Hi everbody!
First, thanks to FSMLabs and all others who did efforts to make
progresses to give linux real-time behavior. You did a very
great work!
I'd like to announce the availabilty of RTL2.0 in the
printed issue of the german Linux-Magazin, but 2.0 doesn't
seem to be as rock stable as RTL1.x is. Thus it would
be a reputational demage for RTL if it would be announced
to early ...
While trying to port may (rather simple) RT-Applications
to 2.0, i discovered two problems:
1.
the "hello"-example sometimes does a
"RTL ERROR: pthread_setperiod crosses cpus"
When looking into rtl/schedulers/rtl_sched.c at
pthread_make_periodic_np(), then it doesn't seem
to be allowed to perform this function within
a rt-task (like it is in the hello-example)?
2.
When "insmod"ing the 3dfx-Modul, the machine crashes with
>>>>
Kernel panic: nmi!
In swapper task - not syncing
Kernel panic: mtrr: timed out waiting for other CPUs
<<<<
Ok, you now might say that it doesn't make much sence
to play q3test on a Voodoo3 with RT-Linux :-)
(using Device3Dfx-2.2-3.src.rpm - the 3dfx-modul plays
around with the mtrr-register ... maybe that's the
problem?).
3.
With RTL1.x, it was possible to change the periodicity
of a running task within the task by simply doing a
my_task.period=new_period;
rt_task_wait();
since the rt_task_wait() invoked the scheduler which took
care about the new value.
Is something like this still possible in RTL2.0?
pthread_setperiod_np() doesn't seem to be allowed,
since the same problem as in (1.) occurs.
Might it help to fix the rt-task to one CPU? how is that done?
Configuration:
RedHat 6.1/RTL2.0 on Kernel 2.2.13
Gigabyte BXD with two Celeron300A on MSI 6905 Rev 1.0 Socket370->Slot1-Adapter.
Maybe that Celerons are the Problems, since the were never
intended to run in dual-mode, but as i didn't had any such
problems before ...
BTW: im using reiserfs-journaling-beta. Realy cool stuff!
if the machine hangs due to a kernel/rtl-level programming
mistake, just push the reset-button and the machine
will only take 5-15 secondes to recover the filesystem,
instead of e2fs-checking the whole partition.
(Im rather sure that the reiserfs-patch doesn't break
rtl since reiserfs touches completly different files in
the kernel tree - it's only filesystem related stuff).
any comments?
Bernhard
--- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
----
For more information on Real-Time Linux see:
http://www.rtlinux.org/~rtlinux/
