ec_slave[slave].state operational when slave is lost

OpenEtherCATsociety / SOEM

Simple Open Source EtherCAT Master

Other

1.36k stars 688 forks source link

ec_slave[slave].state operational when slave is lost #144

Closed kajusK closed 6 years ago

kajusK commented 6 years ago

Hi, I've been fighting with the simple_test example on windows. There are several issues when running setup with several slaves:

expected WKC calculation - when using 2 slaves (with both inputs and outputs configured), expected WKC is calculated as 6, real WKC returned during communication is 5. I'm pretty new to ethercat, but shouldn't be the correct calculation of expected wkc like:
```
(ec_group[0].outputsWKC*2 - 1)+ec_group[0].inputsWKC;
```

Anyway, that's not a blocker, the more serious issue I hit is state when a slave is lost (disconnected). Error detection works as expected when I disconnect slave directly connected to the master, but when I disconnect a different slave, the ec_slave[slave].state remains EC_STATE_OPERATIONAL for all connected slaves, any idea why? simple_test starts printing OK : all slaves resumed OPERATIONAL.

Thanks, Jakub

nakarlsson commented 6 years ago

About WKC, check chapter 2.4 in https://download.beckhoff.com/download/document/io/ethercat-development-products/ethercat_esc_datasheet_sec1_technology_2i3.pdf. I guess you execute LRW and would get a read & write +3 giving 2 *3 = 6.

Are you running linux/simple_test/simple_test.c?

nakarlsson commented 6 years ago

Could you provide a console log snippet and wireshark log when it happens? Also do a dump via slaveinfo.

kajusK commented 6 years ago

Hi, aha, I found the issue with wkc, the second slave was not initialized correctly, now it works as expected. Thank you.

Yes, I'm running linux/simple_test/simple_test.c. Although I running on windows due to the project requirements (I'm linux only user otherwise), the linux version works just fine. I've also tried to compile win32 version and run simple_test again, but with the same result.

The requested logs are below. First, the second slave was disconnected, reconnected and then the same procedure happened with the first slave.

slaveinfo.txt simple_test.txt wireshark.zip

nakarlsson commented 6 years ago

I'm out of office but I think I have an idea what might be the trouble, can you please try the following fix, I haven't had a chance to compile or test so bear with me. I think the issue is that we miss the "islost" due to the fact that the other slave(s) is/are still up and running as expected and trigger a bug in readstate, readstate should identify that one slave is missing and try to read the states individually.

ethercatmain.zip

kajusK commented 6 years ago

Thank you for your fast reply. The fix you sent works perfectly.

nakarlsson commented 6 years ago

Thanks for testning, I’ve to make this official and use this Issue.