[Emc-developers] Shared memory access + emcmot locks up

[Lorenzo M. <l.m...@lo...>]

I'm trying to make emc2-cvs work but I've got some problems

First of all, the hw is compatible, since with the ubuntu-cd everything works fine. As a reference is an AMD Athlon Classic core clocked a 1GHz, and a VIA based southbridge (something like an ASUS V7 board...). Machine control is (for now) on the parport using a xylotex-like connection (there is a master amplifier enable system and external estop connected too).

What I've got now (on a non-ubuntu base. An ubuntu distro is NOT an option for me):
- Kernel 2.6.15 (same as the cd, minus the eventual Debian stuff)
- RTAI-magma from CVS (not the same as the cd, since the `math' option is not a module)
- EMC2 from CVS

In particular I don't have the SELinux extension, nor libcap. Or maybe I missed something during kernel configuration. I'm not using initrd and it's all built in since it's intended for a fixed platform. Only RTAI stuff is loaded as module. BTW it boots REALLY fast now...

I updated the ini and hal files for the new emc version (nothing spectacular, just the syntax in the HAL and some parameter name in the INI).

There are now 2 problems:
1) If I run emc from a user (not root) it say it can't access the shared memory. What kind of shared does it means? RTAI, SysV or Posix? (or simply a mmapped file :P)
2) External estop, spindle control and jogging work fine, so I suppose the HAL and the basic stuff are working.

MDI and auto mode don't. axis and tkemc become unresponsive, keystick says timeout.

Further testing with usrmot shown that it hangs when I ask for a motion (ie not a jog, a coordinated motion). It seems that emcmot is stuck waiting for something. Status requests from usrmot also time out after a motion request (before they work fine).

I also noticed that the 'fast task' in the rtai scheduler is no more in status 0x3 (but in 0x2001 IIRC, I'm not in front of the machine right now).

So I can suppose there is SOMETHING in the motion request that puts emcmot in a sleeping state (waiting for a semaphore? I don't know...)

I already used RTAI (emulated an H8 MCU at full speed :) but the IPC architecture in EMC is a little complicated for me :P especially the NML stuff...

Can anyone help me troubleshooting the fault?
Thanks

PS: I even tried 1.6 from Sourceforge but the problem is the same...