We get this compiler warning;
Compile [ 7.9%]: AT_CellularDevice.cpp
[Warning] AT_CellularDevice.cpp@206,18: variable 'err' set but not used [-Wunused-but-set-variable]
Due to the fact, that the code that would actually use this variable
is behind trace flags. Based on review feedback from Antti Kauppila and
Kimmo Väisänen, adding same flagging also for that.
- Reason being - the operation needs to be within the locks.
- When calling socket APIs when socket is not open, NSAPI_ERROR_NO_SOCKET will be
returned instead of generic NSAPI_ERROR_DEVICE_ERROR
- If socket_send() is called when connection is not open, NSAPI_ERROR_NO_CONNECTION
will be returned instead of generic NSAPI_ERROR_DEVICE_ERROR
Some external modems have an internal TLSSocket implementation which can be used
instead of mbedtls based TLSSocket. Using offloaded TLSSocket can result in
significantly reduced ROM usage.
Offloaded TLSSocket can be enabled by enabling "nsapi.offload-tlssocket" and the used
network stack (e.g. cellular modem's CellularStack class) must support the setsockopt's
defined in nsapi_types.h.
Compared to original mbedtls based TLSSocket, offloaded TLSSocket brings in one significant
API limitation. Offloaded TLSSocket requires setting of certificates and keys after open()
and before connect() calls, where mbedtls based TLSSocket allows setting these before open()
call.
QISEND command can respond either SEND OK, SEND FAIL or ERROR.
If response is not SEND OK, sent bytes should not be checked but
error should be reported.
Possible responses for send command are SEND OK<cr><ln>, SEND FAIL<cr><ln> or ERROR<cr><ln>
so normal OK<cr><ln> response check does not work properly.
Multihoming documentation about interface name:
"Two character name string is concatenated with 8 bit value containing index which is incremented on each netif addition"
Cellular uses context id as index and to follow LWIP (LWIP::Interface::get_interface_name), index does not include leading zeros.
A new API `CellularDevice::clear()` to clean-up the modem to a default initial state.
Function is virtual so it can be overridden. The default implementation clears all PDP contexts,
but the the first one if that has APN defined as `nsapi.default-cellular-apn`.
CellularStateMachine calls `clear()` to clean-up the modem on initial `connect()`,
if the flag `cellular.clear-on-connect: true` is defined.
In some multithread cases there is possibility that process_oob function
was called after ATHandler was deleted. Fix is to wait if oob processing
is ongoing.
New ATHandler functions taken into use for rest of the targets (BG96 was updated initially) to reduce code size. This means basically that new functions using variadic list approach are taken into use and with those one can usually write AT commands in single line instead of multiple lines.
Only internal changes and API's are not modified.
When AT+CGML is used to retrieve list of SMS stored in modem inbox,
every message has an associated index. ETSI TS 127 005 v7.0.0 does not
specify what is the allowed range of such indices - all it says is
"integer type; value in the range of location numbers supported by the
associated memory".
Usually, AT modems use positive indexes (starting at 1). Quectel BG96
modem takes a different approach, indexing messages starting at 0.
Current implementation of `AT_CellularSMS::list_messages()` considers
index 0 invalid and ignores such message, effectively making it
impossible to access using mbed-os API.
This commit changes the behavior so that value of 0 is handled as any
other positive message index.
Currently, create_pdu receives a destination address without '+' prefix,
and always sets the "type of address" to "unknown". That means, the
number needs to contain appropriate international number prefix (00/011)
if necessary - which is not the case if the leading + is simply
stripped.
This changes send_sms behavior so that when a SMS is sent to an
international number (indicated by leading +):
- AT+CMGS command receives the number with + prefix,
- created PDU has the "international" flag set.
CellularContext now tries to get an IP address after connect and before
sending NSAPI_STATUS_GLOBAL_UP. Even if we don't the IP address from
the modem we will send NSAPI_STATUS_GLOBAL_UP and return success.
Modem has an ip address but for some reason some modems don't
give it to us.
AT commands used in read_radio_technology() function are not supported
by all boards so moving the function where it's actually used. Usage was
already inside of the #ifdef's.
If delete fails there is nothing we or application can do. There is no point
returning an error for this. This affects to AT_CellularContext::disconnect
not returning an error if context delete fails.
ARMC5 failed to compile the code with debug-profile (!!) as va_list
is getting into std:: namespace when one includes <cstdarg>. Other
compilers seem to be more relaxed, and so is ARMC5 if compiled
with other profiles.
Add the explicit std:: to references of va_list.
While here, remove one extra copy of "#include "PlatformMutex.h""
and a "#include <stdarg.h>" which is kind of duplicate of
"#include <cstdarg>".
Error being fixed:
--8<--8<--8<--
Compile [ 81.8%]: ATHandler.cpp
[Error] ATHandler.h@552,0: #20: identifier "va_list" is undefined
[Error] ATHandler.cpp@1226,0: #147: declaration is incompatible with "void mbed::ATHandler::handle_args(const char *, <error-type>)" (declared at line 552 of "./mbed-os/features/cellular/framework/AT/ATHandler.h")
[ERROR] "./mbed-os/features/cellular/framework/AT/ATHandler.h", line 552: Error: #20: identifier "va_list" is undefined
"./mbed-os/features/cellular/framework/AT/ATHandler.cpp", line 1226: Error: #147: declaration is incompatible with "void mbed::ATHandler::handle_args(const char *, <error-type>)" (declared at line 552 of "./mbed-os/features/cellular/framework/AT/ATHandler.h")
./mbed-os/features/cellular/framework/AT/ATHandler.cpp: 0 warnings, 2 errors
IPV6 and IPV6V4 support is also network dependent not only modem.
Having these properties enabled for a modem requires a fallback
mechanism during PDP context activation. This mechanism is missing
at the moment and that can result in imposibility to establish
succesfull connection when network only supports IPV4 contexts.
Deprecate wait() in favour of acquire(), try_acquire(),
try_acquire_for() and try_acquire_until().
Brings Semaphore more into line with CMSIS-RTOS 2 (which uses "acquire"),
itself (as it has "release"), and other classes having "try", "try for"
and "try until".
Also steps away from vague "wait" term - the primary operation here is
to acquire the semaphore, and this will of course sleep.
This modem is a special case. It uses a given socket ID value rather
than providing one. A naive solution here would be to directly map the
index of a CellularSocket object in the CellularSocket container. But
considering the case where there are multiple sockets being opened (some
sockets being already created at the modem and some yet not created), direct mapping
to indices will not work. As it can happen that the CellularSocket
object is allocated but the socket id is not assigned yet as it is not
actually created on the modem.
In such a case, we check the container and assign the socket id from the
pool if an empty slot was found.
Local modem ip stacks vary in their implementations and the way of
working. Some of the modems may not open a socket until an IP context is
assigned. That's why we came up with a container that stores addresses of
any CellularSocket instances created on-demand by the application. When
the application requests opening a socket we store allocate and store the
premitive in the container however actual socket creation at the modem
may happen at a later stage, e.g., a call to send_to() may result in
actual opening of a socket.
That's why we must not assign socket ids in the CellularSocket object
during construction. It must happen when actual socket is opened and is
alive.
Another implication of the previous model is that we may have multiple
sockets created in our container but the actual socket ids are not
assigned yet, so we cannot directly map the socket id to the container
indices which has been happening previously.
To solve this issue we have promoted the AT_CellularStac::find_socket_index(...) method
to be a protected method rather than being private so that the children
can use the method to determine if the given index in the container
corrsponds to the assigned socket id or not.
We have given up on the socket->created flag and the whole decision
making to actually open a socket on the modem happens on the basis of a
valid socket being assigned or not.
CellularContext member variables were initialized in inheriting class.
Now in base class where they should be initialized so that every inheriting
class don't have to init them.
Socket ID is usually spitted out by the modem, however there are cases
when the modem is actully taken in as an input argument, e.g., in the
case of QUECTEL M26 modem. This minor knit clarifies that the
CellularSocket::id can be an input argument.
Disconnect was sent to all CellularContext classes even it concerned
one specific context. Some disconnect events are still sent to all
context classes. These event are coming from network and ment for all
context classes or event did not specify cid.
CellularDevice::get_default_instance() is a weak method and is overriden
by either a default construction provided in the code or by application
at some stage. This method needs to be flagged otherwise using another
driver will be hindered by the default overriding of this driver.
Kimmo Vaisanen
mudassar-ublox
Martin Kojtal
Janne Kiiskila
Mirela Chirica
Ari Parkkila
int_szyk
Syed Fahimuddin Alavi
Hugues Kamba
Teppo Järvelin
George Psimenos
Antti Kauppila
Marcin Radomski
Anna Bridge
Seppo Takalo
Reda Maher
Kevin Bracey
jeromecoutant
Tero Jääskö
Sebastian Stockhammer
Laurent Meunier
Chris Trowbridge
Bilal Qamar
Hasnain Virk
Daniel Lee
Pekka Pesu
kevin.ong
Martin Kojtal