mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Reinette Chatre <reinette.chatre@intel.com>
To: Tony Luck <tony.luck@intel.com>, Ben Horgan <ben.horgan@arm.com>,
	"James Morse" <james.morse@arm.com>,
	Dave Martin <Dave.Martin@arm.com>,
	Babu Moger <babu.moger@amd.com>,
	Drew Fustini <fustini@kernel.org>,
	Fenghua Yu <fenghuay@nvidia.com>, Chen Yu <yu.c.chen@intel.com>,
	Jinjie Ruan <ruanjinjie@huawei.com>
Cc: Borislav Petkov <bp@alien8.de>,
	Thomas Gleixner <tglx@linutronix.de>,
	"Dave Hansen" <dave.hansen@linux.intel.com>,
	Peter Newman <peternewman@google.com>,
	"x86@kernel.org" <x86@kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	<reinette.chatre@intel.com>
Subject: [RFC v3] arm,x86,fs/resctrl: Generic schema description Proof of Concept
Date: Thu, 19 Nov 2026 21:04:26 -0800	[thread overview]
Message-ID: <b06e1fc2-87f1-4004-8102-c8cc8928345c@intel.com> (raw)

Hi Everybody,

I believe that discussions around resctrl's upcoming support for multiple
controls are starting to settle. This RFC aims to capture all the remaining
items in implementation as well as documentation. Please point out to me if
there is anything I missed.

RFC v3 of this PoC can be found at:
git://git.kernel.org/pub/scm/linux/kernel/git/reinette/linux.git branch resctrl/controls_rfc_v3

Please refer to the documentation patch for an overview of the planned interface
changes as well as user space contracts/expectations:

https://git.kernel.org/pub/scm/linux/kernel/git/reinette/linux.git/commit/?h=resctrl/controls_rfc_v3&id=95ff9ff7430095cb7264946c944c750315db73d9

Remaining opens:
---------------
- Ben: I do have a question about the formula used to determine the size of
  a scalar cache control. Please see patch changelog of:
  https://git.kernel.org/pub/scm/linux/kernel/git/reinette/linux.git/commit/?h=resctrl/controls_rfc_v3&id=8695aab0997d5aa60e104374fae9701f51156c18
- MPAM only compile tested.
- The same x86 "SAMPLE" code as RFC v2 to demonstrate emulated controls.
- Add relationships between controls. For example, cannot set "MIN" control
  value to be larger than "MAX. Unclear how to do this in resctrl fs when
  considering that the relationship may span between emulated and backing
  controls. This is potentially something that needs to be managed by architecture
  and thus potentially without a consistent failure to user space.

Changes since RFC v2:
--------------------
RFC v2: https://lore.kernel.org/lkml/d258a32f-12d5-464d-abe9-4720fb3e44b3@intel.com/
        git://git.kernel.org/pub/scm/linux/kernel/git/reinette/linux.git branch resctrl/controls_rfc_v2

User visible changes:
- Make scope a property of the control again, now in addition to a separate
  allocation scope of the resource. Expose the control scope via a new
  per-control "scope" file (L2, L3, or NODE) that tells user space what a
  domain ID in the schemata file represents. Needed to support AMD's GLBE
  and CPU-less memory bandwidth allocation. (Ben and Babu)
- Name controls according to the scheme agreed during RFC v2 discussions:
  <resource>_<resource_scope><control_scope>_<control>, with the control
  scope dropped when it matches the resource scope. MB and SMBA carry their
  L3 scope only implicitly, so their new controls show it explicitly, for
  example MB_L3_MAX, or MB_L3NODE for a default control at NODE scope.
  https://lore.kernel.org/all/bccb1249-e722-451a-9c75-8aa846713045@intel.com/
- Report "max" per domain in preparation for systems with resource
  instances of different capacity. A bitmap's "max" no longer has a "0x"
  prefix. (Drew and Tony)
- Drop the "min_bits" file. The "min" file of a bitmap control now reports
  the minimum number of consecutive bits.
- Only offer, and only allow switching to, "native" control mode when the
  resource has emulated controls. (Ben)
- Only allow cache pseudo-locking via the default cache control.
- Display the "size" of scalar cache controls. (Ben)
- Stop reporting scalar control parsing errors as being about "MB". (Ben)
- Document the control naming, the "scope" file, that control values may be
  reset on a control mode switch, that legacy info files return no content
  in "native" mode, and that non-legacy resources need not have a default
  control.

Other changes:
- Require a control and the controls it is emulated by to have the same
  scope.
- Require only legacy resources to start in legacy control mode.
- Generate control schema suffix once during initialization.
- Fix initialization of the list of emulating controls (Ben), clearing of
  staged configurations of backing controls (Chenyu), and Haswell control
  initialization (Chenyu).
- Reword changelogs about CDP, allocation size and the bitmap control so
  they no longer suggest resctrl will only ever support one cache control.
- Fix checkpatch.pl issues, subject prefixes and changelog typos.

Series structure:
- Rebase on:
  - x86/cache branch of tip that is based on v7.4-rc1 with the queued
    resctrl cleanups
  - MPAM domain separation that was extracted from RFC v2 and submitted
    separately:
    https://lore.kernel.org/lkml/cover.1794283825.git.reinette.chatre@intel.com/
- Squash four patches into the patches they support:
  - "fs/resctrl: Provide control to cache allocation"
  - "x86,fs/rectrl: Provide control when requesting arch to update a single
    domain", which also keeps MPAM building at every patch in the series.
  - "fs/resctrl: Expose bitmap properties via new files in schemata folder"
  - "fs/resctrl: Include emulated controls when determining control field
    width"
- New patch from Ben: "fs/resctrl: Display size for scalar cache controls".
- Move IO allocation control selection, and the helpers that display the
  new control properties, earlier in the series.

Any feedback is appreciated.

Reinette

                 reply	other threads:[~2026-11-20  5:04 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=b06e1fc2-87f1-4004-8102-c8cc8928345c@intel.com \
    --to=reinette.chatre@intel.com \
    --cc=Dave.Martin@arm.com \
    --cc=babu.moger@amd.com \
    --cc=ben.horgan@arm.com \
    --cc=bp@alien8.de \
    --cc=dave.hansen@linux.intel.com \
    --cc=fenghuay@nvidia.com \
    --cc=fustini@kernel.org \
    --cc=james.morse@arm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=peternewman@google.com \
    --cc=ruanjinjie@huawei.com \
    --cc=tglx@linutronix.de \
    --cc=tony.luck@intel.com \
    --cc=x86@kernel.org \
    --cc=yu.c.chen@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®