Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@config/boards/ayn-odin3.csc`:
- Around line 84-90: Update the executable script and initramfs hook installs in
the board configuration from mode 655 to 755, including
usb-gadget-initramfs-hook, usb-gadget-initramfs-premount, dropbear, and
kill-dropbear, so the owner retains execute permission; leave the 644 service
install unchanged.
- Around line 32-34: Replace the fragile rootfs_image_uuid pipeline with
validated extraction from the actual fstab entry: locate a non-comment line
containing a UUID= field and extract its value without assuming the first line
is valid. Fail the build immediately with a clear error if no valid UUID is
found, before rootfs_image_uuid is used in the boot --cmdline.

In `@config/sources/families/sm8750.conf`:
- Around line 16-23: Add a `current)` clause to the `case $BRANCH` block
alongside `edge)` and initialize `KERNEL_MAJOR_MINOR`, `KERNELBRANCH`, and
`KERNEL_GIT_CACHE_TTL` with the appropriate current-kernel values so
`KERNEL_TEST_TARGET="current"` builds receive all required settings.

In `@packages/bsp/ayn-odin3/HiFi.conf`:
- Line 1: Update the header comment in HiFi.conf to identify the configuration
as intended for AYN Odin3 instead of AYN Odin2; leave the audio routing
configuration unchanged.

In `@packages/bsp/ayn-odin3/rocknix_abl/SM8750/flash_abl.sh`:
- Around line 5-6: Harden the ABL toolchain: in
packages/bsp/ayn-odin3/rocknix_abl/SM8750/flash_abl.sh (lines 5-6), enable
fail-fast handling and verify abl_signed-SM8750.elf with
abl_signed-SM8750.elf.sha256 before either abl_a or abl_b dd write; in
backup_abl.sh (lines 5-6) and restore_backup_abl.sh (lines 5-6), ensure every dd
failure terminates the script; update the README to document checksum
verification. The checksum file at
packages/bsp/ayn-odin3/rocknix_abl/SM8750/abl_signed-SM8750.elf.sha256 (line 1)
requires no direct change, but must be referenced by flash_abl.sh.

In `@packages/bsp/ayn-odin3/zz-update-abl-kernel`:
- Around line 3-11: Update the kernel artifact references in the postinst script
to use the ABI version passed as $1, selecting /boot/vmlinuz-$1,
/boot/initrd.img-$1, and the matching sm8750 DTB instead of wildcard paths.
Remove the unused source /boot/armbianEnv.txt statement while preserving the
existing gzip and mkbootimg flow.

In
`@patch/kernel/archive/sm8750-7.1/0028-drm-panel-Add-panel-driver-for-Chipone-ICNA35XX-base.patch`:
- Around line 121-141: Update icna35xx_get_current_mode() to check whether
pinfo->connector is NULL before accessing connector->state, returning the
default mode index 0 when no connector is available. Preserve the existing
checks and mode-matching behavior for initialized connectors.

In
`@patch/kernel/archive/sm8750-7.1/0030-leds-Add-driver-for-HEROIC-HTR3212.patch`:
- Around line 288-290: Update the sdb acquisition error handling in the probe
function around devm_gpiod_get so every error pointer is returned, while
preserving -EPROBE_DEFER propagation. Treat the mandatory sdb descriptor as
valid only when the call succeeds, preventing subsequent
gpiod_set_value_cansleep and regulator-related paths from receiving an ERR_PTR.

In
`@patch/kernel/archive/sm8750-7.1/0031_input--Add-driver-for-RSInput-Gamepad.patch`:
- Around line 483-492: Balance regulator enables on probe failures in both
drivers: update the RSInput probe flow after regulator_enable(drv->vdd) and the
HTR3212 probe flow after regulator_enable(priv->vdd) so every listed later error
path disables the supply before returning, preferably by using
devm_regulator_get_enable() where compatible. In the HTR3212 paths, also drive
sdb low before returning. Apply changes in
patch/kernel/archive/sm8750-7.1/0031_input--Add-driver-for-RSInput-Gamepad.patch
lines 483-492 and
patch/kernel/archive/sm8750-7.1/0030-leds-Add-driver-for-HEROIC-HTR3212.patch
lines 298-301; cover RSInput failures from serdev_device_open, baudrate setup,
input allocation/registration, and rsinput_init_commands, plus HTR3212 failures
from regmap initialization, htr3212_init_regs, and htr3212_parse_dt.
- Around line 298-370: Update handle_cmd_status to validate that payload_length
covers every field accessed through FRAME_POS_DATA_14 before reading status
data. Reject shorter payloads through the existing invalid-length path, while
preserving the current key and axis reporting behavior for complete frames.

In `@patch/kernel/archive/sm8750-7.1/0501-ROCKNIX-fix-wifi-and-bt-mac.patch`:
- Around line 116-141: Update qca_uart_setup so generate_bdaddr_from_serial
failure does not return early or disable Bluetooth; instead, preserve the
existing qca_check_bdaddr fallback path for unavailable or malformed
qcom_serial_number values. Only apply qca_set_bdaddr and update hdev addresses
after successful serial derivation, while retaining the current error handling
for failures setting a successfully generated address.

In `@patch/kernel/archive/sm8750-7.1/0503-ROCKNIX-battery-name.patch`:
- Around line 10-22: Revert the `.name` change in `sm8550_bat_psy_desc` so the
shared qcom_battmgr power supply continues using `qcom-battmgr-bat`; do not
rename it to `battery`. If a board-specific name is required, implement it
through a separate board-scoped path rather than modifying this shared
descriptor.

In
`@patch/kernel/archive/sm8750-7.1/0504-Enable-64-bit-processes-to-use-compat-input-syscalls.patch`:
- Around line 44-52: Update the condition in input_ff_effect_from_user() to
select the compat ff_effect_compat conversion whenever current->compat_input or
in_compat_syscall() is true, removing the !COMPAT_USE_64BIT_TIME gate. Preserve
the existing size validation and conversion logic.

---

Nitpick comments:
In `@config/boards/ayn-odin3.csc`:
- Around line 65-73: Update the curl invocation in the ALSA archive download
block to include the fail-on-HTTP-error option while preserving the existing
redirect-following and output behavior.
- Around line 36-40: Remove the dead commented-out wget line and make the
mkbootimg patch step validate that the expected import exists before modifying
/tmp/mkbootimg; fail immediately with a clear error if the installed script no
longer contains that exact line, while preserving the existing copy and
executable-permission flow.
- Around line 46-52: The boot cmdline is duplicated and does not consume
BOOTIMG_CMDLINE_EXTRA, allowing the build image and kernel-update hook to
diverge. In config/boards/ayn-odin3.csc lines 46-52, derive the --cmdline from
BOOTIMG_CMDLINE_EXTRA or a shared helper instead of hardcoding it; in
packages/bsp/ayn-odin3/zz-update-abl-kernel lines 12-19, source that same
generated value through the existing armbianEnv.txt mechanism or equivalent
rather than duplicating the string.

In `@patch/kernel/archive/sm8750-7.1/0500-ROCKNIX-set-boot-fanspeed.patch`:
- Line 1: Document the coupling between the boot PWM value and cooling-levels
table: in the boot fan-speed change from 0500, note that 70 corresponds to
cooling-levels index 4 defined by 0601; in the table change from 0601, note that
index 4’s value of 70 is required by the hardcoded boot-time PWM in 0500.
Preserve the existing values and behavior.
- Around line 1-13: Replace the hardcoded 70 passed to set_pwm in pwm_fan_probe
with a documented, named value derived from the intended cooling-levels boot
state, such as the existing state-index-4 value. Keep the generic driver
behavior unchanged for platforms without that configuration and ensure future
cooling-levels changes cannot silently desynchronize the boot PWM.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In
`@patch/kernel/archive/sm8750-7.1/0030-leds-Add-driver-for-HEROIC-HTR3212.patch`:
- Around line 283-284: Update htr3212_parse_dt() to reject child counts greater
than HTR3212_CHANNELS before indexing priv->leds, and allocate the driver state
with sizeof(*priv) rather than struct_size using the unbounded count. Preserve
normal parsing for valid counts and return an appropriate error for oversized
device trees.

---

Nitpick comments:
In
`@patch/kernel/archive/sm8750-7.1/0030-leds-Add-driver-for-HEROIC-HTR3212.patch`:
- Around line 157-158: Separate the enable-bit mask from the register-channel
divisor in the initialization using distinct constants or macros. Update the
GENMASK expression to use the enable-width symbol and the num_regs calculation
to use an explicit channels-per-control-register value, preserving the resulting
mask and register count.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In
`@patch/kernel/archive/sm8750-7.1/0031_input--Add-driver-for-RSInput-Gamepad.patch`:
- Around line 300-365: Move prev_states from function-local static storage into
the per-device struct rsinput_driver, then update the button-state comparison
and assignment in the report path to use drv->prev_states. Keep button
transition tracking independent for each RSInput device.

---

Outside diff comments:
In
`@patch/kernel/archive/sm8750-7.1/0031_input--Add-driver-for-RSInput-Gamepad.patch`:
- Around line 410-440: The rsinput_rx handler must reassemble the UART stream
instead of treating each callback as one frame: append incoming bytes to
persistent RX state, process and checksum each complete frame individually,
retain incomplete trailing bytes for the next callback, and handle overflow
without discarding valid buffered data. Update rsinput_rx and the driver’s
RX-buffer state accordingly, and return count to report all received bytes
consumed.
- Around line 494-513: Update the GPIO acquisition checks in the probe path for
boot_gpio, enable_gpio, and reset_gpio to detect ERR_PTR values and immediately
return the corresponding PTR_ERR result, preserving -EPROBE_DEFER. Keep absent
optional GPIOs as NULL, and only continue to initialization and later
gpiod_set_value_cansleep calls after all acquisitions succeed.