mediatek: add ubootmod layout for cudy tr3000 v1

This allows us to use the full size of nand, which increases ubi size
from 64M to 122.25M.

If you are at factory firmware, please refer commit 63b8d98dd0d2 ("mediatek: add support for Cudy TR3000 v1")
to boot into OpenWrt initramfs (stock layout).

Flash instructions:
1. Login into the device and backup everything, especially 'Factory' part.
1. Unlock mtd partitions:
   apk update && apk add kmod-mtd-rw
   insmod mtd-rw i_want_a_brick=1
3. Write new BL2 and FIP
   mtd write openwrt-mediatek-filogic-cudy_tr3000-v1-ubootmod-preloader.bin BL2
   mtd write openwrt-mediatek-filogic-cudy_tr3000-v1-ubootmod-bl31-uboot.fip FIP
4. Set static IP on your PC:
   IP 192.168.1.254/24, GW 192.168.1.1
5. Serve OpenWrt initramfs image using TFTP server.
6. Cut off the power and re-engage, wait for TFTP recovery to complete.
7. After OpenWrt has booted, perform sysupgrade.

Tested-by: Yangyu Chen <[email protected]>
Signed-off-by: Tianling Shen <[email protected]>

umbim: add support for wwan device class

Some MBIM devices can exist on an MHI bus (over PCIe) instead of being presented as USB devices.

In such cases the interface name lookup needs to be done from /sys/class/wwan/ instead of /sys/class/usbmisc/

Add another readlink call in case the first lookup fails.

This allows the MBIM protocol to find the interface name and then work with both type of devices provided that /etc/config/network has the right device for MBIM interface (such as /dev/wwan0mbim0 in case of MHI)

Signed-off-by: Antti Seppälä <[email protected]>

generic: 6.12: backport upstream v6.16 brcmnand patches

This is needed to restore compatibility with legacy brcmnand controllers.

3bfb22cecfe6 mtd: rawnand: brcmnand: legacy exec_op implementation
528b541b71cf mtd: nand: brcmnand: fix NAND timeout when accessing eMMC
56fce7547004 mtd: rawnand: brcmnand: remove unused parameters

Signed-off-by: Álvaro Fernández Rojas <[email protected]>

wifi-scripts: use consistent names for Qualcomm devices

The device manufacturer name for Qualcomm Atheros is listed
as "Qualcomm, Atheros" for PCI devices but "Qualcomm Atheros" under
"compatible". This gives inconsistent results in `iwinfo` and

```
Hardware: 0x168c:0x0046 0x168c:0xcafe [Qualcomm, Atheros QCA9984]
Hardware: embedded [Qualcomm Atheros IPQ8074]
```

Signed-off-by: Sean Khan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18039
Signed-off-by: Robert Marko <[email protected]>

wifi-scripts: ucode: iwinfo: fix 'unknown' typo

The 'unknown' string was misspelled as 'unknonw'.

Signed-off-by: Sean Khan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18039
Signed-off-by: Robert Marko <[email protected]>

wifi-scripts: iwinfo: show units for signal and noise

Add 'dBm' units to 'Signal' and 'Noise' to match what non-ucode iwinfo shows.

Before:
  Signal: -49  Noise: -103

After:
  Signal: -49 dBm  Noise: -103 dBm

Signed-off-by: Sean Khan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18039
Signed-off-by: Robert Marko <[email protected]>

wifi-scripts: iwinfo: don't show bitrate unit when unknown

When bitrate is unknown, the units shouldn't be displayed. This is
consistent with other "unknown" fields and non-ucode iwinfo.

Signed-off-by: Sean Khan <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18039
Signed-off-by: Robert Marko <[email protected]>

bcm27xx: refresh 6.12 patches

Commit 438ed53577ad ("generic: 6.12: backport MediaTek Ethernet PHY changes")
backported commits to generic 6.12, but missed refreshing bcm27xx patches.

Fixes: 438ed53577ad ("generic: 6.12: backport MediaTek Ethernet PHY changes")
Signed-off-by: Álvaro Fernández Rojas <[email protected]>

realtek: align declaration/static/extern function definitions for dsa driver

In 2023 upstream has tightened compiler checks with this patch
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit
/scripts/Makefile.extrawarn?h=v6.15&id=0fcb70851fbfea1776ae62f67c503fef8f0292b9

A consistent function definition is now needed. I.e.

- functions must be either declared with "static" or
- functions need an additional declaration (in the header file)

Before upgrading the Realtek target to 6.12 clean the code so that the main
6.12 PR can focus on real issues from the version bump.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18925
Signed-off-by: Robert Marko <[email protected]>

realtek: align declaration/static/extern function definitions for eth driver

In 2023 upstream has tightened compiler checks with this patch
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit
/scripts/Makefile.extrawarn?h=v6.15&id=0fcb70851fbfea1776ae62f67c503fef8f0292b9

A consistent function definition is now needed. I.e.

- functions must be either declared with "static" or
- functions need an additional declaration (in the header file)

Before upgrading the Realtek target to 6.12 clean the code so that the main
6.12 PR can focus on real issues from the version bump.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18925
Signed-off-by: Robert Marko <[email protected]>

realtek: align declaration/static/extern function definitions for phy driver

In 2023 upstream has tightened compiler checks with this patch
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit
/scripts/Makefile.extrawarn?h=v6.15&id=0fcb70851fbfea1776ae62f67c503fef8f0292b9

A consistent function definition is now needed. I.e.

- functions must be either declared with "static" or
- functions need an additional declaration (in the header file)

Before upgrading the Realtek target to 6.12 clean the code so that the main
6.12 PR can focus on real issues from the version bump.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18925
Signed-off-by: Robert Marko <[email protected]>

realtek: align declaration/static/extern function definitions for clock driver

In 2023 upstream has tightened compiler checks with this patch
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit
/scripts/Makefile.extrawarn?h=v6.15&id=0fcb70851fbfea1776ae62f67c503fef8f0292b9

A consistent function definition is now needed. I.e.

- functions must be either declared with "static" or
- functions need an additional declaration (in the header file)

Before upgrading the Realtek target to 6.12 clean the code so that the main
6.12 PR can focus on real issues from the version bump.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18925
Signed-off-by: Robert Marko <[email protected]>

realtek: align declaration/static/extern function definitions for startup code

In 2023 upstream has tightened compiler checks with this patch
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit
/scripts/Makefile.extrawarn?h=v6.15&id=0fcb70851fbfea1776ae62f67c503fef8f0292b9

A consistent function definition is now needed. I.e.

- functions must be either declared with "static" or
- functions need an additional declaration (in the header file)

Before upgrading the Realtek target to 6.12 clean the code so that the main
6.12 PR can focus on real issues from the version bump.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18925
Signed-off-by: Robert Marko <[email protected]>

d1: improve Nezha D1 support

The Nezha board has a PCF8574 that acts as an I/O expander on the
i2c2 bus. Enable the required module for it.

Signed-off-by: Zoltan HERPAI <[email protected]>

d1: enable crypto offload engine

Enable the sun8i_ce engine that's included on the D1 SoC, along
with the respective hash/prng/trng options. The sun8i_ce driver
also requires PM and PPU to be enabled.

[    1.368345] sun8i-ce 3040000.crypto: Set mod clock to 300000000 (300 Mhz) from 400000000 (400 Mhz)
[    1.378592] sun8i-ce 3040000.crypto: will run requests pump with realtime priority
[    1.386894] sun8i-ce 3040000.crypto: will run requests pump with realtime priority
[    1.395114] sun8i-ce 3040000.crypto: will run requests pump with realtime priority
[    1.403197] sun8i-ce 3040000.crypto: will run requests pump with realtime priority
[    1.411110] sun8i-ce 3040000.crypto: Register cbc(aes)
[    1.416235] sun8i-ce 3040000.crypto: Register ecb(aes)
[    1.421347] sun8i-ce 3040000.crypto: Register cbc(des3_ede)
[    1.426898] sun8i-ce 3040000.cryxto: Register ecb(des3_ede)
[    1.432366] sun8i-ce 3040000.crypto: Register md5
[    1.437112] sun8i-ce 3040000.crypto: Register sha1
[    1.441816] sun8i-ce 3040000.crypto: Register sha224
[    1.446821] sun8i-ce 3040000.crypto: Register sha256
[    1.451697] sun8i-ce 3040000.crypto: Register sha384
[    1.456733] sun8i-ce 3040000.crypto: Register sha512
[    1.461673] sun8i-ce 3040000.crypto: Register stdrng
[    1.467096] sun8i-ce 3040000.crypto: CryptoEngine Die ID 0

root@OpenWrt:~# cat /proc/crypto |grep sun8i
driver       : sun8i-ce-prng
driver       : sha512-sun8i-ce
driver       : sha384-sun8i-ce
driver       : sha256-sun8i-ce
driver       : sha224-sun8i-ce
driver       : sha1-sun8i-ce
driver       : md5-sun8i-ce
driver       : ecb-des3-sun8i-ce
driver       : cbc-des3-sun8i-ce
driver       : ecb-aes-sun8i-ce
driver       : cbc-aes-sun8i-ce

Signed-off-by: Zoltan HERPAI <[email protected]>

toolchain: gcc: drop GCC11

There is no practical value in keeping GCC11 around, as even OpenWrt 23.05
uses GCC12 as the default one, so drop it.

Signed-off-by: Robert Marko <[email protected]>

ramips: mt7621: sercomm_na502s: use upstream XR USB serial driver

Instead of including the out-of-tree XR USB serial driver, use the
newly packaged in-tree driver for it.

Link: https://github.com/openwrt/openwrt/pull/18926
Signed-off-by: Robert Marko <[email protected]>

kernel: usb: package MaxLinear/Exar USB serial driver

Currently, MaxLinear/Exar USB serial devices are supported via out-of-tree
usb-serial-xr_usb_serial_common driver which is broken on 6.12.

So since upstream has support for these devices since 6.5 lets package
the in-tree driver to be able to drop the out-of-tree one.

Link: https://github.com/openwrt/openwrt/pull/18926
Signed-off-by: Robert Marko <[email protected]>

mediatek: enable testing kernel 6.12

Set KERNEL_TESTING_PATCHVER to allow testing Linux 6.12.

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

mediatek: mt7629: update config-6.12

Refresh kernel configuration of the mt7629 subtarget for use
with Linux 6.12.

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

mediatek: mt7623: update config-6.12

Refresh kernel configuration of the mt7623 subtarget for use
with Linux 6.12.

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

mediatek: mt7622: update config-6.12

Refresh kernel configuration of the mt7622 subtarget for use
with Linux 6.12.

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

mediatek: filogic: update config-6.12

Refresh kernel configuration of the filogic subtarget for use
with Linux 6.12.

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

kernel/mediatek: 6.12: fix and refresh patches

Fix patches and refresh for Linux 6.12.

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

kernel/mediatek: 6.12: adapt for changed MMC card binding

Adapt patches adding NVMEM or fitblk rootdisk references to MMC
partitions for the changed MMC card binding.

Note that boards in 'dts' folder can only be converted once Linux 6.6,
which still depends on the old binding, has been dropped.

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

kernel/mediatek: 6.12: allow setting OF node for MSDOS partitions

Add patch to allow referencing MSDOS/MBR partitions in device tree to
allow legacy MT7623 boards to continue working with fitblk.

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

kernel/mediatek: 6.12: adapt SPI calibration

Fix build of downstream SPI calibration on Filogic SoCs for Linux 6.12.

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

kernel/mediatek: 6.12: replace downstream files by patches

Replace downstream files by patches, either backports of those
which have already applied or pending patches tracked on patchwork.
This is done to make future maintainance more easy.

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

kernel/mediatek: 6.12: drop patches which have been applied upstream

Drop patches which have been applied upstream in the meantime.

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

kernel/mediatek: 6.12: drop patches which are now part of generic

Drop patches which have been moved into target/linux/generic by commit
f4144d61d2 ("generic: 6.12: backport MediaTek Ethernet PHY changes").

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

kernel/mediatek: Restore kernel files for v6.6

This is an automatically generated commit which aids following Kernel patch
history, as git will see the move and copy as a rename thus defeating the
purpose.

For the original discussion see:
https://lists.openwrt.org/pipermail/openwrt-devel/2023-October/041673.html

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

kernel/mediatek: Create kernel files for v6.12 (from v6.6)

This is an automatically generated commit.

When doing `git bisect`, consider `git bisect --skip`.

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

mediatek: filogic: adapt BananaPi-R4 for use with upstream DT

 * the variant with 2.5G PHY instead of LAN SFP is called '2p5'
   upstream and 'poe' in our downstream Linux 6.6 DT. Use the right
   DTS depending on the kernel version and set an additional
   compatible.

 * drop additional DT overlay for WiFi.
   The final version of the board uses a physical switch for the 12V
   power of the WiFi module and the I2C EEPROM of the module always
   comes empty (instead of with a MAC address).

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

mediatek: dts: bring mt7988a.dtsi closer to upstream

In preparation of using the upstream mt7988a.dtsi when switching
to Linux 6.12 prepare by bringing our downstream version closer to
what went upstream.
 * rename 'xphy' -> 'xsphy'
 * rename 'uart[012]' -> 'serial[012]'
 * only list pinctrl settings directly used in mt7988a.dtsi there,
   leave it to boards to define all additional pinctrl settings
   they need.
 * move fan and thermal-zone to board level

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

mediatek: dts: rename mt7981.dtsi to mt7981b.dtsi

Upstream uses a different filename, so lets rename our downstream
mt7981.dtsi to mt7981b.dtsi and update the device tree of all
MT7981 boards accordingly.
This is to prepare for the switch to Linux 6.12 which is going to
use the upstream mt7981b.dtsi (plus some patches on top).

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

mediatek: mt7622: prepare legacy rtl8367c driver for Linux 6.12

Do the minimum necessary to get the legacy rtl8367c driver to build
with Linux 6.12.

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

generic: 6.12: backport MediaTek Ethernet PHY changes

The MediaTek Ethernet PHY drivers are going to be used by multiple
targets (airoha, mediatek, ramips). Add generic backports of changes
required for recently added Ethernet PHYs.

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

generic: 6.12: mtk_eth_soc: adapt downstream SerDes patch

While it will hopefully be replaced soon by an upstream solution for
now we keep the downstream SerDes PCS implementation for MT7988A.
In order to make it work with upstream mt7988a.dtsi we have to
compensate for the changed start address of topmisc which was
moved from 0x11d10000 to 0x11d10084 to accomodate a future power
domain controller located at 0x11d10000~0x11d10080.
Hence we need to change TOP_MISC_NETSYS_PCS_MUX from 0x84 to 0x0, so
the Ethernet path/mux selection can work with upstream mt7988a.dtsi.

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

generic: 6.12: fix patch adding EEE-support to mtk_eth_soc

Upstream now uses struct ethtool_keee instead of struct ethtool_eee
as parameter to EEE-related functions. Follow that change and modify
the patch accordingly.

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

generic: net: phy: rtl8261n: fix build with Linux 6.12

Mark functions which aren't exported as static to fix build with
Linux 6.12.

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

generic: 6.12: add hack patch for transition to new partition bindings

Commit fa0f130764 ("generic: 6.12: update block NVMEM driver") switched
to the upstream DT bindings for block partitions.

Bring back the old/legacy downstream way to assign the OF node to a
block device or partition in order to allow sharing a single DTS for
both, Linux 6.6 (which exclusively uses the old/legacy binding) and
Linux 6.12 (which will now support both, new/upstream binding and the
old/legacy binding).

Once we drop Linux 6.6 and all boards have been converted to the new
binding we can drop this patch.

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

generic: 6.12: add missing Kconfig symbol

Augmented symbol CONFIG_SND_SOC_MT8365 was missing, add it.

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

realtek: refactor RTL838x mdio serdes read/write functions

There are still a lot of mdio functions scattered around the code.
Move the RTL838x serdes helpers closer to the bus, add the proper
prefix and simplify the functions.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18847
Signed-off-by: Robert Marko <[email protected]>

realtek: simplify RTL8214FC patches for ethtool copper/fiber switching

There is a patch/bug cascade in the realtek target phy code that must be resolved.

1. The phy_driver structure is patched to add features ONLY needed for RTL8214FC

2. The kernel is patched to allow switching fiber/copper port of phys through ethtool
   by calling these new features.

3. With those patches applied the bootup always switches RTL8214FC ports to copper.
   Even if a SFP module was found before and the phy driver switched to fibre before.

3. So another patch is needed that reprobes the SFP module to activate fiber again.

4. Because of the reprobing we need a fourth patch that avoid duplicate devices.

Simplify this by removing all patches and reusing the existing ethtool phy tunable
interface. The command line usage might be counterintuitive but it avoids tons of
problems in the code. In addition, this scenario is not used frequently.

Before:ethtool -s lan25 port fibre/tp
After: ethtool --set-phy-tunable lan25 downshift on/off

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18816
Signed-off-by: Robert Marko <[email protected]>

armsr: remove 'console=tty1' from kernel command line

We have modified the kernel to setup all "default" consoles,
including serial ports and framebuffers/screens, providing
no console= argument is supplied on the kernel command line.

Adding 'console=tty1' caused the 'default' serial port on
device tree systems to break, as the kernel would not carry
over the settings (like baud rate) from the bootloader.

The system administrator can still force the use of a
specific console by adding their own console= arguments.

Signed-off-by: Mathew McBride <[email protected]>
Fixes: c099523d66a3 ("armsr: use console=tty1 to make
console more readily available")
Link: https://github.com/openwrt/openwrt/pull/17012
Signed-off-by: Robert Marko <[email protected]>

armsr: add patch (hack) to enable all "default" consoles

A previous change added 'console=tty1' to the default kernel command
line on armsr, in order to ensure the framebuffer console is enabled
on systems capable of graphics output.

Unfortunately, this change broke boards that used device tree
(DT) firmware with serial consoles, as the serial console
specified by the system firmware (stdout-path) was no longer
setup by the kernel.

A bit of probing determined that the SPCR (serial port console
direction table) on ACPI systems was preventing Linux from setting
up a default framebuffer console on these systems (which is why
console=tty1 was added).

(The affected ACPI systems are usually VMs using QEMU's
'virt' machine and EDK2 firmware. The firmware on these systems
does not remove the SPCR when a screen is present)

So to ensure all possible systems are setup correctly, we modify
the kernel so all "default" console types (serial and screen)
are setup when no console= arguments are specified on the kernel
command line.

Signed-off-by: Mathew McBride <[email protected]>
Fixes: c099523d66a3 ("use console=tty1 to make console more
readily available")
Link: https://github.com/openwrt/openwrt/pull/17012
Signed-off-by: Robert Marko <[email protected]>

ath79: refresh 6.12 kernel patches

Refresh patches to fix the GitHub CI warning.

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18912
Signed-off-by: Nick Hainke <[email protected]>

mediatek: filogic: add support for SNR-CPE-AX2

SOC: MediaTek MT7981b
RAM: 256MB DDR3
FLASH: 128MB SPI-NAND (Winbond W25N01GV)
WIFI: Mediatek MT7981b DBDC 802.11ax 2.4/5 GHz
ETH: MediaTek MT7531 Switch
UART: 3V3 115200 8N1 (Pinout silkscreened / Do not connect VCC)

Interface MAC Algorithm
LAN 8C:AE:DB:2C:xx:xx label
WAN 8C:AE:DB:2C:xx:xx label +1
WLAN 2.4G 8C:AE:DB:2C:xx:xx label +2
WLAN 5G 8C:AE:DB:2C:xx:xx label +3
Installation
-------------------Install openwrt image-------------------------------:

Set a static ip on the ethernet interface of your PC. (ip address:
192.168.1.254, subnet mask:255.255.255.0) .

Download the OpenWrt uboot image
(openwrt-mediatek-filogic-snr_cpe-ax2-bl31-uboot.fip).

SSH/SCP opened by default on the stock firmware (3.0.1).
Username: Admin, default password: Admin. Check it on the bottom of
the router.

Copy uboot image using SCP (WinSCP) to /tmp dir on SNR-CPE-AX2.

Download recovery file.
openwrt-mediatek-filogic-snr_snr-cpe-ax2-initramfs-recovery.itb.

Copy the recovery image to a TFTP server reachable at 192.168.1.254/24.

Open ssh shell to the SNR-CPE-AX2.
Run commands:
mtd write \
/tmp/openwrt-mediatek-filogic-snr_snr-cpe-ax2-bl31-uboot.fip FIP
reboot

Wait until recovery boot.
Open web 192.168.1.1 and do sysupgrade by
openwrt-mediatek-filogic-snr_cpe-ax2-squashfs-sysupgrade.itb

Signed-off-by: Nikolay March <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18700
Signed-off-by: Hauke Mehrtens <[email protected]>

x86: legacy: config: alphabetize config-6.12

Sort the config in alphabetical order

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18897
Signed-off-by: Hauke Mehrtens <[email protected]>

stm32: add missing config option

6.6.92 requires the addition of CONFIG_ARM_SCMI_DEBUG_COUNTERS

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18835
Signed-off-by: Hauke Mehrtens <[email protected]>

rockchip: armv8: add missing config option

6.6.92 requires the addition of CONFIG_ARM_SCMI_DEBUG_COUNTERS

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18835
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 6.6 to 6.6.92

Changelog: https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.6.92

No patches needed a rebase with this release/only archive hash updated.

Build system: x86/64
Build-tested: flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3
Run-tested: flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18835
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 6.6 to 6.6.91

Changelog: https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.6.91

Removed upstreamed:
generic/pending-6.6/730-net-ethernet-mtk_eth_soc-reset-all-TX-queues-on-DMA-.patch[1]

All other patches automatically rebased.

1. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v6.6.91&id=68f29bb97a0e0519156664cdc23e8b1f129e3254

Build system: x86/64
Build-tested: flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3
Run-tested: flogic/xiaomi_redmi-router-ax6000-ubootmod, ramips/tplink_archer-a6-v3

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18835
Signed-off-by: Hauke Mehrtens <[email protected]>

mac80211: ath9k: replace qca,led-sources

Upstream seems to be using led-sources instead of custom properties.

Code mostly taken from mt76.

Changed all(few) users of qca,led_pin to use the new format.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18805
Signed-off-by: Robert Marko <[email protected]>

adb: Switch to mbedtls

Currently adb uses libopenssl for certain authentication tasks between
the host and the target device such as certificate generation, hashing,
base64 encoding and pki signatures.

Add a patch to use functionalities available in mbedtls instead.

Also switch package makefile and dependency to libmbedtls and drop
patches and references to lib{crypto,openssl} as they are no longer
required.

This conserves considerable amount of space on the device as openwrt
ships with libmbedtls by default.

Signed-off-by: Antti Seppälä <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18819
Signed-off-by: Robert Marko <[email protected]>

realtek: relocate mips cpc probing

The MIPS CPC (Cluster Power Controller) is setup during boot and can take
its configuration from the devicetree. This is currently not possible
because the cpc probing happens before dt initialization. Call order
during startup is:

setup_arch()
  prom_init() <- our function
    mips_cpc_probe()
    smp_stuff()
  arch_mem_init()
    device_tree_init() <- our function
      unflatten_and_copy_device_tree()

To avoid ugly hacking and support a clean devicetree relocate the cps/smp
stuff to device_tree_init(). This is basically the same location as in
generic mips initialization.

Signed-off-by: Markus Stockhausen <[email protected]>
Signed-off-by: Sebastian Gottschall <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18888
Signed-off-by: Robert Marko <[email protected]>

realtek: take over boot command line patch into driver

Make it easier to upgrade the kernel in the future. For this remove
the 320-harden-fw_init_cmdline.patch and add the logic into the
startup sequence at the right place.

Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18853
Signed-off-by: Robert Marko <[email protected]>

toolchain: gcc: update GCC14 to 14.3

Remove upstreamed patches:
- 020-MIPS-Include-missing-mips16.S-in-libgcc-lib1funcs.S.patch
- 021-Reuse-scratch-registers-generated-by-LRA.patch

All other patches are automatically refreshed.

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18891
Signed-off-by: Robert Marko <[email protected]>

mxs: drop 6.6 support

Drop configs for Linux 6.6.

Signed-off-by: Zoltan HERPAI <[email protected]>

mxs: use kernel 6.12 by default

Switch to Linux kernel version 6.12.

Signed-off-by: Zoltan HERPAI <[email protected]>

ath79: enable 6.12 testing kernel

The 6.12 testing kernel for ath79 target is ready now.

Tested on AR9344 (ath79/nand) and QCA9531 (ath79/generic).

Signed-off-by: Shiji Yang <[email protected]>

ath79: irqchip/ath79-misc: fix missing prototypes warnings

The newly introduced patch fixes the compilation warning for
kernel module irq-ath79-misc.

Signed-off-by: Shiji Yang <[email protected]>

ath79: refresh 6.12 kernel config files

All kernel symbols are automatically refreshed by
`make kernel_oldconfig CONFIG_TARGET=target` and
`make kernel_oldconfig CONFIG_TARGET=subtarget`.

Signed-off-by: Shiji Yang <[email protected]>

ath79: refresh 6.12 kernel patches

Remove upstreamed:
001-v6.11-gpio-ath79-convert-to-dynamic-GPIO-base-allocation.patch [1]

Manually rebased:
341-wifi-ath9k-obtain-system-gpios.patch
900-unaligned_access_hacks.patch

[1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-6.12.y&id=9a473c2a093e0d1c466bf86073230e2c8b658977
Signed-off-by: Shiji Yang <[email protected]>

ath79: restore kernel 6.6 config files and patches

Copy patches and kernel configs from 6.12 kernel to restore the
default 6.6 kernel support files.

Signed-off-by: Shiji Yang <[email protected]>

ath79: rename patchset and kernel configs to 6.12

This is a preparation for 6.12 kernel support. It can help us
track the files history by using the Git tool.

Signed-off-by: Shiji Yang <[email protected]>

ath79: refresh kernel symbol configs

Manually move the common NVMEM Kconfig symbols to the shared
config file and disable them for the mikrotik subtarget. Other
changes are automatically completed by `make kernel_oldconfig`.
This is a preparation for introducing the 6.12 kernel support.

Signed-off-by: Shiji Yang <[email protected]>

d1: drop 6.6 support

Drop configs and patches for Linux 6.6.

Signed-off-by: Zoltan HERPAI <[email protected]>

d1: use kernel 6.12 by default

Switch to Linux kernel version 6.12.

Signed-off-by: Zoltan HERPAI <[email protected]>

image: only filter out images when ImageBuilder is used

Currently, we are filtering out images if DEFAULT:=n or BROKEN:=y are set,
so if you are building from scratch and want to build custom images that
are stripped down to fit, you must edit the image recipe or its just
filtered out.

So, to allow this behaviour when building from scratch as we can assume
that person doing that knows what they are attempting to do lets just limit
the filtering to ImageBuilder.

Fixes: f060615a78e5 ("image: respect DEFAULT and BROKEN when Default profile is selected")
Signed-off-by: Robert Marko <[email protected]>

realtek: prepare RTL931x for full SMP support

RTL931x devices are dual core with two threads each. That is a total
of 4 VPEs (CPUs) in the kernel. Adapt the kernel config for that.

Signed-off-by: Markus Stockhausen <[email protected]>
Signed-off-by: Sebastian Gottschall <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18889
Signed-off-by: Robert Marko <[email protected]>

intel-microcode: update to 20250512

Debian Changelogs from 20250512:
  * New upstream microcode datafile 20250512
    - Mitigations for INTEL-SA-01153 (ITS: Indirect Target Selection):
      CVE-2024-28956: Processor may incompletely mitigate Branch Target
      Injection due to indirect branch predictions that are not fully
      constrained by eIBRS nor by the IBPB barrier.  Part of the "Training
      Solo" set of vulnerabilities.
    - Mitigations for INTEL-SA-01244:
      CVE-2025-20103: Insufficient resource pool in the core management
      mechanism for some Intel Processors may allow an authenticated user
      to potentially enable denial of service via local access.
      CVE-2025-20054: Uncaught exception in the core management mechanism
      for some Intel Processors may allow an authenticated user to
      potentially enable denial of service via local access.
    - Mitigations for INTEL-SA-01247:
      CVE-2024-43420, CVE-2025-20623: Exposure of sensitive information
      caused by shared microarchitectural predictor state that influences
      transient execution for some Intel Atom and some Intel Core
      processors (10th Generation) may allow an authenticated user to
      potentially enable information disclosure via local access.
      CVE-2024-45332 (Branch Privilege Injection): Exposure of sensitive
      information caused by shared microarchitectural predictor state that
      influences transient execution in the indirect branch predictors for
      some Intel Processors may allow an authenticated user to potentially
      enable information disclosure via local access.
    - Mitigations for INTEL-SA-01322:
      CVE-2025-24495 (Training Solo): Incorrect initialization of resource
      in the branch prediction unit for some Intel Core Ultra Processors
      may allow an authenticated user to potentially enable information
      disclosure via local access (IBPB bypass)
      CVE-2025-20012 (Training Solo): Incorrect behavior order for some
      Intel Core Ultra Processors may allow an unauthenticated user to
      potentially enable information disclosure via physical access.
    - Improved fix for the Vmin Shift Instability for the Intel Core 13th
      and 14th gen processors under low-activity scenarios (sig 0xb0671).
      This microcode update is supposed to be delivered as a system
      firmware update, but according to Intel it should be effective when
      loaded by the operating system if the system firmware has revision
      0x12e.
    - Fixes for unspecified functional issues on several processor models
  * New microcodes or new extended signatures:
    sig 0x000a06d1, pf_mask 0x95, 2025-02-07, rev 0x10003a2, size 1664000
    sig 0x000a06d1, pf_mask 0x20, 2025-02-07, rev 0xa0000d1, size 1635328
    sig 0x000b0650, pf_mask 0x80, 2025-03-18, rev 0x000a, size 136192
    sig 0x000b06d1, pf_mask 0x80, 2025-03-18, rev 0x011f, size 79872
    sig 0x000c0662, pf_mask 0x82, 2025-03-20, rev 0x0118, size 90112
    sig 0x000c06a2, pf_mask 0x82, 2025-03-20, rev 0x0118
    sig 0x000c0652, pf_mask 0x82, 2025-03-20, rev 0x0118
    sig 0x000c0664, pf_mask 0x82, 2025-03-20, rev 0x0118
  * Updated microcodes:
    sig 0x00050657, pf_mask 0xbf, 2024-12-12, rev 0x5003901, size 39936
    sig 0x0005065b, pf_mask 0xbf, 2024-12-12, rev 0x7002b01, size 30720
    sig 0x000606a6, pf_mask 0x87, 2025-01-07, rev 0xd000404, size 309248
    sig 0x000606c1, pf_mask 0x10, 2025-01-07, rev 0x10002d0, size 300032
    sig 0x000706a8, pf_mask 0x01, 2024-12-05, rev 0x0026, size 76800
    sig 0x000706e5, pf_mask 0x80, 2025-01-07, rev 0x00ca, size 115712
    sig 0x000806c1, pf_mask 0x80, 2024-12-01, rev 0x00bc, size 112640
    sig 0x000806c2, pf_mask 0xc2, 2024-12-01, rev 0x003c, size 99328
    sig 0x000806d1, pf_mask 0xc2, 2024-12-11, rev 0x0056, size 105472
    sig 0x000806ec, pf_mask 0x94, 2024-11-17, rev 0x0100, size 106496
    sig 0x000806f8, pf_mask 0x87, 2025-01-28, rev 0x2b000639, size 591872
    sig 0x000806f7, pf_mask 0x87, 2025-01-28, rev 0x2b000639
    sig 0x000806f6, pf_mask 0x87, 2025-01-28, rev 0x2b000639
    sig 0x000806f5, pf_mask 0x87, 2025-01-28, rev 0x2b000639
    sig 0x000806f4, pf_mask 0x87, 2025-01-28, rev 0x2b000639
    sig 0x000806f8, pf_mask 0x10, 2025-01-28, rev 0x2c0003f7, size 624640
    sig 0x000806f6, pf_mask 0x10, 2025-01-28, rev 0x2c0003f7
    sig 0x000806f5, pf_mask 0x10, 2025-01-28, rev 0x2c0003f7
    sig 0x000806f4, pf_mask 0x10, 2025-01-28, rev 0x2c0003f7
    sig 0x00090672, pf_mask 0x07, 2024-12-12, rev 0x003a, size 226304
    sig 0x00090675, pf_mask 0x07, 2024-12-12, rev 0x003a
    sig 0x000b06f2, pf_mask 0x07, 2024-12-12, rev 0x003a
    sig 0x000b06f5, pf_mask 0x07, 2024-12-12, rev 0x003a
    sig 0x000b06f6, pf_mask 0x07, 2024-12-12, rev 0x003a
    sig 0x000b06f7, pf_mask 0x07, 2024-12-12, rev 0x003a
    sig 0x000906a3, pf_mask 0x80, 2024-12-12, rev 0x0437, size 224256
    sig 0x000906a4, pf_mask 0x80, 2024-12-12, rev 0x0437
    sig 0x000906a4, pf_mask 0x40, 2024-12-06, rev 0x000a, size 119808
    sig 0x000906ed, pf_mask 0x22, 2024-11-14, rev 0x0104, size 106496
    sig 0x000a0652, pf_mask 0x20, 2024-11-14, rev 0x0100, size 97280
    sig 0x000a0653, pf_mask 0x22, 2024-11-14, rev 0x0100, size 98304
    sig 0x000a0655, pf_mask 0x22, 2024-11-14, rev 0x0100, size 97280
    sig 0x000a0660, pf_mask 0x80, 2024-11-14, rev 0x0102, size 98304
    sig 0x000a0661, pf_mask 0x80, 2024-11-14, rev 0x0100, size 97280
    sig 0x000a0671, pf_mask 0x02, 2024-12-01, rev 0x0064, size 108544
    sig 0x000a06a4, pf_mask 0xe6, 2025-02-13, rev 0x0024, size 140288
    sig 0x000a06f3, pf_mask 0x01, 2025-02-10, rev 0x3000341, size 1542144
    sig 0x000b0671, pf_mask 0x32, 2025-03-17, rev 0x012f, size 219136
    sig 0x000b0674, pf_mask 0x32, 2025-03-17, rev 0x012f
    sig 0x000b06a2, pf_mask 0xe0, 2025-01-15, rev 0x4128, size 224256
    sig 0x000b06a3, pf_mask 0xe0, 2025-01-15, rev 0x4128
    sig 0x000b06a8, pf_mask 0xe0, 2025-01-15, rev 0x4128
    sig 0x000b06e0, pf_mask 0x19, 2024-12-06, rev 0x001d, size 139264
    sig 0x000c06f2, pf_mask 0x87, 2025-03-14, rev 0x210002a9, size 563200
    sig 0x000c06f1, pf_mask 0x87, 2025-03-14, rev 0x210002a9
  * Removed microcodes (ES/QS steppings):
    sig 0x00050656, pf_mask 0xbf, 2023-07-28, rev 0x4003605, size 38912
    sig 0x000c06f1, pf_mask 0x87, 2025-03-14, rev 0x210002a9 [EXCLUDED]

Build system: x86/64
Build-tested: x86/64
Run-tested: x86/64

Signed-off-by: John Audia <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18869
Signed-off-by: Robert Marko <[email protected]>

qualcommax: ipq50xx: backport upstreamed patches for IPQ5018 tsens support

Use upstreamed v6.16 patches for IPQ5018 tsens support.

Signed-off-by: George Moussalem <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18884
Signed-off-by: Robert Marko <[email protected]>

qualcommax: ipq50xx: backport upstreamed patches for IPQ5018 TCSR support

Use upstreamed v6.16 patch for IPQ5018 TCSR support and setting the
download mode.

Signed-off-by: George Moussalem <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18884
Signed-off-by: Robert Marko <[email protected]>

qualcommax: ipq50xx: backport upstreamed patches for IPQ5018 PCIe support

Use upstreamed patches for IPQ5018 PCIe support, including patches for
phy and controller drivers and dts nodes.

Signed-off-by: George Moussalem <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18884
Signed-off-by: Robert Marko <[email protected]>

tools/mold: remove static only dependencies

It seems that MOLD_MOSTLY_STATIC was required before to avoid packaging
every single dependency. But nowadays, mold uses whatever dependency it
can find and builds its own if it cannot.

Saves a small amount of compilation time.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18881
Signed-off-by: Robert Marko <[email protected]>

tools/mpfr: update to 4.22

Refresh patches.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18857
Signed-off-by: Robert Marko <[email protected]>

mac80211: brcm: update RPi brcmfmac patches

- Reorganize brcmfmac patches.
- Drop outdated RPi brcmfmac patches.
- Add RPi 6.12 brcmfmac patches with "rpi-6.12" name tag (excluding SAE).

Signed-off-by: Álvaro Fernández Rojas <[email protected]>

qualcommax: 6.6: remove completely

Remove all 6.6 only patches, config etc.

Signed-off-by: Robert Marko <[email protected]>

qualcommax: import pending fix for multiple conf

Importing upstream multiple conf patches broke UNIPHY2 TX clock parenting
and thus no traffic could pass via it.

Import pending fix by Christian that fixes this[1].

[1] https://patchwork.kernel.org/project/linux-arm-msm/patch/20250522202600[email protected]/

Fixes: cc50cac8a0fd ("qualcommax: use upstreamed multiple conf clock patches")
Signed-off-by: Robert Marko <[email protected]>

realtek: fix MAC_FORCE_MODE_CTRL register for RTL931x

Ports will not work as expected on the RTL931x devices. This comes
from a typo in the address. Fixed according to the documentation:
https://svanheule.net/realtek/mango/register/mac_force_mode_ctrl

Signed-off-by: Markus Stockhausen <[email protected]>
Signed-off-by: Sebastian Gottschall <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18864
Signed-off-by: Robert Marko <[email protected]>

tools: add cmake dependency to bzip2

It's using cmake.mk so it needs CMake to build.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18880
Signed-off-by: Robert Marko <[email protected]>

sifiveu: add 6.12 kernel as testing

Allow selecting 6.12 as testing kernel on sifiveu.

Runtime-tested:
 - HiFive Unleashed (FU540)
 - HiFive Unmatched (FU740)

Signed-off-by: Zoltan HERPAI <[email protected]>

sifiveu: 6.12: drop upstreamed patches and refresh remaining ones

Drop upstreamed patches from 6.12 and refresh remaining ones.

Signed-off-by: Zoltan HERPAI <[email protected]>

sifiveu: refresh 6.12 kernel config

Refresh kernel config for 6.12.

Signed-off-by: Zoltan HERPAI <[email protected]>

kernel/sifiveu: Restore kernel files for v6.6

This is an automatically generated commit which aids following Kernel patch
history, as git will see the move and copy as a rename thus defeating the
purpose.

For the original discussion see:
https://lists.openwrt.org/pipermail/openwrt-devel/2023-October/041673.html

Signed-off-by: Zoltan HERPAI <[email protected]>

kernel/sifiveu: Create kernel files for v6.12 (from v6.6)

This is an automatically generated commit.

When doing `git bisect`, consider `git bisect --skip`.

Signed-off-by: Zoltan HERPAI <[email protected]>

bcm27xx: drop README patches

This patches are useless for us and can be removed.

Signed-off-by: Álvaro Fernández Rojas <[email protected]>

tools/elftosb: fix compilation with C++17

Newer GCC and Clang default to C++17, which does not support register.
Just remove it.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18859
Signed-off-by: Hauke Mehrtens <[email protected]>

uboot-bcm4908: fix build with GCC14

A lot of warnings were treated as errors after the default compiler
switched to GCC14. It's hard to fix them one by one, and this u-boot
is not maintained by upstream, so let's just silence these warnings.

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18833
Signed-off-by: Hauke Mehrtens <[email protected]>

kernel: bump 6.12 to 6.12.30

Changelog: https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.12.30

All patches automatically rebased.

Signed-off-by: Mieczyslaw Nalewaj <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18877
Signed-off-by: Hauke Mehrtens <[email protected]>

uboot-mvebu: update to version 2025.04

Update package to the latest stable version. All patches automatically
refreshed.

Signed-off-by: Stefan Kalscheuer <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/17813
Signed-off-by: Hauke Mehrtens <[email protected]>

realtek: fix mdio parent/child locking issues

Since the early beginning of the Realtek DSA driver there is an uncovered
locking issue between the standard (parent) mdio bus and the DSA (child)
mdio bus. This comes from the fact that the DSA bus simply links to the
parent read and write functions and calls them directly. This leads to
the following lock issue.

- Child bus calls phy_read/write functions and uses its internal lock
- Parent bus calls phy_read/write functions and uses its internal lock

It becomes clear that critical section can be accessed twice without
knowing that a operation from the other bus is currently active. This
can lead to critical malfunctions because the mdio driver needs a lot of
internal magic to get page selection done right. Effects are:

- The original page is lost after a phy_write/read_paged() call
- dmesg like "Realtek RTL8218B (external) rtl838x slave mii-0:00:
  Expected external RTL8218B, found PHY-ID 6b23"

Other DSA drivers simply use the read/write functions from the parent bus
and thus avoid locking issues. Do it the same way.

Fixes: 2b88563ee5aafd9 ("realtek: update the tree to the latest refactored version")
Signed-off-by: Markus Stockhausen <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18824
Signed-off-by: Robert Marko <[email protected]>

tools/quilt: update to 0.69

Refresh patches.

Signed-off-by: Rosen Penev <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18858
Signed-off-by: Robert Marko <[email protected]>

boot: uboot-armsr: update to U-Boot 2025.04

This U-Boot version is used to generate a 'reference'
U-Boot binary for QEMU's 'virt' machine on both
armv7 and armv8.

It has not been updated since EFI support was merged
into the then-armvirt target, so we should bring it up
to the latest version.

The 'mkeficapsule' tool is disabled due to a host-side
dependency on GnuTLS.

Signed-off-by: Mathew McBride <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18862
Signed-off-by: Robert Marko <[email protected]>

ramips: fix wrg-header image recipe

Before generating the factory image, check if the input file
exists. Fix the build error when sysupgrade image is too big:

[mkwrgimg] *** error: stat failed on /builder/shared-workdir/build/build_dir/target-mipsel_24kc_musl/linux-ramips_rt288x/tmp/openwrt-ramips-rt288x-airlink101_ar670w-squashfs-factory.bin, No such file or directory

Signed-off-by: Shiji Yang <[email protected]>
Link: https://github.com/openwrt/openwrt/pull/18836
Signed-off-by: Robert Marko <[email protected]>

d1: add 6.12 kernel as testing

Allow selecting 6.12 as testing kernel on d1.

Runtime-tested:
 - Nezha D1

Signed-off-by: Zoltan HERPAI <[email protected]>

d1: 6.12: refresh kernel config

Refresh kernel config.

Signed-off-by: Zoltan HERPAI <[email protected]>

d1: 6.12: remove upstreamed patches and refresh remaining ones

Remove patches that have been upstreamed or dropped, and refresh
the remaining ones as required.

Signed-off-by: Zoltan HERPAI <[email protected]>

kernel/d1: Restore kernel files for v6.6

This is an automatically generated commit which aids following Kernel patch
history, as git will see the move and copy as a rename thus defeating the
purpose.

For the original discussion see:
https://lists.openwrt.org/pipermail/openwrt-devel/2023-October/041673.html

Signed-off-by: Zoltan HERPAI <[email protected]>