Reading accurate size of recepts from serial port

Reading accurate size of recepts from serial port

[Laurent Hugé]

Hi,

I feel sorry to annoy you again with my problem, but I can't imagine there is 
no way to know the accurate size of a recept on the serial port.

I'm trying to implement a network driver above this port :
	- I've done a raw read from 0x03f8 but it was not fast enougth (I'm working 
at 115200 bauds) ;
	- I've created a line discipline, but it can't (according to this mailing 
list) give me the real size of what was read.
Does anybody know a way to read from the serial port with speed and accuracy ?

I'm really confused because the serial driver on Windows can do such a thing, 
and, since I've no other way to find the size of PDU (CCSDS segments), I'm 
obliged to rely on the serial port to let me know that size.
Thanks in advance,
-- 
Laurent Hugé.

Re: Reading accurate size of recepts from serial port

[Russell King]

On Tue, Aug 26, 2003 at 10:32:13AM +0200, Laurent Hugé wrote:
> I feel sorry to annoy you again with my problem, but I can't imagine there is 
> no way to know the accurate size of a recept on the serial port.

The serial driver can not and does not know in advance how many characters
the other is going to send.  As far as the serial driver is concerned, it's
just a meaningless stream of characters.

To give an example, if your device sends the following character stream:

  first                                               last
    v                                                   v
    .....................................................

your line discipline may receive this as four separate blocks:

    ................                                       block 1
                    ................                       block 2
                                    .................      block 3
                                                     ....  block 4

It may not receive it like the above - it may be several blocks of 8
characters or whatever, depending on the UARTs FIFO, interrupt load,
etc.

There is just no way for the serial driver itself to batch them back
up.

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: Reading accurate size of recepts from serial port

[Laurent Hugé]

Le Mardi 26 Août 2003 10:47, Russell King a écrit :
> There is just no way for the serial driver itself to batch them back
> up.
Ok. Thanks for your explanation. I'm going to try and find another solution.
-- 
Laurent Hugé.

RE: Reading accurate size of recepts from serial port

[Stuart MacDonald]

From: linux-kernel-owner@vger.kernel.org 
> Le Mardi 26 Août 2003 10:47, Russell King a écrit :
> > There is just no way for the serial driver itself to batch them back
> > up.
> Ok. Thanks for your explanation. I'm going to try and find 
> another solution.

I may be mistaken, but I believe that Windows serial drivers work the
same way; so whatever you meant by your previous comment that you can
get what you want under windows, either you can get the same thing
under linux, or windows doesn't behave like you think it does.

..Stu

Re: Reading accurate size of recepts from serial port

[Laurent Hugé]

Le Mardi 26 Août 2003 16:22, Stuart MacDonald a écrit :
> I may be mistaken, but I believe that Windows serial drivers work the
> same way; so whatever you meant by your previous comment that you can
> get what you want under windows, either you can get the same thing
> under linux, or windows doesn't behave like you think it does.
I actually don't know how it works, because I didn't contrive it (I can only 
rely on what has been told to me by the people whom has done it). I'm only in 
charge of porting it to Linux.
Anyway, I'm on the way to another solution (by using some property of CCSDS 
segments, and see what happens).
-- 
Laurent Hugé.

Re: Reading accurate size of recepts from serial port

[Herbert Pötzl]

On Tue, Aug 26, 2003 at 05:23:04PM +0200, Laurent Hugé wrote:
> Le Mardi 26 Août 2003 16:22, Stuart MacDonald a écrit :
> > I may be mistaken, but I believe that Windows serial drivers work the
> > same way; so whatever you meant by your previous comment that you can
> > get what you want under windows, either you can get the same thing
> > under linux, or windows doesn't behave like you think it does.
> I actually don't know how it works, because I didn't contrive it (I can only 
> rely on what has been told to me by the people whom has done it). I'm only in 
> charge of porting it to Linux.
> Anyway, I'm on the way to another solution (by using some property of CCSDS 
> segments, and see what happens).

hmm, why not do simple framing ...
[length]<data>[length]<data> ....

best,
Herbert

> -- 
> Laurent Hugé.
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: Reading accurate size of recepts from serial port

[Laurent Hugé]

Le Mercredi 27 Août 2003 16:50, Herbert Pötzl a écrit :
> hmm, why not do simple framing ...
> [length]<data>[length]<data> ....
That's impossible. CCSDS is the committee for space date systems and it 
provides standards that I can't overrule (even if I can't really understand 
why they've done it like that !).
-- 
Laurent Hugé.

Re: Reading accurate size of recepts from serial port

[Richard B. Johnson]

On Wed, 27 Aug 2003, Laurent [iso-8859-1] Hugé wrote:

> Le Mercredi 27 Août 2003 16:50, Herbert Pötzl a écrit :
> > hmm, why not do simple framing ...
> > [length]<data>[length]<data> ....
> That's impossible. CCSDS is the committee for space date systems and it
> provides standards that I can't overrule (even if I can't really understand
> why they've done it like that !).
> --
> Laurent Hugé.
>

The transfer frame in your reference specified, contains all
the information necessary for the protocol, even if it's stupid
to use that protocol on a RS-232C link. Nevertheless, there is
a minimim size for the header (5 octets in length). There is
also the 3 octets used for sync, which I'm pretty sure will
not be put onto the RS-232C links. Anyway, you need to read
40 bytes (always), from that, you will learn the length
of the rest of the data. So you use poll()/read() until you
get that header information. Then you will know what the
total read-length should be.

Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (794.73 BogoMips).
            Note 96.31% of all statistics are fiction.