[PATCH v2 0/2] interconnect: qcom: add MSM8x60 NoC driver

[Herman van Hazendonk <github.com@herrie.org>]

v2 addresses Jie's, Sashiko's and Dmitry's review of v1:

Binding (PATCH 1/2):
 - Add the dt-binding schema (qcom,msm8660.yaml).  v1 only shipped
   the ID header (Jie flagged the missing schema).  The schema
   covers all four compatibles, per-fabric clock-name lists
   (bus/bus_a/ebi1/ebi1_a for AFAB; bus/bus_a/smi/smi_a for MMFAB;
   bus/bus_a for SFAB and DFAB) with descriptions for each clock
   (Dmitry), the required qcom,rpm phandle, and
   #interconnect-cells = <1>.

 - Drop the SFPB/CFPB mention from the header commit message;
   those fabrics aren't modelled by the driver (Sashiko Low).

Driver (PATCH 2/2):
 - Fix the bandwidth aggregation in msm8660_icc_set() (Sashiko
   High).  v1 walked provider->nodes and re-summed n->avg_bw
   across all nodes, which double-counts: the framework already
   writes each path's bw to every node it traverses, so summing
   here gives e.g. master_bw + slave_bw for the same flow.  v2
   takes the max per-node rate across the provider (matches the
   qcom_icc_bus_aggregate() convention in icc-rpm.c).

 - Switch to dynamic ICC node IDs via icc_node_create_dyn() and
   pointer-based linking via icc_link_nodes(), matching the
   icc-rpmh-style pattern (Dmitry).  Drop the internal MSM8660_*
   enum entirely; struct msm8660_icc_node now carries a struct
   icc_node *node and a flexible link_nodes[] array of qnode
   pointers.

 - Expand the DEFINE_QNODE macro (Dmitry); each of the 63 qnodes
   is now an explicit static struct definition, with forward
   declarations grouped at the top.

 - Use dev_err_ptr_probe() in msm8660_get_rpm()'s deferred-probe
   paths (Dmitry).

 - Limit the clock-bulk-get fallback to -ENOENT only (Dmitry);
   propagate every other error including -EPROBE_DEFER instead
   of "any non-DEFER continues without clock scaling".

 - Kconfig: depend on MFD_QCOM_RPM=y so a built-in interconnect
   provider can't link against a modular RPM (Sashiko Low).

 - struct msm8660_icc_node: const char *name (Sashiko Low).

On-device validation (HP TouchPad / APQ8060):
  All four fabric providers probe cleanly:

    qnoc-msm8660 soc:interconnect@0: RPM fabric ARB enabled
                 (4 masters, 4 slaves, 2 tiered)             [AFAB]
    qnoc-msm8660 soc:interconnect@1: RPM fabric ARB enabled
                 (17 masters, 9 slaves, 2 tiered)            [SFAB]
    qnoc-msm8660 soc:interconnect@2: RPM fabric ARB enabled
                 (14 masters, 4 slaves, 3 tiered)            [MMFAB]
    qnoc-msm8660 soc:interconnect@3: MSM8660 interconnect provider
                 registered                                  [DFAB]
    mmcc-msm8660: MMSS fabric: unhalted all master ports (0-13)
    msm_hsusb 12500000.usb: USB HS: Setting interconnect
                 bandwidth avg=61440 peak=61440 kBps

  interconnect_summary shows the expected per-node aggregation
  (e.g. slv_ebi_ch0 sums MMC + MDP + DMA + USB consumer requests)
  and msm8660_icc_set() then derives a single fabric clock rate
  from the per-node max rather than the per-node sum (the v1
  double-count is gone in v2).

The companion DTS patches that enable the providers for the
HP TouchPad will be sent separately to the ARM/DTS tree.

Herman van Hazendonk (2):
  dt-bindings: interconnect: qcom: add msm8660 NoC
  interconnect: qcom: add MSM8x60 NoC driver

 .../bindings/interconnect/qcom,msm8660.yaml   |  157 ++
 drivers/interconnect/qcom/Kconfig             |   14 +
 drivers/interconnect/qcom/Makefile            |    2 +
 drivers/interconnect/qcom/msm8660.c           | 1619 +++++++++++++++++
 .../dt-bindings/interconnect/qcom,msm8660.h   |  156 ++
 5 files changed, 1948 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/interconnect/qcom,msm8660.yaml
 create mode 100644 drivers/interconnect/qcom/msm8660.c
 create mode 100644 include/dt-bindings/interconnect/qcom,msm8660.h


base-commit: 944125b4c454b58d2fe6e35f1087a932b2050dff
-- 
2.43.0