* Tony Lindgren | 2014-07-17 00:09:00 [-0700]:

Seems to boot a bit further now with output from serial console
initially, then I'm getting the following error again that's probably
related to clocks not enabled when the registers are accessed:

It is (mostly) the same thing as before. We have additionally 
omap_8250_startup() in the backtrace but it is the same thing.
So you  say I miss a clock? Looking through serial8250_do_startup() I see:
- pm_runtime_get_sync(port->dev); which should get the clocks up
- serial8250_clear_fifos() which does a write at address + 8. Seems to 
  work. 
- serial_port_in(port, UART_LSR); does a read at address + 0x14, seems 
  to work.
- serial_port_in(port, UART_RX); does a read at address + 0. This is 
  probably the bad one.

Now comparing with omap-serial I noticed that I do a 32bit access
instead a 16bit.
Could you please try the following hack:

diff --git a/drivers/tty/serial/8250/8250_core.c b/drivers/tty/serial/8250/8250_core.c
index 2e4a93b..94af5a3 100644
--- a/drivers/tty/serial/8250/8250_core.c
+++ b/drivers/tty/serial/8250/8250_core.c

@@ -420,13 +420,13 @@ static void mem_serial_out(struct uart_port *p, int offset, int value)
 static void mem32_serial_out(struct uart_port *p, int offset, int value)
 {
 	offset = offset << p->regshift;
-	writel(value, p->membase + offset);
+	writew(value, p->membase + offset);
 }
 
 static unsigned int mem32_serial_in(struct uart_port *p, int offset)
 {
 	offset = offset << p->regshift;
-	return readl(p->membase + offset);
+	return readw(p->membase + offset);
 }
 
 static unsigned int io_serial_in(struct uart_port *p, int offset)

besides that, I don't see what could be different.

Regards,

Tony

Sebastian