Ha Thach

hmm, this is weird, somehow I could "approve and run" ci for this PR. something is tricky with github ci. 

Sorry, I am still lacking behind in term of PR, will try this and integrate this as part of stock examples when I have time to test it. Please be...

There is no clues if this is an tinyusb issue. There is too many custom changes from your setup. And there is no way for me to understand and track...

The issue seems to be the **stack's vendor driver**, currently it doesn't guard against race condition as cdc. I haven't updated the vendor driver once introducing the usbd_edpt_claim() for race...

yeah, it is better to fix existing problem with vendor right now. If you could submit PR, I am happy to review. Otherwise, give me a bit of time, I...

> OT: looking onto claim/release functions, they are using mutex with RTOS... Did you consider using [atomic_compare_exchange](https://en.cppreference.com/w/cpp/atomic/atomic_compare_exchange)? It may require chaging that struct to bitfield (I am not very familiar...

currently it can only be safe to call tud_task() if you don't use RTOS. Otherwise calling the mutex API within ISR can cause issues depending on your RTOS. The idea...

This is already asked, but can you put up your board & setup and how you test the throughput and its result. This allow others to join and try with...

@HiFiPhile wow, that is a very great detail, current queuing only 1 transfer per endpoint can also limit the actual throughput. Though expanding it can increase code complexity since there...

have you tried it, and do you have any examples to prove it ?