Skip to content

Fix use-after-free in FA3 deterministic mode. #2063

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

Merged
tridao merged 1 commit into from
Dec 12, 2025
+6 −4
Merged

Fix use-after-free in FA3 deterministic mode. #2063

Show file tree
Hide file tree
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
5 changes: 3 additions & 2 deletions hopper/flash_api.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1535,10 +1535,11 @@ std::tuple<at::Tensor, at::Tensor, at::Tensor, at::Tensor, at::Tensor> mha_bwd(
// Will be zero'ed out in the backward preprocess kernel
at::Tensor dq_semaphore = torch::empty({(seqlen_q + kBlockM - 1) / kBlockM, batch_size, num_heads}, opts.dtype(torch::kInt32));
params.dq_semaphore = dq_semaphore.data_ptr<int>();
at::Tensor dk_semaphore, dv_semaphore;
if (num_heads_k != num_heads && params.deterministic) {
// TODO: maybe also zero'ed out dk_semaphore and dv_semaphore in the backward preprocess kernel
at::Tensor dk_semaphore = torch::zeros({(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, opts.dtype(torch::kInt32));
at::Tensor dv_semaphore = torch::zeros({(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, opts.dtype(torch::kInt32));
dk_semaphore = torch::zeros({(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, opts.dtype(torch::kInt32));
dv_semaphore = torch::zeros({(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, opts.dtype(torch::kInt32));
params.dk_semaphore = dk_semaphore.data_ptr<int>();
params.dv_semaphore = dv_semaphore.data_ptr<int>();
}
Expand Down
5 changes: 3 additions & 2 deletions hopper/flash_api_stable.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1610,10 +1610,11 @@ std::tuple<Tensor, Tensor, Tensor, Tensor, Tensor> mha_bwd(
// Will be zero'ed out in the backward preprocess kernel
Tensor dq_semaphore = torch::stable::new_empty(q, {(seqlen_q + kBlockM - 1) / kBlockM, batch_size, num_heads}, std::make_optional(torch::headeronly::ScalarType::Int));
params.dq_semaphore = static_cast<int*>(dq_semaphore.data_ptr());
Tensor dk_semaphore, dv_semaphore;
if (num_heads_k != num_heads && params.deterministic) {
// TODO: maybe also zero'ed out dk_semaphore and dv_semaphore in the backward preprocess kernel
Tensor dk_semaphore = torch::stable::new_zeros(q, {(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, std::make_optional(torch::headeronly::ScalarType::Int));
Tensor dv_semaphore = torch::stable::new_zeros(q, {(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, std::make_optional(torch::headeronly::ScalarType::Int));
dk_semaphore = torch::stable::new_zeros(q, {(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, std::make_optional(torch::headeronly::ScalarType::Int));
dv_semaphore = torch::stable::new_zeros(q, {(seqlen_k + kBlockN - 1) / kBlockN, batch_size, num_heads_k}, std::make_optional(torch::headeronly::ScalarType::Int));
params.dk_semaphore = static_cast<int*>(dk_semaphore.data_ptr());
params.dv_semaphore = static_cast<int*>(dv_semaphore.data_ptr());
}
Expand Down