Re: Random system freezes with Linux 6.12 + Xenomai 3.3.1 on Intel Ultra 5 125H

刘杨 <[email protected]> Wed, 29 Jul 2026 16:04:44 +0800
Newsgroups dev.linux.lists.xenomai
Message-ID <ed7c43ef85386e7d9b7bfcd2895307dc2948cb81.2cdf3bbf.6288.45f6.b5d6.9760f144d99c@feishu.cn>
Thank you for your reply. Regarding kernel ftrace, do you have any suggesti=
ons on which parts I should mainly trace? If I enable tracing for everythin=
g, the output file will become extremely large

> From: "Jan Kiszka"<[email protected]>
> Date:=C2=A0 Wed, Jul 29, 2026, 15:54
> Subject:=C2=A0 Re: Random system freezes with Linux 6.12 + Xenomai 3.3.1 =
on Intel Ultra 5 125H
> To: "=E5=88=98=E6=9D=A8"<[email protected]>
> Cc: "xenomai"<[email protected]>, "=E5=AE=8B=E5=81=A5=E7=8E=AE"<son=
[email protected]>
> On 29.07.26 08:57, =E5=88=98=E6=9D=A8 wrote:
> > Hi,
> > =C2=A0=C2=A0 We have already conducted long-term tests on the following=
 hardware platforms using kernel 6.12.55 with Xenomai 3.3.1: N100, N97, J64=
12, i5-12400, and AMD Ryzen 7 8845HS. I am using the 6.12.y-cip-dovetail-re=
base=C2=A0branch, and the latest kernel version available in this branch is=
 6.12.90. I can use the xenomai-
>=C2=A0
> Ups, I guess I need to re-sync the v6.12.y-cip-dovetail to -rebase soon..=
.
>=C2=A0
> > image=C2=A0provided officially by Xenomai. Could you please provide a d=
ownload link for the xenomai-image=C2=A0corresponding to the 6.12 kernel ve=
rsion?
>=C2=A0
> We do not offer pre-built images, but it should be rather simple to
> generate one yourself, just following the README.
>=C2=A0
> > If we completely disable the NVIDIA driver, our test programs may not r=
un properly. I will disable the nvidia_uvm=C2=A0driver later for further ve=
rification.
> > For the 6.18 kernel, I compiled it from the 6.18.y-dovetail-rebase=C2=
=A0branch. The 6.18 kernel exhibits the same issue: the system hangs after =
running for a certain period. However, when using the 6.18 kernel, SSH rema=
ins accessible even after the hang.
> >=C2=A0
>=C2=A0
> Then you could try to let tracing run during the test and extract an
> event trace after the hang. Maybe that reveal how the core(s) lock up,
> over which exact workload.
>=C2=A0
> Jan
>=C2=A0
> --=C2=A0
> Siemens AG, Foundational Technologies
> Linux Expert Center
>=C2=A0