ng_slip: initial import

[miri64]

Updated version of #1454. I removed the board_uart0 support for now, to keep it simple.

Depends on #2575. (merged)

Needs #2901 to build properly (merged)

[haukepetersen]

this will not work (you will lose chars). Use uart_write_blocking() when you use the uart like this. Better (and I really would to this here): use the transmit interrupt and a transmit buffer!

static ringbuffer_t *out_buf;

_slip_send(..) {
    ...
    default:
    ringbuffer_add_one(&out_buf, data[i]);
    uart_tx_begin(uart);
    ...
}

// and
int _slip_rx_cb(void *arg) {
    if (_out_buf.avail > 0) {
        char c = (char)ringbuffer_get_one(&out_buf);
        uart_write(uart, c);
        return 1;
    }
    return 0;
}

[miri64]

you mean tx_cb, right?

[miri64]

The problem I have with this is, that we than need a second buffer of at least size IPV6_MTU.

[haukepetersen]

yes

[haukepetersen]

and true for the buffer size. But the blocking use of the UART interface is very slow. This basically means, that for sending one full IPv6 MTU, you controller basically blocks for about 10ms...

[haukepetersen]

You should spend the SLIP interface a device descriptor to make it possible to run more than one instance at a time! This could be interesting for setups, where one interface is connected to linux, and another one is connected to another RIOT based board - some of our boards have up to 8 UART interfaces :-)

[miri64]

I was thinking about it, but than thought it might be an interesting challenge for future generations :D

[haukepetersen]

hm, why not do it right in the first place? So others don't even get the idea to implement drivers with global variables (and the SLIP device would be in line with all new readio and device drivers...).

[miri64]

You saw that I did it, right?

[haukepetersen]

Ups, did not. But the new version does not work -> you will have data garbage as to shared uart buffers... I would really do it the same way we do any other device driver: declare the device descriptor in the header file and include the uart buffers with full size in this struct. Then it is up to the user of a slip device to allocate this device descriptor (not on the stack!) and just give the pointer to this memory to the init function.

[miri64]

The uart buffers are not shared, they are uart local, but I can make the struct public. As for the full size data buffers: I chose very carefully not to include the size of the ringbuffers into this PR, simply because it mainly depends on the network layer what size is required.

[miri64]

Done

[miri64]

Sorry, somehow there was a foreign commit in my rebase

[miri64]

Rebased to current master

[miri64]

Rebased to current master

[miri64]

Fixed a whitespace error

[miri64]

Rebased to current master

[miri64]

Rebased to current master and #3005.

[miri64]

Another day another rebase :D

[miri64]

Rebased to current master… can we FINALLY merge this (this still needs a re-ACK) (no it doesn't)

[miri64]

Travis fails in unrelated applications. What's happening oO @phiros @OlegHahm @x3ro @gebart?

[miri64]

Ah, a rebase error was the culprit. Let's see if it works now.

[miri64]

This PR still fails due to the license checker… I added a pattern file which is to my knowledge correct, but it still fails… @LudwigOrtmann @OlegHahm @mehlis help!

[miri64]

Okay… don't know exactly why, but it seems to be fixed... Now cppcheck complaining about the Contiki code… do I have to fix those?

[jnohlgard]

Ignore the cppcheck warnings/errors in tunslip6 (Contiki). We'll fix it in a follow up PR.

[miri64]

Ok, then I'll merge

[miri64]

\o/

[OlegHahm]

Sorry to being a bit late: but this is copy pasta, right?

[miri64]

Yes... Should I open a PR or did you already?

[OlegHahm]

Nope, I didn't, so feel free.