nvt
|
539ff33104
|
Turn off debugging
|
2011-02-13 21:36:07 +01:00 |
|
nvt
|
0df450f0dc
|
Add the NEIGHBOR_ATTR prefix to a macro name
|
2011-02-13 21:34:59 +01:00 |
|
nvt
|
5ec2f19dc9
|
Remove only the worst parent if the table is full when processing a DIO.
|
2011-02-13 21:33:47 +01:00 |
|
Adam Dunkels
|
432928d000
|
Changed the name of macros defined in neighbor-info.h to have the NEIGHBOR_INFO prefix.
|
2011-02-13 21:29:59 +01:00 |
|
Adam Dunkels
|
acb5480aa9
|
Changed the name of macros defined in neighbor-info.h to have the NEIGHBOR_INFO prefix.
|
2011-02-13 21:21:14 +01:00 |
|
Adam Dunkels
|
ed3995960d
|
Changed the name of macros defined in neighbor-info.h to have the NEIGHBOR_INFO prefix.
|
2011-02-13 21:21:13 +01:00 |
|
Adam Dunkels
|
15bf73ba08
|
Use DAG_RANK() macro to compute rank
|
2011-02-13 19:01:15 +01:00 |
|
Adam Dunkels
|
e980072817
|
Don't add neighbor as a potential parent if its rank is too high.
|
2011-02-13 18:05:28 +01:00 |
|
Adam Dunkels
|
2a96835c05
|
Don't remove all parents that have a worse rank than an incoming DIO: we may inadvertently remove our preferred parent. Instead, remove parents with a rank that is worse than our preferred parent. If we are lucky, this will give us enough space for the new parent.
|
2011-02-13 14:16:15 +01:00 |
|
Adam Dunkels
|
f1ac94e415
|
Slight code complexity reduction
|
2011-02-13 14:14:49 +01:00 |
|
Adam Dunkels
|
3a824d0ed6
|
Added a hysteresis when choosing a new parent: don't switch parent unless the ETX is significantly better (1.5 ETX better).
|
2011-02-13 14:14:16 +01:00 |
|
Joakim Eriksson
|
48fb9dda9e
|
fixed of0 to respect min hop rank increase
|
2011-02-13 11:25:53 +01:00 |
|
nvt
|
20171985c6
|
Split the RPL header into a private and a public header.
|
2011-02-11 16:21:17 +01:00 |
|
Adam Dunkels
|
706045120f
|
Changed the variable name local_confidence to etx to better reflect what it was used for
|
2011-02-11 14:18:57 +01:00 |
|
Adam Dunkels
|
ba6bf7dd6b
|
Rewrote the parent selection to favour parents with a low ETX and a good rank. The old parent selection would always favour parents with a lower rank, regardless of their ETX, which in some cases lead to suboptimal parent selections. Also, the old code was buggy in that it would always pick the worst parent (the one with the highest ETX).
|
2011-02-11 14:17:26 +01:00 |
|
Adam Dunkels
|
bb8377dfee
|
Set the end-of-stream flag on the TCP FIN packet
|
2011-02-10 22:47:45 +01:00 |
|
Joakim Eriksson
|
d24c63c0f1
|
fixed so that minhoprankinc is configurable
|
2011-02-10 19:44:12 +01:00 |
|
Adam Dunkels
|
8d55a68ebe
|
Added a PACKETBUF_ATTR_STREAM_END packet flag
|
2011-02-08 22:20:07 +01:00 |
|
Adam Dunkels
|
46348d9013
|
Added the low-power streaming mode where frequent channel checks are performed when streaming is on. Enabling this will (by setting WITH_STREAMING to 1), however, significantly slow down Cooja/MSPsim simulations.
|
2011-02-08 21:57:46 +01:00 |
|
Joakim Eriksson
|
d518c84471
|
made rank calculation depend on selected parent when joining DAG
|
2011-02-08 19:59:50 +01:00 |
|
Nicolas Tsiftes
|
aeca16d01f
|
Just update the node rank and let rpl_process_parent_event take care of the rest
|
2011-02-07 17:29:23 +01:00 |
|
Nicolas Tsiftes
|
e3638d2791
|
Updated to the minrank-hysteresis OF
|
2011-02-04 15:46:01 +01:00 |
|
joxe
|
389a4f9220
|
added rpl-all-nodes multicast address to nodes mcast addresses
|
2011-01-25 22:41:30 +00:00 |
|
adamdunkels
|
09c223f549
|
Updated Contiki X-MAC to use hardware acks, which allows for tighter timing and results in a lower duty cycle. Hardware acks also make X-MAC more robust against packet losses, as there is a direct way for the CSMA layer to detect loss of data packets. Before, Contiki X-MAC only detected loss of strobes.
|
2011-01-25 14:31:09 +00:00 |
|
adamdunkels
|
d51e99a3dd
|
Updated streaming mode to use only about 20% radio. Made guard time more conservative due to problem report by Ricklef Wohlers.
|
2011-01-25 14:29:46 +00:00 |
|
adamdunkels
|
fa524fabd7
|
Bugfix: q->buf was not tested for NULL, as it should be. Fix by Simon Duquennoy.
|
2011-01-25 14:24:38 +00:00 |
|
joxe
|
e79c0ab1ee
|
fixed rpl so that it sends and receives on the rpl-all-nodes multicast address
|
2011-01-25 09:55:12 +00:00 |
|
joxe
|
3a643fa0c6
|
added address macros for rpl-all-nodes
|
2011-01-25 09:54:03 +00:00 |
|
adamdunkels
|
4b4b8fad56
|
More recent packet entries. Allow 4 queue entries to be dedicated to generated packets.
|
2011-01-18 16:05:53 +00:00 |
|
adamdunkels
|
808c21fbfa
|
Bugfix: default should be 3 MAC-level retransmissions if collissions or loss occurs
|
2011-01-18 16:03:57 +00:00 |
|
adamdunkels
|
affba785b5
|
Enabled phase-lock only for channel check rates of 32 Hz or lower
|
2011-01-18 16:03:19 +00:00 |
|
adamdunkels
|
26e3b6fdc0
|
Bugfix by Simon Duquennoy: don't transmit packet until previous packet has been sent by the RDC layer.
|
2011-01-13 19:06:22 +00:00 |
|
adamdunkels
|
81d23a3a36
|
Minor bugfixes pertaining to queue handling: if a packet was explicitly dropped, the packet would later be acknowledged despite it never being received; added a limit on how many queue entries that can be used for packets that are being forwarded, to allow for originated packets even if the forwarding queue is filled.
|
2011-01-10 15:10:43 +00:00 |
|
adamdunkels
|
70f9f075a0
|
Safeguard against NULL arguments
|
2011-01-10 15:08:52 +00:00 |
|
adamdunkels
|
148bd60fe1
|
Bugfix: the defensive retransmission timer setting must be made before the packet is transmitted, otherwise the retransmission timer will always be for the defensive retransmit
|
2011-01-09 23:49:27 +00:00 |
|
adamdunkels
|
50a3e6ea23
|
Increased the max ETX to allow for longer paths even with the initial conservative ETX estimate; made collect use announcements by default
|
2011-01-09 23:48:33 +00:00 |
|
adamdunkels
|
acf45ec648
|
Bugfix: memory wasn't freed when a neighbor was dropped from the list. Also increase the max age to 180 minutes to allow for regular beacons to refresh the neighbor before it is dropped.
|
2011-01-09 21:20:05 +00:00 |
|
adamdunkels
|
90dc0ceebf
|
Replaced default link estimator with an exponential weighted moving average, which reduces memory consumption per neighbor.
|
2011-01-09 21:14:22 +00:00 |
|
adamdunkels
|
4e91d9303e
|
Added safeguard timer if the MAC layer would fail to call back after transmission.
|
2011-01-09 21:13:20 +00:00 |
|
adamdunkels
|
a0960154ba
|
Cleaned up parts of the code; added comments
|
2011-01-09 21:07:01 +00:00 |
|
nvt-se
|
30f5e29db7
|
dos2unix conversion
|
2011-01-05 17:01:03 +00:00 |
|
joxe
|
65048c519c
|
removed some void * casts in uip code
|
2011-01-04 22:11:37 +00:00 |
|
nvt-se
|
7a43b7d832
|
Configuration of the objective function made easier. Updated OF0 to the new rank calculation.
|
2011-01-04 20:43:28 +00:00 |
|
joxe
|
b2296e462b
|
removed void * + changed ux_t to uintx_t
|
2011-01-04 20:22:21 +00:00 |
|
dak664
|
3c3cd868cd
|
Allow shorter UIP_CONF_WAIT_TIMEOUT for more aggressive power saving modes.
|
2010-12-24 00:39:04 +00:00 |
|
dak664
|
322d835fee
|
Fix avr-gcc compiler warnings
|
2010-12-18 22:12:53 +00:00 |
|
nvt-se
|
8f76c82433
|
Improved DAO management.
|
2010-12-17 15:24:25 +00:00 |
|
adamdunkels
|
4914739e5d
|
Revived the timesynch module, which has been switched off since the introduction of the new radio API after Contiki 2.4. The new code uses explicit timestamp messages instead of piggybacking timestamps on every transmission. It is also significantly better in terms of exactness.
|
2010-12-16 22:47:38 +00:00 |
|
adamdunkels
|
10fb99a89d
|
Reduced the value of the max depth, so that a count-to-infinity loop will not count too long before hitting the maximum value. This is important when a network loses its sink, so that packets are not forwarded around the network for ever.
|
2010-12-16 22:46:34 +00:00 |
|
adamdunkels
|
19e930e6f2
|
Changed the behaviour of retransmissions to work better with the in-order CSMA code. The retransmission timer is now set when the packet is known to be sent by the MAC layer.
|
2010-12-16 22:45:15 +00:00 |
|