dovetail kernel maintenance

dovetail kernel maintenance

[Jan Kiszka]

Hi Philippe,

people started to ask about updates for our kernel trees, and you
mentioned to me before that you would focus on the last two LTS kernels.
Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
then also in 5.10 (all referring to the -rebase branches). Before taking
over 5.10 and 5.15, I would like to clarify their state. But I would
also like to know what is delaying the tagging of 6.1 and 6.6.

Thanks,
Jan

-- 
Siemens AG, Technology
Linux Expert Center

Re: dovetail kernel maintenance

[Jan Kiszka]

On 21.02.24 19:30, Jan Kiszka wrote:
> Hi Philippe,
> 
> people started to ask about updates for our kernel trees, and you
> mentioned to me before that you would focus on the last two LTS kernels.
> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
> then also in 5.10 (all referring to the -rebase branches). Before taking
> over 5.10 and 5.15, I would like to clarify their state. But I would
> also like to know what is delaying the tagging of 6.1 and 6.6.
> 

Gentle ping for this question.

Jan

-- 
Siemens AG, Technology
Linux Expert Center

Re: dovetail kernel maintenance

[Philippe Gerum]

Jan Kiszka <jan.kiszka@siemens.com> writes:

> On 21.02.24 19:30, Jan Kiszka wrote:
>> Hi Philippe,
>> 
>> people started to ask about updates for our kernel trees, and you
>> mentioned to me before that you would focus on the last two LTS kernels.
>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>> then also in 5.10 (all referring to the -rebase branches). Before taking
>> over 5.10 and 5.15, I would like to clarify their state. But I would
>> also like to know what is delaying the tagging of 6.1 and 6.6.
>> 
>
> Gentle ping for this question.
>

Well, my posts to the mailing list from my usual address have been
rejected a couple of weeks ago due to smtp authentication issues it
seems, which are now (mostly) sorted out, I thought you did receive this
reply as a direct recipient though. Looks like you did not. So here it
is.

PS: Since this reply was written, the x86 status with respect to the
Dovetail port to v6.8-rc7 has improved. Still in the process of
rewriting the arm64/fpsimd stuff.

> From: Philippe Gerum <rpm@xenomai.org>
> Subject: Re: dovetail kernel maintenance
> To: Jan Kiszka <jan.kiszka@siemens.com>
> Cc: Xenomai <xenomai@lists.linux.dev>
> Date: Wed, 21 Feb 2024 21:00:17 +0100
> 
> 
> Jan Kiszka <jan.kiszka@siemens.com> writes:
> 
> > Hi Philippe,
> >
> > people started to ask about updates for our kernel trees, and you
> > mentioned to me before that you would focus on the last two LTS kernels.
> > Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
> > then also in 5.10 (all referring to the -rebase branches). Before taking
> > over 5.10 and 5.15, I would like to clarify their state.
> 
> Dovetail ports to 5.10 and 5.15 are not maintained anymore, although I'm
> still upgrading 5.10 from time to time for work-related reasons, but
> nobody should expect this to last for long.
> 
> > But I would
> > also like to know what is delaying the tagging of 6.1 and 6.6.
> >
> 
> Those two have been tagged this afternoon. Dovetail maintenance takes
> time I must divert from my day-to-day workload where the real-time
> network stack for xenomai4 has precedence. The delay was mostly due to
> lack of basic testing until recently. Btw, the situation is going to get
> worse with 6.8 which requires a significant rewrite of the arm64/fpu
> support for Dovetail, not to speak of breaking changes in the generic
> entry path x86 uses.
> 
> 
> 
> -- 
> Philippe.
> 

Re: dovetail kernel maintenance

[Jan Kiszka]

On 06.03.24 09:41, Philippe Gerum wrote:
> 
> Jan Kiszka <jan.kiszka@siemens.com> writes:
> 
>> On 21.02.24 19:30, Jan Kiszka wrote:
>>> Hi Philippe,
>>>
>>> people started to ask about updates for our kernel trees, and you
>>> mentioned to me before that you would focus on the last two LTS kernels.
>>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>>> then also in 5.10 (all referring to the -rebase branches). Before taking
>>> over 5.10 and 5.15, I would like to clarify their state. But I would
>>> also like to know what is delaying the tagging of 6.1 and 6.6.
>>>
>>
>> Gentle ping for this question.
>>
> 
> Well, my posts to the mailing list from my usual address have been
> rejected a couple of weeks ago due to smtp authentication issues it
> seems, which are now (mostly) sorted out, I thought you did receive this
> reply as a direct recipient though. Looks like you did not. So here it
> is.
> 
> PS: Since this reply was written, the x86 status with respect to the
> Dovetail port to v6.8-rc7 has improved. Still in the process of
> rewriting the arm64/fpsimd stuff.
> 

Let us know where exactly help could be provided. I just want to avoid
duplicate attempts, so the development should be coordinated publicly.

>> From: Philippe Gerum <rpm@xenomai.org>
>> Subject: Re: dovetail kernel maintenance
>> To: Jan Kiszka <jan.kiszka@siemens.com>
>> Cc: Xenomai <xenomai@lists.linux.dev>
>> Date: Wed, 21 Feb 2024 21:00:17 +0100
>>
>>
>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>>
>>> Hi Philippe,
>>>
>>> people started to ask about updates for our kernel trees, and you
>>> mentioned to me before that you would focus on the last two LTS kernels.
>>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>>> then also in 5.10 (all referring to the -rebase branches). Before taking
>>> over 5.10 and 5.15, I would like to clarify their state.
>>
>> Dovetail ports to 5.10 and 5.15 are not maintained anymore, although I'm
>> still upgrading 5.10 from time to time for work-related reasons, but
>> nobody should expect this to last for long.
>>

I can take over those two, but we should coordinate that to avoid
parallel updates, at least of 5.10. As 5.15 is dropped already, I would
grab it now officially. Still need to decide whether I continue the
rebase branches, though.

>>> But I would
>>> also like to know what is delaying the tagging of 6.1 and 6.6.
>>>
>>
>> Those two have been tagged this afternoon. Dovetail maintenance takes
>> time I must divert from my day-to-day workload where the real-time
>> network stack for xenomai4 has precedence. The delay was mostly due to
>> lack of basic testing until recently. Btw, the situation is going to get
>> worse with 6.8 which requires a significant rewrite of the arm64/fpu
>> support for Dovetail, not to speak of breaking changes in the generic
>> entry path x86 uses.
>>

Jan

-- 
Siemens AG, Technology
Linux Expert Center

Re: dovetail kernel maintenance

[Philippe Gerum]

Jan Kiszka <jan.kiszka@siemens.com> writes:

> On 06.03.24 09:41, Philippe Gerum wrote:
>> 
>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>> 
>>> On 21.02.24 19:30, Jan Kiszka wrote:
>>>> Hi Philippe,
>>>>
>>>> people started to ask about updates for our kernel trees, and you
>>>> mentioned to me before that you would focus on the last two LTS kernels.
>>>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>>>> then also in 5.10 (all referring to the -rebase branches). Before taking
>>>> over 5.10 and 5.15, I would like to clarify their state. But I would
>>>> also like to know what is delaying the tagging of 6.1 and 6.6.
>>>>
>>>
>>> Gentle ping for this question.
>>>
>> 
>> Well, my posts to the mailing list from my usual address have been
>> rejected a couple of weeks ago due to smtp authentication issues it
>> seems, which are now (mostly) sorted out, I thought you did receive this
>> reply as a direct recipient though. Looks like you did not. So here it
>> is.
>> 
>> PS: Since this reply was written, the x86 status with respect to the
>> Dovetail port to v6.8-rc7 has improved. Still in the process of
>> rewriting the arm64/fpsimd stuff.
>> 
>
> Let us know where exactly help could be provided. I just want to avoid
> duplicate attempts, so the development should be coordinated publicly.
>

Addressing [1] comes to mind. As I said in this thread, one could use
the same logic which allows x86 to support preemption of the helpers for
in-kernel fpu by the real-time core.  However, I have to upgrade the
basic arm64/fpsimd port to v6.8 before someone else can handle this
task. ETA, hopefully next week.

>>> From: Philippe Gerum <rpm@xenomai.org>
>>> Subject: Re: dovetail kernel maintenance
>>> To: Jan Kiszka <jan.kiszka@siemens.com>
>>> Cc: Xenomai <xenomai@lists.linux.dev>
>>> Date: Wed, 21 Feb 2024 21:00:17 +0100
>>>
>>>
>>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>>>
>>>> Hi Philippe,
>>>>
>>>> people started to ask about updates for our kernel trees, and you
>>>> mentioned to me before that you would focus on the last two LTS kernels.
>>>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>>>> then also in 5.10 (all referring to the -rebase branches). Before taking
>>>> over 5.10 and 5.15, I would like to clarify their state.
>>>
>>> Dovetail ports to 5.10 and 5.15 are not maintained anymore, although I'm
>>> still upgrading 5.10 from time to time for work-related reasons, but
>>> nobody should expect this to last for long.
>>>
>
> I can take over those two, but we should coordinate that to avoid
> parallel updates, at least of 5.10. As 5.15 is dropped already, I would
> grab it now officially. Still need to decide whether I continue the
> rebase branches, though.

I don't plan to merge the generic support for oob networking (in
net/core/) which underlies the UDP implementation for xenomai4. So I'm
fine with handing over the maintenance of v5.10 asap.

[1] https://lore.kernel.org/xenomai/ff580dec-096d-45cb-b906-527446c4fa87@163.com/T/#t

-- 
Philippe.

Re: dovetail kernel maintenance

[Philippe Gerum]

Philippe Gerum <rpm@xenomai.org> writes:

> Jan Kiszka <jan.kiszka@siemens.com> writes:
>
>> On 06.03.24 09:41, Philippe Gerum wrote:
>>> 
>>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>>> 
>>>> On 21.02.24 19:30, Jan Kiszka wrote:
>>>>> Hi Philippe,
>>>>>
>>>>> people started to ask about updates for our kernel trees, and you
>>>>> mentioned to me before that you would focus on the last two LTS kernels.
>>>>> Right now, there are to-be-tagged updated sleeping in 6.1 and 6.6 but
>>>>> then also in 5.10 (all referring to the -rebase branches). Before taking
>>>>> over 5.10 and 5.15, I would like to clarify their state. But I would
>>>>> also like to know what is delaying the tagging of 6.1 and 6.6.
>>>>>
>>>>
>>>> Gentle ping for this question.
>>>>
>>> 
>>> Well, my posts to the mailing list from my usual address have been
>>> rejected a couple of weeks ago due to smtp authentication issues it
>>> seems, which are now (mostly) sorted out, I thought you did receive this
>>> reply as a direct recipient though. Looks like you did not. So here it
>>> is.
>>> 
>>> PS: Since this reply was written, the x86 status with respect to the
>>> Dovetail port to v6.8-rc7 has improved. Still in the process of
>>> rewriting the arm64/fpsimd stuff.
>>> 
>>
>> Let us know where exactly help could be provided. I just want to avoid
>> duplicate attempts, so the development should be coordinated publicly.
>>
>
> Addressing [1] comes to mind. As I said in this thread, one could use
> the same logic which allows x86 to support preemption of the helpers for
> in-kernel fpu by the real-time core.  However, I have to upgrade the
> basic arm64/fpsimd port to v6.8 before someone else can handle this
> task. ETA, hopefully next week.
>

Dovetail and EVL ports to v6.8 are now available from the usual repos
for all the supported architectures. Only lightly tested so far, but
things look good (arm64/fpsimd is back on track).

If CONFIG_NET_OOB is set, there are some warnings about missing function
declarations in the net/ core which are now turned on for ARM. We may
have to live with them until the generic changes required for enabling
the UDP support in the xenomai4 netstack are merged. Significant changes
are pending in this area.

-- 
Philippe.