Due to unknown reason on some platforms (e.g. 918+) the first serial port is not functional after boot. It's unknown if this affects other platforms and/or other serial ports.
The affected serial port will work for the kernel earlycon but it will fail as soon as the proper 8250 driver
initializes. Effectively this means you cannot set console= boot option to ttyS0 as it will go silent after a
short boot log. If console= specifies another port (e.g. ttyS2 which is free, unlike ttyS1 used by the PMU)
the boot process will continue as usual with the terminal appearing on such port.
It is unknown how/why is the ttyS0 broken. It may be an intentional change to prevent console access to a real box,
thus improving security and at least partially mitigating supply-chain attacks. We spaculate that if intentional it's
most likely hidden in plain sight in an overbloated 8250_core.c, maybe under MY_DEF_HERE constant.
While this is just a minor annoyance during debugging, for consistency a decent loader should fix it and have the
console on ttyS0 like a real box. We cannot be sure if some programs running in the OS don't assume ttyS0 is the
console.
See https://github.com/RedPill-TTG/redpill-lkm/blob/master/shim/uart_fixer.c for details.