Real Time Kernel 4.19.25-rt16 for 32 and 64 bit is Available
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Tried at compiling git latest machinekit ,linuxcnc pre 2.8 or 2.9 too..it hang on the same looping /usr/lib//tcl8.6/tclConfig.sh: line 2: dpkg-architecture: command not found... start to think linux is secretly sponsored by coffee corps conglomerate.....
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
Hello Revolverve!
a slightly newer version of a full real time kernel, 4.19.82-rt30 SMP PREEMPT RT
http://www.murga-linux.com/puppy/viewto ... 81#1042981
**
a slightly newer version of a full real time kernel, 4.19.82-rt30 SMP PREEMPT RT
http://www.murga-linux.com/puppy/viewto ... 81#1042981
**
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Hi rockedge,
it boot, its fast,faster?
got for one machine to compile input-wacom 0.42...peebee latest kernel does to ,i think ,build his now with latest 0.44 which wont enable finger touch on this tablet..
Works now...
Did not read about one success at running linuxcnc uspace on bionic yet...
just hope where not trying to pie in the sky here..
it for sure will be tricky..
And the main features changed are?
Nice work! I've been told its easy to compile kernel with woof-ce kit,i did tried...,so ,plus a rt patch,bravo!
it boot, its fast,faster?
got for one machine to compile input-wacom 0.42...peebee latest kernel does to ,i think ,build his now with latest 0.44 which wont enable finger touch on this tablet..
Works now...
Did not read about one success at running linuxcnc uspace on bionic yet...
just hope where not trying to pie in the sky here..
it for sure will be tricky..
And the main features changed are?
Nice work! I've been told its easy to compile kernel with woof-ce kit,i did tried...,so ,plus a rt patch,bravo!
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Hey rockedge,
Step one...it configured finally...
It seems that ./configure completed successfully. #
# This means that RT is properly installed #
No wonder you cant find online the solution i applied to get there!!!
of course make bump in tcl.h... a dozen symlink and it compiled OK
I get the exact same error as with the pet i made from binairies..
but learn some more about compiling ...and wrote notes..
root# linuxcnc
LINUXCNC - 2.7.14
Machine configuration directory is '/mnt/sda6/cnc/linuxcnc-2.7.14.tar.gz.extracted/linuxcnc-2.7.14/configs/sim/axis'
Machine configuration file is 'axis.ini'
Starting LinuxCNC...
Found file(lib): /mnt/sda6/cnc/linuxcnc-2.7.14.tar.gz.extracted/linuxcnc-2.7.14/lib/hallib/core_sim.hal
Refusing to run as root without fallback UID specified
---see zip,but not much more in log ,only pidof error ,yes linuxcncsvr got to be kill...
back to the middle ..of the sea of priviledges..
your rt preempt run fine here!
Whats net.ifnames=0 for?
Step one...it configured finally...
It seems that ./configure completed successfully. #
# This means that RT is properly installed #
No wonder you cant find online the solution i applied to get there!!!
of course make bump in tcl.h... a dozen symlink and it compiled OK
I get the exact same error as with the pet i made from binairies..
but learn some more about compiling ...and wrote notes..
root# linuxcnc
LINUXCNC - 2.7.14
Machine configuration directory is '/mnt/sda6/cnc/linuxcnc-2.7.14.tar.gz.extracted/linuxcnc-2.7.14/configs/sim/axis'
Machine configuration file is 'axis.ini'
Starting LinuxCNC...
Found file(lib): /mnt/sda6/cnc/linuxcnc-2.7.14.tar.gz.extracted/linuxcnc-2.7.14/lib/hallib/core_sim.hal
Refusing to run as root without fallback UID specified
---see zip,but not much more in log ,only pidof error ,yes linuxcncsvr got to be kill...
back to the middle ..of the sea of priviledges..
your rt preempt run fine here!
Whats net.ifnames=0 for?
- Attachments
-
- bionic-linuxc2714test-rt-uspace.tar.gz
- (859 Bytes) Downloaded 195 times
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
Hello Revolverve,
That is good news! I am going to follow your lead and get LinuxCNC as far as you have.
Starting with Bionicpup the network card name uses a different naming convention. In Puppy Bionic it is optional but since I copy and paste to the menu.lst and use the same Grub4Dos code over and over again, I just keep the line in the boot menu code.
I am looking forward to a working LinuxCNC on a Bionic64...perhaps making a remaster of a working system for a PuppyCNC64!
That is good news! I am going to follow your lead and get LinuxCNC as far as you have.
what this line does is to make sure that the network devices are named starting with "eth0" and "wlan0". It is optional but when I work with WeeDog (Void Linux) or Bionicpups, the network device names have a different naming system, my scripts that auto-connect the computer to a network are insured to work using the same device name all the time.Whats net.ifnames=0 for?
Starting with Bionicpup the network card name uses a different naming convention. In Puppy Bionic it is optional but since I copy and paste to the menu.lst and use the same Grub4Dos code over and over again, I just keep the line in the boot menu code.
I am looking forward to a working LinuxCNC on a Bionic64...perhaps making a remaster of a working system for a PuppyCNC64!
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Hi rockedge,
Still studying linuxcnc run-in-place dev-environment which is new to me,it may lead me solving some more problems.
There is a lot more learning steps on my side before a puppy of its own will born!! woof-ce being the last one ...but its a long way to there for now.
I am curious, what does in the 10mb stuffing you include in your new rt-preempt make it more linuxcnc friendly ?
edit;oops ,mistmatch rt kernels..your latest is smaller...
Should have ask;did you change latest kernel config?
just asking, they both look stable here.
me too...I am looking forward to a working LinuxCNC on a Bionic64...
Still studying linuxcnc run-in-place dev-environment which is new to me,it may lead me solving some more problems.
There is a lot more learning steps on my side before a puppy of its own will born!! woof-ce being the last one ...but its a long way to there for now.
I am curious, what does in the 10mb stuffing you include in your new rt-preempt make it more linuxcnc friendly ?
edit;oops ,mistmatch rt kernels..your latest is smaller...
Should have ask;did you change latest kernel config?
just asking, they both look stable here.
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Hi rockedge,
Do you mind if ,being dependent of your rtpreempt efforts, i keep the linuxcnc subjet into this topic for a while?
Linuxcnc.pet from binaries +bionicpup64+rtpreempt might get sooner then i expected to a plausible solution simply called ;spot...right before i was tempt "chown'& mess it all...
got a linuxcnc sim to run fine as spot..time to hook it up on cnc-realtime,tomorrow maybe..
Thank for compiling rt kernels.
Do you mind if ,being dependent of your rtpreempt efforts, i keep the linuxcnc subjet into this topic for a while?
Linuxcnc.pet from binaries +bionicpup64+rtpreempt might get sooner then i expected to a plausible solution simply called ;spot...right before i was tempt "chown'& mess it all...
got a linuxcnc sim to run fine as spot..time to hook it up on cnc-realtime,tomorrow maybe..
Thank for compiling rt kernels.
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
Hello Revolverve,
Yes definitely keep the focus on LinuxCNC! I am very excited about the progress you are making. I still have not had a successful ./configure completion as of yet...but I am keeping at it, using the notes you are providing.
The RT patches were more involved but the configuration changes are basically the same.
I am looking forward to your report on using "spot". We should remember that since I use the spot user or webuser:webgroup (www-data sometimes) to run packages that require anybody but the root user. All my web servers and mysql servers run under other users than root.
Yes definitely keep the focus on LinuxCNC! I am very excited about the progress you are making. I still have not had a successful ./configure completion as of yet...but I am keeping at it, using the notes you are providing.
The RT patches were more involved but the configuration changes are basically the same.
I am looking forward to your report on using "spot". We should remember that since I use the spot user or webuser:webgroup (www-data sometimes) to run packages that require anybody but the root user. All my web servers and mysql servers run under other users than root.
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
Hi rockedge,
Reading about spot i found old spot script topic where jamesbond http://murga-linux.com/puppy/viewtopic.php?t=86526 which point a few facts,one state;
You can read a lot on working mint linuxcnc uspace success ,debdog-rtai work great as its a shrink adapted "clone" of the "native" linuxcnc wheezy rtai iso and not at all an all purposes iso ..
Its one thing to be able to get the linuxcnc simulator as spot to run but another when machine hooked to cnc using parport like i do.
An all purpose +linuxcnc-cnc-non-simulation working rtpreempt puppy is one big project,for the "surface scratching end-user" i am..but i will tweak some more, as i dont care breaking everything but i try to keep a slightly more safe mater on priviledges side..
About the "solution" i applied to pass tcl tk ./configure problem come from a redhat tcl8.6-dev rpm,oops,just realize its also in deb package(/usr/lib/x86_64-linux-gnu/tcl8.6/tclConfig.sh)...proof the non-coder i am can try every find..work for compiling run-in-place.
-replace both /usr/lib/tcl8.6/tclConfig.sh and relevant tkConfig.sh (with dpkg-architecture ) for those in /usr/lib/x86_64-linux-gnu/tcl8.6/tclConfig.sh
it should pass tcl/tk configure glitch
-for make, edit for each errors their relevants:
e.g.; hal/utils/halsh.c:20:10: fatal error: tcl.h: No such file or directory
#include <tcl.h>
to
#include <tcl/tcl.h>
I end up creating symlinks, to those not found make complains;objects/hal/utils/halsh.o+ emc/usr_intf/emcsh.cc+/emc/usr_intf/emcsh+ more which i flushed the notes about...my bad,and cannot recreate the errors ,i would have to reload my clean"dev-savefile" to do that,
if you have problems with make i could do that,probably sooner or later i will mess it up and do it anyhow
It still would be interesting to have the point of view of experience spot scripts devs/users/creators about feasibility of a working linuxcnc rtpreempt thing !
Lets test some more..
Glad to see your interest about spot, linuxcnc and rt.
Reading about spot i found old spot script topic where jamesbond http://murga-linux.com/puppy/viewtopic.php?t=86526 which point a few facts,one state;
As its the case with linuxcnc, i think,its more tweaks than i am able to fix..now..or in a near future.. well,using a bionicpup64 based one.; If you run a program under spot, then "all-aspects-of-it" will be running under spot. A program that needs to run under multiple-privilege levels (sometimes as root, sometimes as spot) cant' be run as "spot", it must always be run as root first, and it is up to the program to downgrade its access level to a non-root user. There are other ways of achieving this, but until aufs supports extended attribute this alternative method isn't gonig to work on Puppies (or Fatdog, for that matter)
You can read a lot on working mint linuxcnc uspace success ,debdog-rtai work great as its a shrink adapted "clone" of the "native" linuxcnc wheezy rtai iso and not at all an all purposes iso ..
Its one thing to be able to get the linuxcnc simulator as spot to run but another when machine hooked to cnc using parport like i do.
An all purpose +linuxcnc-cnc-non-simulation working rtpreempt puppy is one big project,for the "surface scratching end-user" i am..but i will tweak some more, as i dont care breaking everything but i try to keep a slightly more safe mater on priviledges side..
About the "solution" i applied to pass tcl tk ./configure problem come from a redhat tcl8.6-dev rpm,oops,just realize its also in deb package(/usr/lib/x86_64-linux-gnu/tcl8.6/tclConfig.sh)...proof the non-coder i am can try every find..work for compiling run-in-place.
-replace both /usr/lib/tcl8.6/tclConfig.sh and relevant tkConfig.sh (with dpkg-architecture ) for those in /usr/lib/x86_64-linux-gnu/tcl8.6/tclConfig.sh
it should pass tcl/tk configure glitch
-for make, edit for each errors their relevants:
e.g.; hal/utils/halsh.c:20:10: fatal error: tcl.h: No such file or directory
#include <tcl.h>
to
#include <tcl/tcl.h>
I end up creating symlinks, to those not found make complains;objects/hal/utils/halsh.o+ emc/usr_intf/emcsh.cc+/emc/usr_intf/emcsh+ more which i flushed the notes about...my bad,and cannot recreate the errors ,i would have to reload my clean"dev-savefile" to do that,
if you have problems with make i could do that,probably sooner or later i will mess it up and do it anyhow
It still would be interesting to have the point of view of experience spot scripts devs/users/creators about feasibility of a working linuxcnc rtpreempt thing !
Lets test some more..
Glad to see your interest about spot, linuxcnc and rt.
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
hey rockedge,
in the cutting edge here i do have the blade pretty dull..
How do you compile a single kernel module??
/usr/src/linux/drivers/char/ppdev.c
make drivers/char/ppdev.ko
yes ,module is there..
depmod -a
modprobe ppdev
modprobe: FATAL: Module ppdev not found in directory /lib/modules/4.19.82-rt30
eeee..i guess copy/paste is not enough ..
Thanks
in the cutting edge here i do have the blade pretty dull..
How do you compile a single kernel module??
/usr/src/linux/drivers/char/ppdev.c
make drivers/char/ppdev.ko
yes ,module is there..
depmod -a
modprobe ppdev
modprobe: FATAL: Module ppdev not found in directory /lib/modules/4.19.82-rt30
eeee..i guess copy/paste is not enough ..
Thanks
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
down load the kernel source file. it is an sfs.....load it with SFS-Load-on-the-fly
http://rockedge.org/kernels/data/kernel ... cpup64.sfs 143.433728 megabytes
get the source code for the module and follow the compile instructions or look for autogen.sh or configure ...or it may just use make
we need to see what is included in the module source package.
When you compile the module and it is finished make install I would guess and then unload the kernel source .sfs file
something along these steps should work
also you can use the kernel source sfs file to compile other kernel modules like the one needed for virtualbox to run or WiFi adapters firmware drivers
http://rockedge.org/kernels/data/kernel ... cpup64.sfs 143.433728 megabytes
get the source code for the module and follow the compile instructions or look for autogen.sh or configure ...or it may just use make
we need to see what is included in the module source package.
When you compile the module and it is finished make install I would guess and then unload the kernel source .sfs file
something along these steps should work
also you can use the kernel source sfs file to compile other kernel modules like the one needed for virtualbox to run or WiFi adapters firmware drivers
- Revolverve
- Posts: 255
- Joined: Sat 08 Nov 2008, 21:01
- Location: 45°17'28.8"N 72°16'08.8"W_avatar/ mira.ca
hey rockedge,
Got ppdev module to compile,well had to compiled the whole char drivers directory,anyhow it work ,thank to google,and you...
So,bionicpup64+rtpreempt+linuxcnc_uspace+spot+hours testing=cnc COMMUNICATE with both parallels ports..more than i expected..
no miracle, first there is latency problem with parport to investigate ...and lots more,
have to find the way to turn friendly spot to the "serial" child process killer i need..chilling...will keep that for later..
Debian rtai/linuxcnc can run the cnc dangerously fast,rtpreempt should be able to do ok,with parallel ports..
, its getting somewhere!
OH,dear devs&compilers,
i bow to your patience and "devotion" in this world of ephemere knowledges...
a thought inspired while checking in the blue kernel menuconfig ...
edit:there aint that much latency as there is ,were and will be weird parallel ports cards adapter behaviours( i end up with 4 of these over the years) ,struggling either in m$ and/or linux,
which here again was my problem...after finding ppdev one..
May it could only be a proof linuxcnc-uspace can run on rtpreempt with an actual machine if im the only user/tester/spot explorer/etc..
In case you recompile another full rt you could look at linuxcnc-stretch-uspace- kernel config for compatibility,e.g. support for ethernet to any i/o you could imagine fpga board...for your zoneminders stuff maybe?
I think a possible full workinglinuxcnc-uspace-rtpreempt deserve its topic....and serve as sweet notes i wont loose ..or the neighbor dogs wont eat..
Got ppdev module to compile,well had to compiled the whole char drivers directory,anyhow it work ,thank to google,and you...
So,bionicpup64+rtpreempt+linuxcnc_uspace+spot+hours testing=cnc COMMUNICATE with both parallels ports..more than i expected..
no miracle, first there is latency problem with parport to investigate ...and lots more,
have to find the way to turn friendly spot to the "serial" child process killer i need..chilling...will keep that for later..
Debian rtai/linuxcnc can run the cnc dangerously fast,rtpreempt should be able to do ok,with parallel ports..
, its getting somewhere!
OH,dear devs&compilers,
i bow to your patience and "devotion" in this world of ephemere knowledges...
a thought inspired while checking in the blue kernel menuconfig ...
edit:there aint that much latency as there is ,were and will be weird parallel ports cards adapter behaviours( i end up with 4 of these over the years) ,struggling either in m$ and/or linux,
which here again was my problem...after finding ppdev one..
May it could only be a proof linuxcnc-uspace can run on rtpreempt with an actual machine if im the only user/tester/spot explorer/etc..
In case you recompile another full rt you could look at linuxcnc-stretch-uspace- kernel config for compatibility,e.g. support for ethernet to any i/o you could imagine fpga board...for your zoneminders stuff maybe?
I think a possible full workinglinuxcnc-uspace-rtpreempt deserve its topic....and serve as sweet notes i wont loose ..or the neighbor dogs wont eat..
- Attachments
-
- config-4.9.0-8-rt-amd64.tar.bz2
- (40.56 KiB) Downloaded 183 times
[url=https://www.mira.ca/en/donate]Help mira.ca[/url]
@rockedge
Just tripped over a problem with the configuration of the kernel sources sfs for (at least one of) your rt kernels.
I'm running Bionic64 with the4.19.25-rt16 kernel.
devx & kernel-sources sfs are loaded.
I'm trying to install virtualbox, which requires we compile some modules using the supplied
/usr/lib/virtualbox/vboxdrv.sh
That script calls another to check that we have everything it needs:
/usr/lib/virtualbox/check_module_dependencies.sh
The critical bit that fails first from the latter is
If we run this command alone in a terminal, output is
If we check down the tree, we see the following (nb `ll` is an alias for `ls -l`):
Just tripped over a problem with the configuration of the kernel sources sfs for (at least one of) your rt kernels.
I'm running Bionic64 with the4.19.25-rt16 kernel.
devx & kernel-sources sfs are loaded.
I'm trying to install virtualbox, which requires we compile some modules using the supplied
/usr/lib/virtualbox/vboxdrv.sh
That script calls another to check that we have everything it needs:
/usr/lib/virtualbox/check_module_dependencies.sh
The critical bit that fails first from the latter is
Code: Select all
test -d "/lib/modules/`uname -r`/build/include" && HAVE_HEADERS=yes
Code: Select all
root# test -d "/lib/modules/`uname -r`/build/include"
root# echo $?
1
Code: Select all
root# ll /lib/modules/$(uname -r)/build/include/
ls: cannot access '/lib/modules/4.19.25-rt16/build/include/': No such file or directory
root# ll /lib/modules/$(uname -r)/build/
ls: cannot access '/lib/modules/4.19.25-rt16/build/': No such file or directory
root# ll /lib/modules/$(uname -r)/build
lrwxrwxrwx 1 root root 84 13 Mar 2019 23:22 /lib/modules/4.19.25-rt16/build -> /initrd/mnt/dev_save/woof-out_x86_64_x86_64_ubuntu_bionic64/kernel-kit/linux-4.19.25
I have set up a Bionic64 and am running virtualbox 6.1
I also can not compile the vboxdrv module.....I can't even get the build script to see any linux-headers for the kernel and of course the original set is somewhere in gigabytes of storage
I loaded the same devx and kernel sources but the headers are missing for me. I may have to recompile the kernel to test this.
I also can not compile the vboxdrv module.....I can't even get the build script to see any linux-headers for the kernel and of course the original set is somewhere in gigabytes of storage
I loaded the same devx and kernel sources but the headers are missing for me. I may have to recompile the kernel to test this.
So far as I can see, the problem lies in the "build" link in the /lib/modules/4.19.25-rt16/ directory which points to a non-existent directory in dev_save.
The link comes from mounting the kernel sources sfs I think?
It looks like it might be a hangover from your build harness & has got into the kernel sources sfs in error.
(You have probably realised this already, alternatively I've got hold of the wrong end of the stick entirely)
The link comes from mounting the kernel sources sfs I think?
It looks like it might be a hangover from your build harness & has got into the kernel sources sfs in error.
(You have probably realised this already, alternatively I've got hold of the wrong end of the stick entirely)
Hello paulh177,
you're totally correct, it does appear that the symlink somehow stayed from the build and was written that way to the sfs.
I was able to compile the vboxdrv module for virtualbox 6.1 after doing several things. First I changed the symlink build to /usr/src/linux. But now it gets fuzzy as what made it work. I reloaded the 4.19.25 sources sfs but then loaded the kernel 4.19.23 sources from the original Bionic64-v8 kernel.and I reloaded binutils for BOTH 64bit and 32 bit and then ran
and the vboxdrv module built correctly using the kernel 4.19.25-rt16
got lucky
you're totally correct, it does appear that the symlink somehow stayed from the build and was written that way to the sfs.
I was able to compile the vboxdrv module for virtualbox 6.1 after doing several things. First I changed the symlink build to /usr/src/linux. But now it gets fuzzy as what made it work. I reloaded the 4.19.25 sources sfs but then loaded the kernel 4.19.23 sources from the original Bionic64-v8 kernel.and I reloaded binutils for BOTH 64bit and 32 bit and then ran
Code: Select all
/etc/init.d/vboxdrv setup
got lucky
- Attachments
-
- capture23984.png
- (141.99 KiB) Downloaded 231 times
F U Z Z Y !
hahhhhhhhhhhhhhhhhhaaaaaa!
I used ( a very long time ago) to run a software house tech support group, and I bashed into them the mantra "ONLY CHANGE ONE THING AT A TIME THEN TEST THEN YOU'LL KNOW WHAT FIXED IT".
Unfortunately, like you I don't take my own good advice
I'll look at it again later, thanks for checking.
hahhhhhhhhhhhhhhhhhaaaaaa!
I used ( a very long time ago) to run a software house tech support group, and I bashed into them the mantra "ONLY CHANGE ONE THING AT A TIME THEN TEST THEN YOU'LL KNOW WHAT FIXED IT".
Unfortunately, like you I don't take my own good advice
I'll look at it again later, thanks for checking.
now that is time to feed the cats and Bundesliga is showing a game on TV
I'm ready to un-fuzz what I did and instead of starting to throw the Out house sink at it I will do the correct troubleshooting so even I know how to solve the fix exactly and how to make the fix permanent in the sources sfs without recompiling..which I now I am running into problems applying the RT patches using the kernel-kit....
I'm ready to un-fuzz what I did and instead of starting to throw the Out house sink at it I will do the correct troubleshooting so even I know how to solve the fix exactly and how to make the fix permanent in the sources sfs without recompiling..which I now I am running into problems applying the RT patches using the kernel-kit....