On Tuesday 22 December 2009 16:53:58 Chris Radek wrote:
> You are right that the first one is real,

Yeah I reproduced it twice with almost identical results (except the followup 
oopses. Which
is why I think we can ignore these).

> and it is obviously an emc process.

Well, no. That is coincidence. This oops happens in the APIC timer interrupt,
so "comm" is meaningless. This is is not a direct oops in some EMC or RTAI code,
but I think some bug in EMC or RTAI _triggers_ it.
Rather hard to track down...

> abilities.  I don't understand your fighty response here.

Well, basically because your response was rather useless, in a technical 
aspect. ;)

> To the best of my knowledge, nobody else is running EMC on a 64 bit
> SMP rtai kernel though.  I'm not sure anyone is actively using even
> a 64 bit UP rtai kernel.

So is isolcpus commonly used? Because that hides the problem.

> I understand that you're not blaming anyone for the bug.  But in my
> limited but nontrivial experience building rtai kernels, when I get
> panics loading EMC, I try a different kernel version and rtai patch,
> and the problem goes away.  For UP i386 it's pretty easy to find a
> working setup.  For SMP back when I tried it, it was much harder.  I
> can only assume that x86_64 + SMP is harder still.

Well, I'm certain it's not a miscompilation or something like that,
because isolcpus hides it. So I'm fairly certain that this is a race
condition that is hidden, if all CPUs but one is isolated.

> Have you exercised rtai with other, non-EMC, software?

No.

> Can you describe what tests you have done?  

Well, the "test" basically is fireing up and shutting down EMC for several
times. After the tenth retry or something like that it will oops (for me),
unless isolcpus is used.

> > I am running rtai 3.7.1 and latest EMC stable release.
> 
> I do not see a 2.6.29.6 x86_64 patch in rtai 3.7.1.  Where do we get
> the patch you used?

Well, there's a .4 patch.

-- 
Greetings, Michael.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Reply via email to