Re: PPMC & rtlinux

[Paul Corner <paul.corner-at-tesco.net>]

Hi Jon.

 Comment between snips.

On Tue, 12 Dec 2000, Jon Elson wrote:
> > As I understand it, for later versions of rtlinux, the message handling is
> > no longer done through shared memory. On the 2.2.13-rtl2.0 kernel, I don't
> > have any shared memory set aside. An advantage at present, as the hard disk is
> > being swapped between a couple of machines with differing amounts of memory.
> 
> Hmm, maybe that solves the problem.  So, your machine is reliable?
> I'm running a 2.0.36 kernel.

Is Linux reliable, swapping the disk between machines ? - Yes.
Is emc reliable ? I don't know at present, as I do not have it set up on this
drive at present.

> > I cannot get the current CVS source to build on any kernel/rtl combinations at
> > present. But I would assume that later kernels no longer require the shared
> > memory.
> Hmm, well, this is a problem.  I don't mind changing EMC versions, because it
> is pretty straightforward to go back to one that works.  But, changing kernels
> is a bigger undertaking.

I had a problem with the original Linux install - My 20G drive was some how
configured as 8G. A bug with the SuSe install tool I believe. Using fdisk and
mkfs to correct the problem, I now have a virgin system to work with. To date,
the only combination I have got to run is 2.2.13-rtl2.0 and emc-Aug-11. All
the kernels (2.2.14  2.217) and RT patches (2.2 2.3 & 3.0) have compiled with
no more than a couple of warnings. Building the rcslib from the CVS tree
generates quite a few warnings - some potentially serious, but none fatal.
The non-realtime side of emc compiles a treat, but when it comes to the
realtime - No joy.
It could just be a simple error on my part that I've overlooked. But I will
take a couple of Linux boxes to the office, and go through it all over
christmass.

>From my limited experience with tkemc and the 2.2.13-rtl2.0 kernel, I have not
noticed a degredation of perfomance. Despite shutting down emc and restarting
incessantly during the course of writing and debugging the new backplotter.

Regards, Paul.