Rather than having a single tx frame completion topic that all transmit frame completions are published to, struct can_tx_frame_s will have a pointer to a topic on which its completion message will be published.
frame_pool will move from struct can_tx_queue_s into struct can_instance_s, and stage_head will be removed from struct can_tx_queue_s. An API will be provided to atomically allocate a list of N struct can_tx_frame_s from unlocked thread context, as well as an API to atomically commit a list of frames to be sent and an API to free a list of frames. This will simplify the tx queue and minimize the time spent in system locked state.
jschall
commented
6 years ago
I'm going to make the following changes:
struct can_tx_frame_swill have a pointer to a topic on which its completion message will be published.frame_poolwill move fromstruct can_tx_queue_sintostruct can_instance_s, andstage_headwill be removed fromstruct can_tx_queue_s. An API will be provided to atomically allocate a list of Nstruct can_tx_frame_sfrom unlocked thread context, as well as an API to atomically commit a list of frames to be sent and an API to free a list of frames. This will simplify the tx queue and minimize the time spent in system locked state.