mt76: update to Git HEAD (2025-10-03)

9ee788a38251 wifi: mt76: mt7996: fix null pointer deref in mt7996_conf_tx()
ec3f05480e5b tools: CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

libnl-tiny: update to Git HEAD (2025-10-03)

feca1d341d4b CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

uclient: update to Git HEAD (2025-10-03)

dc909ca71bc8 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

udebug: update to Git HEAD (2025-09-28)

d3be5474f6e6 udebug-cli: ignore zero-length messages in logstream
c79f02d899df ucode: fix skipping lines where the timestamp cannot be parsed
5327524e7153 cmake: bump minimum required version to 3.13

Signed-off-by: Christian Marangi <[email protected]>

ustream-ssl: update to Git HEAD (2025-10-03)

5a81c108d20e CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

odhcp6c: update to Git HEAD (2025-10-03)

96d9e0b6e813 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

uhttpd: update to Git HEAD (2025-10-03)

ebb92e6b339b CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

unetd: update to Git HEAD (2025-10-03)

2f67f6faa08a CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

fstools: update to Git HEAD (2025-10-03)

12858e2878ef CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

fwtool: update to Git HEAD (2025-10-03)

04cd252e4e93 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

procd: update to Git HEAD (2025-10-03)

ebfe7d0f2844 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

rpcd: update to Git HEAD (2025-10-03)

cfb93f105ca2 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

ubox: update to Git HEAD (2025-10-03)

c75525a58a7e CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

ucert: update to Git HEAD (2025-10-03)

57270b247c91 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

urngd: update to Git HEAD (2025-10-03)

f17e33d99427 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

usign: update to Git HEAD (2025-10-03)

c4c72b1b0794 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

ugps: update to Git HEAD (2025-10-03)

9a351d411113 CMakeLists: update cmake minimum required version to 3.10

Signed-off-by: Christian Marangi <[email protected]>

realtek: rtl93xx: Ignore STP for per port TX

If transmissions are done outside of the DSA switch (directly from the CPU
port), the STP state must not block the transmission. Otherwise, STP frames
are not correctly submitted and the STP frames cannot correctly detect
loops before switching a port in the forwarding state.

The same applies for the LLDP frames. These must be submitted independent
of the STP state to identify neighbors or configure POE limits.

It is not necessary to filter specific destination mac addresses because
the transmission was done outside the bridge/switch in the first place. The
transmission is therefore forced.

Signed-off-by: Issam Hamdi <[email protected]>
Co-developed-by: Sven Eckelmann <[email protected]>
Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20184
Signed-off-by: Robert Marko <[email protected]>

realtek: rtl93xx: dsa: Add support for port based mirroring

The RTL930X and RTL931X SoCs support port-based, flow-based, and
RSPAN-based mirroring. Like for other SoCs from the realtek target, only
the port based port mirroring can be exposed using Linux's tc subsystem.

The port_mirror_add() implementation was updated with the following
considerations for RTL93xx SoCs:

* mirrored packets must pass through the TX pipeline of the mirroring
  port, so they are subject to configuration such as VLAN tagging,
  remarking, and EVC
* when a packet hits both source ports (SPM) and destination port (DPM) of
  a mirror group, the egress port traffic will be mirrored

The port_mirror_del() function doesn't require any modifications.

Signed-off-by: Sharadanand Karanjkar <[email protected]>
Co-developed-by: Sven Eckelmann <[email protected]>
Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20264
Signed-off-by: Robert Marko <[email protected]>

realtek: dsa: Keep HW specific mirror code in SoC helper

Instead of using a lot of if-else blocks in the port mirror code, provide
SoC specific function which calculates the SoC specific portions. The
generic part of the port mirroring code can then simply operate on the
calculated register addresses and values.

Suggested-by: Markus Stockhausen <[email protected]>
Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20264
Signed-off-by: Robert Marko <[email protected]>

realtek: Refresh kernel patches

I just ran: make target/{clean,refresh} V=99

Fixes: 272a392fe19e ("realtek: rtl93xx: replace pending I2C patches with upstreamed patches")
Signed-off-by: Hauke Mehrtens <[email protected]>

package: generate sha256sums by default

Right now the sha256sums are only created for the targets/ folder (i.e.
firmware images) and only the buildbot generates those sha256sums. Instead, let
the build system create the sha256sums directly.

Signed-off-by: Paul Spooren <[email protected]>

kernel: modules: fix mlxreg dependency to avoid deferred probe

The mlxreg-hotplug platform driver was failing to probe due to a missing
I²C multiplexer dependency. Previously, only kmod-i2c-mux-mlxcpld was
declared, but mlxreg also requires kmod-i2c-mux-reg to initialize its
I²C mux functionality.

Without this dependency, the driver probe was deferred:

  [   24.902331] platform mlxreg-hotplug: deferred probe pending: (reason unknown)

With kmod-i2c-mux-reg added, the probe succeeds and the expected
multiplexed I²C buses are registered:

  [   15.793884] i2c i2c-1: Added multiplexed i2c bus 2
  [   15.799391] i2c i2c-1: Added multiplexed i2c bus 3
  [   15.804830] i2c i2c-1: Added multiplexed i2c bus 4
  ...

Signed-off-by: Til Kaiser <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20274
Signed-off-by: Robert Marko <[email protected]>

rockchip: add Radxa ROCK 4D support

Hardware
--------
RockChip RK3576 ARM64 (8 cores)
2/4/8/16GB LPDDR5 RAM
1x 1000 Base-T (with optional PoE support)
2x LEDs (POWER / USER)
eMMC/UFS connector
Micro-SD Slot
HDMI OUT
PCIe FPC connector
2x USB 2.0 Port
2x USB 3.0 Port
USB Type-C PD Power

Installation
------------
Uncompress the OpenWrt sysupgrade and write it to a micro SD card or
internal eMMC using dd.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

uboot-rockchip: add Radxa ROCK 4D support

Add support for the Radxa ROCK 4D board.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

uboot-rockchip: backport RK3576 support

Backport core support for RK3576 SoCs.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

rkbin: pack rk3576 tpl blob

Currently there's no usable open source TPL implementation for
rk3576 SoCs, so pack the prebuilt firmware from the vendor.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

arm-trusted-firmware-rockchip: Update to 2.13

Add RK3576 ATF build.

Changelog: https://trustedfirmware-a.readthedocs.io/en/v2.13.0/change-log.html#id1

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

rockchip: backport dts updates for rk3576

Backport core dts updates for rk3576.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

rockchip: backport driver updates for rk3576

Backport clk/phy/rng/ufs/usb driver updates for rk3576.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

rockchip: use per-SoC defined kernel loadaddr

The FIT loadaddr on RK3308/RK3566/RK3568/RK358x is 0x02000000
instead of 0x02080000, while on RK3576 it's 0x42000000, which is
quite different from the former SoCs and incompatible with current
kernel loadaddr value.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

rockchip: add DEVICE_DTS_DIR definition

Avoid duplicating `rockchip/` prefix when specifying dts path.

Signed-off-by: Tianling Shen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20041
Signed-off-by: Hauke Mehrtens <[email protected]>

realtek: rtl93xx: replace pending I2C patches with upstreamed patches

Replace the pending I2C backport patches for RTL93XX added in
44655c97bb with the upstreamed variants. The patches have been accepted
upstream in the meantime and are included in v6.17 or v6.18.

Signed-off-by: Jonas Jelonek <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20273
Signed-off-by: Robert Marko <[email protected]>

wifi-scripts: retry setup of wireless devices that show up late

When a phy appears after setup has already been attempted, tell netifd
to retry setup for all failed wireless devices.

Signed-off-by: Felix Fietkau <[email protected]>

netifd: send event after running wireless handler

This can be used by other services to trigger reconfiguration, or detect when
PHY renaming has been performed.

Signed-off-by: Felix Fietkau <[email protected]>

procd: update to Git HEAD (2025-10-02)

ace5f272e139 service: fix use-after-free on service data update
94bfb9cf4238 service: implement set_data ubus method
cb40d082b126 service: send ubus notifications for event triggers

Signed-off-by: Felix Fietkau <[email protected]>

wifi-scripts: ucode: only set a hostapd config if it has interfaces

Do not pass invalid configs to hostapd.

Signed-off-by: Felix Fietkau <[email protected]>

netifd: always call setup for disabled radios

Ensures that renaming is handled properly. For disabled radios, setup is
performed with an empty list of interfaces.

Signed-off-by: Felix Fietkau <[email protected]>

realtek: rtl931x: Add SPI_CTRL0 as pinmux

The RTL931x has next to its SPI flash controller a SPI master interface. It
is connected to

* SPI_CS#[1,0]: AH22 , AK22 (aka: GPIO 12, 11)
* SPI_CLK:      AL23 (aka: GPIO 8)
* SPI_MISO:     AM23 (aka: GPIO 9)
* SPI_MOSI:     AL22 (aka: GPIO 10)

It is not the same as the SPI flash controller which uses pins:

* SPI_CS#[1,0]: B24, A24
* SPI_SCLK:     A23
* SPI_SDI/SIO0: B21
* SPO_SDO_SIO1: B21
* SPI_SIO2:     A22
* SPI_SIO3:     B22
* SPI_RSTN:     B23

As shown above, the SPI master controller shares its pin with GPIO 8, 9,
10, 11, 12. In some upcoming devices (like the Plasma Cloud PSX28/ESX28),
they will be used for SFP cage signaling. These pins must therefore be
switched manually to the GPIO mode.

The SPI_CTRL0 register provides all necessary configuration to enforce the
GPIO mode of the pins. And until more requirements (and a correct driver)
for the SPI master controller arise, it is therefore possible to use
pinctrl-single to configure it using the devicetree.

Previously the ethernet driver did configure the SPI master controller for
31.25 MHz. It is unknown for which kind of device this was originally made
and what was actually connected there. But this manual write to the
register conflicts potentially with the write of the pinctrl driver to the
same register. Luckily, we don't need this SPI speed configuration in the
ethernet driver. Still, to allow this device an easy migration, the
`spi0-31mhz` configuration was already prepared.

Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20263
Signed-off-by: Robert Marko <[email protected]>

mediatek: fix polarity of user button on openwrt-one

Signed-off-by: Felix Fietkau <[email protected]>

mac80211: ath: improve ath10k "failed to flush transmit queue" errors

Currently, in busy environments, ath10k logs "failed to flush transmit
queue" errors and have a spiking CPU usage, making the wireless barely
usable.
With this patch, taken from https://patchwork.kernel.org/project/linux-wireless/patch/20250806070005[email protected]/
this does not occur in normal operation and the wifi is much more stable

Supersedes #19427
Ref: https://github.com/openwrt/openwrt/issues/15929#issuecomment-2227391307
Ref: https://github.com/openwrt/openwrt/issues/14965

Signed-off-by: Florian Maurer <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19529
Signed-off-by: Hauke Mehrtens <[email protected]>

libunwind: enable it to all targets

This commit will enable building for riscv64
and maybe for other platforms such as mipsel64.

Suggested-by: Rosen Penev <[email protected]>
in https://github.com/openwrt/openwrt/pull/3871#issuecomment-778987028

Signed-off-by: Josef Schlehofer <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20193
Signed-off-by: Hauke Mehrtens <[email protected]>

libunwind: avoid using $(PKG_NAME) in PKG_SOURCE_URL

If there is used $(PKG_NAME) in PKG_SOURCE_URL,
then it can not be copy&pasted to the browser's address bar.

Let's remove $(PKG_NAME) and use hardcoded project name
in the PKG_SOURCE_URL

Signed-off-by: Josef Schlehofer <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20193
Signed-off-by: Hauke Mehrtens <[email protected]>

libunwind: update to version 1.8.3

Release notes:
https://github.com/libunwind/libunwind/releases/tag/v1.8.2
https://github.com/libunwind/libunwind/releases/tag/v1.8.3

Signed-off-by: Josef Schlehofer <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20193
Signed-off-by: Hauke Mehrtens <[email protected]>

firmware-utils: revert commit breaking x86 MBR images

 0782d24 Revert "ptgen: do not create stub partition to fill a gap if gap caused by alignment"

Fixes: #20254
Fixes: c5ece837a3 ("firmware-utils: update to git HEAD")
Signed-off-by: Daniel Golle <[email protected]>

netifd: update to Git HEAD (2025-09-30)

c3bf8fd913a4 interface: fix reload for devices that point to vlan aliases
22216cac7c94 bridge: fix reload when ports refer to aliased vlans on another bridge
ecca21ca07dd system: add logging wrappers for basic system functions

Signed-off-by: Felix Fietkau <[email protected]>

wifi-scripts: ucode: add support for setting Multi-AP profile/vlanid

Required for Profile-2 support.

Signed-off-by: Felix Fietkau <[email protected]>

wpa_supplicant: report interface events and support querying interface status

Makes information like Multi-AP status (including VLAN ID) accessible

Signed-off-by: Felix Fietkau <[email protected]>

wpa_supplicant: include multi-ap status in interface state

This can be used to read the Multi-AP profile and default VLAN ID

Signed-off-by: Felix Fietkau <[email protected]>

wpa_supplicant: do not wait for notify completion

There is no callback, so no use tracking the request status.

Signed-off-by: Felix Fietkau <[email protected]>

realtek: rtl930x: Fix Plasma Cloud PSX8/PSX10 copper phy-mode

The RTL8224 used by Plasma Cloud PSX8/PSX10 is not using USXGMII but
USXGMII 10G-QXGMII mode. The correct phy-mode string for this is
"10g-qxgmii".

Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20239
Signed-off-by: Robert Marko <[email protected]>

realtek: Work around missing 10g-qxgmii PHY mode

The current SerDes implementation for RTL931x handles 10G-QXGMII via the
"usxgmii" PHY mode. This is not 100% correct because it is not a single
port with 10G (max) but 4 ports with 2.5G each.

To allow setting of the "10g-qxgmii" phy mode, just change the code for now
to use the same codepaths as USXGMII. This has to be cleaned up further
during the SerDes driver rewrites.

Suggested-by: Markus Stockhausen <[email protected]>
Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20239
Signed-off-by: Robert Marko <[email protected]>

realtek: rtl931x: Enable REALTEK_PHY for RTL8224 support

The Plasma Cloud PSX28 and ESX28 are using RTL8224 as ethernet PHY. This
phy works perfectly fine on PSX8/PSX10 (RTL930x) but failed to establish a link
on rtl931x because the upstream realtek phy driver was not enabled.

Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20239
Signed-off-by: Robert Marko <[email protected]>

hostapd: skip add call on already present MLD interfaces

Avoid tearing down running interfaces

Signed-off-by: Felix Fietkau <[email protected]>

uboot-mediatek: rax3000m: use ddr3-1866mhz built in the dependency

The u-boot of cmcc_rax3000m_*-ddr3 declares a dependency on
trusted-firmware-a-mt7981-*-ddr3-1866mhz, while actually packages
the normal BL2. This causes build failure if the the non-1866m
variant isn't built.
According to the previous commit, it's intended to use the lower
frequency variant to fix device stability issues. Correct
BL2_DDRTYPE as intended.

Fixes: 028050da69 ("mediatek: CMCC RAX3000Me: fix stability issues")
Signed-off-by: Chuanhong Guo <[email protected]>

include: make APK packing mtime reproducible

APK kindly stores the mtime of each containing file in created packages,
breaking reproducibility. As a fix, touch all files of the package with the
timestamp of PKGSOURCE_DATE_EPOCH, which contains the timestamp based on the
last package modification.

Over at OPKG, something similar is done by setting mtime in the tar command,
see the `ipkg-build` script.

To tackle this in APK directly, some changes are suggested. However until this
is merged, we should fix it downstream.
https://gitlab.alpinelinux.org/alpine/apk-tools/-/merge_requests/348

Signed-off-by: Paul Spooren <[email protected]>

rpcd: update to git HEAD

 d4fa147 rpc-sys: packagelist: handle ABI versions in apk world properly

Signed-off-by: Daniel Golle <[email protected]>

ath79: replace various mtd_get_mac_text

Deprecated. Replaced with nvmem.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20176
Signed-off-by: Robert Marko <[email protected]>

ath79: dir-8x5: use nvmem

Userspace handling is deprecated.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20176
Signed-off-by: Robert Marko <[email protected]>

ath79: dir-505: use nvmem

Userspace handling is deprecated.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20176
Signed-off-by: Robert Marko <[email protected]>

ath79: tew-673gru: use nvmem

Userspace handling is deprecated.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20176
Signed-off-by: Robert Marko <[email protected]>

realtek: backport ECC driver

Upstream will get support for the Realtek ECC engine with 6.18.
To make use of this in Openwrt

- backport upstream patches
- change config so that ECC will be built for nand subtargets
- define ECC engine in RTL93xx DTS.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/19746
Signed-off-by: Robert Marko <[email protected]>

realtek: rtl931x: set hash_msb based on VLAN ID when adding a new L2 entry

During testing, we discovered that when adding a new offload FDB rule
on certain VLANs and then delete it, does not work as expected.

Steps to Reproduce:

* Create VLAN 4094 on the port lan1:

      bridge vlan add vid 4094 dev lan1 pvid

* Add a new FDB entry on port lan1 for VLAN 4094:

      bridge fdb add 00:01:02:22:33:44 dev lan1 vlan 4094 master permanent

* Delete the new FDB entry on port lan1 for VLAN4094

      bridge fdb del 00:01:02:22:33:44 dev lan1 vlan 4094 master permanent

Root Cause:

The failure occurs because the hash_msb flag is not set correctly
based on the VLAN ID when adding a new L2 entry.

Signed-off-by: Issam Hamdi <[email protected]>
Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20183
Signed-off-by: Robert Marko <[email protected]>

realtek: Switch booleans in rtl838x_l2_entry to single bits

In upstream kernel, it is not well received to use a lot of simple booleans
in structs. It is preferred to use 1-bit bitfields [1] and consolidate the
booleans together.

[1] https://www.kernel.org/doc/html/v6.16/process/coding-style.html#using-bool

Signed-off-by: Sven Eckelmann <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20183
Signed-off-by: Robert Marko <[email protected]>

realtek: add phy-handle for Zyxel GS1900-10HP SFP slots

Align GS1900-10HP dts with other realtek devices to reduce the risk of device
specific regressions with the upcoming driver cleanup/rewrite.

Signed-off-by: Bjørn Mork <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20228
Signed-off-by: Robert Marko <[email protected]>

realtek: fix Zyxel GS1900-10HP SFP slots

Parse the pcs-handle property regardless of phy-handle

Signed-off-by: Bjørn Mork <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20228
Signed-off-by: Robert Marko <[email protected]>

ath79: mikrotik: check RouterBOOT version matching sysupgrade image

Check if the uploaded image matches the version of RouterBOOT before
proceeding with sysupgrade on MikroTik devices with NOR flash.

Signed-off-by: Daniel Golle <[email protected]>

ath79: mikrotik: generate a RouterBOOT v7 NOR compatible sysupgrade

RouterBOOT v7 for NOR devices does not support the historic yaffs
"kernel" ELF boot method.
Generate a compatible kernel

Signed-off-by: John Thomson <[email protected]>

ramips: mikrotik: generate a RouterBOOT v7 NOR compatible sysupgrade

RouterBOOT v7 for NOR devices does not support the historic yaffs
"kernel" ELF boot method.
Generate a compatible kernel

Signed-off-by: John Thomson <[email protected]>

ipq40xx: mikrotik: generate a RouterBOOT v7 NOR compatible sysupgrade

RouterBOOT v7 for NOR devices does not support the historic yaffs
"kernel" ELF boot method.
Generate a compatible kernel

Signed-off-by: John Thomson <[email protected]>

image: adapt mikrotik yafut to allow switch to bootimage YAFFS path

Adjust the YAFFS file path written depending on the sysupgrade
filename.
Default to kernel (for ELF), switch to bootimage (for NPK) if image name
has v7.

Signed-off-by: John Thomson <[email protected]>

build: add Mikrotik NOR RouterBOOT v7 image build

Add build step 'kernel-pack-npk' which uses 'npk_pack_kernel' which is now
part of firmware-utils to enable wrapping the kernel inside a MikroTik NPK
package.

Signed-off-by: John Thomson <[email protected]>
Signed-off-by: Daniel Golle <[email protected]>

firmware-utils: update to git HEAD

 7e6f69b npk_pack_kernel: add tool for creating MikroTik NPK kernel packages
 d3f8b6e zynsig: add new tool for creating images for the ZyXEL GS1920 series
 9fa340d ptgen: create separate images for gpt data structure
 78d8084 ptgen: allow image generation for a specified disk size
 3f2d148 ptgen: allow non-default placement of gpt entry table
 f1f98b0 ptgen: do not create stub partition to fill a gap if gap caused by alignment
 6ea8b6d ptgen: fix protective MBR partition size
 febfef7 ptgen: use long long instead of long for sizes
 996dc48 ptgen: fix misprint and simplify calculation a bit
 0725c3d build: require CMake >= 3.5 due to dropped legacy support

Signed-off-by: Daniel Golle <[email protected]>

ucode: update to Git HEAD (2025-09-29)

f642f65c3d00 rtnl: optimize string attribute encoding
1946c571eb25 nl80211: rework string attribute encoding
6b16cd37373a nl80211: properly handle decoding numeric struct members
5a342b86c010 nl80211: support sending requests on listener socket
cd2850eec8ca lib: fix `rindex()` return value when needle at the start of the haystack
23b21eb2e8cb ubus: add connection functions to global scope
a5b206f60f5d struct: align `X` and `Z` length & truncation semantics with `*`
387880348c89 nl80211: read all pending event messages
470bc3ebffe0 uloop: fix potential use-after-free when destroying uloop entities

Signed-off-by: Felix Fietkau <[email protected]>

kernel: netdevices: fix typo for Aeonsemi AS21xxx description

Fix a typo on defining the PHY package kmod description that reference
phy-aeonsemi-as21x1x package instead of phy-aeonsemi-as21xxx.

Fixes: 08a616b2163e ("generic: backport support for Aeonsemi AS21xxx PHY")
Signed-off-by: Christian Marangi <[email protected]>

hostapd: add new ubus method for performig channel switch

This is a replacement for the bss based function, which cannot be used for MLO.

Signed-off-by: Felix Fietkau <[email protected]>

scripts/patch-kernel.sh: do not try to check files after patch

Since we are not using patch -b, *.orig files are only created when
there are conflicts, or never according to posix patch.

As such, it doesn't really make sense to always delete *.orig files
presuming they are patch backups, even if they are patch backups.
Doing so is both deleting potentially useful information for failed
patch applications and creating hard to diagnose bugs [1].

In a similar vein, checking for *.rej files does not add any value
since we're already checking the patch command's return code.

[1]: https://github.com/openwrt/packages/issues/27485

Signed-off-by: George Tsiamasiotis <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20141
Signed-off-by: Robert Marko <[email protected]>

hostapd: fix condition for re-creating MLD interfaces

Check the correct interface in order to avoid tearing down an existing
itnerface.

Signed-off-by: Felix Fietkau <[email protected]>

wpa_supplicant: MLO bringup order

Do not allow hostapd phy state update to bring up links as long between mld_set
and mld_start calls. Configuration on other PHYs could still be pending.

Signed-off-by: Felix Fietkau <[email protected]>

wifi-scripts: ucode: add start_disabled for AP mode

Enable it, whenever the AP needs to wait for a supplicant instance.

Signed-off-by: Felix Fietkau <[email protected]>

hostapd: fix an AP+STA corner case on MLO APs

Bring up AP interfaces, even if no frequency update was provided.
Fixes bringup when a MLO STA on the same radios connects to fewer links
than available, or to a non-MLD AP.

Signed-off-by: Felix Fietkau <[email protected]>

airoha: fix scuclk node for an7581

Add missing syscon compatible for scuclk node.
Fix the unit-address of the scuclk node.
This fixes the pcs driver error:

airoha-pcs 1fa08000.pcs: probe with driver airoha-pcs failed with error -22
airoha-pcs 1fa09000.pcs: probe with driver airoha-pcs failed with error -22

Fixes: c3d70b1 ("airoha: en7581: Add support for external PHY")
Signed-off-by: Aleksander Jan Bajkowski <[email protected]>
Signed-off-by: Chukun Pan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20190
Signed-off-by: Christian Marangi <[email protected]>

airoha: an7581: refresh kernel config

Enable MEDIATEK_GE_SOC_PHY driver and disable NET_DSA_MT7530_MDIO.
Refresh kernel configs with 'make kernel_oldconfig CONFIG_TARGET=subtarget'.

Signed-off-by: Chukun Pan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20190
Signed-off-by: Christian Marangi <[email protected]>

airoha: fix pinctrl driver function bug

The pinctrl driver for airoha was expecting a function name that was
not a string, but was passed one. Removing #string fixed this issue.

Fixes: c5b12fc ("airoha: Introduce support for Airoha AN7583 SoC")
Signed-off-by: Chukun Pan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20190
Signed-off-by: Christian Marangi <[email protected]>

airoha: remove duplicate AN7581 PHY patch

This patch already exists in generic/backport-6.12.

Fixes: 122135b ("airoha: an7581: add support for kernel 6.12")
Signed-off-by: Chukun Pan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20190
Signed-off-by: Christian Marangi <[email protected]>

airoha: set default DEVICE_DTS/DEVICE_DTS_DIR

All devices under airoha use dts directory under the target,
so update the default DEVICE_DTS_DIR. Also set the default
DEVICE_DTS based on the SoC name for non-dev boards.

Signed-off-by: Chukun Pan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20190
Signed-off-by: Christian Marangi <[email protected]>

kernel: rtl8261n: add support for Serdes TX swap

Add support for swapping the Serdes TX line on RTL8261N PHYs.
This is used on an Arcadyan Mozart board where the Serdes TX is swapped
on the PHY (instead of on the Soc) to permit support of SFP module by
using toggling the integrated MUX.

Link: https://github.com/openwrt/openwrt/pull/20227
Signed-off-by: Christian Marangi <[email protected]>

ramips: mt76x8: fix 02_network typo

Remove unnecessary tab which breaks 02_network script with syntax error.

Fixes: f29bc8736aac ("ramips: CREALITY BOX WB01")
Signed-off-by: Simonas Tamošaitis <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20195
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 6.6 to 6.6.108

Changelog: https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.6.108

Signed-off-by: Goetz Goerisch <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/20214
Signed-off-by: Hauke Mehrtens <[email protected]>

iwinfo: drop @!WIFI_SCRIPTS_UCODE dependency

Make the package empty if ucode wifi scripts are being used (since iwinfo is
provided by wifi-scripts in that case). This fixes packages that explicitly
select iwinfo

Signed-off-by: Felix Fietkau <[email protected]>

wifi-scripts: force wmm_enabled to 1 in AP mode, remove option

There are no supported drivers where it even makes sense to disable WMM
anymore, since so much depends on it.

Signed-off-by: Felix Fietkau <[email protected]>

odhcpd: update to Git HEAD (2025-09-27)

bc9f9d93d4d6 dhcpv4: allow lease takeover (bugfix)
b1547922257e odhcpd: bump minimum cmake version
86c0200f8067 odhcpd: apply RFC9096 recommended lifetimes

Signed-off-by: Ted Hess <[email protected]>

ath79: wd,mynet-nxxx: use nvmem

Userspace handling is deprecated.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/16285
Signed-off-by: Hauke Mehrtens <[email protected]>

expat: Update to 2.7.3

Fixes:
    CVE-2025-59375
    Regression from: CVE-2024-8176

Changelog: https://github.com/libexpat/libexpat/blob/R_2_7_3/expat/Changes

Signed-off-by: Ted Hess <[email protected]>

build: remove iwinfo dependency

Do not always try to include iwinfo in the images when wpa supplicant or
Broadcom nas is also included. iwinfo is incompatible with current
default configuration.

iwinfo is only build when CONFIG_WIFI_SCRIPTS_UCODE is not set. If
CONFIG_WIFI_SCRIPTS_UCODE is not set kmod-cfg80211 depends on iwinfo,
so it should be included in all images with wifi drivers.

The CONFIG_WIFI_SCRIPTS_UCODE option was recently changed to be active
by default.

This should fix the current buildbot build failures.

This reverts commit 6435b8bb27ea ("build: include iwinfo by default
if nas or wpad(-mini) is selected")

Fixes: 04e9929c47e9 ("wifi-scripts: enable ucode scripts by default")
Link: https://github.com/openwrt/openwrt/pull/20211
Signed-off-by: Hauke Mehrtens <[email protected]>

octeon: set kernel 6.12 as default and remove support for 6.6

Get the Octeon target ready for the next OpenWrt release.

Signed-off-by: Rui Salvaterra <[email protected]>

octeon: update Linux 6.12 kconfig

Update with make kernel_oldconfig CONFIG_TARGET=target. Also disable
CONFIG_POSIX_MQUEUE, as it's available in the OpenWrt configuration.

Signed-off-by: Rui Salvaterra <[email protected]>

octeon: fix octeon build with CONFIG_KERNEL_WERROR

Add a patch fixing a bunch of missing prototype errors, scattered all over the
tree.

Signed-off-by: Rui Salvaterra <[email protected]>

octeon: fix command line hack for Linux 6.12

Since strlcpy has been removed in Linux 6.8, replace it with strscpy.

Signed-off-by: Rui Salvaterra <[email protected]>

octeon: refresh 6.12 patches

All but one needs (automatic) rebasing.

Signed-off-by: Rui Salvaterra <[email protected]>