Message-ID: <20230411185411.23360-1-Jonathan.Cameron@huawei.com>
Date:   Tue, 11 Apr 2023 19:54:06 +0100
From:   Jonathan Cameron <Jonathan.Cameron@...wei.com>
To:     Liang Kan <kan.liang@...ux.intel.com>, <linux-cxl@...r.kernel.org>,
        <peterz@...radead.org>
CC:     <mingo@...hat.com>, <acme@...nel.org>, <mark.rutland@....com>,
        <will@...nel.org>, <dan.j.williams@...el.com>,
        <linuxarm@...wei.com>, <linux-perf-users@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>,
        Davidlohr Bueso <dave@...olabs.net>,
        Dave Jiang <dave.jiang@...el.com>
Subject: [PATCH v5 0/5] perf: CXL 3.0 Performance Monitoring Unit support

Was CXL 3.0 Performance Monitoring Unit support.
Renamed to highlight the driver has moved to drivers/perf

Dan Williams has expressed that he is happy to take this through the CXL
tree once perf reviewers are happy with those parts.

Thanks to Kan and Dan for reviews.

Major changes since v4: (Smaller changes described in individual patches)
- Move driver from drivers/cxl/cpmu.c to drivers/perf/cxl_pmu.c
- Rename devices to associate with cxl/devices/memX etc. Scheme
  extends to CPMU instances on Upstream and Downstream CXL ports.
- Rename CXL core files to simply pmu.* and simplify code structure
  naming and similar to cxl_pmu* CXL_PMU* etc

Patch 1/5 is also present in the series:
[PATCH 00/32] Add parents to struct pmu -> dev
which may merge first.

Update introduction.

The CXL rev 3.0 specification introduces a CXL Performance Monitoring
Unit definition. CXL components may have any number of these blocks. The
definition is highly flexible, but that does bring complexity in the
driver. All instances are self describing, though for vendor defined
events we expect mapping from numeric values to counter names to be
performed in userspace tooling. In common with many CXL features,
this driver precedes any announced hardware (that I'm aware of anyway!).  

Supported features are:
- Devices that allow counters to be written when frozen (allows
  a single register write to start / stop all counters).
- Fixed purpose counters
- Configurable counters
- CXL specification defined events + HDM filters.

This initial series covers a useful subset of functionality and is
expected to be followed by patch series addressing:
* Discoverability beyond fine grained events. This will cover both
  vendor defined events and providing perf tool with the visibility
  to be able construct 'summed' events.  For example if a single
  counter can cover all host to device read traffic.
  Perf tool patches will use this information and appropriate schema
  to present a richer set of countable events.
* CXL PMU instances on Upstream and Downstream CXL switch ports
  and root ports.
* Free running counters. Often used for vendor defined debug type
  events and error counters.  Likely to appear on real devices.
* Devices without interrupt support for overflow.
* Devices that don't support freeze (counters need to be enabled
   individually).

Exact priority order for the above features will depend on early
hardware though (a) will definitely be top of the list as any
likely hardware will be able to take advantage of that feature.

Notes.

1) The QEMU model against which this was developed needs tidying up.
   Latest tree at https://gitlab.com/jic23/qemu branch cxl-2023-02-28
   It's backed up behind other series that I plan to upstream first.
2) V1 led to a discussion of how to handle the self describing
   and extensible nature of the CPMU counters.  That is likely to involve
   a description in the "caps" sysfs directory and perf tool code that is
   aware of the different event combinations that make sense and can
   establish which sets are available on a given device.
   That is likely to take a while to converge on, so as the driver is useful
   in the current state, I'm looking to upstream this first and deal with
   the more complex handling later.
3) There is a lot of other functionality that can be added in future
   include allowing for simpler hardware implementations that may not
   support the minimum level of features currently required by the driver
   (freeze, overflow interrupts etc).
4) Adding support for ports will require solving msi/msix vector requests
   from portdrv and how to pass those to the CXL port drivers. (or some
   other way to instantiate the CXL PMU devices.) RFC on that to follow.

CXL rev 3.0 specification available from https://www.computeexpresslink.org

Jonathan Cameron (5):
  perf: Allow a PMU to have a parent
  cxl: Add functions to get an instance of / count regblocks of a given
    type
  cxl/pci: Find and register CXL PMU devices
  perf: CXL Performance Monitoring Unit driver
  docs: perf: Minimal introduction the the CXL PMU device and driver

 Documentation/admin-guide/perf/cxl.rst   |  68 ++
 Documentation/admin-guide/perf/index.rst |   1 +
 MAINTAINERS                              |   7 +
 drivers/cxl/Kconfig                      |  13 +
 drivers/cxl/core/Makefile                |   1 +
 drivers/cxl/core/core.h                  |   1 +
 drivers/cxl/core/pmu.c                   |  69 ++
 drivers/cxl/core/port.c                  |   2 +
 drivers/cxl/core/regs.c                  |  75 +-
 drivers/cxl/cxl.h                        |  16 +
 drivers/cxl/cxlpci.h                     |   1 +
 drivers/cxl/pci.c                        |  26 +-
 drivers/cxl/pmu.h                        |  28 +
 drivers/perf/Kconfig                     |  13 +
 drivers/perf/Makefile                    |   1 +
 drivers/perf/cxl_pmu.c                   | 984 +++++++++++++++++++++++
 include/linux/perf_event.h               |   1 +
 kernel/events/core.c                     |   1 +
 18 files changed, 1301 insertions(+), 7 deletions(-)
 create mode 100644 Documentation/admin-guide/perf/cxl.rst
 create mode 100644 drivers/cxl/core/pmu.c
 create mode 100644 drivers/cxl/pmu.h
 create mode 100644 drivers/perf/cxl_pmu.c

-- 
2.37.2

Powered by blists - more mailing lists