fs problems with lxrt?

Paolo Mantegazza mantegazza at aero.polimi.it
Wed Oct 13 10:45:48 CEST 2004


root wrote:
> Some days ago I started a discussion thread about file system problems
> seemingly due to LXRT. That thread then ran out without any conclusion.
> 
> Now I would like to add the results of some experiments I've made
> working with a kernel 2.4.22 patched with rtai-24.1.12 (adeos, but the
> problem is present with legacy rthal too).
> 
> With newlxrt scheduler set, if I:
> 1) boot 2.4.22-adeos;
> 2) login as root (X, blackbox window manager);
> 3) bring up an xterm and issue 'modprobe rtai_sched';
> 4) issue 'lsmod' and see that rtai and rtai_sched are there;
> 5) don't absolutely run my code nor any other application, not even rtai
> examples/tests;
> 6) issue 'modprobe -r rtai_sched';
> 7) issue 'lsmod' and see that rtai and rtai_sched aren't there any more;
> 8) issue 'reboot',
> 
> during shutdown I always get messages complaining that /dev/pts is busy
> and that umount failed on /dev/hda4 (my /). Usually reboot works fine,
> but occasionally I am asked to run fsck manually.
> 
> If I do exactly the same operations (1-8) but with the up scheduler set
> instead of newlxrt, I never see any complaint at shutdown and never need
> to run fsck at reboot.
> 
> The file system was reiser for these tests, but I had similar results
> with ext3 (both journalling file systems - could this mean?)
> 
> Neither the hardware seems important, as I got similar results on three
> different machines (all i386 architecture though).
> 
> Also, I got similar results with rtai-3.1 over 2.4.24.
> 
> I can send to the willing my kernel and rtai configurations (looks like 
> it's not possible to attach them here).
> 
> IMHO in this moment while Paolo is trying to persuade everybody to go
> LXRT (and I fully share his opinion) it would be a very, very welcome
> thing to have this problem clarified.
> 

Well Daniele, for me 24.1.12 is prehistory :-).

Notice also that I'm pushing LXRT in relation to the latest 3.x 
development lane.

With such a version the problem you report happens just if you reboot 
with the modules installed. Since, if you are able to type "reboot", it 
means the machine is sane, there should be no problem rmmoding the RTAI 
stuff you have installed and avoid the dev busy problem.

In any case when LXRT will be the only scheduler in use I'll care of it 
a bit more.

Paolo.





More information about the Rtai mailing list