[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20250716035500.3101551-1-txpeng@tencent.com>
Date: Wed, 16 Jul 2025 11:55:00 +0800
From: Tianxiang Peng <luminosity1999@...il.com>
To: Borislav Petkov <bp@...en8.de>
Cc: Tianxiang Peng <txpeng@...cent.com>,
puwen@...on.cn,
tglx@...utronix.de,
mingo@...hat.com,
hpa@...or.com,
x86@...nel.org,
linux-kernel@...r.kernel.org,
Hui Li <caelli@...cent.com>,
Hao Peng <flyingpeng@...cent.com>
Subject: Re: [PATCH RESEND] x86/cpu/hygon: fix missing resctrl_cpu_detect() in bsp_init helper
Thank you for your feedback, I'm sorry for the delayed response.
On Mon, 23 Jun 2025 15:31:17 +0200, Borislav Petkov wrote:
> > Since upstream commit 923f3a2b48bdccb6a1d1f0dd48de03de7ad936d9
> > ("x86/resctrl: Query LLC monitoring properties once during boot"),
>
> This commit is in 5.7.
>
> You notice the issue only now because you were on an old kernel before and
> moved to 5-something, which is >= 5.7?
The issue surfaced when we migrated our 5.15-based kernel to new Hygon machines
supporting X86_FEATURE_CQM*. Our older Hygon hardware lacks this capability
(manually verified).
To confirm the behavior:
- On new Hygon machines, both 5.15 and 6.15 kernels trigger the div-zero fault.
- Applying the patch resolves the issue consistently.
On Mon, 23 Jun 2025 15:31:17 +0200, Borislav Petkov wrote:
> > resctrl_cpu_detect() has been moved from common cpu initialization
> > code to vendor-specific bsp init helper, while hygon didn't put
> > that call in their code.
> >
> > This triggers div-zero fault during early booting stage on our
> > machines with X86_FEATURE_CQM* supported, where
> > get_rdt_mon_resources() tries to calculate mon_l3_config with
> > uninitialized boot_cpu_data.x86_cache_occ_scale.
> >
> > Fix the missing resctrl_cpu_detect() in hygon bsp init helper.
> >
> > Signed-off-by: Tianxiang Peng <txpeng@...cent.com>
>
> Your From: is Tianxiang Peng <luminosity1999@...il.com> and that will appear
> in git as author.
>
> Do you want it to be the email address you have in the Signed-off-by instead?
> I.e., your company.
Apologies for the gmail sender address, corporate network constraints require this
workaround. For the git author, I'd appreciate using my company email(txpeng@...cent.com)
instead.
Regards,
Tianxiang Peng
Powered by blists - more mailing lists