Fix CVE-2025-3730 for PyTorch 2.5.1 (rhbz#2360874)

Signed-off-by: Alexander F. Lent <lx@xanderlent.com>

01f226bfb8f2c343f5c614a6bbf685d91160f3af.patch

@@ -0,0 +1,96 @@
+From 01f226bfb8f2c343f5c614a6bbf685d91160f3af Mon Sep 17 00:00:00 2001
+From: zeshengzong <zesheng.zong@outlook.com>
+Date: Mon, 14 Apr 2025 07:24:30 +0000
+Subject: [PATCH] Add check for ctc_loss targets param (#150981)
+
+Fixes #150835
+
+## Test Result
+
+```python
+# cuda
+>>> import torch
+>>> import torch.nn.functional as F
+>>> device = "cuda" # "cpu" is fine
+>>> num_classes = 4
+>>> log_probs = torch.rand(0, 0, num_classes, device=device)
+>>> targets = torch.tensor([], device=device, dtype=torch.long)
+>>> input_lengths = torch.tensor([], device=device, dtype=torch.long)
+>>> target_lengths = torch.tensor([], device=device, dtype=torch.long)
+>>> result = F.ctc_loss(log_probs, targets, input_lengths, target_lengths, reduction='none')
+
+Traceback (most recent call last):
+  File "<stdin>", line 1, in <module>
+  File "/home/zong/code/pytorch/torch/nn/functional.py", line 3079, in ctc_loss
+    return torch.ctc_loss(
+           ^^^^^^^^^^^^^^^
+RuntimeError: log_probs tensor must not be empty
+
+# cpu
+>>> device = "cpu"
+>>> num_classes = 4
+>>> log_probs = torch.rand(0, 0, num_classes, device=device)
+>>> targets = torch.tensor([], device=device, dtype=torch.long)
+>>> input_lengths = torch.tensor([], device=device, dtype=torch.long)
+>>> target_lengths = torch.tensor([], device=device, dtype=torch.long)
+>>> result = F.ctc_loss(log_probs, targets, input_lengths, target_lengths, reduction='none')
+Traceback (most recent call last):
+  File "<stdin>", line 1, in <module>
+  File "/home/zong/code/pytorch/torch/nn/functional.py", line 3079, in ctc_loss
+    return torch.ctc_loss(
+           ^^^^^^^^^^^^^^^
+RuntimeError: log_probs tensor must not be empty
+
+```
+Pull Request resolved: https://github.com/pytorch/pytorch/pull/150981
+Approved by: https://github.com/eqy
+---
+ aten/src/ATen/native/LossCTC.cpp     | 1 +
+ aten/src/ATen/native/cuda/LossCTC.cu | 1 +
+ test/test_nn.py                      | 9 +++++++++
+ 3 files changed, 11 insertions(+)
+
+diff --git a/aten/src/ATen/native/LossCTC.cpp b/aten/src/ATen/native/LossCTC.cpp
+index 1513e756c71d7..46b9397a008c4 100644
+--- a/aten/src/ATen/native/LossCTC.cpp
++++ b/aten/src/ATen/native/LossCTC.cpp
+@@ -126,6 +126,7 @@ std::tuple<Tensor, Tensor, size_t, std::vector<int64_t>> ctc_loss_allocate_outpu
+ // the alphas from the user by only returning the loss.
+ template<typename scalar_t, ScalarType target_scalar_type>
+ std::tuple<Tensor, Tensor> ctc_loss_cpu_template(const Tensor& log_probs, const Tensor& targets, IntArrayRef input_lengths, IntArrayRef target_lengths, int64_t BLANK) {
++  TORCH_CHECK(log_probs.numel() > 0, "log_probs tensor must not be empty");
+   // log_probs: input_len x batch_size x num_labels
+   // targets [int64]: batch_size x target_length OR sum(target_lengths)
+   constexpr scalar_t neginf = -std::numeric_limits<scalar_t>::infinity();
+diff --git a/aten/src/ATen/native/cuda/LossCTC.cu b/aten/src/ATen/native/cuda/LossCTC.cu
+index e597b64c0c175..b5908cc0abcfc 100644
+--- a/aten/src/ATen/native/cuda/LossCTC.cu
++++ b/aten/src/ATen/native/cuda/LossCTC.cu
+@@ -219,6 +219,7 @@ ctc_loss_log_alpha_gpu_kernel(scalar_t* __restrict__ log_alpha_data,
+ // backward. The dispatch function will only return the loss.
+ template<typename scalar_t, ScalarType target_scalar_type>
+ std::tuple<Tensor, Tensor> ctc_loss_gpu_template(const Tensor& log_probs, const Tensor& targets, IntArrayRef input_lengths, IntArrayRef target_lengths, int64_t BLANK) {
++  TORCH_CHECK(log_probs.numel() > 0, "log_probs tensor must not be empty");
+   // log_probs: input_len x batch_size x num_labels
+   // targets [int64]: batch_size x target_length OR sum(target_lengths)
+   CheckedFrom c = "ctc_loss_gpu";
+diff --git a/test/test_nn.py b/test/test_nn.py
+index 32b0efd40aff1..3a2a89a92ba6c 100644
+--- a/test/test_nn.py
++++ b/test/test_nn.py
+@@ -11532,6 +11532,15 @@ def test_ctc_loss_cudnn_tensor(self, device):
+         grad_cudnn, = torch.autograd.grad(loss_cudnn, log_probs, grad_out)
+         self.assertEqual(grad_cudnn, grad_native, atol=1e-4, rtol=0)
+ 
++    @expectedFailureMPS
++    def test_ctc_loss_error(self, device):
++        log_probs = torch.rand(0, 0, 4, device=device)
++        targets = torch.tensor([], device=device, dtype=torch.long)
++        input_lengths = torch.tensor([], device=device, dtype=torch.long)
++        target_lengths = torch.tensor([], device=device, dtype=torch.long)
++        with self.assertRaisesRegex(RuntimeError, "log_probs tensor must not be empty"):
++            F.ctc_loss(log_probs, targets, input_lengths, target_lengths, reduction='none')
++
+     @expectedFailureMPS  # RuntimeError: LSTM with projections is not currently supported with MPS.
+     @dtypesIfCUDA(torch.half, torch.float, torch.double)
+     @dtypes(torch.float)

python-torch.spec

@@ -103,6 +103,10 @@ Patch101:      0001-cuda-hip-signatures.patch
 # https://github.com/pytorch/pytorch/issues/145608
 Patch102:      0001-torch-paper-over-c-assert.patch
 
+# Fix CVE-2025-3730
+# source: https://github.com/pytorch/pytorch/commit/01f226bfb8f2c343f5c614a6bbf685d91160f3af
+Patch201:     01f226bfb8f2c343f5c614a6bbf685d91160f3af.patch
+
 ExclusiveArch:  x86_64 aarch64
 %global toolchain gcc
 %global _lto_cflags %nil