SCC used for WAN connection

[hugh.mcdonald at nokia.com]

Hi John & Dan,
   See Below for comments....

>Subject: Re: SCC used for WAN connection
>
>
>john zhan wrote:
>
>> me too.  in fact,   also, I need generic x.25 running on it.
>
>So, write it.
>
>
>>
>> >    What is needed is a new version of 8xx_io/uart.c that
>sets up and drives
>
>No, what you want is an SCC synchronous driver. Period.  Don't put
>stuff like this in the uart driver, it is a big enough mess the
>way it is.  I am close to having a really configurable uart driver,
>so you can select what SCC/SMCs you want.  The uart driver is a uart
>driver, not an SCC driver.

Sorry, my words were ambiguious, I did not actually intend to add to uart.c.
Dan is correct , sync does not belong in a uart driver.
I planned to write a new driver (based on uart.c skelaton). Also I was
planning to make it message/frame based not character based which also has
big impact on the design/functionality.

>
>We will probably have to create some other functions for managing the
>baud rate generators, but for synchronous I/O that clock should be
>provided externally.
>
>> > other possible solutions....
>> I think we can  work together .
>> If you like,let's discuss it later , privately.

john,
  I would like to further discuss with you.  I am on a training course for
the next few days so maybe sometime after that...
>
>I would strongly suggest not going off and making a big mess, then
>sending me a patch and hope it gets applied to the kernel sources.
>
  Agreed

>I have done lots of synchronous SCC, HDLC, and other things as custom
>proprietary drivers for many customers.  Typically, the driver does
>very little, just buffers the data for mmap()'ed applications that
>do the real work.  I will see if I can provide some framework from
>past work I have done, but some of these folks are quite protective
>of this.
>
Any code snippets would be more than welcome!

>These drivers are not hard to write, and you can use examples from
>Motorola's NetComm site to get started.  You may need to do some
>uncached memory buffering tricks in Linux or just use cache management
>functions to ensure cache coherency.
>
>Don't use old silicon, like before Rev. C 860.  Most of these functions
>don't work as advertised on SCC3 and SCC4, only on SCC1 and SCC2.  Make
>sure you pay close attention to the external hardware, as clock and
>data transition edges are criticaly to this working correctly.

Thanx for the info I shall check my hardware revs.

>
>	-- Dan
>

Cheers
    Hugh McD

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/