As per discussion with Nicolas Sornin, any device time synchronization
value must be handled as is (monotonically increasing GPS epoch time guaranteed to be monotonic with granularity 1 second).
Once we receive a response to DevTimeReq (i.e., DevTimeAns is received)
we immediately set our local gps time stamp and take a snapshot of our
local monotonic (ticker or equivalent) clock. This is achieved by
calling 'set_gps_time(...)' API in the LoRaWANTimer module.
After that an event is generated to let the application know that a
device time has been synched.
LoRaWANTimer class is responsible for providing time base for the
LoRaWAN stack in Mbed OS. It derives its core monotonic clock from the
EventQueue which is essentially a millisecond ticker or equivalent (in
tickless mode). LoRaWAN network can assist the device for device level
time synchronization using ClockSynchronization protocol (for v1.0.2) or
by replying to DevTimeReq MAC command (for v1.0.3 and above). The time
base used for device synchronization using above methods is purely
based on GPS epoch and needs no conversion to UTC or TAI etc (no
adjustment for leap seconds either).
Therefore we have extended the LoRaWANTimer APIs to account for GPS time
base if assisted by the network.
get_gps_time() is the API to use to acquire GPS time base from the
stack. If the network never assisted the device with a time stamp, a
value of zero is returned which means GPS time base is not available.
set_gps_time(time) API is used to set a given GPS time-stamp when the
network sends it.
These APIs are hooked via LoRaMAC to LoRaWANInterface which exposes the
access to the application.
The application can use add_device_time_request() API to attach
DeviceTimeRequest mac command for the next outgoing packet.
It is a non-stick mac command, so if there is no response, the app can
retry. The API indicates if the mac command was queued successfully or
not. The actual indication of successful time and date update will come
through an event 'DEVICE_TIME_SYNCHED'.
While post processing a transmission in case of no reception, we already
take care of frame counters. We must not kick post processing step once
again.
* While starting timer for rejoin request type 0, we should multiply max
time with 1000 as the timer APIs take ms values as parameters.
* RJCountX are incremented every time a Rejoin request is sent. For MIC
calculation we need to take the previous RJCnt value in account.
* Rejoin process should start against an event otherwise it will meddle
with state machine and any ongoing traffic.
* If a Type 1 rejoin is ongoing, we should not trigger a Type 0 rejoin.
* Some bug fixes after Triage with Antti.
LoRaWAN specification 1.1.0 is incorrect in handling of FOpts encrytpion.
A CR was made by STMicroelectronics which correctly addresses the issue.
We have taken that CR and implemented in our code base.
BE to LE fixes, missing MLME types added
LoRaWAN 1.1 Features added (Some LoRaPhy impl missing still + some TODOs in code)
- MLME confirm handling refactored
- Rejoin handling missing
- new CF_LIST mechanism missing (+resets involved)
- NVM handling missing
Rejoin logic added
We provide now downlink channel frequency and time on air for the
received frame in the RX metadata.
Previously the channel information in both TX and RX metada contained
the index number of the channel. That information wasn't very useful
except the index numbers of default channels. To make more sense of the
meta data, we now store the channel frequency in the channel parameter
rather than the index number of the channel.
RX time on air is collected from the radio driver and it is assumed that
the downlink frame had 8 downlink preamble symbols (plus 4.25 of the
preambles added by the chip) for LoRa modulation.
This commit also include a bit of tidying of RX frequency storage in rx
configuration parameters storage. Previously we were missing filling in
the RX1 frequency correctly.
Previously we had been incrementing UL frame counter for a CONFIRMED
message only when the transmission was deemed successful i.e., we would
have received an ack before all the retries would have exhausted.
Now we have opted to increment the frame counter if all the retries are
exhausted considering the fact that we essentially treat the next
message after retry exhaustion as a new packet so we should also
increment the frame counter.
The idea behind the method post_process_no_reception() is to post
process any outgoing TX but we shouldn't do that if a CONFIRMED message
is outgoing and there are still some retries left.
After transmission we should change the state before invoking opening of
slots as we may start receiving in the rx slots and the state would
suddenly change from SENDING to RECEIVING without going through the
ACK_WAIT state (in case of CONFIRMED messages). Tests show that after
this slight adjustment, our number of ack retries have significantly
reduced.
LinkADRReq mac command can be used by the network server to set a
certain level of QOS using NbTrans field which is applicable to
Unconfirmed traffic only for 1.0.2 spec.
This commit introduces mechanisms to facilitate this QOS. It means to
repeat an outgoing unconfirmed message NbTrans times without changing
its frame counter.
For class C, we have retired the ack_expiry_timer_for_class_c and have
replaced it with another timer which mimics the RX2 closure as in Class
A but doesn't actually close RX2 window. It's just a mechanism by which
the state machine is informed that the you can proceed forward, we have
not received anything in RX2 window either. This is needed as RX2
doesn't timeout in class C (i.e., the radio remains in continuous mode).
In addition to that we need to close any pending timers for Receive
windows after the MIC has passed and the Duplicate counter check has
also been passed.
Before going after an automatic uplink, we should check if there was an
automatic uplink already ongoing, i.e., the ack for the previous
automatic uplink cycle has not been received.
If there is we shouldn't queue the new automatic uplink and wait for
the previous Ack cycle to complete.
If the packet is already handed over to the PHY layer, we shouldn't be
able to cancel that particular transmission. In addition to that if the
backoff timer is either not applied or has been deactivated, should end
up in no-op rather than having normal termination. A new error code has
been introduced to cover no-op cases. This error code replaces the
compliance test related error code which is no longer relevant.
clear_tx_pipe() does nothing if:
- The stack cannot cancel TX (already handed over to PHY)
- The backoff timer is not active at all
- The event is disaptched to schedule
stop_sending() will only post process ongoin TX if the pipe was
definitely cleared.
The idea behind the method post_process_no_reception() is to post
process any outgoing TX but we shouldn't do that if a CONFIRMED message
is outgoing and there are still some retries left.
After transmission we should change the state before invoking opening of
slots as we may start receiving in the rx slots and the state would
suddenly change from SENDING to RECEIVING without going through the
ACK_WAIT state (in case of CONFIRMED messages). Tests show that after
this slight adjustment, our number of ack retries have significantly
reduced.
LinkADRReq mac command can be used by the network server to set a
certain level of QOS using NbTrans field which is applicable to
Unconfirmed traffic only for 1.0.2 spec.
This commit introduces mechanisms to facilitate this QOS. It means to
repeat an outgoing unconfirmed message NbTrans times without changing
its frame counter.
For class C, we have retired the ack_expiry_timer_for_class_c and have
replaced it with another timer which mimics the RX2 closure as in Class
A but doesn't actually close RX2 window. It's just a mechanism by which
the state machine is informed that the you can proceed forward, we have
not received anything in RX2 window either. This is needed as RX2
doesn't timeout in class C (i.e., the radio remains in continuous mode).
In addition to that we need to close any pending timers for Receive
windows after the MIC has passed and the Duplicate counter check has
also been passed.
Before going after an automatic uplink, we should check if there was an
automatic uplink already ongoing, i.e., the ack for the previous
automatic uplink cycle has not been received.
If there is we shouldn't queue the new automatic uplink and wait for
the previous Ack cycle to complete.
While performing compliance tests with an industry tester, we realized
that there was no need for any extra handling code for the compliance
tests. The tests would run fine, if we only have a handling application.
However, in normal operation we wouldn't like the network to send us any
traffic on compliance testing port. To mitigate that, on the reception
path we filter out any traffic on compliance testing port if compliance
testing is not underway. User should define LORAWAN_COMPLIANCE_TEST macro
in mbed_app.json to enable traffic on compliance test port.
When application requests link check request with add_link_check_request()
it was added to MAC command buffer. Again when actual TX was started,
stack cherry-picked the request into MAC buffer again resulting in a duplicate
entry in the buffer.
Functionality is changed now so that add_link_check_request() does not add
the request into MAC command buffer but it is cherry-picked into the buffer when
TX is issued.
In order have a consistent return value for all methods in case of system is uninitialized
now also send and receive methods can return LORAWAN_STATUS_NOT_INITIALIZED.
Thix commit fixes a bug where stack was resending join-accept already
after RX1 if received join-response was invalid.
Stack should wait for RX2 before resending the join-accept.
When node sends a CONFIRMED message and gateway sends ACK in RX1 window but the
message gets corrupted during the transmission (e.g. MIC fails), currently
our stack already checks for retransmission after RX1 and if retries attemps are
exhausted, TX_ERROR event is sent to application. This is wrong as MAC layer
will still attempt reception in RX2 window.
This commit fixes the behaviour so that TX_ERROR is not sent until RX2 window
has been closed.
Roman Volosatovs
Unknown
Hasnain Virk
Antti Kauppila
Kimmo Vaisanen
Kevin Bracey
Martin Kojtal
Cruz Monrreal