Keyboard handler stops being called for brief periods while typing

[KyleCardoza]

I'm sure it's something I'm doing or not doing, but as I tinker with Circle, I have noticed that as I type on the rpi 400's keyboard continuously, for brief, ~1-2 second periods, it stops responding, then starts up again. No debug information is printed on the serial console when this happens. I was wondering if anyone else had seen this happen, and what could be done about it?

[KyleCardoza]

Never mind, I figured it out -- it turns out even a DMA burst setting of "1" is sufficient to monopolize the bus enough to mess with USB keyboards -- so it's back to updating the framebuffer with memcpy(), I guess.

[rsta2]

Have you tried a DMA burst length of 0? That's a valid parameter.

[KyleCardoza]

Fascinating. It failed with burst length 0 as well, until I randomly tried initializing the CDMAChannel instance with DMA_CHANNEL_NORMAL instead of DMA_CHANNEL_EXTENDED. With DMA_CHANNEL_NORMAL it works fine at burst 0, 1, and 2. I haven't tested higher with DMA_CHANNEL_NORMAL, since you said it usually causes problems.

So what, exactly, is the distinction between DMA_CHANNEL_NORMAL and DMA_CHANNEL_EXTENDED that might be the cause?

[rsta2]

DMA_CHANNEL_NORMAL uses the older 32-bit DMA engine, while DMA_CHANNEL_EXTENDED uses the 64-bit DMA4 engine, which was introduced with the RPi 4. Please read section 4.6. in this document for details on the DMA4 engine.