🐛 Describe the bug

Heap corruption in grid_sampler_3d_backward.

Example to reproduce

import torch

grad_output = torch.full((1, 1, 1, 4, 4,), 1, dtype=torch.float64, requires_grad=True)
input = torch.full((5, 5, 5, 5, 5,), 3.5e+35, dtype=torch.float64, requires_grad=True)
grid = torch.full((1, 1, 1, 4, 4,), 1, dtype=torch.float64, requires_grad=True)
interpolation_mode = 0
padding_mode = 0
align_corners = True
res = torch.grid_sampler_3d(input, grid, interpolation_mode, padding_mode, align_corners)
grad_out = torch.zeros_like(res)
torch.autograd.backward(res, grad_tensors=grad_out)

Result

sysmalloc: Assertion (old_top == initial_top (av) && old_size == 0) || ((unsigned long) (old_size) >= MINSIZE && prev_inuse (old_top) && ((unsigned long) old_end & (pagesize - 1)) == 0)' failed`
Program aborts due to heap corruption.

Expected behavior

Heap does not get corrupted and program does not abort.

Note

This bug was discovered using fuzz testing.

Versions

PyTorch version: 1.9.0a0+git4d36e60
Is debug build: True
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: Debian GNU/Linux 11 (bullseye) (x86_64)
GCC version: (Debian 10.2.1-6) 10.2.1 20210110
Clang version: 11.0.1-2
CMake version: version 3.22.1
Libc version: glibc-2.31

Python version: 3.9.7 (default, Sep 16 2021, 13:09:58) [GCC 7.5.0] (64-bit runtime)
Python platform: Linux-5.10.0-11-amd64-x86_64-with-glibc2.31
Is CUDA available: False
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A

Versions of relevant libraries:
[pip3] numpy==1.22.2
[pip3] torch==1.9.0a0+git4d36e60
[conda] mkl 2022.0.1 h06a4308_117
[conda] mkl-include 2022.0.1 h06a4308_117
[conda] numpy 1.19.5 pypi_0 pypi
[conda] torch 1.9.0a0+git4d36e60 dev_0

cc @ezyang @gchanan @zou3519 @albanD @mruberry @jbschlosser @walterddr @kshitij12345