Wolfgang Denk wrote: 
In message <48936FD4.3010802@harris.com> you wrote:
  
I have verified that the Sequoia (440EPx) does not have its UARTs
properly configured.  The attached patch corrects this by setting three
bits in SDR0_PFC1 to enable 4-wire mode, and to select cts/rts
functionality for the UARTs.  Also, I modified the GPIO settings for
    

We definitely do NOT want any hardware handshake on the serial
console. Never.

Best regards,

Wolfgang Denk

Perhaps my comments were not clear.  Please let me try again:  The schematic for the Sequoia shows two uarts.  U-boot leaves the SDR0_PFC1[U0IM] bit cleared to 0, which means that the 440EPx will only have one uart, operating in 8-wire mode.  So, U-boot does not set the CPU to the correct mode to enable two uarts.  This is independent of whether you want RTS/CTS or not, and as far as I can see, it must be fixed if both uart ports are going to work.

This also applies to the GPIOs.  They are not set correctly to connect the uarts to the I/O pins - the wrong functions and polarities are selected.  This too is dictated by the schematic.  The wires go where they go, and the GPIOs should be configured to match the schematic (or the schematic should be changed to match the software, but we know that isn't going to happen).  :-)

The remaining point is the SDR0_PFC1[U0ME] and SDR0_PFC1[U1ME] bits, which determine whether the control signals are RTS/CTS or DCD/DSR.  These signals are wired to the DB-9 jacks on pins 7 and 8.  According to the RS232 standard, these pins are RTS/CTS, and there is nothing software can do to change their meaning.

So, I believe my patch should be applied as written, so that the software and hardware agree as to function.

That said, what we choose to do with the lines is up to us.  If we don't want to enable RTS/CTS processing in the drivers, that is fine.  We can leave the function disabled.  But the hardware registers must be configured to at least match the schematic wiring.  Anything else makes no sense.

    Steve