/dev/ttyAMA0 doesn't go away when those things are rmmod'd # cat /sys/class/tty/ttyAMA0/device/uevent DRIVER=uart-pl011 OF_NAME=serial OF_FULLNAME=/soc/serial@7e201000 OF_COMPATIBLE_0=brcm,bcm2835-pl011 OF_COMPATIBLE_1=arm,pl011 OF_COMPATIBLE_2=arm,primecell OF_COMPATIBLE_N=3 OF_ALIAS_0=serial1 OF_ALIAS_1=uart0 AMBA_ID=00241011 MODALIAS=amba:d00241011 I'm not finding an 'amba' kernel module, maybe that means it's compiled in. This looks like it's the source code: https://github.com/torvalds/linux/blob/master/drivers/tty/serial/amba-pl011.... I found a quick summary on this page, which rehashes stuff from the linux tree: https://cateee.net/lkddb/web-lkddb/SERIAL_AMBA.html If the serial port is straight on the arm chip, which I guess makes sense, it seems unlikely that it would be a serial port issue. More likely the chip connected to the serial port, which would mean the problem would be resolvable inside hciuart. Part of me is considering trying to put an oscilloscope on the pins. I have a little cheap pocket oscilloscope. -- But most likely the problem here is going to be that the chip is in a nonresponsive state that is documented in its specs manual, and can be reset by providing some pattern to the port. This seems really doable now. But also labor intensive. I guess the chip it is is described by that bcm43xx string, but it might make sense to look up the specific model number.