Skip to content

[Bug] Fix integer overflow in activation_kernels.cu pointer arithmetic #42861

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
dparikh79 wants to merge 1 commit into from
Closed

[Bug] Fix integer overflow in activation_kernels.cu pointer arithmetic #42861

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 20 additions & 6 deletions csrc/activation_kernels.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,9 +79,16 @@ __global__ void act_and_mul_kernel(
scalar_t* __restrict__ out, // [..., d]
const scalar_t* __restrict__ input, // [..., 2, d]
const int d, const float limit) {
const scalar_t* x_ptr = input + blockIdx.x * 2 * d;
// `blockIdx.x` is `unsigned int`; multiplying by `d` (`int`) keeps the
// result in 32 bits, which overflows once `blockIdx.x * 2 * d` exceeds
// INT_MAX (about 2.15 billion). For large hidden sizes this corrupts the
// pointer arithmetic and reads/writes the wrong memory. Promote the index
// to int64_t first; matches the pattern already used in
// `swigluoai_and_mul_kernel` below. See issue #42860.
const int64_t token_idx = blockIdx.x;
const scalar_t* x_ptr = input + token_idx * 2 * d;
const scalar_t* y_ptr = x_ptr + d;
scalar_t* out_ptr = out + blockIdx.x * d;
scalar_t* out_ptr = out + token_idx * d;

if constexpr (use_vec) {
using cuda_t = typename CUDATypeConverter<scalar_t>::Type;
Expand Down Expand Up @@ -313,9 +320,12 @@ template <typename scalar_t, typename packed_t,
__global__ void act_and_mul_kernel_with_param(
scalar_t* __restrict__ out, const scalar_t* __restrict__ input, const int d,
const float param) {
const scalar_t* x_ptr = input + blockIdx.x * 2 * d;
// See `act_and_mul_kernel` above for the int64_t-promotion rationale
// (overflow at `blockIdx.x * 2 * d` for large hidden sizes). #42860.
const int64_t token_idx = blockIdx.x;
const scalar_t* x_ptr = input + token_idx * 2 * d;
const scalar_t* y_ptr = x_ptr + d;
scalar_t* out_ptr = out + blockIdx.x * d;
scalar_t* out_ptr = out + token_idx * d;

if constexpr (use_vec) {
using cuda_t = typename CUDATypeConverter<scalar_t>::Type;
Expand Down Expand Up @@ -517,8 +527,12 @@ __global__ void activation_kernel(
scalar_t* __restrict__ out, // [..., d]
const scalar_t* __restrict__ input, // [..., d]
const int d) {
const scalar_t* in_ptr = input + blockIdx.x * d;
scalar_t* out_ptr = out + blockIdx.x * d;
// See `act_and_mul_kernel` near the top of this file for the
// int64_t-promotion rationale (`blockIdx.x * d` overflows int when
// `blockIdx.x * d > INT_MAX`). #42860.
const int64_t token_idx = blockIdx.x;
const scalar_t* in_ptr = input + token_idx * d;
scalar_t* out_ptr = out + token_idx * d;

if constexpr (use_vec) {
// Fast path: 128-bit/256-bit vectorized loop
Expand Down
Loading