Re: [PATCH net-next v2 00/14] net: enable TC956x support
flat view
From: Alex Elder <hidden>
Date: 2026-08-11 13:46:41
Also in:
bpf, linux-arm-kernel, linux-arm-msm, linux-devicetree, linux-gpio, lkml
On 6/4/26 8:00 PM, Alex Elder wrote:
(A fairly detailed list of changes since v1 is included below. Because there are so many changes listed, they are not repeated in the individual patches that follow. Note also that this series depends on two prerequisite commits, as explained below.)
We are now ready to send version 3 of this series. It has undergone a complete restructuring, though many detailed bits of code remain the same as before. Because of this we intend to use a new strategy in posting the code for review. I describe that here; if you think there are things we could do differently to improve this process, please say so. I can provide a longer explanation for how the new model works if that is requested, but here is a summary of the process: - There will be no "version 3" of the original series. - Instead, the TC9564 will be treated as an SoC that consists of a set of distinct components (clock, reset, interrupts, GPIO, etc.). - Support for each of these components will be provided by distinct platform device drivers. - These platform drivers will be upstreamed independently. I.e., the clock driver will be sent to clock maintainers for review, the reset driver will be sent to reset maintainers, and so on. - One of these platform drivers will provide support for the XGMAC 10 Gbit Ethernet interface (whose support was the the original, primary goal of this series). - The anchor driver is a PCI endpoint/function driver, which implements the framework necessary to support the pci-ep-bus that all the other devices use for their MMIO. We will send the code for all of these parts out for review separately. Upstreaming each can proceed independently, and when all are eventually accepted (including devicetree updates), the originally-desired XGMAC functionality will be supported. Here is a list of the "components": - PCI driver (drivers/misc/tc9564-pci.c) - Clock driver (drivers/clk/clk-tc9564.c) - Reset driver (drivers/reset/reset-tc9564.c) - Interrupt controller driver (drivers/irqchip/irq-tc9564-msigen.c) - Address translation unit driver (drivers/misc/tc9564-translate.c) - XGMAC driver (drivers/net/ethernet/stmicro/stmmac/dwmac-tc9564.c) There was previously a GPIO driver included. That has been reimplemented to be accessed via I2C rather than PCIe BARs. The new version will be associated with the PCI pwrctl driver, pci-pwrctrl-tc9563.c, and will be posted for review separately. -Alex
This series introduces stmmac driver support for the Toshiba TC9564
(also known as Qualcomm QPS615). This is an Ethernet-AVB/TSN bridge IC
that provides a high-speed connection between a host SoC and Ethernet
devices on a network. It incorporates a PCIe switch, and implements
two 10 Gbps capable Ethernet MACs (along with other IP blocks), and
is essentially a small and highly-specialized SoC. The TC9564 is a
member of a family of similar chips, and the driver code uses "tc956x"
to reflect this.
TC956x chips incorporate a PCIe gen 3 switch, with one upstream and
three downstream ports. Its PCIe functionality is already supported
upstream, including a power control driver that performs some early
configuration of the PCI ports ("pci-pwrctrl-tc9563.c").
One of the PCIe switch's downstream ports has an internal PCIe endpoint,
which implements two PCIe functions, each of which has an Ethernet MAC
(eMAC) subsystem. The eMAC is composed of a Synopsis Designware XGMAC
combined with an XPCS and PMA. Each MAC is capable of operating at
10M/100M/1G/2.5G/5Gps and 10Gps. The initial target platform is the
Qualcomm RB3gen2, which supports a 10Gbps Marvell PHY on port A, and
a 2.5Gbps Qualcomm PHY on port B. (The Marvell PHY is not populated
on all RB3gen2 boards, and only the 2.5 Gbps Qualcomm PHY is supported
initially.)
TC956x chips also implement several other blocks of functionality,
including a GPIO controller, interrupt controllers (MSIGEN), I2C
and SPI, a UART, and an Arm Cortex M3 CPU with 128KB SRAM. The GPIO
interface exposes several lines to manage external resets. The
interrupt controllers are used internally by the eMAC functions. The
UART, SPI, microcontroller, and SRAM are currently unused.
----------------------------------
| Host |
------+...+----------+........+---
|i2c| | PCIe |
----------------+...+----------+........+------
| TC956x |I2C| |upstream| |
| ----- --+--------+--- |
| ----- ------ ------- | PCIe switch | |
| |SPI| |GPIO| |reset| | | |
| ----- ------ |clock| | DS3 DS2 DS1 | |
| ------- ---++--++--++-- |
| ----- ------ downstream// \\ \\ | downstream
| |MCU| |SRAM| /==========/ \\ \===== PCIe port 1
| ----- ------ //PCIe port 3 \\ |
| || \======= downstream
| ----+-----------++-----------+---- | PCIe port 2
| | M | internal PCIe endpoint | M | |
| | S |------------------------| S | ------ |
| | I | PCIe | | PCIe | I | |UART| |
| | G |function 0| |function 1| G | ------ |
| | E |----++----| |----++----| E | |
| | N | eMAC 0 | | eMAC 1 | N | |
--------+.......+------+.....+-----------------
|USXGMII| |SGMII|
--+.......+-- --+.....+--
| ARQ113C | | QEP8121 |
| PHY | | PHY |
------------- -----------
The primary objective for this series is to support the Ethernet
functionality provided by the TC956x. The code providing this
support has been structured into three distinct modules.
- A "chip" driver, associated with the PCIe functions
- A driver for the GPIO controller
- Code enabling the TC956x-specific eMAC/MSIGEN hardware
To manage the common functionality (including configuring address
translation and controlling internal reset and clock signals), a
"chip" driver is implemented. The chip driver binds to PCI
functions 0 and 1, and creates a shared data structure describing
the common chip elements shared by the two driver instances. PCIe
function 1 defers probing until after PCIe function 0 has created
this shared data structure.
Common/shared functionality is handled by PCIe function 0. This
common code uses PCIe function 0's mapped BARs to configure the
hardware. If the TC9564 GPIO controller is needed (i.e., defined
in a node in devicetree), it is created as an auxiliary device
when PCIe function 0 is probed.
Both PCIe function 0 and function 1 implement an Ethernet
controller as well, though one or the other might be unused on
a given platform. If a PCIe function node defines an ethernet
subnode in devicetree, the corresponding eMAC is enabled, and
the PCIe function driver will create an auxiliary device to
represent it.
The GPIO driver is implemented separately because in some hardware
configurations, these GPIO lines are used to manage resets for
external Ethernet PHYs. We describe these PHYs via devicetree,
where the GPIO-based reset signals are defined using phandles.
The code for eMAC and MSIGEN consists of a new source file that
populates hardware-specific details about the two MACs, and integrates
with the existing stmmac driver. This also required implementing some
enhancements to the core stmmac driver, described further below.
The version of the Synopsys MAC IP is 3.01, which is largely compatible
with version 2.20. The core stmmac driver required several changes to
enable support for the TC956x.
- A change to dwxgmac2 support changes the interrupt mode when
multi_msi_en is enabled.
- While most support for version 3.01 simply uses the 2.20 code,
an erratum related to the RX ring length is implemented for
3.01 DMA operations.
- Having the PCIe device be separate from an auxiliarly device
implementing the eMAC required allowing a distinct DMA device
to be maintained for an stmmac interface.
In addition:
- A new source file provides memory-mapped access to XPCS using
regmap. The alignment of the TC956x MDIO registers aren't
suitable for using simple MMIO.
- Two additional XPCS changes are implemented that provides
support for the XPCS as implemented in the TC956x.
NOTE:
This series is based on net-next/main, but depends on these
two commits found in linux-next/master:
68d801eabda52 gpio: regmap: Support sparsed fixed direction
806e7acf7f331 gpio: regmap: Don't set a fixed direction line
-Alex (and Daniel)
This series is available here:
https://github.com/riscstar/linux/tree/tc956x/stmmac-v2
Changes between version 1 and version 2:
NEW
- The qca808x.c PHY driver now supports enabling up to four optional
regulators
GENERAL
- The order of patches has been rearranged a bit. One consequence
is that the TOSHIBA_TC956X_PCI Kconfig symbol is now defined
before it's used.
PATCH 00/12
- MAINTAINERS entries have been added for all new files
PATCH 01/12
- devm_xpcs_regmap_register() now ensures the xpcs_destroy() is
called when the device is unbound (suggested by Sashiko).
PATCH 02/12
- Reworked how the quirky behavior was handled. Now:
- A soft reset is issued in xpcs_pre_config() *before* calling
xpcs_switch_interface_mode() calls txgbe_xpcs_switch_mode().
It does this now for an SGMII interface, not just if the
need-reset flag is set
- xpcs_switch_interface_mode() calls txgbe_xpcs_switch_mode(),
preserving previous behavior
- For other (non-Wangxun) platforms, if the XPCS supports
10Gbase-R, xpcs_switch_interface_mode() programs the interface
operating mode to use a reserved value rather than the default
PATCH 03/12
- This change has been removed; the default_an_inband platform
data flag is set instead (in what was previously patch 10)
PATCH 04/12
- Two issues were pointed out by Sashiko, both of which were
problems related to DMA handling. They were pre-existing
issues, and I plan to send separate patches to address
them.
- A mask local variable (to shorten a long line) is no longer
created. Checkpatch no longer complains about that line.
PATCH 05/12
- The description was clarified a bit in response to a comment
from Andrew
PATCH 06/12
- No comments received, no changes
PATCH 07/12
- Sashiko suggested that exporting some symbols from the stmmac
core might not be desirable, and suggested using the standard
DMA API to accomplish the same thing. For now we are igoring
this suggestion--not finding a way to do what was suggested.
PATCH 08/12
- The description has been revised to focus on the hardware and
not the "node".
- The ASCII art (copied from below) has been incorporated into the
binding
- The file name has been changed to "toshiba,tc9564-dwmac.yaml" and
the $id value matches the file name
- Only one example is provided now
- The references and unevaluatedProperties have been moved later
in the file
- We no longer reference "pci-bus-common.yaml". We are describing
a device, not a bridge.
- Subnodes have been added to represent the GPIO controller (on
PCIe function 0) and the Ethernet interfaces (possible on PCIe
functions 0 and 1).
PATCH 09/12
- The GPIO controller is now implemented using regmap-gpio
- The Kconfig entry for the GPIO driver has been updated to select
GPIO_REGMAP, and to define its default value as "m" unconditionally
- The Kconfig entry is now defined in the memory-mapped GPIO section
- The regmap use has been adjusted so it's clear it is MMIO-based,
which normally does not require error checking.
- We no longer call drv_get_drvdata()
PATCH 10/12
- The name of the IRQ domain now includes the MAC ID (0 or 1)
- SP_SEL_SGMII_2500M is now named SP_SEL_2500BASEX
- A typo in the commit has been fixed, and the error check after creating
the RX IRQ mapping now checks the right pointer for an error
- The Kconfig entry now selects GENERIC_IRQ_CHIP
- The PHY interface is now taken into account in the fix_max_speed
callback
- We now check the correct pointer for an error when creating the RX
IRQ mappings
- We are not implementing a shutdown driver callback
- Xilin has other discussion related to WoL, which we should save
and use, but we're not acting on it now.
- In four function, Sashiko noted that a 32-bit dma_addr_t can't hold
TC956X_SLV00_SRC_ADDR, so the type of local variable has been changed
to be u64: tc956x_dma_init_rx_chan(), tc956x_dma_init_tx_chan(),
tc956x_desc_set_addr(), and tc956x_desc_set_sec_addr().
- tc956x_mac_disable() is now called if stmmac_dvr_probe() fails;
suggested by Sashiko.
- Sashiko noted the MDIO device tree node was being leaked. This
has been resolved.
- Sashiko also pointed out that IRQ mappings were not being freed,
and this too is fixed.
PATCH 11/12
- The reset and clock functions are no longer defined inline
- The REFCLK clock name has been renamed REFCLKO, and a comment
has been added to explain it is an (unused) output clock.
- If the PCI device has no DT node it is reported with dev_err().
- <linux/bitfield.h> is now included, to avoid errors in various
non-Arm64 builds.
- The devicetree Ethernet node is no longer leaked in adev_device_add()
(noticed by Sashiko).
- The Kconfig entry now depends on ARM64 or COMPILE_TEST
- The Kconfig entry now selects REGMAP and REGMAP_MMIO
- Sashiko reported that the reference count on the PCIe slot
for function 0 was being leaked. This has been fixed.
- Sashiko questioned whether the cleanup action to delete
a created device link was associated with the wrong device.
It was, and this has been fixed.
- Sashiko also suggested that the allocated PCI IRQ vector
was not properly getting removed. It now is.
- Sashiko suggested that function 1 might proceed with probing
even though there was still a chance that function 0's probe
failed. The platform_data pointer that function 1 uses to
decide it's OK to probe is now assigned later.
PATCH 12/12
- The PHY interrupt is now defined as a level interrupt.
- A typo in the PHY name has been corrected in the patch description
Here is the initial version of the series.
https://lore.kernel.org/netdev/20260501155421.3329862-1-elder@riscstar.com/ (local)
Alex Elder (3):
net: stmmac: dma: create a separate dma_device pointer
misc: tc956x_pci: add TC956x/QPS615 support
gpio: tc956x: add TC956x/QPS615 support
Daniel Thompson (11):
dt-bindings: net: qca,qca808x: Add regulator properties
net: phy: qcom: qca808x: Add regulator management
net: pcs: pcs-xpcs-regmap: support XPCS memory-mapped MDIO bus via
regmap
net: pcs: xpcs: re-order xpcs_pre_config() to update after the reset
net: pcs: pcs-xpcs: select operating mode for 10G-baseR capable PCS
net: stmmac: dwxgmac2: Add multi MSI interrupt mode
net: stmmac: dwxgmac2: Add XGMAC 3.01a support
net: stmmac: dwxgmac2: export symbols for XGMAC 3.01a DMA
dt-bindings: net: toshiba,tc9654-dwmac: add TC9564 Ethernet bridge
net: stmmac: tc956x: add TC956x/QPS615 support
arm64: dts: qcom: qcs6490-rb3gen2: enable TC9564 with a single QCA8081
phy
.../devicetree/bindings/net/qca,qca808x.yaml | 14 +
.../bindings/net/toshiba,tc9564-dwmac.yaml | 120 +++
MAINTAINERS | 12 +
arch/arm64/boot/dts/qcom/qcs6490-rb3gen2.dts | 48 +
drivers/gpio/Kconfig | 12 +
drivers/gpio/Makefile | 1 +
drivers/gpio/gpio-tc956x.c | 130 +++
drivers/misc/Kconfig | 12 +
drivers/misc/Makefile | 1 +
drivers/misc/tc956x_pci.c | 741 ++++++++++++++++
drivers/net/ethernet/stmicro/stmmac/Kconfig | 14 +
drivers/net/ethernet/stmicro/stmmac/Makefile | 2 +
.../net/ethernet/stmicro/stmmac/chain_mode.c | 12 +-
.../ethernet/stmicro/stmmac/dwmac-tc956x.c | 818 ++++++++++++++++++
.../net/ethernet/stmicro/stmmac/dwxgmac2.h | 12 +
.../ethernet/stmicro/stmmac/dwxgmac2_core.c | 1 +
.../ethernet/stmicro/stmmac/dwxgmac2_descs.c | 1 +
.../ethernet/stmicro/stmmac/dwxgmac2_dma.c | 79 +-
.../net/ethernet/stmicro/stmmac/ring_mode.c | 12 +-
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 1 +
.../net/ethernet/stmicro/stmmac/stmmac_main.c | 58 +-
.../net/ethernet/stmicro/stmmac/stmmac_xdp.c | 2 +-
drivers/net/pcs/Makefile | 4 +-
drivers/net/pcs/pcs-xpcs-regmap.c | 219 +++++
drivers/net/pcs/pcs-xpcs.c | 93 +-
drivers/net/phy/qcom/qca808x.c | 30 +
include/linux/pcs/pcs-xpcs-regmap.h | 20 +
include/linux/stmmac.h | 1 +
include/soc/toshiba/tc956x-dwmac.h | 84 ++
29 files changed, 2478 insertions(+), 76 deletions(-)
create mode 100644 Documentation/devicetree/bindings/net/toshiba,tc9564-dwmac.yaml
create mode 100644 drivers/gpio/gpio-tc956x.c
create mode 100644 drivers/misc/tc956x_pci.c
create mode 100644 drivers/net/ethernet/stmicro/stmmac/dwmac-tc956x.c
create mode 100644 drivers/net/pcs/pcs-xpcs-regmap.c
create mode 100644 include/linux/pcs/pcs-xpcs-regmap.h
create mode 100644 include/soc/toshiba/tc956x-dwmac.h
base-commit: 07c4bca9d92e51ab6b46797a86c6ea559812351a