On 17-1-2013 22:57, Kent A. Reed wrote: > On 1/17/2013 2:59 AM, Bas Laarhoven wrote: >> I think that running linuxcnc is mandatory for the lockup. After a dozen >> runs, it looks like I can reproduce the lockup with 100% certainty >> within one hour. >> Using the JTAG interface to attach a debugger to the Bone, I've found >> that once stalled the kernel is still running. It looks like it won't >> schedule properly and almost all time is spent in the cpu_idle thread. >> >> The kernel with extra diagnostics produces these messages: >> >> [ 3480.386342] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" >> disables this message. >> [ 3480.395913] INFO: task axis:799 blocked for more than 120 seconds. >> [ 3480.406643] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" >> disables this message. >> [ 3600.408670] INFO: task hal_manualtoolc:788 blocked for more than 120 >> seconds. > Hi, Bas. Hi Kent, > > Like Michael I haven't been running LinuxCNC continuously but I'm doing > it now to see if I can reproduce your lockup on my BeagleBone. > > I'm using a Rev. A6 BeagleBone powered solely through the microUSB port > (I've tried unsuccessfully to induce power problems; there's a 2a supply > in the drawer if I need it) One difference will be the clock speed the processor is running on. Can you use a power supply to match my configuration ? > > I'm running off Michael's updated root filesystem. Both the > vmlinuz-3.2.21-xenomai+ kernel and the rootfs are loaded from the > microSD card (e.g., TFTP and NFS were not used to bring up the system > and DHCP was used only to provision the kernel ip settings, not the > U-Boot settings). [1] That's also a difference, let's find out if that matters. You're not using the cpsw driver with this setup, that driver had problems in the past when I tried a preemptive kernel. So if you're not getting lockups with this configuraton, it would be interesting to also test with a NFS mounted root FS ! (Myself I'm not using TFTP at all, only NFS for both the kernel and filesystem). > > I'm talking to the BeagleBone via ssh -X over my LAN so I don't have to > futz with DISPLAY and permissions settings. [2] I'm doing exactly the same (-X or -Y). > > Since the microUSB port is connected I'm logged in over the serial port > and running 'top' to see how I'm doing. Same as my setup. > > So far, I've had LinuxCNC up for more than an hour with the 3-axis mill > simulator running in Axis, repeatedly executing and reopening the > 3d_chips.ngc file within Axis. > > Granted I haven't automated this; every time I notice the simulator has > finished the G-Code I manually reopen and execute the file. > Nevertheless, the processes Axis, milltask, hal_manualtoolc, and halui > remain loaded, active, and the big cpu consumers according to 'top'. As > an aside, so is sshd, which is consuming 25 percent (yikes) of CPU as > long as it's updating the backplot. I'm often reducing the feedrate so I don't have to restart that often : ) > > So, to help me understand better what you are seeing, could you please > add some detail to your report? > > -What rev board (ok, so I think this an unlikely problem, but at some > point you said you've been working with your BeagleBone for months so I > wonder what's different between yours and mine)? I now realise I haven't used this particular board, it's brand new because my A6 had problems with the NIC. First thing tomorrow is to try one of my older boards, although I don't have reason (yet) to suspect the processor or board. > > -are you running a simulator or are you exercising some PRUSS code or what? It's running linuxcnc with all the default settings from the distribution Michael built. So start linuxcnc, confirm default configuration selection. On axis, power on, and just start (engraving linuxcnc). > > -what do you mean by "running" LinuxCNC? Does LinuxCNC have to be > actively running a program without interruption for the fault to occur > or does it fault when run the way I did my own test? The fault reproduces when I run linuxcnc, mostly between 50-60 minutes of uptime. I'm not 100% certain that it will only lockup when running linuxcnc, but that's how I trigger it right now. > > Finally, with respect to your next email message, how universal is your > statement that TI "...decided not to support the PRUSS anymore." Does > this refer to the TI documentation and software situation or does it > mean TI has decided they can't make money developing SoCs with PRUSS > and, hence, the current AM335x series is the end of the road? I could > get depressed:-( Disclaimer: First of all, some of my conclusions here may be wrong, but it's all based on what I saw, heard and discovered during my experiments with the PRUSS over the last 12 months. I think the PRUSS is of strategic importance for this SOC family, and wouldn't expect TI to abandon it. Unless it's causing silicon yield problems or something like that. The story is that some people at TI decided they didn't want to spend time supporting the PRUSS to the open source community. Another part of TI didn't agree and that resulted in the crippled pru package on github. TI spend a lot of time (money) rewriting documentation, strip the assembler source of the new features, change to name from PRUSS into ICCS, etc. etc. The real reason for this still being unclear. Is the silicon defective? Are some of the new features not working properly? I did find some problems with the built-in multipliers. I can imagine they only want to support a select group op developers who make the commercial solutions (e.g. Ethercat) for them and probably work under NDA. But that's all speculation and maybe we'll never know the real reason(s). In the meantime, the PRUSS subsystem is much more powerfull than currently documented. If you want to catch a glimpse you'll have to find you a copy the rev. C TRM. BTW: My fork of the pru assembler supports a large part of these extensions, in case you want to experiment with them : )
Cheers, Bas --- > > Regards, > Kent > > > PS - I promise I'll update Michael's wiki page concerning items [1] and [2] > > ------------------------------------------------------------------------------ > Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS, > MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current > with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft > MVPs and experts. ON SALE this month only -- learn more at: > http://p.sf.net/sfu/learnmore_122712 > _______________________________________________ > Emc-developers mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/emc-developers ------------------------------------------------------------------------------ Master HTML5, CSS3, ASP.NET, MVC, AJAX, Knockout.js, Web API and much more. Get web development skills now with LearnDevNow - 350+ hours of step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122812 _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
