On Wed, 30 Jan 2002 [EMAIL PROTECTED] wrote: > When you build Linux/RTLinux are you configuring for an AMD, enabling > any MTRR ?
My kernel and RTLinux modules were built for i486, because that's what an AMD 5x86 essentially is. CONFIG_MTRR is not set. I also tried to rebuild the whole kernel and all RTLinux modules for 5x86, which I believe refers to other CPUs called 5x86 that are more Pentium-like. This time I built everything with gcc-2.95.3. The problem is still there, though. > Is this a patched kernel from RTLinux.org? > Does it have any other patches in it? It is a standard 2.4.4 kernel from kernel.org with only the RTLinux 3.1 patch on it. > > > > RTLinux 3.1, Linux 2.4.4, CPU AMD 5x86, gcc 2.96 (from RedHat 7.2). > > > > I have now tried using nanosleep() and clock_nanosleep() but the result is > > the same. But when I set the timer in periodic mode my thread ran for over > > twelve hours before I interrupted it so it seems as if rtl_time sometimes > > fails to re-program the timer chip. > > > > Some more observations: > > > > When my task is frozen, everything else timer related is frozen too. > > 'date' returns the time when the task froze, 'sleep 1' blocks forever. > > > > If I leave the stopped task in memory and don't touch the system it will > > start again after some time, typically it seems to take about 30 min. Another observation: When the task freezes and is left alone, it wakes up automatically after 30min +/- one or two seconds. Since 30 min = 2^31 ticks of the 8253 timer, this tells me that the timer accidentally gets programmed with the msb set or something similar. Is that at all possible? Thanks again. Kris -- [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/
