Bluetooth: Host: Invoke tx callbacks when channel is deleted #3367
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When bt_l2cap_send_pdu() succeeds, it transfers buffer ownership to the stack, which must eventually invoke the provided callback. This contract is honored in all paths where transmission becomes impossible: - Normal transmission: callback invoked with err=0 after HCI Number of Completed Packets event (tx_notify_process) - Send errors (after tx allocated): callback invoked with err=-ESHUTDOWN via conn_tx_destroy - Send errors (before tx allocated): callback invoked with the specific error code in send_buf error_return path - Connection disconnect: callbacks invoked with err=-ESHUTDOWN via process_unack_tx -> conn_tx_destroy for all PDUs in tx_pending However, when a channel is deleted (l2cap_chan_del), PDUs remaining in the tx_queue are dropped without invoking their callbacks, violating the ownership contract. Fix this by extracting and invoking any non-NULL callbacks from the closure stored in buf->user_data before releasing the buffers. The callback is invoked with err=-ESHUTDOWN, making this path analogous to process_unack_tx: both drain queues of unsent PDUs when transmission becomes impossible due to external events (channel deletion vs connection disconnect). The only difference is the buffer lifecycle stage - in l2cap_chan_del, PDUs are still in tx_queue (closure in buf->user_data), while in process_unack_tx, they've progressed to tx_pending (callback in bt_conn_tx struct). Note: conn_tx_destroy() cannot be used here because no bt_conn_tx struct has been allocated yet - the closure is still in buf->user_data. Signed-off-by: Aleksander Wasaznik <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When bt_l2cap_send_pdu() succeeds, it transfers buffer ownership to the stack, which must eventually invoke the provided callback. This contract is honored in all paths where transmission becomes impossible:
However, when a channel is deleted (l2cap_chan_del), PDUs remaining in the tx_queue are dropped without invoking their callbacks, violating the ownership contract.
Fix this by extracting and invoking any non-NULL callbacks from the closure stored in buf->user_data before releasing the buffers. The callback is invoked with err=-ESHUTDOWN, making this path analogous to process_unack_tx: both drain queues of unsent PDUs when transmission becomes impossible due to external events (channel deletion vs connection disconnect). The only difference is the buffer lifecycle stage - in l2cap_chan_del, PDUs are still in tx_queue (closure in buf->user_data), while in process_unack_tx, they've progressed to tx_pending (callback in bt_conn_tx struct).
Note: conn_tx_destroy() cannot be used here because no bt_conn_tx struct has been allocated yet - the closure is still in buf->user_data.