DCCo802.11

Official repository of all of the stuff mentioned into master thesis.

There are different folders:

• Patches: patches used for Rock and iPerf tweaking.
• Thesis tests: python scripts used for plotting (they are available, along with numeric results, into a proper repository).
• Various tests scripts: every script, used or not, is in this folder (along with some that are not used).
• iperf-2.2.0-patched: source code of iPerf v2.2.0 already patched.

iPerf w/ AC

A customized iPerf with Access Categories (BK, BE, VI, VO). Currently extremely work in progress. Because patch file comes from a different iPerf version, it must be applied manually (but in this case is already applied).

Rock3a tweaking

Rock3a ath9k debug mode

git clone --depth=1 --branch=main https://github.com/armbian/build 

cd build/

vim config/kernel/linux-rockchip64-current.config 

edit config file by adding:

CONFIG_ATH9K_HTC_DEBUGFS=y
CONFIG_ATH9K_HWRNG=y
CONFIG_ATH9K_DEBUGFS=y

Enable queues

cd ~/build/patch/kernel/archive/rockchip64-6.6

cp "Armbian patch"/00550-ac.patch .

More useful stats

cd ~/build/patch/kernel/archive/rockchip64-6.6

cp "Armbian patch"/00551-ac.patch .

N.b.: Then, you have to multiplicate cc->cycles with div = common->clockrate * 1000 if needed:

if (cc->cycles > 0) ret = cc->rx_busy * 100 / cc->cycles;

Look here.

Kernel building

Compile just the kernel with:

./compile.sh kernel ARTIFACT_IGNORE_CACHE='yes' BOARD=rock-3a BRANCH=current

After a long time, copy the *.deb files found in output/deb into rock and installing by *dpkg -i .deb (can be necessary, after, to launch apt install -f to fix possibily missing dependences).

Setting rock-3a Wi-Fi interface

sleep 15

ip link set dev wlp1s0 down
# imposta la tipologia come ad-hoc
iw wlp1s0 set type ocb
# accende l'interfaccia
ip link set dev wlp1s0 up

iw wlp1s0 ocb join 2462 10MHz
ip addr add 192.168.100.26/24 dev wlp1s0
ip route add default via 192.168.100.26 dev wlp1s0

Setting an IEEE 802.11p physical data rate/modulation

Normally, the user can set the physical data rate by leveraging the iw tool, and specifying:

iw dev wlan0 set bitrates legacy-5 <double of the desired bitrate value in Mbit/s>

This is also the procedure implemented in iw_startup to set an initial data rate to 3 Mbit/s, and this command can also be used in standard OpenWrt installations to set a desidered data rate with, for instance, other 802.11n or 802.11ac cards.

Unfortunately, we found out that this is not always possible with 802.11p and OCB mode, at least for certain desidered physical data rates, due to iw returning "Invalid argument (-22)" errors. These errors, and thus the selectable physical data rates, seem also to be dependant on the actual hardware and chipset revision, and on how it "reacts" after the physical data rate change request.

Concerning the DHXA-222 cards, iw dev wlan0 set bitrates legacy-5 lets the user select any of the mandatory physical data rates for IEEE 802.11p, as specified in IEEE 802.11-2020, i.e., either 3 Mbit/s (iw dev wlan0 set bitrates legacy-5 6), or 6 Mbit/s (iw dev wlan0 set bitrates legacy-5 12), or 12 Mbit/s (iw dev wlan0 set bitrates legacy-5 24).

With other cards, however, this is not always guaranteed to work. If it happens that you card does not let you select at least the mandatory rates, you should be able to rely on a workaround, which involves forcing a fixed data rate index in the Minstrel rate adapation algorithm used by Linux.

You can set a desired fixed data rate with:

echo <index> > /sys/kernel/debug/ieee80211/phy0/rc/fixed_rate_idx

Finding the right indeces for the IEEE 802.11p rates may require a bit of work (the values are not so well documented, and may also be driver dependant - we are, however, investigating this point!). The following are some values we found, with the corresponding IEEE 802.11p data rates, for some AR9642-based mPCIe cards we tested (i.e., Atheros AR5B22):

  |INDEX|  -> |PHYS. DATA RATE|
4294967288 ->     3 Mbit/s
4294967289 ->   4.5 Mbit/s
4294967290 ->     6 Mbit/s
4294967291 ->     9 Mbit/s
4294967292 ->    12 Mbit/s
4294967293 ->    18 Mbit/s
4294967294 ->    24 Mbit/s

Final tests

• C0: Caso senza congestione: c1 e c2 spenti
• C1: Caso traffico stradale: c1 e c2 con cam ogni 5 ms in broadcast
• C2: Caso congestione severa: c1 client e c2 server con trasmissione udp in flooding
• T1: Caso t1 e t2 fanno cose 1: iperf senza qos con una connessione con due flussi tcp (t1 client e t2 server)
• T2: Caso t1 e t2 fanno cose 2: come T1 ma usando VO e BK (t1 client e t2 server)

I test da eseguire sono:

1. T1 in contemporanea a C0
2. T1 in contemporanea a C1
3. T1 in contemporanea a C2
4. T2 in contemporanea a C0
5. T2 in contemporanea a C1
6. T2 in contemporanea a C2

Evaluation tests

Test 1 (iPerf powered)

The Test 1 consists in the running of multiple instances of iPerf; one device have to operate as a "server" and the other ones as a "client". We can use different class of priorities.

• Server:

  iperf -s -u -i 1

• Client (VO - VI - BK):

  iperf -c 224.0.67.67 -t 600 -u -b 10m -A VI

Results

See test1.xlsx.

Test 2 (ncat powered)

Be aware! To use -q 0 parameter you must install netcat-openbsd instead of netcat-traditional (this one lacks of it).

The Test 2 consists in the running of multiple instances of netcat, in order to try to create congestion on the medium. Then, it tunes the number of packets transmitted accordly to a sort of pioneering DCC implementation.

We use on_your_mark.sh script to be able to launch the script on every devices at the same time; the scripts startup are triggered by go.sh script launch.

• Nodes:

  "Usage: ./on_your_marks.sh <max_concurrent_messages> <message_size>"

  ./"Test Scripts"/on_your_marks.sh 1 5 350

• Start:

  ./"Test Scripts"/go.sh

Results

See test2.xlsx.

Lost packets test

Test to check the number of lost packets.

• Server:

  ./"Tests scripts"/test2_1.sh

• Client:

  "Usage: $0 <delay> <number_of_packets> <packet_size>"

  ./"Comparison Scripts"/receiver.sh 1 5 530

• Comparison:

  To make comparison between tx and rx packets.

  ./"Comparison Scripts"/comparison.sh

Results

Currently work in progress...