So, there's this somewhat suboptimal situation:

In a project on a ZynqMP / Ultrascale+, running Linux on the CPU and the FPGA doing application specific stuff (of course), we have a running system so far.

I do application software and some bare metal MCU programming, then there's an FPGA guy. Neither of us more than superficially touched Linux kernel stuff.
The guy who was a bit of a Linux kernel / driver guy is no longer with us, though.
While there were no plans to change the custom logic stuff where we'd have reckoned with needing a replacement for him (not easy to find), we added an extra 16550 UART, which should just use a mainline kernel driver and all is good.

Well, we did that and the kernel halts during boot.
We passed, via u-boot, kernel args like "debug" and "ignore_loglevel". The extra information shown in the log did not help.
Basically there were a few lines of "ZynqMP DMA driver Probe sucess" and then the last line "ttyS0 at MMIO ... (irq = 55, base_baud = ...) is a 16550A",
and then it simply halts.

How would one proceed from here to find out what exactly the kernel doesn't like about the HW change?
Is full blown source level kernel debugging sensible here? It seems a bit involved, to simply get an idea about the decision to halt (if that was a decision and not some wild crash).

If there is no kernel stack trace, then it doesn't seem like a software problem in the kernel. Since it's crashing right where you enable the UART, you may have a hardware issue. Perhaps it is coming up in an incorrect state.

1 members found this post helpful.

But would one gain any insight about that "coming up in an incorrect state" by engagnig any sort of debugging mechnisms?

I did notice now that there is an entry about 8250 (also counting for 16550, apparently) UARTS to have the max number of 4, while the system now apparently has 4 that I surely know are of that type, and 2 of ATM unknown to me type, which cuold be that. Though AFAIK only one is new, and the system worked then apparently with 5 UARTS before. Well, maybe only this one new is also of the type.
Would Linux then simply halt (and not just warn) at boot, if there are 5 such UARTS and the max number is set to 4? (e.g. CONFIG_SERIAL_8250_NR_UARTS=4)