epics-modules / modbus

EPICS support for communication with Programmable Logic Controllers (PLCs) and other devices via the Modbus protocol over TCP, serial RTU, and serial ASCII links

An EPICS module that supports communication with Programmable Logic Controllers (PLCs) and other devices via the Modbus protocol over TCP, serial RTU, and serial ASCII links.

The modbus software provides a layer between standard EPICS asyn device support, and the EPICS asyn TCP/IP or serial port driver.

Additional information:

• Documentation
• Release notes

modbus's Issues

String support

Support for absolute Modbus addressing.

Currently each driver is limited to addressing at most 125 registers (read operations) or 123 registers (write operations), and the asyn "addr" field specified an offset relative to the modbusStartAddress passed as the fifth argument to drvAsynModbusConfigure().

This is inconvenient when many Modbus registers need to be written to and these are widely scattered over the 16-bit Modbus address space.

Support for modbus function 17 (Feature request)

We are in the process of migrating LabVIEW to EPICS for beamline equipment protection system that consists of Modicon PLCs. All PVs are translated except status registers that use modbus function 17. It would be great if the module has support for this.

Reference:

• Modicon Modbus Protocol Reference Guide - Page 48, 49, 54
• Modicon 984 series manual

Thanks,
Najm

MODBUS/TCP does not make use of transaction ID

The MODBUS/TCP transaction ID is set to 1 on writes and ignored on reads. It would be more robust if a unique ID were used for each transaction and replies with invalid IDs ignored -- especially for MODBUS/UDP links. My proposed change is attached.

modbusTIDpatch.txt

Need to support 64-bit integers and byte-swapped versions of 32-bit and 64-bit datatypes

Currently this driver supports big-endian and little-endian versions of 32-bit integer numbers, and big-endian and little-endian versions of 32 and 64-bit floating point numbers. However, some vendors also swap the bytes within the 16-bit words for these data types.

I looked at the Modbus Slave Emulator program I have installed on my PC. It handles the byte-swapped versions of all 32-bit and 64-bit data types, i.e. both INT and FLOAT.

Now that EPICS supports int64in and int64out records 64-bit integer values should also be supported.

We thus need to add support for 10 more data types:

• INT32_LE_BS
• INT32_BE_BS
• INT64_LE
• INT64_LE_BS
• INT64_BE
• INT64_BE_BS
• FLOAT32_LE_BS
• FLOAT32_BE_BS
• FLOAT64_LE_BS
• FLOAT64_BE_BS

Modbus driver continuously prints error messages when the device is disconnected

The Modbus driver continuously prints error messages when the device is disconnected. It should just put the error string in pasynUser->errorMessage and pass it back to device support.

Modbus exceptions need to be documented

The documentation for this driver needs to discuss Modbus exceptions. They are not currently mentioned.

segmentation fault

I occasionally get a SIGSEGV in the modbus IOC. Now for the first time it happened when running in gdb. The backtrace is

2016/06/15 14:42:52.928 drvModbusAsyn::doModbusIO port block1316 error calling writeRead, error=127.0.0.1:5502 timeout: Resource temporarily unavailable, nwrite=6/6, nread=0

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff42c3700 (LWP 315)]
lockPort (pasynUser=0x0) at ../../asyn/asynDriver/asynManager.c:1677
1677 port *pport = puserPvt->pport;
(gdb) bt

0 lockPort (pasynUser=0x0) at ../../asyn/asynDriver/asynManager.c:1677

1 0x00007ffff7bd28ae in readPoller (pPlc=0x6ae7b0) at ../drvModbusAsyn.c:1319

2 0x00007ffff61fb07f in start_routine (arg=0x6b0d10)

at ../../../src/libCom/osi/os/posix/osdThread.c:385

3 0x00007ffff5379454 in start_thread (arg=0x7ffff42c3700) at pthread_create.c:334

4 0x00007ffff5677edd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:109

I will leave the gdb session open so that I can give more feedback as required.

String support

The Modbus driver has not supported string data type. Some vendors use Modbus registers to contain string data. The driver should support reading and writing strings to Modbus devices.

drvModbusAsynConfigure should accept strings for the dataType argument

The drvInfo string uses strings like INT32_LE for the datatype of a specific record. However, drvModbusAsynConfigure only allows integer values, which are the modbusDataType_t enums, for the dataType argument.

drvModbusAsynConfigure should accept either numbers or the same strings used by the records for the dataType argument. If the first character of the string is a digit then it should be assumed to be the enum value, for backwards compatibility. If it is not a digit then it must be one of the datatype strings defined in drvModbusAsyn.h. The comparison should be case-insensitive.

This has been implemented in the datatype_string branch.

asynInterposeEos should be enabled for low-level drivers

The current modbus documentation and example IOCs do not load the asynInterposeEos interface (they set noProcessEos=1), except for Modbus serial.

This works OK with real Modbus TCP devices because the response will be returned in a single packet that should contain the requested number of bytes. But for devices with a serial interface, including those with an Ethernet terminal server, the response may be broken into multiple packets. In that case the asynInterposeEos layer is needed so that it does not return until the requested number of bytes have been received, or there is a timeout.

It does not hurt to load the asynInterposeEos interface even if it is not strictly needed. The documentation and example IOCs should be changed to set noProcessEos=0.

Minor: Still relies on a modbusInclude.dbd file

An EPICS user wanted to run QSRV on the modbus IOC. My initial instructions on how to do that assumed that the IOC's DBD file was generated in the Makefile. I recommend replacing the modbusInclude.dbd file with the more modern approach.

Making this change might be a good Codeathon project.

The report() function does not show the default datatype of the port

The report function previously did not report the default datatype for that port.

This has now been added in the datatype_string branch.

It shows the default datatype, both as the integer enum value, and as the corresponding string. For example:

modbus port: A300_In_Word
    initialized:        true
    asynOctet server:   sim1
    modbusSlave:        0
    modbusFunction:     3
    modbusStartAddress: 0454
    modbusLength:       0120
    absoluteAddressing: false
    dataType:           22 (FLOAT32_LE_BS)
    plcType:
    I/O errors:         0
    Read OK:            1
    Write OK:           0
    pollDelay:          0.100000
    Time for last I/O   11 msec
    Max. I/O time:      11 msec
    Time per hist. bin: 1 msec