rk3528.dtsi and rk3528-pinctrl.dtsi are label-identical between 7.2 and
mainline v7.3-rc1 (checked directly against a sparse clone of
torvalds/linux at that tag); the only difference is a cosmetic 'sfc'
label that Armbian's own rk3528-02-arm64-dts-rockchip-Add-SFC-node
patch reintroduces anyway and that this board doesn't reference. The
one rk3528-specific patch touching ethernet PHY reset handling
(nanopi-zero2-fix-ethernet-phy-reset) matches the snps,reset-gpio
pattern this board's dts already uses. So the board dts carries over
unchanged from 7.2.
Verified: compiles cleanly against a sparse checkout of
torvalds/linux v7.3-rc1 (arch/arm64/boot/dts/rockchip +
include/dt-bindings); not yet build-tested against the real
Armbian-patched 7.3 tree (oras kernel-git-shallow-7.3 bundle isn't
published yet) and not wired into KERNEL_TARGET.
Same fix as the 6.12 counterpart: stable 6.18.49 backported the
DOUBLE_REPORT_ID recursion hardening (__magicmouse_raw_event() + 'nested'
flag), which was absent in 6.18.48. The T2 raw_event split in
4001-asahi-trackpad.patch then leaves the "if (nested) return 0;" guard in
the split-out magicmouse_raw_event_usb() (no such parameter), failing the
build with "'nested' undeclared".
Move the recursion guard into the input_ops dispatcher, as done for
uefi-x86-6.12 and uefi-x86-7.2 (#10575). kernel-x86-current builds 6.18.49
and would hit this once it reaches the compile step.
Verified by replaying 4001/4004/4005 over upstream hid-magicmouse.c at
v6.18.49: all apply with no rejects, 'nested' survives only inside the
dispatcher, and magicmouse_raw_event_usb has no undefined symbol.
Signed-off-by: Igor Pecovnik <igor@armbian.com>
kernel-x86-legacy (6.12.108) fails to build:
drivers/hid/hid-magicmouse.c:813:21: error: 'nested' undeclared
Stable 6.12.108 backported the DOUBLE_REPORT_ID recursion hardening: the
report parser became __magicmouse_raw_event() carrying a 'nested' flag,
wrapped by a thin magicmouse_raw_event(), and the wrapped-report branch
refuses to recurse a second time.
The T2 series (4001-asahi-trackpad.patch) splits that parser into a
per-bus handler plus an input_ops dispatcher. Applied unchanged onto
6.12.108 it cuts __magicmouse_raw_event() in the wrong place, leaving the
"if (nested) return 0;" guard behind in the split-out body, which is now
magicmouse_raw_event_usb() and has no such parameter - hence the build
error - while __magicmouse_raw_event() keeps 'nested' but discards it.
Rebase the split so the recursion guard moves into the dispatcher, the
single choke point every report passes through (entry with nested=false,
DOUBLE_REPORT_ID recursion with nested=true), checking
"nested && data[0] == DOUBLE_REPORT_ID" there with a size guard - i.e. the
upstream check one frame deeper. The handlers keep the 4-argument raw_event
signature that sub-patches 4004/4006 build on.
This is the 6.12 counterpart of the same fix already landed for uefi-x86-7.2
(#10575). Verified by replaying 4001/4004/4006 over upstream
hid-magicmouse.c at v6.12.108: all apply with no rejects, 'nested' survives
only inside the dispatcher, and magicmouse_raw_event_usb has no undefined
symbol.
Signed-off-by: Igor Pecovnik <igor@armbian.com>
board-firefly-rk3399-dts.patch re-assigns three RK808 LDOs away from
what upstream rk3399-firefly.dts has:
LDO_REG1 vcc1v8_dvp 1.8V -> vcca1v8_codec 1.8V
LDO_REG2 vcc2v8_dvp 2.8V -> vcca1v8_hdmi 1.8V
LDO_REG7 vcca1v8_codec 1.8V -> vcca0v9_hdmi 0.9V
That is the layout used by rk3399-roc-pc / rock-pi-4 / rock960, not by
the Firefly-RK3399. On this board LDO7 is the analog 1.8V supply of the
on-board Realtek ALC5640 (rt5640) codec, so driving it at 0.9V leaves
the codec unpowered. It never acks on I2C1 and probe fails with
rt5640 1-001c: Device with ID register 0x0 is not rt5640/39
so no analog sound card is registered at all: the headphone and
microphone jacks are dead. i2cdetect shows nothing on the whole bus.
Booting Firefly's own Ubuntu image (BSP 4.4 kernel, whose device tree
maps LDO7 to vcca1v8_codec at 1.8V) on the same board gives working
headphone audio, confirming which mapping matches the hardware.
Restore the upstream mapping and drop the two &hdmi avdd supplies that
were pointing at the now-misnamed regulators; upstream does not set
them on this board.
With LDO7 back at 1.8V the codec is detected, the rt5640 card comes up
with its full mixer (120 controls) and playback works at 44.1 and
48 kHz. Verified on a Firefly-RK3399 (4 GB) running Armbian 6.18.45.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Add some extra patches from upstream to fix some dmesg warnings.
Reorder legacy and current. Delete 6.1 as it's no longer used.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
- `1001-fdt_fixup_ethernet-add-logs.patch` is already in v2026.07's root, just drop
- move dt/u-boot-dt/defconfig (as-is)
- drop ramboot/471/472 bins as it's handled by `EXT=rkusboot` now
- fancy: LwIP, mbedTLS, wget
Linux 7.2 changes interfaces used by Q8B display, Iris, PCIe, and
thermal drivers. Refresh the stack, drop upstreamed or reverted patches,
and include tested vendor fixes plus the ADC thermal backport.
Signed-off-by: SuperKali <hello@superkali.me>
Pin edge to the stable v7.2.3 tag instead of following the moving
7.1.y branch. Regenerate the config against the final Q8B patch stack.
Signed-off-by: SuperKali <hello@superkali.me>
Drop 23 Waveshare patches already in Linux v7.2.3 and rebase the other 17.
Keep the downstream audio, A702 firmware, media carrier and DT overlay changes.
Signed-off-by: SuperKali <hello@superkali.me>
Move Arduino UNO Q from Linux 7.1 to v7.2.3 and refresh the edge config.
Pin the kernel tag so builds remain reproducible.
Signed-off-by: SuperKali <hello@superkali.me>
Pick up fifteenhex/xradio#29, which advertises MFP_CAPABLE for the
XR819. This lets mac80211 expose the management frame cipher suites
required for 802.11w Protected Management Frames.
Without this capability, wpa_supplicant cannot install the IGTK on
networks that require PMF. The existing timer_delete_sync compatibility
patch remains applicable because the upstream change only adds the
hardware capability flag.
Tested on an Orange Pi Zero with XR819 and Linux
6.18.48-current-sunxi. The device connects to a PMF-required network
and remains connected.
Signed-off-by: Mayur Panchal <mayur@canva.com>
Christoph König
Igor Pecovnik
vidplace7
nyanmisakaandvidplace7
JiaY-shi
Yaozhong LiandClaude Opus 5
EvilOlaf
Rosen Penev
igorpecovnik
Ricardo Pardini
SuperKali
dependabot[bot]
Mayur Panchal