The key press response is slightly delayed and sometimes nonexistent, especially when a key is pressed twice in a row or with a gap of one character. I tried with 3 different hubs, one of which has a power supply.
USB connection and disconnection seem to be registering alright (no error), when checked with dmesg --follow
Perhaps there is something consuming the CPU which runs the driver. Does serial console do this as well? The serial console has far fewer drivers involved, but would use the same CPU core. You might also post the output of “cat /proc/interrupts”. Perhaps run htop (or top if you don’t have that and don’t want to install it during the error condition) and see if it says the first core is maxed out.
The CPU usage is pretty low, including the X server, so that isn’t an issue. Can you copy the “/proc/interrupts” after typing some, and post that without the screenshot? If serial console works, then you could just log and run “cat /proc/interrupts” from the host PC.
Does ssh login work, and if so, does it have delayed characters?
ssh login works and there is no delay with ssh - a useful workaround, thanks!
I would still like to keep looking for a fix for using the keyboard directly connected to the tegra though.
I have attached copies of /proc/interrupts before and after typing some. I logged and ran “cat /proc/interrupts” over ssh, with some typing on the keyboard connected to the tegra. Does this make sense?
I upgraded to the 4.6.3 release of Jetpack using SDK manager (verified with “apt-cache show nvidia-jetpack”) . Hasn’t helped though.
Is there anything I’m missing?
I don’t see anything interesting in how IRQs run. If you run “dmesg --follow” from ssh, and then plug in the keyboard, what do you see as a result (try to use the keyboard a bit too and see if it adds to the dmesg)?