[PATCH v19 6/8] i3c: hub: Add support for the I3C interface in the I3C hub
Alexandre Belloni
alexandre.belloni at bootlin.com
Mon Sep 21 14:15:13 PDT 2026
On 16/09/2026 12:37:26+0530, Lakshay Piplani wrote:
> Add virtual I3C bus support for the hub and provide interface to enable
> or disable downstream ports.
>
> Signed-off-by: Aman Kumar Pandey <aman.kumarpandey at nxp.com>
> Signed-off-by: Vikash Bansal <vikash.bansal at nxp.com>
> Signed-off-by: Lakshay Piplani <lakshay.piplani at nxp.com>
> Reviewed-by: Frank Li <Frank.Li at nxp.com>
>
> ---
> Changes in v19:
> - No change
>
> Changes in v18:
> - Rebased onto v7.3-rc1; no intended driver behaviour changes
>
> Changes in v17:
> - Move the lockdep-only routing-lock class assignment into a dedicated
> helper, leaving hub context initialization unconditional when
> CONFIG_LOCKDEP is disabled
> - Document that hub_dev must already be registered and remove the
> WARN_ON_ONCE checks for its descriptor and parent controller
> - Export i3c_hub_master_ops as a const object and reference it directly
> instead of returning it through an accessor
> - Document that downstream ports share the parent controller's dynamic
> address domain and that independent per-port DAA domains are not
> provided
> - Document that a broadcast CCC on a virtual bus reaches the parent bus
> and devices behind other connected ports; directed CCCs must be used
> for per-device scope
> - Document that fixed-DT address reservations remain allocated for the
> lifetime of the parent bus, are reused across driver reprobes, and
> that downstream hot-join support would also require reserving static
> addresses
>
> Changes in v16:
> - Fix the lockdep "recursive locking" false positive on the forwarding path
> (holding the virtual bus lock while taking the parent bus lock, both
> sharing one class). Introduce per-nesting-depth lock_class_key arrays for
> the bus lock and the routing mutex, computed via
> i3c_hub_controller_depth(), assigned in i3c_hub_master_bus_init() and
> i3c_hub_init()
> - Fix the parent-bus dynamic-address reservation: reserve by
> "assigned-address" regardless of the static address (the old
> static_addr != assigned_addr skip left addresses free and risked ENTDAA
> collisions). Reserve only when the slot is currently FREE, under the
> parent bus maintenance lock
> - Make the reattach info update atomic w.r.t. asynchronous IBI: update
> parent_desc.info under i3c_bus_maintenance_lock(&parent->bus) and reject
> reattach with -EBUSY (WARN_ON_ONCE) while parent_desc.ibi is live
> - Document and check the detach-time IBI lifecycle invariant:
> parent_desc.ibi must already have been cleared through i3c_hub_free_ibi()
> before the parent-facing descriptor is detached and freed. Add
> WARN_ON_ONCE() to detect a violation. The underlying generic I3C-core
> unregister/IBI-quiesce behaviour is a known I3C-core limitation
> - Document that DAA is intentionally run on the parent controller because
> downstream devices share the parent's dynamic-address space; broadcast
> RSTDAA is intentionally not forwarded (would reset the hub address)
>
> Changes in v15:
> - Replace temporary descriptor reparenting and hub-address switching with
> a permanent parent-facing descriptor for each downstream I3C device
> - Keep the logical device descriptor associated with the virtual hub
> controller while using the parent-facing descriptor for physical
> controller operations
> - Implement downstream device attach, reattach and detach using the new
> controller-only helpers
> - Forward IBI slot recycling to the physical parent controller
> - Remove the helper that temporarily changed the logical descriptor's
> master pointer
>
> Changes in v14:
> - Add hub route serialization around DAA, CCC, private transfer and IBI paths
> - Add comments explaining that downstream devices behind hub target ports
> share the parent controller address space
> - Use i3c_bus_maintenance_lock()/unlock() instead of raw parent bus lock
> operations
> - Hold the parent maintenance lock across temporary hub address reattach,
> transfer and address restore
> - Rework IBI request/free forwarding, so downstream devices are temporarily
> attached to the parent controller for IBI setup
> - Clean up the preallocated virtual IBI object and workqueue before forwarding
> the IBI request to the parent controller
> - Avoid NULL dereference in the core IBI free path when the forwarding
> callback has already cleared dev->ibi
>
> Changes in v13:
> - Reserve parent I3C bus address slots for downstream devices with matching static and
> assigned addresses by parsing target-port DT nodes
> - Keep broadcast RSTDAA blocked and add explanatory comment
> - Add TODO for empty IBI slot recycle callback
> - Add comments for intentional no-op callbacks
>
> Changes in v12:
> - Drop redundant depends on I3C from config I3C_HUB
> - Return -EOPNOTSUPP for unsupported I2C transfers instead of reporting
> success.
>
> Changes in v11:
> - Fix virtual hub address reattach handling
> - Fix parent bus locking during address reattach operations
> - Improve transfer and IBI request error handling
>
> Changes in v10:
> - Remove unnecessary ibi_lock handling in request/enable/disable/free
> IBI APIs
> - Remove redundant parent pointer from struct i3c_hub and derive upstream
> master from hub_dev
>
> Changes in v9:
> - No change
>
> Changes in v8:
> - No change
>
> Changes in v7:
> - Convert Kconfig option to tristate
> - Fix signedness issue in return value
> - Fix kernel-doc warnings
>
> Changes in v6:
> - Add support for the generic I3C interface in the I3C Hub
> ---
> ---
> MAINTAINERS | 2 +
> drivers/i3c/Kconfig | 14 +
> drivers/i3c/Makefile | 1 +
> drivers/i3c/hub.c | 832 ++++++++++++++++++++++++++++++++++++++++
> include/linux/i3c/hub.h | 92 +++++
> 5 files changed, 941 insertions(+)
> create mode 100644 drivers/i3c/hub.c
> create mode 100644 include/linux/i3c/hub.h
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 514cda5088ba..142b0f7636c1 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -19766,8 +19766,10 @@ L: linux-kernel at vger.kernel.org
> L: linux-i3c at lists.infradead.org
> S: Maintained
> F: Documentation/devicetree/bindings/i3c/nxp,p3h2840.yaml
> +F: drivers/i3c/hub.c
> F: drivers/mfd/p3h2840.c
> F: drivers/regulator/p3h2840_i3c_hub_regulator.c
> +F: include/linux/i3c/hub.h
> F: include/linux/mfd/p3h2840.h
This should be moved to 7/8 so it can be applied independently of the
p3h2840 driver itself. I'm ready to take 1,2 and 6 but to be able to
apply the rest, I need an ack from Lee on 4/8.
--
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
More information about the linux-i3c
mailing list