (4 instances) NUMA node(s): 1 NUMA node0 CPU(s): 0-23 Vulnerability Gather data sampling: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Retbleed: Mitigation; untrained return thunk; SMT enabled with STIBP protection Vulnerability Spec rstack overflow: Mitigation; safe RET Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Mitigation; Retpolines, IBPB conditional, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

Versions of relevant libraries: [pip3] flake8==6.1.0 [pip3] mypy==1.3.0 [pip3] mypy-extensions==1.0.0 [pip3] numpy==1.26.1 [pip3] pytorch-lightning==2.0.7 [pip3] torch==2.1.0 [pip3] torchmetrics==1.1.0 [pip3] torchvision==0.15.2a0 [conda] Could not collect

cc @jbschlosser @ptrblck

[malfet]

Not yet adding oncall: jit as I want to debug it a bit further. @gorilux alas, this snippet is entirely sufficient to debug the problem. Would you object sharing the entire code? (Privately if you prefer)

[gorilux]

Not yet adding oncall: jit as I want to debug it a bit further. @gorilux alas, this snippet is entirely sufficient to debug the problem. Would you object sharing the entire code? (Privately if you prefer)

The project is currently private here on github. I can share it with you right now until I make it all public.

[gorilux]

@malfet There is no issue, the problem is within my code. Closing it.

[malfet]

@gorilux thank you, I was about to mention that it looks like that. Please do not hesitate to hesitate to remove me from the project.

[SebastianBruijns]

I have the same issue now, what was the problem in your case?

[gorilux]

I have the same issue now, what was the problem in your case?

In my case, I solved it by clearing the vector with the pytorch models before the application exited. I suspect it was a concurrency problem.