XIO: fatal IO error 11

[gdargaud]

Hello all,

OK, this is gonna be a tough question...

I never noticed this problem before, only since the upgrade to the latest patch of CVI for Linux.

After a few days of use one of my CVI programs crashes with the following message:

XIO:  fatal IO error 11 (Resource temporarily unavailable) on X server ":0"
  after 2322 requests (2322 known processed) with 0 events remaining.

Part of the problem is that it does not generate a core file, although it should.

So anybody at NI has any idea what this message is about ? Searching for it on the web leads to widely scattered problems, none of which seem relevant.

[Basil-B]

Hello,

Can you provide the distro of linux and version that you are using. What version of CVI Run Time for Linux are you using? This sounds like an OS issue.

Regards,

Basil

[gdargaud]

Hello Basil,

I'm on Scientific Linux 6.4 32, with CVI 13.0.0-30.i386.

I'm currently investigating if the crash is due to using Xinerama (3 monitors) with an NVidia driver. It hasn't crashed since last week.

[gdargaud]

Well, that wasn't it. The crash still happens without Xinerama. Not having a core dump, I'm at a loss to find a cause. It doesn't happen on my development machine which has an older Scientific Linux.

[hnesmith]

gdargaud, 

what type of tasks are you completing? does it crash reliably at the same location? and does the crash still occur if you run an example program? 

[gdargaud]

It's a control/command/monitoring app, with tens of panels (hundreds if you count tabs and child panels) and thousands of numerics, dials, strip charts, etc updated every second. There's no performance issue. And 'where' it crashes is the issue: I have absolutely no hint where it happens. The program runs unnatended so it's not an issue of 'it crashes when I do that'. And then some mornings we find it closed with only this 'XIO' message: no core dump, nothing in the program log, nothing in the system log, nothing in the X11 log...

I've run it with valgrind so I know there's no memory leak in _my_ code (and if there was, there'd be a core dump). Valgrind didn't help with diagnosing this problem. I'm currently running with ltrace, we'll see...

[hnesmith]

did ltrace provide any insight? 

[gdargaud]

Still running... hasn't crashed yet.

[gdargaud]

OK, I still have this ellusive and difficult to trace problem (it crashes only every 3~7 days).

I've tried many things:

- running with valgrind - nothing interesting

- running with mcheck - no problem found

- running with ltrace - nothing interesting

- catching all possible signal kills - none found

- placing an atexit() - it is indeed being called before the program exits unexpectedly. But called by whom ? I still need to try and trace that.

Since XIO is related to X11, I tried:

- removing Xinerama - no change

- removing 3rd monitor - no change

- removing 2nd NVidia card - no change

- removing nvidia driver and using Nouveau driver - no change.

- running other graphic intesinve programs (non-CVI, for instance glxgears): they don't crash

The only conclusions I can draw from this is that either:

- the problem is in Scientific Linux 6.4, since I never had this problem with 6.2

- the problem is in the patch nicvirte-13.0.0-30.i386.rpm

What I can still try:

- write a simpler graphic intensive program and see if it crashes

- try older versions of Scientific Linux

Has anyone at NI ever seen this XIO problem ?

[hnesmith]

Trying a simpler graphic intensive program is a great next step. You mentioned never having the problem on 6.2. I'm assuming you're running the same application on both machines and both have the patch?