should the kernel be initializing the uarts on the Au1550?

should the kernel be initializing the uarts on the Au1550?

[Clem Taylor]

It seems that the kernel (2.6.10) isn't properly initializing the
Au1550 serial ports. All three of the serial ports work, just not at
the same time. Linux seems to need yamon to configure the serial port
first. Out of the box yamon uses UART0 & UART3 and ttyS0 & ttyS2
(UART3, the 1550 doesn't have a UART2) work in linux, but ttyS1
doesn't. If I switch yamon to use UART1 & UART3, then ttyS0 doesn't
seem to work in linux. The serial port that isn't configured by yamon
will hang in an ioctl() on calling tcsetattr().

Before I just cheat and add a third serial port to yamon, should the
kernel be initializing the UARTs or does it really expect the yamon to
initialize them first? Is anyone using all 3 serial ports on an
Au1550?

                                  Thanks,
                                  Clem

Re: should the kernel be initializing the uarts on the Au1550?

[Dan Malek]

On Apr 26, 2005, at 3:58 PM, Clem Taylor wrote:

> It seems that the kernel (2.6.10) isn't properly initializing the
> Au1550 serial ports.

What do you mean by "initializing"?

> .... Linux seems to need yamon to configure ....

Linux assumes yamon for the Db1xxx and Pb1xxx
boards will configure nearly all of the IO pins
properly.  This is quite the opposite of most embedded
systems that assume nothing and rely on the Linux
board startup functions to do this.

It is just historical, and could be changed.  First, we
need to add sufficient configuration information so
we know what IO is used (serial ports in this case).
Second we need the functions to do the proper
set up based upon the selected configuration.

Then, the discussions (arguments) start about what
should be further assumed about initialization :-)
My preference is that if we should do all of the
initialization and assume nothing, because we could
probably eliminate lots of board specific set up
files and use more common software.

Thanks.


	-- Dan

Re: should the kernel be initializing the uarts on the Au1550?

[Pete Popov]

On Tue, 2005-04-26 at 15:58 -0400, Clem Taylor wrote:
> It seems that the kernel (2.6.10) isn't properly initializing the
> Au1550 serial ports. All three of the serial ports work, just not at
> the same time. Linux seems to need yamon to configure the serial port
> first. Out of the box yamon uses UART0 & UART3 and ttyS0 & ttyS2
> (UART3, the 1550 doesn't have a UART2) work in linux, but ttyS1
> doesn't. If I switch yamon to use UART1 & UART3, then ttyS0 doesn't
> seem to work in linux. The serial port that isn't configured by yamon
> will hang in an ioctl() on calling tcsetattr().
> 
> Before I just cheat and add a third serial port to yamon, should the
> kernel be initializing the UARTs or does it really expect the yamon to
> initialize them first? Is anyone using all 3 serial ports on an
> Au1550?

Yes, it should. And we should really rewrite the serial driver from
scratch. 

Pete