[PATCH v6 0/2] perf: Add Raspberry Pi AXI PMU driver

[Ian Rogers]

This patch series adds an uncore Performance Monitoring Unit (PMU) driver
for Broadcom AXI system and VideoCore VPU performance monitors found on
Raspberry Pi SoCs (BCM2835 through BCM2712 / Raspberry Pi 1 through 5).

Motivation & Background
------------------------
Currently, Linux lacks a standard perf-API compatible driver for the
Broadcom AXI performance counter blocks on Raspberry Pi platforms. Prior
out-of-tree vendor solutions relied on custom debugfs nodes and ad-hoc
kthreads, preventing integration with standard Linux perf tooling (`perf stat`,
`perf list`, etc.).

This driver implements standard `struct pmu` hardware uncore callbacks
under `drivers/perf/`, exposing human-readable sysfs event aliases, unit
scaling (`Bytes`), and bus filtering directly to user space.

Key Architectural Improvements & Features
------------------------------------------
1. Standard Linux Perf Integration:
   - Exposes uncore AXI interconnect events via `/sys/bus/event_source/devices/rpi_axi_pmu/`.
   - Supports event sampling and hardware counter accumulation (`local64_add`),
     automatically managing 31-bit hardware counter wraparound across high-bandwidth
     interconnect transfers.

2. CPU Hotplug Support (`cpuhp`):
   - Registers dynamic CPU hotplug notifiers (`CPUHP_AP_ONLINE_DYN`).
   - Automatically migrates PMU context (`perf_pmu_migrate_context`) to an online
     CPU core when a designated CPU goes offline, avoiding stale uncore state.

3. Hybrid Memory-Mapped & Mailbox Work Queue Architecture:
   - System Monitor (MMIO): Performs fast atomic-safe memory reads (~15ns)
     directly mapped over ARM physical memory space (`MON__SYSTEM`).
   - VPU Monitor (Mailbox IPC): For Broadcom BCM2835-BCM2711 platforms (RPi 1-4),
     VideoCore VPU monitor IPC calls are offloaded to process context via a dedicated
     workqueue (`vpu_work`) and serialized under `vpu_mutex`. This avoids atomic
     sleeps or blocking in timer/interrupt context.

4. PREEMPT_RT & Safety Hardening:
   - Uses HRTIMER_MODE_REL_SOFT for timer callbacks to execute in softirq context,
     ensuring spinlock acquisitions are 100% PREEMPT_RT safe.
   - Sets suppress_bind_attrs = true to prevent unsafe manual sysfs unbinding while
     active perf events exist.

5. SoC Generation Support:
   - Patch 1 adds core driver support for Broadcom BCM2835-BCM2711 (RPi 1-4).
   - Patch 2 expands support for Broadcom BCM2712 (Raspberry Pi 5), adding PCIe RP1
     Southbridge links, HEVC decoder, HVS display engine, and Cortex-A76 DSU L3
     interconnect monitoring.

Hardware Validation
-------------------
The driver has been rigorously validated on physical hardware across multiple 
SoC generations:
- Raspberry Pi 400 (BCM2711): Validated System MMIO and VideoCore VPU firmware
  mailbox IPC performance counters. Explicitly mechanically verified precisely zero 
  lockup behaviors across 100Hz aggressive asynchronous timer interrupts. Profiling 
  the `h264_v4l2m2m` pipeline hardware isolated >11.1MB in pure AXI VPU hardware matrix reads.
- Raspberry Pi 5 (BCM2712): Validated live byte throughput across HVS display
  refresh cycles, Cortex-A76 DSU L3 interconnect memory traffic, and PCIe RP1
  Southbridge transfers.

Changes in v6
-------------
- Reorganized commit structure rigorously into two cleanly bisectable patches
  (BCM2835 core in Patch 1, BCM2712 additions in Patch 2).
- Resolved unmet Kconfig dependency warnings when MAILBOX=n by cleanly relying
  on 'depends on RASPBERRYPI_FIRMWARE || (COMPILE_TEST && !RASPBERRYPI_FIRMWARE)'
  rather than forcefully using 'select' (Patch 1).
- Resolved catastrophic PREEMPT_RT/softirq deadlocks associated with atomic
  spinlocks globally preempting 'hrtimer_cancel()' routines by permitting the
  timer loop to detect 0 active events and cleanly self-terminate (Patch 1).
- Purged unsafe 'spin_unlock_irqrestore' lock dropping maneuvers occurring
  abruptly inside 'rpi_axi_pmu_read()', ensuring atomic contexts and missing
  flags no longer inadvertently trigger kernel panics (Patch 1).
- Eliminated silent ABA VPU slot data-corruption and event hardware leaks by
  decoupling 'vpu_work_handler' loops safely onto a 'vpu_disable_pending' matrix.
  Deleted VPU watchers are safely disabled exactly prior to newly allocated slots 
  executing initialization within the same asynchronous worker loop (Patch 1).
- Resolved lockless data races across VPU mailbox hardware mutation checks by fully
  protecting 'monitor_running' and 'enabled' state modifications structurally inside 
  the 'rpi_axi_pmu_start' and 'rpi_axi_pmu_vpu_work_handler' core locks.
- Disentangled rigid 'cancel_work_sync' usages back to safe 'flush_work' variants in
  teardown unloading contexts so dynamic Mailbox IPC teardown flows successfully 
  execute without aborting directly in mid-flight (Patch 2).
- Sharpened the sysfs 'is_visible' string-masking logic to properly accommodate
  Raspberry Pi 5 aliases, resolving an overzealous 'strstr' bug directly hiding
  "bcm2712_peripheral_" aliases on RPi 5 hardware (Patch 2).
- Rewrote component destruction sequencing logic completely onto safe CPU hotplug 
  '_nocalls' handlers to strictly avoid User-After-Free events from persisting 
  against unregistered/defunct driver models during 'perf_pmu_unregister' (Patch 1).
- Eradicated cross-execution bus tracker leaks across failed/zombied workflows where
  'monitored' mask properties remained structurally trapped inside invalid state
  loops by surgically injecting manual 'enabled = false' boolean teardowns into 'del()'.
- Inserted explicit comments explaining the critical 31-bit counter-masking logic
  regarding standard U32_MAX hardware exception signaling (Patch 1).
- Scrubbed all erroneous 'TAG=agy' trailing metadata from commit descriptions.

Ian Rogers (2):
  perf: Add Raspberry Pi BCM2835 AXI PMU driver
  perf: Add Raspberry Pi 5 (BCM2712) AXI PMU support

 drivers/perf/Kconfig       |    9 +
 drivers/perf/Makefile      |    1 +
 drivers/perf/rpi_axi_pmu.c | 2119 ++++++++++++++++++++++++++++++++++++
 3 files changed, 2129 insertions(+)
 create mode 100644 drivers/perf/rpi_axi_pmu.c

-- 
2.47.0