[PATCH 0/5] perf/core: add AUX buffer kernel-consumer API

[Kunwu Chan]

ARM SPE and other tracing PMUs write records directly to perf AUX
buffers.  Unlike ordinary sampling PMUs, they do not necessarily
generate a perf_event_overflow() callback for each record.  A kernel
consumer therefore needs to own and drain an AUX buffer rather than
rely on the overflow path.

Userspace obtains AUX storage through perf_event_open() and mmap().
Kernel consumers created through perf_event_create_kernel_counter()
currently have no interface to allocate and manage AUX buffers.

This series adds the minimal perf-core interfaces needed for that use
case.  It was previously posted as RFC [1]; this version incorporates
feedback from that posting and is intended for review toward merging.
Patches 1-2 add the AUX owner interface, lifecycle helpers, and
cursor/copy operations without exposing perf's internal page layout.
Patches 3-5 exercise the new interfaces with KUnit, a userspace
regression test, and a selftest script.  The first consumer is a DAMON
ARM SPE backend, but this series is independent of DAMON and adds no
consumer-specific policy to perf.

The API follows the existing non-overwrite AUX ring-buffer protocol.
Setup and release serialise buffer lifetime transitions through
event->mmap_mutex.  Cursor reads/writes use the same ordering rules as
the userspace mmap consumer, and copy validates ranges in the absolute
cursor domain before applying the ring mask.

Testing
=======

The series builds cleanly at each step.

arm64 (Kunpeng 920, 256 CPUs, SPEv1.1), Image built from this series
and booted on the hardware:

KUnit: pass:39 fail:0 skip:0.  Userspace regression (arm_spe_0): full
pass; the producer head advances with live SPE traffic.  Selftest
script: 92 passed, 0 failed, 2 skipped (both environment skips).
perf record with arm_spe_0 captures data; dmesg stays clean.

x86_64 (vmlinux build, qemu/kvm): KUnit pass:39 fail:0 skip:0, also
under KASAN + lockdep (PROVE_LOCKING) and 15 repeated boots with no
splat.  Userspace regression (Intel PT): full pass.  The selftest
script also runs from installed selftest locations.

Signed-off-by: Kunwu Chan <kunwu.chan@xxxxxxxxx>

Changes since RFC [1]:

- Serialize AUX teardown under event->mmap_mutex and fix lifetime
  handling around ring_buffer_attach().
- Improve KUnit cleanup, concurrent-release coverage, and buffer-full
  assertions.
- Extend userspace and selftest coverage, including environment
  handling fixes and installed selftest support.
- Clarify AUX lifecycle, ordering, and cursor validation
  documentation.

[1] https://lore.kernel.org/linux-perf-users/20260814144927.489172-1-kunwu.chan@xxxxxxxxx/

Kunwu Chan (3):
  perf/core: add AUX buffer ownership for kernel events
  perf/core: add AUX ring accessors for kernel consumers
  selftests/perf_events: add AUX kernel API selftest script

Lian Wang (2):
  perf/core: add KUnit tests for AUX kernel-consumer API
  selftests/perf_events: add userspace AUX regression test

 .../userspace-api/perf_ring_buffer.rst        |   62 +-
 include/linux/perf_event.h                    |   18 +
 kernel/events/Makefile                        |    1 +
 kernel/events/aux_kernel_test.c               | 1152 +++++++++++++++++
 kernel/events/core.c                          |  164 +++
 kernel/events/internal.h                      |    6 +
 kernel/events/ring_buffer.c                   |  211 ++-
 lib/Kconfig.debug                             |   14 +
 tools/testing/selftests/perf_events/Makefile  |    2 +
 .../selftests/perf_events/aux_kernel.sh       |  631 +++++++++
 .../perf_events/aux_kernel_usermode.c         |  627 +++++++++
 tools/testing/selftests/perf_events/config    |    2 +
 12 files changed, 2885 insertions(+), 5 deletions(-)
 create mode 100644 kernel/events/aux_kernel_test.c
 create mode 100755 tools/testing/selftests/perf_events/aux_kernel.sh
 create mode 100644 tools/testing/selftests/perf_events/aux_kernel_usermode.c


base-commit: 6350de8671b94afb7691d110f63bcda42f658a69
-- 
2.43.0