About the increasement of the queue size.

[Hugues-Mao]

Hello,

According to the mainpage _rpmsg_lite.md of v5.1.0, the version 'Increased the queue size to (2 * RL_BUFFER_COUNT) to cover zero copy cases.'. I want to know what are the 'zero copy cases'? I think a size of 1 * RL_BUFFER_COUNT would be enough for the rx queue.

[MichalPrincNXP]

Hello @Hugues-Mao ,
thanks for your question. As you can read from https://github.com/nxp-mcuxpresso/rpmsg-lite/blob/release/24.12.00/template_application/rpmsg_config.h#L48 the RL_BUFFER_COUNT represents the amount of buffers for one direction, the total amount of buffers used for the communication between the master and the slave is RL_BUFFER_COUNT * 2. Thus it could happen (under certain conditions) that one endpoint receive queue can have RL_BUFFER_COUNT * 2 messages.
Regards
Michal

[Hugues-Mao]

Hi @MichalPrincNXP ,

Thanks for your answer.

But I think only when an endpoint receives data from the other endpoint, the received element could be pushed into the queue. Considering the amount of buffers for RX direction is RL_BUFFER_COUNT, I still think 1 * RL_BUFFER_COUNT would be enough.

As I see you are taking TX direction into consideration, do you mean under certain conditions we may push TXed element into the RX
queue via callbacks or some other interfaces?

Regards,
Mao

[MichalPrincNXP]

Hello @Hugues-Mao ,
I do not remember details of the case when we faced the issue with not-enough-long queue, let me run the test suite with reverted changes ... will get back to you then, ok? Anyway, thank you for raising this question ...
Regards
Michal