Re: uart(4) not working in QEMU

In article <49910C3D.90709_at_incunabulum.net> you write:
>Hi,
>
>I have been trying to test my kernel code in QEMU as it saves a lot of 
>time and effort.
>However, I have noticed since returning to my current project, that 
>sio(4) was deprecated in favour of uart(4).
>
>Whilst I updated my kernel configs to reflect this, I've noticed a lot 
>of problems with I/O and QEMU -- in particular, the kernel will log 
>messages over uart(4) just fine, but when the kernel runs init, I can't 
>get any I/O out of the uart(4) at all, apart from a single 'c' or 'F' 
>character.
>
>The kernel continues to log messages OK to the uart0/ttyu0 device 
>regardless of what's going on in userland.
>
>If I configure ttyv0 in the QEMU virtual machine up via /etc/ttys to run 
>a getty there, I can get in, and see that the getty for ttyu0 is 
>running. However, echo'ing or cat'ing data to /dev/ttyu0 won't work, 
>even if I kill the getty process first. I just don't see anything 
>appearing in my QEMU serial console.
>
>I've tried a lot of combinations of 3wire.115200 vs std.9600, 
>boot.config options, loader.conf options, none of which have solved the 
>problem (mostly working from the threads on this list from when the 
>changes were made).
>
>I have also tried other bindings for the QEMU serial device -- e.g. tcp 
>ports, nmdm(4), and always see the same effects.   I do have INVARIANTS 
>enabled -- could this be an issue?

I dunno if INVARIANTS changes the behaviour of the uart driver (my guess
is it doesn't), but I see there have been commits to qemu's hw/serial.c
since the versions in ports so you could try a more recent svn snapshot
like the one posted here:
	http://lists.freebsd.org/pipermail/freebsd-emulation/2009-February/005650.html

 If that doesn't help and you feel like debugging this maybe uncommenting
the DEBUG_SERIAL #define in hw/serial.c helps...

 Good luck,
	Juergen