On Sun, Aug 25, 2024 at 03:09:20PM +0100, Peter Maydell wrote:
> On Sun, 25 Aug 2024 at 12:35, Jason A. Donenfeld <ja...@zx2c4.com> wrote:
> >
> > On Fri, Aug 23, 2024 at 07:28:43AM +0100, Stafford Horne wrote:
> > > Also, I will wait to see if Jason has anything to say.
> >
> > So long as this doesn't change the assignment of the serial ports to
> > device nodes in Linux, I don't think this should interfere with much.
> > You might want to try it, though.
>
> It looks like this board already creates the fdt /aliases/
> node and puts uart0, uart1, etc, so that part should be OK.
>
> However I notice that the openrisc_sim_serial_init() code
> will always set the /chosen/stdout-path, so this means
> (unless I'm misreading the code -- I haven't tested) that
> the last UART we create will be the stdout-path one. Before
> this patch, that would be serial_hd(0), but after this it
> will not be. So I think we probably need to fix this too
> in the same patch, so that we only set stdout-path for uart0,
> rather than setting it and then overwriting it on all the
> subsequent calls. This patch on its own will change the
> stdout-path value I think.
Hi Peter,
I suspected the same and tested the theory. Now when running linux with
or1k-sim machine we get no stdout output from qemu. Upon debugging and
looking at dmesg via gdb I can see the wrong uart is getting setup in
Linux:
[ 0.080000] workingset: timestamp_bits=30 max_order=17 bucket_order=0
[ 0.100000] Serial: 8250/16550 driver, 4 ports, IRQ sharing disabled
[ 0.110000] printk: legacy console [ttyS0] disabled
[ 0.110000] 90000300.serial: ttyS0 at MMIO 0x90000300 (irq = 2,
base_baud = 1250000) is a 16550A
[ 0.120000] printk: legacy console [ttyS0] enabled
[ 0.120000] 90000200.serial: ttyS1 at MMIO 0x90000200 (irq = 2,
base_baud = 1250000) is a 16550A
[ 0.130000] 90000100.serial: ttyS2 at MMIO 0x90000100 (irq = 2,
base_baud = 1250000) is a 16550A
[ 0.130000] 90000000.serial: ttyS3 at MMIO 0x90000000 (irq = 2,
base_baud = 1250000) is a 16550A
[ 0.150000] NET: Registered PF_PACKET protocol family
[ 0.160000] clk: Disabling unused clocks
I will amend the patch.
-Stafford
