Re: Serial Flow Control

[Rocker Dude <life.does.rock@gmail.com>]

Hi Alan,

Thank you very much for the prompt reply!

I debugged out the problem though. I will describe it, even if it
seems childish!

I was using the flow control on the console serial. As a result, any
data being sent to the powerpc from the x86 was also being read by the
console driver, and being responded to. As a result, many of times, my
actual serial read application would lose bytes and sit idle. I
changed the console port and observed data validity perfectly.

However, I have some few queires.

1. To prove flow control, I sleep for 5 seconds after every read.
Although my logic says that both the xmitter and the reciever in this
case should complete the process within a few time difference, I find
that my xmitter exits out after sending the chunk; while my receiver
keeps on receiving it for a few more time. And this time is almost a
minute or something. I did find out the xmitter 4K page buffers, but
it was too misleading for me to believe this!

Alan, I will once again go through the discipline drivers more before
I shoot may doubts!

Thanx a lot for the help and replies,

Thanx!
Sundar

On Tue, Mar 17, 2009 at 2:27 PM, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> On Tue, 17 Mar 2009 11:04:36 +0530
> Rocker Dude <life.does.rock@gmail.com> wrote:
>
>> Hi All,
>>
>> I am using a powerpc8548e custom hardware and testing flow control on
>> it. I have cross-wired the RTS/CTS of the board with my host terminal.
>> I find that when I am reading from the powerpc, it does not change the
>> CTS signal when it is running out of buffers and hence it is dropping
>> bytes.
>>
>> I referred to something similar at http://lkml.org/lkml/2007/7/26/46.
>>
>> Can anyone please tell me if this flow control needs to be done by the SW?
>
> Flow control needs to be enabled by the user space software (if not
> enabled already), managed by the line discipline and implemented by the
> device driver (or automatically in hardware in some cases)
>
> Basically; App "set CRTSCTS"
> Ldisc: buffers full/not full logic for throttling
> Driver: throttling events -> flow control, honour flow on write
>