Peter M. Groen / libmodbus

Thanks to @lipengin1975 for reporting this issue.

Signed-off-by: Michael Heimpold <mhei@heimpold.de>

According to the Modbus specification
(http://www.modbus.org/docs/Modbus_over_serial_line_V1_02.pdf, section 2.1)
a Modbus RTU master can send a broadcast to all of it's slaves. This
broadcasts can only be write requests as otherwise collisions could
occur, eg. on a RS-485 bus.

When receiving such a broadcast, the slave should process the request as
usual, but must not reply anything, neither a normal response nor an
exception reply in case of an error.

Adjust the unit test for this case, too.

Signed-off-by: Michael Heimpold <mhei@heimpold.de>

Do not allow raw request length longer than the PDU size plus
the additional requested slave address byte.
Without this check modbus_send_raw_request could be used to
trigger a buffer overflow on the stack since the parameter
is passed unchecked to memcpy.

Thanks to Hanno Neuer for spotting this security flaw.

Signed-off-by: Michael Heimpold <mhei@heimpold.de>

Signed-off-by: Michael Heimpold <mhei@heimpold.de>

This replaces the lengthy license text headers with a short
and standardized license tag. See http://spdx.org for details.

This is useful e.g. for license compliance tools which scan
through files and generate a report of the licenses used
in a project.

Signed-off-by: Michael Heimpold <mhei@heimpold.de>

This change has introduced by 83c34102.
Fix unit tests on some systems.

Function modbus_receive_from has no implementation in the library and is not used anywhere.

Protects against crafted write requests with a large quantity but a small
byte count. If address + quantity was in the mapping space of the server
and quantity greater than the response size, it was possible to crash
the server.

The sleep/flush sequence improves the handling of following requests.

Thank you Petr Gladkiy and MarjanTomas.
Sorry for the delay...

From Michael Heimpold:
Just a nitpick: the INADDR_* macros are defined in host byte order,
compare INADDR_LOOPBACK. So it would be reasonable to wrap this with
htonl(INADDR_ANY). However, every experienced socket programmer should
know that INADDR_ANY equals 0.0.0.0, passing it through htonl does not
have any effect.

- protect against NULL IP address BTW
- update documentation

- allow an empty string or NULL for node argument
- protect against NULL in service argument
- new test for NULL service
- update documentation

Thanks to Marjan Tomas

Change API of function for libmodbus v3.2.0

Thanks to Perry Kundert for bug report and initial patches.

- add byte timeout tests (TCP backend only)
- update and improve documentation
- long timeout values are now uint32_t so it changes the API
  to disable byte timeout

Thank you AlexMaz.

https://www.securecoding.cert.org/confluence/display/seccode/DCL37-C.+Do+not+declare+or+define+a+reserved+identifier
Not applied to libmodbus constants for now...