Re: v6.1.y-dovetail-rebase

[Philippe Gerum <rpm@xenomai.org>]

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

> On 09.08.23 15:11, Philippe Gerum wrote:
>> 
>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>> 
>>> On 09.08.23 07:52, Jan Kiszka wrote:
>>>> On 01.08.23 15:26, Philippe Gerum wrote:
>>>>>
>>>>> Florian Bezdeka <florian.bezdeka@siemens.com> writes:
>>>>>
>>>>>> On Mon, 2023-07-10 at 08:59 +0200, Florian Bezdeka wrote:
>>>>>>> On Thu, 2023-07-06 at 21:00 +0200, Philippe Gerum wrote:
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>>>>>>>>
>>>>>>>>> Hi Philippe,
>>>>>>>>>
>>>>>>>>> quick feedback on latest rebasing:
>>>>>>>>>
>>>>>>>>> https://source.denx.de/Xenomai/linux-dovetail/-/commit/8f155631c8b51731dfc2ae5145d4dcc68f2ed7ed
>>>>>>>>>
>>>>>>>>> This was likely accidentally folded, right?
>>>>>>>>>
>>>>>>>>
>>>>>>>> It looks so. This conditional statement appeared upstream during the
>>>>>>>> v6.0 cycle it seems, and was unfortunately left unfixed in the dovetail
>>>>>>>> patch for this entire series.
>>>>>>>>
>>>>>>>>> Furthermore, I will factor this change out into a separate commit in 
>>>>>>>>> merging v6.1.y-dovetail, for better traceability:
>>>>>>>>>
>>>>>>>>> diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
>>>>>>>>> index ce8bdd804bc7..c843bf352999 100644
>>>>>>>>> --- a/arch/x86/mm/fault.c
>>>>>>>>> +++ b/arch/x86/mm/fault.c
>>>>>>>>> @@ -1466,7 +1466,7 @@ void do_user_addr_fault(struct pt_regs *regs,
>>>>>>>>>  
>>>>>>>>>  	/* The fault is fully completed (including releasing mmap lock) */
>>>>>>>>>  	if (fault & VM_FAULT_COMPLETED)
>>>>>>>>> -		return;
>>>>>>>>> +		goto out;
>>>>>>>>>  
>>>>>>>>>  	/*
>>>>>>>>>  	 * If we need to retry the mmap_lock has already been released,
>>>>>>>>>
>>>>>>>>> To my understanding, this was a dovetail fix that applied to previous 
>>>>>>>>> versions as well.
>>>>>>>>
>>>>>>>> Only to v6.0, however this series is not maintained dovetail-wise
>>>>>>>> anymore, so this stayed under the radar.
>>>>>>>
>>>>>>> FYI: This change fixes a real issue here. The 6.1-rebase branch made it
>>>>>>> through a 24h stress test while the "merge" branch failed. Happily
>>>>>>> waiting for the next release...
>>>>>>
>>>>>> Do we have a idea / plan when the next 6.1 dovetail release will
>>>>>> happen? My internal customer is waiting for that and I want to avoid
>>>>>> backporting to 6.1.34.
>>>>>>
>>>>>
>>>>> There is a brewing rebase to v6.1.42 pending tests here, no challenging
>>>>> source conflict. I pushed the tree untagged to the v6.1.y-rebase branch
>>>>> if you want to give it a try before Jan updates the stable branch.
>>>>>
>>>>
>>>> We are testing against -rebase anyway (but that needs a trigger which I
>>>> just provided). The merging branch is only updated once there is a
>>>> release tagged.
>>>>
>>>
>>> Passed with the Xenomai 3 tests:
>>>
>>> https://source.denx.de/Xenomai/xenomai-images/-/pipelines/17252
>>>
>>> But can we update to 6.1.44 directly? Florian mentioned that there are
>>> Downfall-related fixes in that release.
>>>
>>> Jan
>> 
>> yes, .44 is actually brewing here already.
>> 
>
> Now we are .45 even. Will this one be tagged? CI looks good so far.
>

Nope, .44 was broken due to [1]. .45 looks better with this fix in
place. This issue triggered on kvm/arm64 w/ PROVE_LOCKING enabled.

[1] https://source.denx.de/Xenomai/linux-dovetail/-/commit/60ae6e4ea6a074698be4d999e96e9c3697b18ea1

-- 
Philippe.