SIGSEGV due to insufficient return value checking for PyFrame_GetLocals #148273
Comments
|
Correction: Staring at the assembly a bit more, the problem seems that the lookup of "self" in a python locals dictionary returns null, e.g. the code assumes that self is never null? |
|
@thomasdullien Are you able to provide a runnable snippet to repro this error or give more insight into the python code that was called to lead to this? |
|
Will readd highpri when a repro is available! |
|
@thomasdullien would you be able to provide a stack trace by any chance? |
|
Same issue here. I also get a SIGSEGV when profiling my PyTorch code. I guess these lines of code cause this problem: pytorch/torch/csrc/autograd/profiler_python.cpp Lines 885 to 886 in f7798d8 I'm running PyTorch 2.7.0 with Python 3.13.3. |
|
I can reproduce the issue with this simple profiling script: import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
from torchvision import datasets, transforms
from torch.profiler import tensorboard_trace_handler
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Dataset
x = torch.linspace(0, 2 * 3.1415926, steps=1000).unsqueeze(1) # [1000,1]
y = torch.sin(x)
dataset = torch.utils.data.TensorDataset(x, y)
trainloader = torch.utils.data.DataLoader(dataset, batch_size=32, shuffle=True)
# Model
model = nn.Linear(1, 1).to(device)
# Optimizer
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr=0.01)
with torch.profiler.profile(
schedule=torch.profiler.schedule(
wait=2,
warmup=2,
active=6,
repeat=1),
on_trace_ready=tensorboard_trace_handler('./log/profiler'),
with_stack=True
) as profiler:
for step, data in enumerate(trainloader):
inputs, labels = data[0].to(device), data[1].to(device)
outputs = model(inputs)
loss = criterion(outputs, labels)
optimizer.zero_grad()
loss.backward()
optimizer.step()
profiler.step()
print(f"step:{step}, loss:{loss.item():.4f}")
if step >= 15:
break
print("Profiling completed. Run 'tensorboard --logdir=./log/profiler' to view.")With Python 3.12.10, this script runs fine. With Python 3.13.3, SIGSEGV is received. |
🐛 Describe the bug
I'm getting a SIGSEGV when running some Torch code locally. It appears to be a null pointer dereference caused by insufficient return value checking of PyFrame_GetLocals (which, starting from more recent Python versions, can in theory return NULL -- but all the code calling it blindly assumes it'll return a valid pointer, and happily dereferences it).
Below is the GDB trace:
Versions
cc @ezyang @gchanan @zou3519 @kadeng @msaroufim @albanD
The text was updated successfully, but these errors were encountered: