Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Thiago Macieira <thiago.macieira@intel.com>
To:     Borislav Petkov <bp@alien8.de>, "Luck, Tony" <tony.luck@intel.com>
Cc:     "Joseph, Jithu" <jithu.joseph@intel.com>,
        "hdegoede@redhat.com" <hdegoede@redhat.com>,
        "markgross@kernel.org" <markgross@kernel.org>,
        "tglx@linutronix.de" <tglx@linutronix.de>,
        "mingo@redhat.com" <mingo@redhat.com>,
        "dave.hansen@linux.intel.com" <dave.hansen@linux.intel.com>,
        "x86@kernel.org" <x86@kernel.org>, "hpa@zytor.com" <hpa@zytor.com>,
        "gregkh@linuxfoundation.org" <gregkh@linuxfoundation.org>,
        "Raj, Ashok" <ashok.raj@intel.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "platform-driver-x86@vger.kernel.org" 
        <platform-driver-x86@vger.kernel.org>,
        "patches@lists.linux.dev" <patches@lists.linux.dev>,
        "Shankar, Ravi V" <ravi.v.shankar@intel.com>,
        "Jimenez Gonzalez, Athenas" <athenas.jimenez.gonzalez@intel.com>,
        "Mehta, Sohil" <sohil.mehta@intel.com>
Subject: Re: [PATCH v2 12/14] platform/x86/intel/ifs: Add current_batch sysfs entry
Date:   Sat, 12 Nov 2022 18:35:23 -0800
Message-ID: <2687702.9iZYToFQE1@tjmaciei-mobl5>
Organization: Intel Corporation
In-Reply-To: <B12A4934-AD7A-4F8E-A2FB-229542C1A098@intel.com>
References: <20221021203413.1220137-1-jithu.joseph@intel.com> <Y2/z0yY3zcKmR5BN@zn.tnic> <B12A4934-AD7A-4F8E-A2FB-229542C1A098@intel.com>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Saturday, 12 November 2022 15:32:47 PST Luck, Tony wrote:
> > Because if this is going to be run during downtime, as Thiago says, then
> > you can just as well use debugfs for this. And then there's no need to
> > cast any API in stone and so on.
>=20
> Did Thiago say =E2=80=9Cduring downtime=E2=80=9D? I think
> he talked about some users opportunistic
> use of scan tests. But that=E2=80=99s far from only
> during downtime. We fully expect CSPs to
> run these scans periodically on production
> machines.

Let me clarify. I did not mean full system downtime for maintenance, but I =
did=20
mean that there's a gap in consumer workload, for both threads of one or mo=
re=20
cores. As Tony said, it should have little observable effect on any other c=
ore,=20
meaning an IFS run can be scheduled *as* any other workload (albeit a=20
privileged one) for a subset of the machine, while the rest of the system=20
remains in production. This allows them a lot of flexibility and is the rea=
son=20
I am talking about containers, with the implied constraint that the=20
container's view of the filesystem is narrower than the kernel's.

There'll be some coordination required to get all cores to have run all tes=
ts,=20
but it should be doable over a period of time, and I'm thinking days, not=20
years. This should still be short enough to reveal if the system can detect=
 a=20
defect or wear-out before any real workload is impacted by it.

If an issue is detected, the admin can decide whether to offline the core(s=
)=20
reporting problems but keep the rest serving workloads and generating reven=
ue,=20
or offline the entire machine for full maintenance and to run more invasive=
 and=20
time-consuming tests.

=2D-=20
Thiago Macieira - thiago.macieira (AT) intel.com
  Cloud Software Architect - Intel DCAI Cloud Engineering