#linuxcnc-devel Logs

Oct 03 2019

#linuxcnc-devel Calendar

12:32 PM drdoc: Hi, all
12:35 PM drdoc: I've been trying to build the current RTAI packages, according to the wiki page
12:35 PM drdoc: http://wiki.linuxcnc.org/cgi-bin/wiki.pl?Building_RTAI_5.2_Packages
12:35 PM drdoc: First on Debian 10, where I stopped because of dependencies that aren't in 10, then Debian 9
12:35 PM drdoc: After getting a clean build without errors, the rtai modules will not load, with an "Operation not permitted" error.
12:35 PM drdoc: APIC is loaded and verified in dmesg output
12:37 PM drdoc: build box is a Dell 9020 i7 with a Quadro K2200 & plenty of RAM
12:39 PM drdoc: All the build .debs were installed, then when the modules errored out I purged them - "dpkg -P blah" - booted into the stock kernel and reinstalled them
12:39 PM drdoc: same issue
12:42 PM drdoc: there's one oddity I don't understand; the kernel packages get named like this:
12:42 PM drdoc: linux-image-4.14.134-rtai-amd64_4.14.134-rtai-amd64-2_amd64.deb
12:44 PM drdoc: but "dpkg -l" returns what I'd expect:
12:44 PM drdoc: ii linux-image-4.14.134-rtai-amd64 4.14.134-rtai-amd64-2 amd64
12:44 PM drdoc: I'd really appreciate any ideas
12:50 PM cradek: have you checked dmesg after you fail to load a module?
12:59 PM drdoc: Just did:
12:59 PM drdoc: [ 302.314277] RTAI[hal]: RTAI CONFIGURED WITH LESS THAN NUM ONLINE CPUS.
12:59 PM drdoc: Really? do I just need to trun off HT?
12:59 PM drdoc: turn*
01:00 PM drdoc: that output is after running "latency-test" and getting the load error for rtai_hal
01:01 PM drdoc: back in a few
01:04 PM drdoc: and that did it.
01:05 PM drdoc: I turned off hyperthreading and latency-test is running
01:06 PM drdoc: doc@box205:~$ lsmod | grep rtai
01:06 PM drdoc: rtai_math 102400 0
01:06 PM drdoc: rtai_sched 196608 2 rtapi,rtai_math
01:06 PM drdoc: rtai_hal 487424 3 rtapi,rtai_math,rtai_sched
01:06 PM drdoc: cradek: thanks!
01:34 PM jthornton: I'm hoping you took good notes...
01:35 PM drdoc: jthornton: I may have, actually
01:35 PM jthornton: cool
01:35 PM cradek: gosh what a lucky guess
01:36 PM * jthornton heads back to the shop to do something
01:36 PM drdoc: cradek: not really - it's an obvious first diagnostic
01:37 PM drdoc: I had looked at dmesg a couple of times, but not at the post-fail output
01:37 PM cradek: I HAVE SOLVED ALL THE PROBLEMS OF THE WORLD oh ok yeah it's pretty basic isn't it
01:37 PM drdoc: holy crap it's hot in here
01:38 PM drdoc: an i7 and a Core 2 Quad running full out
04:29 PM rene_dev_: drdoc why do you want to use rtai?
04:40 PM drdoc: rene_dev_: I wanted to compare rtai to preempt_rt on apples-to-apples installs
04:40 PM drdoc: then I just wanted to make it work because I got mad.
04:42 PM rene_dev_: there you go, rtai is painful, and rt preempt is almost in mainline ;D
04:42 PM drdoc: I expect that jthornton's notes on building preempt_rt on Buster will work on Stretch
04:43 PM drdoc: so that gives me a common base for comparison
04:44 PM drdoc: I think I've got the target system's hardware squared away - 3 hours in I'm not getting those 30-40uSec spikes
04:45 PM drdoc: Huh
04:45 PM drdoc: the unstable tree is still Sid?
04:46 PM drdoc: that's a blast from the past
05:32 PM CaptHindsight: drdoc: besides killing all power management in BIOS you also need to turn off any virtualization and hyperthreading
05:33 PM CaptHindsight: drdoc: you can run whatever kernel on whatever Debian distros
05:33 PM CaptHindsight: it's not the distro that effects latency, it's the kernel version
05:38 PM CaptHindsight: all this info is in the README, what do we have to do for people to read and comprehend this???
05:39 PM CaptHindsight: https://github.com/NTULINUX/RTAI/blob/master/README.BIOS
05:39 PM CaptHindsight: https://github.com/NTULINUX/RTAI/blob/master/README.INSTALL
05:39 PM CaptHindsight: 4.14.132 references are out of date and need to be updated to 4.14.134
05:40 PM CaptHindsight: I don't know what else to say, RTFM
07:06 PM drdoc: CaptHindsight: you're telling me that the amount of stress that system software incurs is irrelevant?
07:06 PM drdoc: read both of those, by the way, before I started trying to build it.
08:35 PM CaptHindsight: drdoc: IF clock throttling has been turned off
08:36 PM CaptHindsight: sorry to sound like I might be picking nits but why did you have hyperthreading on?
08:42 PM drdoc: The CAD box is pretty new to me. I hadn't had any performance problems, so I hadn't dug through the BIOS
08:42 PM drdoc: IOW: hadn't thought about it.