[PATCH v14 00/15] accel/rocket: RK3576 NPU (RKNN) enablement

This adds the RK3576 NPU to accel/rocket. The RK3576 has two cores of
the RKNN block the driver supports on RK3588, with more clocks, two
power domains per core, one reset instead of two, and a sixteen-bit task
number in PC_TASK_CON.

1/15 is Igor Paunovic's patch and also 05/11 of his DVFS series;
whichever lands first, the other drops it. His 06/11 edits the same
binding as 6/15.

Changes since v13:

  3/15   The comment notes that a core still up with runtime PM disabled
         is left unmasked (Sashiko). The message quotes Igor's re-runs
         on v13; his earlier summary is behind the Links. Igor, say if
         you want it back.
  7/15   Heiko's Reviewed-by.
  9/15   Message trimmed (Ulf), keeping one sentence on need_regulator.
  10/15  One reset, not an array (Philipp), and the pointer is cleared
         under pmu->mutex before it is put (Sashiko). Abel, the code
         under your Reviewed-by changed; say if the tag should go.
  12/15  Drops a claim that its test was byte exact; it was within one
         count. Its comment on PC_TASK_CON now says the count clear
         lands inside the task number and inflates it, which is what
         happens.
  13/15, 14/15
         v13's 13/14, split as Heiko asked. 14/15 gives 786 MHz at 750
         mV as 11 to 25 wrong words a pass; v12 and v13 said 13 to 20
         wrong rows, which no record supports.

The other messages, apart from 1/15, are shorter.

On Heiko's question, 9/15's delay at 0 makes the first power-on of the
NPU domain take an async SError, also with vdd_npu_s0 always-on; at 15
us, 536 cold power-ons were clean. Details are in the v13 9/14 thread:
https://lore.kernel.org/all/20260924090824.38493-1-gahing@gahingwoo.com/ (local)

Routing: Ulf offered to take 7, 9 and 10. 1 to 5, 11 and 12 go through
accel, 13 to 15 through rockchip, 6 with accel and 8 with the DT patches
that use it. 15/15 needs 9, 10 and 12 to 14 first; without 9, 10 and 13
the NPU domain comes up with no settle delay or reset pulse and the
board takes an SError at boot.

Still open:

  4/15   whether you want a fix for the asynchronous put, and which
  5/15   moves the line that ZhaoJinming's "accel/rocket: Fix
         iommu_group leak and unsafe IRQ register access" changes;
         whichever lands first, the other conflicts
  14/15  both NPU domains per core, as tested, or one
  --     rocket_core_fini() puts the IOMMU group before it cancels the
         timeout worker: fix it here or on its own
  15/15  hold it a cycle, or coordinate the three trees

Link to v13: https://lore.kernel.org/all/20260915104328.45901-1-gahing@gahingwoo.com/ (local)

Igor Paunovic (1):
  accel/rocket: request the core clocks by name

Jiaxing Hu (14):
  accel/rocket: take the completion register writes under job_lock
  accel/rocket: wait for a running IRQ handler before resetting a core
  accel/rocket: let the core suspend after a reset
  accel/rocket: factor the completion tail out of the IRQ handler
  dt-bindings: npu: rockchip: add rockchip,rk3576-rknn-core
  dt-bindings: power: rockchip: allow resets in a power domain node
  dt-bindings: iommu: rockchip: describe the RK3576 NPU MMU
  pmdomain: rockchip: add optional per-domain power-on settle delay
  pmdomain: rockchip: cycle an optional power-domain reset on power-on
  accel/rocket: select the per-core clock and reset counts from match
    data
  accel/rocket: add RK3576 NPU (RKNN) support
  arm64: dts: rockchip: add NPU core domain clocks and resets to rk3576
  arm64: dts: rockchip: add NPU (RKNN) nodes to rk3576
  arm64: dts: rockchip: enable the NPU on rk3576-rock-4d

 .../bindings/iommu/rockchip,iommu.yaml        |  28 ++++
 .../npu/rockchip,rk3588-rknn-core.yaml        |  47 +++++-
 .../power/rockchip,power-controller.yaml      |   8 +
 .../boot/dts/rockchip/rk3576-rock-4d.dts      |  22 +++
 arch/arm64/boot/dts/rockchip/rk3576.dtsi      |  86 +++++++++-
 drivers/accel/rocket/rocket_core.c            |  32 +++-
 drivers/accel/rocket/rocket_core.h            |  11 +-
 drivers/accel/rocket/rocket_device.c          |   7 +-
 drivers/accel/rocket/rocket_drv.c             |  42 ++++-
 drivers/accel/rocket/rocket_drv.h             |   2 +
 drivers/accel/rocket/rocket_job.c             | 157 +++++++++++++++---
 drivers/pmdomain/rockchip/pm-domains.c        |  89 +++++++---
 12 files changed, 467 insertions(+), 64 deletions(-)


base-commit: 1a1de54f7369cd2b5bac0f265910e60ad3a6b4c3
-- 
2.43.0