Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Alexander Sharabanov]

Hi All,

I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch. I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. Could you please help me to resolve that.

Thanks,
Alexander

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Jan Kiszka]

On 22.04.19 15:36, Alexander Sharabanov via Xenomai wrote:
> Hi All,
> 
> I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch. I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. Could you please help me to resolve that.

The primary error seems that your userspace is still on 2.5 while the kernel 
part is now on 3.0.

Regarding the kernel and ppc: 3.16 will be dead soon as well (early next year), 
and we do not test against such old kernels anymore. We do not have an official 
ppc ipipe release newer than 4.4.71 or 4.9.51. The 4.9 kernel is no longer 
regularly updated, but 4.4 is - specifically via the STLS CIP kernel - and will 
continue to be.

However, ppc is not tested by me, not even build-tested anymore. If you validate 
ipipe-4.4.y[-cip] from https://gitlab.denx.de/Xenomai/ipipe/branches, I can do 
the tagging also for ppc (like I'm doing for arm if users report it works).

Steven, are you planning to refresh the kernel patches beyond that? How about 
adding CI build for ppc? We would need support on that topic. Otherwise, it's 
likely time to bury that arch, like it's happening all around.

Jan

-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc]]

Hi guys.

I released a 4.14.36 ipipe patch last June. I used the tags mechanism that Philippe did. The latest tag is ipipe-core-4.14.36-ppc32-1.1.

You can find this stuff at: https://gitlab.denx.de/Xenomai/ipipe-ppc32 or https://xenomai.org/downloads/ipipe/v4.x/ppc32/ for a quick download link.

We made ipipe-ppc32. The ppc64 arch is now removed as there was an RFC and nobody spoke up. I have no hardware to test on it.

We are moving forward with xenomai/ppc32 at NASA and our arinc653 implementation. The current thinking is while it may not fly, it can at least be a viable (and free) development platform for arinc-653 applications that can then be moved to commercial proprietary platforms with certification if need-be.

Steven

________________________________
From: Jan Kiszka <jan.kiszka@siemens.com>
Sent: Wednesday, April 24, 2019 1:46 AM
To: Alexander Sharabanov; xenomai@xenomai.org; Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc]
Subject: Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

On 22.04.19 15:36, Alexander Sharabanov via Xenomai wrote:
> Hi All,
>
> I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch. I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. Could you please help me to resolve that.

The primary error seems that your userspace is still on 2.5 while the kernel
part is now on 3.0.

Regarding the kernel and ppc: 3.16 will be dead soon as well (early next year),
and we do not test against such old kernels anymore. We do not have an official
ppc ipipe release newer than 4.4.71 or 4.9.51. The 4.9 kernel is no longer
regularly updated, but 4.4 is - specifically via the STLS CIP kernel - and will
continue to be.

However, ppc is not tested by me, not even build-tested anymore. If you validate
ipipe-4.4.y[-cip] from https://urldefense.proofpoint.com/v2/url?u=https-3A__gitlab.denx.de_Xenomai_ipipe_branches&d=DwICaQ&c=ApwzowJNAKKw3xye91w7BE1XMRKi2LN9kiMk5Csz9Zk&r=r63S9M7Jhsu-Ginn_ewiQeYiA39ideOTJunKLVvbEVs&m=2HxaHR_Z81H2hWU2fVH4D7v_N6tFRXgEfKqqJID9-HA&s=qZatnoFMBt8oA7MdrLnOFzrcfiJp9Q6-hpCfYusFmq8&e=, I can do
the tagging also for ppc (like I'm doing for arm if users report it works).

Steven, are you planning to refresh the kernel patches beyond that? How about
adding CI build for ppc? We would need support on that topic. Otherwise, it's
likely time to bury that arch, like it's happening all around.

Jan

--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc]]

Sorry, but what is CI build?

Steven
________________________________
From: Jan Kiszka <jan.kiszka@siemens.com>
Sent: Wednesday, April 24, 2019 1:46 AM
To: Alexander Sharabanov; xenomai@xenomai.org; Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc]
Subject: Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

On 22.04.19 15:36, Alexander Sharabanov via Xenomai wrote:
> Hi All,
>
> I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch. I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. Could you please help me to resolve that.

The primary error seems that your userspace is still on 2.5 while the kernel
part is now on 3.0.

Regarding the kernel and ppc: 3.16 will be dead soon as well (early next year),
and we do not test against such old kernels anymore. We do not have an official
ppc ipipe release newer than 4.4.71 or 4.9.51. The 4.9 kernel is no longer
regularly updated, but 4.4 is - specifically via the STLS CIP kernel - and will
continue to be.

However, ppc is not tested by me, not even build-tested anymore. If you validate
ipipe-4.4.y[-cip] from https://urldefense.proofpoint.com/v2/url?u=https-3A__gitlab.denx.de_Xenomai_ipipe_branches&d=DwICaQ&c=ApwzowJNAKKw3xye91w7BE1XMRKi2LN9kiMk5Csz9Zk&r=r63S9M7Jhsu-Ginn_ewiQeYiA39ideOTJunKLVvbEVs&m=2HxaHR_Z81H2hWU2fVH4D7v_N6tFRXgEfKqqJID9-HA&s=qZatnoFMBt8oA7MdrLnOFzrcfiJp9Q6-hpCfYusFmq8&e=, I can do
the tagging also for ppc (like I'm doing for arm if users report it works).

Steven, are you planning to refresh the kernel patches beyond that? How about
adding CI build for ppc? We would need support on that topic. Otherwise, it's
likely time to bury that arch, like it's happening all around.

Jan

--
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Jan Kiszka]

On 24.04.19 08:11, Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc] 
wrote:
> Sorry, but what is CI build?
> 

https://gitlab.denx.de/Xenomai/xenomai/blob/master/.travis.yml
https://travis-ci.com/xenomai-ci/xenomai/

Jan

> Steven
> --------------------------------------------------------------------------------
> *From:* Jan Kiszka <jan.kiszka@siemens.com>
> *Sent:* Wednesday, April 24, 2019 1:46 AM
> *To:* Alexander Sharabanov; xenomai@xenomai.org; Seeger, Steven D. 
> (GSFC-582.0)[Embedded Flight Systems, Inc]
> *Subject:* Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.
> On 22.04.19 15:36, Alexander Sharabanov via Xenomai wrote:
>> Hi All,
>> 
>> I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch.  I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE 
> disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. 
> Could you please help me to resolve that.
> 
> The primary error seems that your userspace is still on 2.5 while the kernel
> part is now on 3.0.
> 
> Regarding the kernel and ppc: 3.16 will be dead soon as well (early next year),
> and we do not test against such old kernels anymore. We do not have an official
> ppc ipipe release newer than 4.4.71 or 4.9.51. The 4.9 kernel is no longer
> regularly updated, but 4.4 is - specifically via the STLS CIP kernel - and will
> continue to be.
> 
> However, ppc is not tested by me, not even build-tested anymore. If you validate
> ipipe-4.4.y[-cip] from 
> https://urldefense.proofpoint.com/v2/url?u=https-3A__gitlab.denx.de_Xenomai_ipipe_branches&d=DwICaQ&c=ApwzowJNAKKw3xye91w7BE1XMRKi2LN9kiMk5Csz9Zk&r=r63S9M7Jhsu-Ginn_ewiQeYiA39ideOTJunKLVvbEVs&m=2HxaHR_Z81H2hWU2fVH4D7v_N6tFRXgEfKqqJID9-HA&s=qZatnoFMBt8oA7MdrLnOFzrcfiJp9Q6-hpCfYusFmq8&e=, 
> I can do
> the tagging also for ppc (like I'm doing for arm if users report it works).
> 
> Steven, are you planning to refresh the kernel patches beyond that? How about
> adding CI build for ppc? We would need support on that topic. Otherwise, it's
> likely time to bury that arch, like it's happening all around.
> 
> Jan
> 
> -- 
> Siemens AG, Corporate Technology, CT RDA IOT SES-DE
> Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Jan Kiszka]

On 24.04.19 08:10, Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc] 
wrote:
> Hi guys.
> 
> I released a 4.14.36 ipipe patch last June. I used the tags mechanism that 
> Philippe did. The latest tag is ipipe-core-4.14.36-ppc32-1.1.
> 
> You can find this stuff at: https://gitlab.denx.de/Xenomai/ipipe-ppc32 or 
> https://xenomai.org/downloads/ipipe/v4.x/ppc32/ for a quick download link.
> 
> We made ipipe-ppc32. The ppc64 arch is now removed as there was an RFC and 
> nobody spoke up. I have no hardware to test on it.

Ah, I was indeed looking at powerpc, not ppc32. Should this be updated to a 
current 4.14? And for LTS, you will likely want to look at 4.19 and/or 
reactivate 4.4.

> 
> We are moving forward with xenomai/ppc32 at NASA and our arinc653 
> implementation. The current thinking is while it may not fly, it can at least be 
> a viable (and free) development platform for arinc-653 applications that can 
> then be moved to commercial proprietary platforms with certification if need-be.

That will surely be a valuable result, likely not even bound to ppc32.

As I only learned recently (from a vendor that is still selling a lot of ppc32 
silicon but no longer developing it) the toolchain situation is also getting 
worse: gcc discontinued ppc with release 8 and will drop it from 9. So, it will 
be "fun" to build for that arch in the not so far future, I'm afraid.

Jan

> 
> Steven
> 
> --------------------------------------------------------------------------------
> *From:* Jan Kiszka <jan.kiszka@siemens.com>
> *Sent:* Wednesday, April 24, 2019 1:46 AM
> *To:* Alexander Sharabanov; xenomai@xenomai.org; Seeger, Steven D. 
> (GSFC-582.0)[Embedded Flight Systems, Inc]
> *Subject:* Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.
> On 22.04.19 15:36, Alexander Sharabanov via Xenomai wrote:
>> Hi All,
>> 
>> I am upgrading the linux kernel from linix kernel v2.6.30 to linux 3.16.61 (long tern support). Also I have upgraded the Xenomai and ipipe. I upgraded the xenomai to stable /v3.0.x (instead the Xenomai-2.4.10 that I used in old kernel), and ipipe-core-3.16-powerpc-1.patch.  I have got the next error:  Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE 
> disabled. (modprobe xeno_native?). I do not see this options in new Xenomai. 
> Could you please help me to resolve that.
> 
> The primary error seems that your userspace is still on 2.5 while the kernel
> part is now on 3.0.
> 
> Regarding the kernel and ppc: 3.16 will be dead soon as well (early next year),
> and we do not test against such old kernels anymore. We do not have an official
> ppc ipipe release newer than 4.4.71 or 4.9.51. The 4.9 kernel is no longer
> regularly updated, but 4.4 is - specifically via the STLS CIP kernel - and will
> continue to be.
> 
> However, ppc is not tested by me, not even build-tested anymore. If you validate
> ipipe-4.4.y[-cip] from 
> https://urldefense.proofpoint.com/v2/url?u=https-3A__gitlab.denx.de_Xenomai_ipipe_branches&d=DwICaQ&c=ApwzowJNAKKw3xye91w7BE1XMRKi2LN9kiMk5Csz9Zk&r=r63S9M7Jhsu-Ginn_ewiQeYiA39ideOTJunKLVvbEVs&m=2HxaHR_Z81H2hWU2fVH4D7v_N6tFRXgEfKqqJID9-HA&s=qZatnoFMBt8oA7MdrLnOFzrcfiJp9Q6-hpCfYusFmq8&e=, 
> I can do
> the tagging also for ppc (like I'm doing for arm if users report it works).
> 
> Steven, are you planning to refresh the kernel patches beyond that? How about
> adding CI build for ppc? We would need support on that topic. Otherwise, it's
> likely time to bury that arch, like it's happening all around.
> 
> Jan
> 
> -- 
> Siemens AG, Corporate Technology, CT RDA IOT SES-DE
> Corporate Competence Center Embedded Linux


-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc]]

Unfortunately NASA has seen fit to require us to use this horrible outlook 365 webmail thing, and apparently it does not support the typical inline quoting that we use on this list. I will have to try and figure out a solution but of course I'm already extremely busy and that is just another thing for the to-do list.

You are correct that the arinc653 work will not be tied to ppc32 at all. It could all be part of noarch. However, we have a todo item here to work out with our lawyers if the licensing issues on the arinc653 specification are such that this work can be made public. I'd like it to be, for sure. Not to compete or take money away from any commercial implementation, but to aid groups like mine in various organizations out there who want to examine arinc-653 and do prototyping without the high cost of entry. Stay tuned for news on this. FYI, most of arinc-653 part 1 required services is already implemented. I have one more batch of work to do on the kernel side but so far it's looking pretty good. We have already taken some applications written against arinc-653 and run them on Xenomai with SCHED_TP.

I should definitely update to a more recent 4.14. I see noarch bits for 4.14.110, so I just rebased all my work on top of that and it was pretty clean. (Any conflicts were not in specific IPIPE bits.) I don't have my boards with me right now, but I will be working with them later this week and can test. I will try and get the patch out by the end of the week.

ppc32 is what we are calling it to prevent confusion with the now-dead ppc terminology, as ppc64 is no longer supported by Xenomai.

As for 4.4, I believe that was pre-noarch days. Philippe and I had decided that any future work on ppc would be done based on the noarch 4.14 branch. I assume at some point there will need to be a noarch branch for another kernel. I've not been as involved with the Xenomai project or mailing list as I'd like, but my time is limited in this area right now.

As for 4.19, I figured I would wait until noarch is stable there. Right now I see a wip branch. Would you like me to look into a 4.19 ipipe patch on top of that? It may not be that hard for PPC.

I am sad to see what you wrote about ppc being dropped from GCC. I was not aware of that. I know that PPC is still very much in use in the aerospace industry. I guess a commercial player will need to pay some folks to support GCC and other FOSS projects for the architecture. There probably isn't any reason that the same tools we have all used for years can't still be used for years, however. Who knows.

I know that most of you guys are doing ARM these days. I almost had a chance to do some arm work here, but I'm apparently grouped in with the wrong set of folks so I didn't get to. Maybe someday.

Steven

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Lennart Sorensen]

On Wed, Apr 24, 2019 at 08:28:05AM +0200, Jan Kiszka via Xenomai wrote:
> That will surely be a valuable result, likely not even bound to ppc32.
> 
> As I only learned recently (from a vendor that is still selling a lot of
> ppc32 silicon but no longer developing it) the toolchain situation is also
> getting worse: gcc discontinued ppc with release 8 and will drop it from 9.
> So, it will be "fun" to build for that arch in the not so far future, I'm
> afraid.
> 
> Jan

GCC has NOT dropped powerpc support.

They have dropped this:

Support for the powerpc*-*-*spe*

PowerPC SPE is the signal processing version implemented in the e500
core for example (good riddance as far as I am concerned).  It has an
incompatible FPU design with different instructions that conflict with
altivec and some other extensions.  Very few CPUs seem to have ever been
designed with the SPE extension and hence why it sees very little use
or testing.

Real PowerPC is doing fine in GCC.

-- 
Len Sorensen

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Lennart Sorensen]

On Wed, Apr 24, 2019 at 07:09:10AM +0000, Seeger, Steven D. (GSFC-582.0)[Embedded Flight Systems, Inc] via Xenomai wrote:
> I am sad to see what you wrote about ppc being dropped from GCC. I was not aware of that. I know that PPC is still very much in use in the aerospace industry. I guess a commercial player will need to pay some folks to support GCC and other FOSS projects for the architecture. There probably isn't any reason that the same tools we have all used for years can't still be used for years, however. Who knows.

Don't worry, PPC is fine in GCC.  Just people misreading a probably not
very clear statement from GCC.

-- 
Len Sorensen

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Jan Kiszka]

On 24.04.19 21:51, Lennart Sorensen wrote:
> On Wed, Apr 24, 2019 at 08:28:05AM +0200, Jan Kiszka via Xenomai wrote:
>> That will surely be a valuable result, likely not even bound to ppc32.
>>
>> As I only learned recently (from a vendor that is still selling a lot of
>> ppc32 silicon but no longer developing it) the toolchain situation is also
>> getting worse: gcc discontinued ppc with release 8 and will drop it from 9.
>> So, it will be "fun" to build for that arch in the not so far future, I'm
>> afraid.
>>
>> Jan
> 
> GCC has NOT dropped powerpc support.
> 
> They have dropped this:
> 
> Support for the powerpc*-*-*spe*
> 
> PowerPC SPE is the signal processing version implemented in the e500
> core for example (good riddance as far as I am concerned).  It has an
> incompatible FPU design with different instructions that conflict with
> altivec and some other extensions.  Very few CPUs seem to have ever been
> designed with the SPE extension and hence why it sees very little use
> or testing.
> 
> Real PowerPC is doing fine in GCC.
> 

Yes, I meanwhile understood that this information I got first-hand from a 
company that still dominants the ppc32 silicon market was inaccurate.

Still, this won't change the fact that ppc32 is a dying arch. There is no new 
silicon design coming up anymore, major distros are moving away from it (not to 
speak of those who never jumped on it), and the arch is - in contrast to MIPS or 
specifically RISC-V - still a proprietary one, thus without hope of a second 
life in the free world. Internally, I haven't received support request for it 
for almost a decade.

It's history, let's face reality.

Jan

-- 
Siemens AG, Corporate Technology, CT RDA IOT SES-DE
Corporate Competence Center Embedded Linux

Re: Xenomai: native skin or CONFIG_XENO_OPT_PERVASIVE disabled.

[Lennart Sorensen]

On Wed, Apr 24, 2019 at 10:21:45PM +0200, Jan Kiszka wrote:
> Yes, I meanwhile understood that this information I got first-hand from a
> company that still dominants the ppc32 silicon market was inaccurate.
> 
> Still, this won't change the fact that ppc32 is a dying arch. There is no
> new silicon design coming up anymore, major distros are moving away from it
> (not to speak of those who never jumped on it), and the arch is - in
> contrast to MIPS or specifically RISC-V - still a proprietary one, thus
> without hope of a second life in the free world. Internally, I haven't
> received support request for it for almost a decade.
> 
> It's history, let's face reality.

That is certainly true.

I wonder how long the products I used to work at using Xenomai on powerpc
will stick around.  They will likely stay in production for years still
but will they decide to move to xenomai 3 from 2.6 at some point I don't
know (given I don't work there anymore and since it is working maybe
there is no point messing with it).  But it means at least a few Siemens
products are running xenomai on powerpc still.

-- 
Len Sorensen