System environment: an embedded system, SMP quad-core processors, RTP is mainly used in equipment control, embedded TCP communication, communication data about 1 MB/s, multiple Socket connection communication
Problem phenomenon: the program run for 24 hours, to verify the stability, to 5 days, RTP VxWorks interface program last print information is TCP select send timeout, task gets stuck, unable to connect to the Target Host, Ctrl X is invalid, the system outage, nor to generate the Core Dump, during the network hardware should be no problem, I don't know you have encountered similar problems, greatly RTP program logic code read it, should be no problem, looks like a network communication problems in surface, the system gets stuck situation is generally what circumstances trigger?

Program invocation API is mainly socket, taskSpawn, taskWait semTake, semGive, serial ports, speaking, reading and writing

CodePudding user response:

There may be memory occupied without release slowly, finally has no memory is available

CodePudding user response:

No memory leak problem, memShow view has been very stable, the current screening to a suspected position, in the call of VxWorks socket function library of the send API, there are some problems with the code, not in front of the calling select check to write, after pressure test (constantly send) can easily reproduce this (system), stop after the send system network cannot restore, the whole system network paralysis unable to ping, CTRL + X downtime, the code is under win32 won't cause system problems, now change under continuous observation equipment status, there are no exception

CodePudding user response:

System check each node, the data communication in addition to the board, are there any other node?

This edition is communication is only a piece of CPU card data? FPGA?

Software level check

You are in a multitasking environment kao machine? Statistics of high-priority task CPU?

Check the fifo, data flow have producers is greater than the consumers,

CodePudding user response:

Data stream consumption should be OK, network data monitoring is a maximum of 100 KB/s, but the client and the server network communication could disconnect, I don't know whether VxWorks is similar to a Unix process, called send if meet network disconnection, will produce SIGPIPE, lead to withdraw from the process or task, or because the abnormal signal, this may still need to test,

CodePudding user response:

For now since before the send added Select, did not appear abnormal or task