In order to have the maximum throughput for the hardware crypto driver, I would like the next request waiting in the queue to be send to the driver as soon as I have a input-queue (low/zero) interrupt. That way the hardware can process a continuous flow. At the moment it seems to me that the next request is only started After I flag the result as processed and the engine is “momentarily” idle.

Can I use the generic crypto-engine framework for this (and how to trigger) or do I have to write my own queue handling??

I need way more info!

I thought input a lot of info already. Anyway reading the crypto engine code it seems it’s hard coded to wait until an idle state. So I either have to patch the code or write my own queue handler. The last seems the best to do at the moment. For now Im debugging my code to have it woeling with the crypto engine. Replacing the calls for my own should not be to much throuble. I’m not sure if I can actually measure the improvement compared to wait until my hardware writes the result before processing the next request. I just feel that there is room for performance improvement.