patch maintenance: rewrite various patch sets against their current major.minor.patch release - #10329
patch maintenance: rewrite various patch sets against their current major.minor.patch release#10329EvilOlaf wants to merge 7 commits into
Conversation
giving it one last cleanup
📝 WalkthroughWalkthroughThis PR updates Meson, Rockchip, and Allwinner kernel patch archives. It adds hardware drivers and device-tree support, adjusts USB, PCIe, GPU, media, power, and networking behavior, and refreshes patches for kernel 6.18 and 7.1. ChangesMeson platform patches
Rockchip platform patches
Allwinner platform patches
Estimated code review effort: 5 (Critical) | ~120 minutes Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
github-actions
Bot
added
Needs review
EvilOlaf
changed the title
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (44)
patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch (1)
81-85: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winRelease the
pc->of_irqreference in both driver variants.Both patches store
of_find_compatible_node()results inpc->of_irqbut do not runof_node_put()or a device-managed OF cleanup in probe failure or remove paths, so the node reference remains leaked.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch` around lines 81 - 85, Release the OF node reference stored in pc->of_irq for both driver variants by adding cleanup in probe-failure and remove paths, or using device-managed OF cleanup consistently. Apply the corresponding fix in patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch at lines 81-85 and patch/kernel/archive/meson64-7.1/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch at lines 81-85, covering both amlogic,meson-gpio-intc and amlogic,meson-gpio-intc-ext lookup results.patch/kernel/archive/sunxi-6.18/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch (3)
210-215: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick winKeep the MCP2515 SPI clock at or below 10 MHz.
spi-max-frequency = <12500000>exceeds the MCP2515 10 MHz SPI clock limit. Use10000000or a lower value.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch` around lines 210 - 215, Update the MCP2515 device node’s spi-max-frequency property to 10000000 or a lower value, ensuring the configured SPI clock does not exceed the MCP2515’s 10 MHz limit.Source: MCP tools
196-220: 🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick winDeclare the MCP2515 interrupt.
Add
interruptsto thecannode so the SPI GPIO interrupt has a defined source and trigger type, as required by the MCP2515 device-tree binding. Ifcan0_pin_irqconfigures that GPIO, add its concrete pin selector too.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch` around lines 196 - 220, Add the required interrupt declaration to the MCP2515 `can` node, using the GPIO controller and trigger configuration represented by `can0_pin_irq`; include the concrete pin selector if that pinctrl entry does not already define it. Keep the existing `interrupt-parent` and MCP2515 node configuration unchanged.Source: MCP tools
146-150: 🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick winUse a hyphen in the node name.
mcp2515_clockis used as the node name, but labels may keep underscores while DTS node names should use only lowercase letters, digits, and dashes. Rename the node tomcp2515-clock; references to the label&mcp2515_clockdo not need to change.Proposed fix
-mcp2515_clock: mcp2515_clock { +mcp2515_clock: mcp2515-clock {🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch` around lines 146 - 150, Rename the fixed-clock node declared alongside the mcp2515_clock label from mcp2515_clock to mcp2515-clock, while preserving the mcp2515_clock label and all references to it.Source: MCP tools
patch/kernel/archive/sunxi-6.18/patches.armbian/drv-staging-media-sunxi-cedrus-add-H616-variant.patch (1)
19-38: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winAssociate H616 with
sun50i_h616_cedrus_variant.The patch defines an H616-specific variant with a 600 MHz module rate, but Line 37 selects
sun50i_h6_cedrus_variant. H616 therefore uses the wrong variant data and the new definition is unused.Proposed fix
{ .compatible = "allwinner,sun50i-h616-video-engine", - .data = &sun50i_h6_cedrus_variant, + .data = &sun50i_h616_cedrus_variant, },🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/drv-staging-media-sunxi-cedrus-add-H616-variant.patch` around lines 19 - 38, Update the H616 entry in cedrus_dt_match to reference the newly defined sun50i_h616_cedrus_variant instead of sun50i_h6_cedrus_variant, ensuring H616 uses its 600 MHz variant data.patch/kernel/archive/sunxi-7.0/patches.armbian/0997-i2c-mv64xxx-pin-runtime-pm.patch (1)
48-51: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winPropagate failed runtime resumes out of both probes.
The I2C and MMC patches call
pm_runtime_resume_and_get()and then continue after a negative return value. Return the probe error path instead so device probe does not proceed when runtime resume cannot activate the controller; apply the same behavior to both0997-i2c-mv64xxx-pin-runtime-pm.patchand0998-sunxi-mmc-pin-runtime-pm.patch.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-7.0/patches.armbian/0997-i2c-mv64xxx-pin-runtime-pm.patch` around lines 48 - 51, Update the probe paths calling pm_runtime_resume_and_get() to return the negative error immediately instead of only warning and continuing. Apply this in patch/kernel/archive/sunxi-7.0/patches.armbian/0997-i2c-mv64xxx-pin-runtime-pm.patch lines 48-51 and patch/kernel/archive/sunxi-7.0/patches.armbian/0998-sunxi-mmc-pin-runtime-pm.patch lines 48-51, preserving successful probe behavior when the resume succeeds.patch/kernel/archive/sunxi-6.18/patches.armbian/drv-staging-rtl8723bs-AP-bugfix.patch (1)
14-22: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winInitialize the complete
station_infostructure before callingcfg80211_new_sta().
struct station_infocontains pointer fields such aspertidandgeneration; sincesinfois stack allocated,cfg80211_new_sta()can read uninitialized values. Initialize the whole structure instead of clearing onlypertid. The same uninitializedsinfopattern should be applied to bothsunxi-6.18andsunxi-7.0copies of this patch.Proposed fix
- struct station_info sinfo; + struct station_info sinfo = {}; ... - sinfo.pertid = 0;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/drv-staging-rtl8723bs-AP-bugfix.patch` around lines 14 - 22, In rtw_cfg80211_indicate_sta_assoc, zero-initialize the entire stack-allocated station_info structure before assigning its fields and calling cfg80211_new_sta(), rather than initializing only pertid. Apply the same correction to both the sunxi-6.18 and sunxi-7.0 copies of this patch.patch/kernel/archive/sunxi-6.18/patches.armbian/drv-bluetooth-hci-sprd-broken-park-link-quirk-v6.16-plus.patch (1)
22-26: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick winQualify the
ttyBTmatch to Spreadtrum hardware.
hdev->manufacturer == 0xffffcan apply to other UART HCI devices that do not report a specific manufacturer, and this condition does not exclude a non-sprd BT driver from the scope of the broken Park link quirk. Tighten the match to the known SPRD Uwe5622/sprd path before applying the quirk.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.armbian/drv-bluetooth-hci-sprd-broken-park-link-quirk-v6.16-plus.patch` around lines 22 - 26, The condition applying HCI_QUIRK_BROKEN_PARK_LINK_STATUS must restrict the ttyBT match to the known Spreadtrum Uwe5622/sprd driver path, not merely manufacturer 0xffff. Update the quirk logic around hci_set_quirk to validate the appropriate SPRD driver identifier before applying it, while preserving the existing quirk behavior for matching Spreadtrum devices.patch/kernel/archive/sunxi-7.0/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch (1)
196-221: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick winAdd the MCP2515 interrupt binding.
Line 197 defines
can0_pin_irqwithoutpins, so it cannot configure an IRQ pin. Lines 210-221 definemcp2515@0without aninterruptsproperty. When a board enables this node, the driver cannot request its IRQ and cannot operate.Set the actual GPIO in
can0_pin_irq. Add matchinginterrupt-parentandinterruptsproperties tomcp2515@0.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-7.0/patches.armbian/arm64-dts-sun50i-h616-bigtreetech-cb1.patch` around lines 196 - 221, Update the can0_pin_irq pinctrl node to specify the actual GPIO through its pins property, then update the mcp2515@0 node to declare the matching interrupt-parent and interrupts properties so the MCP2515 driver can request the configured IRQ.patch/kernel/archive/sunxi-7.0/patches.armbian/drv-staging-media-sunxi-cedrus-add-H616-variant.patch (1)
35-38: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winBind H616 to its declared Cedrus variant.
Line 37 maps
allwinner,sun50i-h616-video-enginetosun50i_h6_cedrus_variant. This leavessun50i_h616_cedrus_variantunreachable. H616 then receives the H6 capability set and, after the H6 clock-rate patch, a 648 MHz module rate instead of its declared 600 MHz rate.Change
.datato&sun50i_h616_cedrus_variant.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-7.0/patches.armbian/drv-staging-media-sunxi-cedrus-add-H616-variant.patch` around lines 35 - 38, Update the compatible entry for allwinner,sun50i-h616-video-engine to reference sun50i_h616_cedrus_variant via its .data field instead of sun50i_h6_cedrus_variant, ensuring H616 uses its declared capabilities and clock rate.patch/kernel/archive/sunxi-6.18/patches.megous/goodix-6.18/0002-Input-goodix-Add-option-to-power-off-the-controller-.patch (1)
58-81: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winRestore regulator state on resume failure.
If
vddioenable,goodix_reset(), orgoodix_request_irq()fails, this path returns withavdd28enabled. The later failure paths can also leave both regulators enabled.Add cleanup labels that disable the regulators enabled by this resume attempt before returning the error. This keeps the failure state consistent with the power-off suspend state.
Proposed fix
error = regulator_enable(ts->vddio); if (error) { dev_err(dev, "Regulator vddio enable failed.\n"); - return error; + goto disable_avdd28; } error = goodix_reset(ts); if (error) { dev_err(dev, "Controller reset failed.\n"); - return error; + goto disable_regulators; } error = goodix_request_irq(ts); if (error) - return error; + goto disable_regulators; return 0; + +disable_regulators: + regulator_disable(ts->vddio); +disable_avdd28: + regulator_disable(ts->avdd28); + return error;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/sunxi-6.18/patches.megous/goodix-6.18/0002-Input-goodix-Add-option-to-power-off-the-controller-.patch` around lines 58 - 81, Update the poweroff-on-resume branch around goodix_reset() and goodix_request_irq() to unwind regulators on every failure: disable vddio when its subsequent operations fail, and disable avdd28 as well when vddio enable or later steps fail. Use cleanup labels that preserve and return the original error, while keeping the successful path unchanged.patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-02-core.patch (3)
307-314: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winRestore the user's LED switch state after resume.
mcu_joypad_leds_suspend()setsled_switchto zero, but resume unconditionally sets it to one. A user who disabled the LEDs will have them enabled after resume.Save the pre-suspend value and restore it during resume.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-02-core.patch` around lines 307 - 314, Update the MCU joypad suspend/resume flow around mcu_joypad_leds_suspend() and the shown resume block to save the pre-suspend led_switch value before clearing it, then restore that saved value during resume instead of unconditionally assigning one; keep led_dirty set so the restored state is transmitted.
99-116: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winHonor each ring's brightness setting.
mcu_joypad_encode_leds()reads brightness only frommc_leds[0]and uses it to scale both rings. The registration code creates two independent multicolor LEDs (left::joystick-leftandright::joystick-right), and setting either ring's brightness triggersmcu_led_mc_set()without storing the caller-supplied value.Encode per-ring brightness if the MCU protocol supports it. Otherwise use one shared brightness control for both rings.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-02-core.patch` around lines 99 - 116, Update mcu_led_mc_set() and mcu_joypad_encode_leds() so brightness changes are stored and applied independently for each registered ring, using each ring’s led_cdev.brightness when scaling its RGB components. If the MCU protocol only supports a single brightness value, replace the independent controls with one shared brightness control and ensure both rings use it consistently.
251-254: 🚀 Performance & Scalability | 🟠 Major | ⚡ Quick winReject zero or too-small polling intervals.
poll-interval-ms = 0becomes a zero-delay workqueue item; becausemcu_joypad_poll()reschedules itself after each transfer, this can create tight polling. The same value also controls the calibrationmsleep()delays. Validate the property before storing it: reject0, enforce a reasonable minimum, and apply an upper bound if the protocol requires it.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-02-core.patch` around lines 251 - 254, Validate the poll-interval-ms value in the initialization flow before assigning it to mjp->poll_interval_ms: reject zero and values below the driver’s reasonable minimum, and cap values at the protocol’s required maximum if one exists. Preserve MCU_DEFAULT_POLL_MS when the property is absent or invalid, and ensure the validated value is used for both polling and calibration delays.patch/kernel/archive/rockchip64-6.18/general-driver-tm16xx-led-driver.patch (3)
541-585: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winFree all direct-attach state on failure.
linedisp_attach()can allocatelinedisp->mapandlinedisp->message, butout_free_buffrees onlylinedisp->buf. A failedcreate_attachment(),linedisp_display(), ordevice_add_groups()therefore leaks state.Use the same cleanup as
linedisp_detach()on every failure path.Proposed cleanup
out_free_buf: + kfree(linedisp->map); + kfree(linedisp->message); kfree(linedisp->buf);🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/general-driver-tm16xx-led-driver.patch` around lines 541 - 585, Update the failure cleanup in linedisp_attach() to perform the same direct-attach state cleanup as linedisp_detach(), including freeing linedisp->map and linedisp->message along with linedisp->buf. Ensure every failure path through out_del_attach, out_del_timer, and out_free_buf releases all allocated state without duplicating or skipping cleanup.
594-608: 🩺 Stability & Availability | 🔴 Critical | ⚡ Quick winPreserve attachment lifetime until all consumers release.
Both cleanup paths free the
linedisp_attachmentwhile sysfs/release paths still need it.
linedisp_detach(): removelinedisp_groupsbefore callingdelete_attachment(), so sysfs callbacks can still lookuplinedispwhile group removal is in flight.linedisp_unregister(): do not remove the false direct attachment before the lastdevice_put()runs. Either keep the mapping until release, or usecontainer_of()inlinedisp_release()and attribute callbacks for child devices.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/general-driver-tm16xx-led-driver.patch` around lines 594 - 608, Update linedisp_detach() to remove linedisp_groups before calling delete_attachment(), keeping the linedisp mapping available while sysfs callbacks complete. In linedisp_unregister() at patch/kernel/archive/rockchip64-6.18/general-driver-tm16xx-led-driver.patch:656-661, retain the false direct attachment until the final device_put() invokes linedisp_release(), or instead make linedisp_release() and child-device attribute callbacks obtain the object with container_of(); do not free the attachment while those consumers may still access it.
1788-1808: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winUse
reg[0]as the digit array index.
tm16xx_display_value()writesbuf[i]todisplay->digits[i], while parsing stores digit nodes in DT enumeration order and only reads theirregproperty. Nodes with non-sequential, out-of-order, missing, or duplicateregvalues can cause characters to display on the wrong digits or be rejected instead of mapped correctly.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/general-driver-tm16xx-led-driver.patch` around lines 1788 - 1808, The digit parsing loop around tm16xx_display_value must index display->digits using each child’s parsed reg[0] value rather than the enumeration counter i. Validate that the index is within the digit array bounds and reject duplicate indices, while preserving the existing segment mapping and hardware maximum calculations.patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch (6)
1353-1357: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winCheck the chip-ID read error before masking it.
sgm4154x_hw_chipid_detect()returns either a register value or a negative errno, but probe appliesSGM4154x_PN_MASKto both. For example,-ENODEVcan produce the expected masked bits and allow probe to continue with an unusable device.Return negative errors before comparing the part ID.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 1353 - 1357, Update the chip-ID handling after sgm4154x_hw_chipid_detect() to return immediately when ret is negative, before applying SGM4154x_PN_MASK; only mask and compare successful register values against SGM4154x_PN_ID, preserving the existing -ENODEV path for mismatched IDs.
1073-1083: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winPropagate initialization write failures.
The three
regmap_update_bits()calls in this block ignore return values, so probe can succeed while charger interrupt and control configuration remain unmodified. Store each result and jump toerr_outon failure.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 1073 - 1083, Update the initialization block containing the three regmap_update_bits() calls to capture each return value and immediately jump to err_out when any call fails, ensuring probe does not continue with incomplete charger configuration.
1258-1279: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winPropagate OTG regulator registration failures.
sgm4154x_vbus_regulator_register()can return-ENXIOordevm_regulator_register()errors, but probe ignores the return value. A successful probe can then complete without exposing the regulator required byvbus5v0_typecconsumers. Check the return value and callof_node_put(np)on theregulatorsnode when registering the OTG regulator.Also applies to line 1409.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 1258 - 1279, Update the probe path calling sgm4154x_vbus_regulator_register() to check and propagate its return value, preventing successful probe without the required OTG regulator; apply the same handling at the additional call site near line 1409. In sgm4154x_vbus_regulator_register(), call of_node_put(np) before returning after the regulator registration attempt, including error paths, while preserving the existing -ENXIO and devm_regulator_register() error codes.
892-923: 🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy liftSerialize charger state and register updates.
s->lockonly protectssgm4154x_charger_get_property().sgm4154x_irq_handler_thread(),sgm_charger_work(), regulator callbacks,sgm4154x_charger_set_property(), andsgm4154x_hw_init()all read/writesgm->stateor charger registers outside that lock. Wrap the full read-modify-write sequences with one defined lock to avoid overlapping charger enable/input-limiting/OTG register writes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 892 - 923, Define one shared lock in the sgm4154x device and use it to serialize all charger state and register access. Update sgm4154x_irq_handler_thread, sgm_charger_work, regulator callbacks, sgm4154x_charger_set_property, and sgm4154x_hw_init to hold the lock across each complete read-modify-write sequence, while retaining protection in sgm4154x_charger_get_property; ensure charger enable, input-current limiting, and OTG updates cannot overlap.
859-862: 🩺 Stability & Availability | 🔴 Critical | ⚡ Quick winRemove or manage the manual
registerssysfs attribute.
device_create_file()is not cleaned up by the driver, andsgm4154x_charger_remove()does not calldevice_remove_file(). Theregisters_show/registers_storecallbacks still dereference the devm-freedsgm4154x_device, so driver unbind can leave a dangling sysfs access.Use managed group creation, or add
device_remove_file()to every failure/unbind/shutdown path. Also check the creation return value at the call sites.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 859 - 862, Replace the unmanaged device_create_file call in sgm4154x_create_device_node with a managed sysfs attribute group, or explicitly track and remove the attribute in every probe failure, sgm4154x_charger_remove, and shutdown path. Propagate and handle creation errors at all call sites so registers_show and registers_store cannot access a freed sgm4154x_device.
1361-1369: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winBalance IRQ wake setup.
enable_irq_wake()is called inside the existing IRQ allocation block, so this does not add a new setup point at lines 1430-1437. Fix the currentenable_irq_wake()call by checking its return value and callingdisable_irq_wake()on failed probe or during remove/error cleanup.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-sgm41542.patch` around lines 1361 - 1369, Update the IRQ setup and cleanup paths around sgm4154x_irq_handler_thread to check and propagate enable_irq_wake() failures, and balance successful wake enabling with disable_irq_wake() during probe error cleanup and device removal. Ensure cleanup only disables wake when it was successfully enabled.patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-03-input.patch (2)
356-386: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winRequire a complete and valid calibration range.
has_dt_rangebecomes true when any one of the four properties is present. If a device tree supplies onlyabs-x-min, the missing endpoints remain zero but the code skips the default range.Require all four endpoints before setting
has_dt_range, and validate that each minimum is below its maximum.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-03-input.patch` around lines 356 - 386, Update the calibration initialization around left_cal and right_cal so has_dt_range is true only when all four corresponding device properties are present and each minimum is strictly less than its maximum. If any property is missing or any range is invalid, set has_dt_range false and initialize all endpoints to MCU_RAW_MIN and MCU_RAW_MAX as the existing fallback does.
76-81: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winRescale the radial deadzone to
MCU_AXIS_MAX.
apply_radial_deadzone()maps[deadzone..max]to[0..max], so full calibrated stick deflection never reachesMCU_AXIS_MAX. For example, with a3000deadzone,32767maps to29767. Scale the output withMCU_AXIS_MAXinstead, and keep thedeadzone == MCU_AXIS_MAXguard safe.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-03-input.patch` around lines 76 - 81, Update apply_radial_deadzone() so the post-deadzone magnitude is rescaled to MCU_AXIS_MAX rather than leaving the maximum at max minus deadzone. Preserve the existing coordinate direction and integer arithmetic, and ensure the deadzone == MCU_AXIS_MAX case remains safely guarded against invalid division or negative scaling.patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch (5)
848-851: 🗄️ Data Integrity & Integration | 🟠 Major | 🏗️ Heavy liftDo not report a fixed 10 Ah battery capacity.
POWER_SUPPLY_PROP_CHARGE_FULLreports10 * 1000 * 1000uAh for every battery. This makes capacity and energy calculations incorrect for boards with a different battery profile, including the stated Anbernic use case.Read the capacity from battery information/profile data, or remove the unsupported property.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch` around lines 848 - 851, Update the CW221X power-supply property handler for POWER_SUPPLY_PROP_CHARGE_FULL and POWER_SUPPLY_PROP_CHARGE_FULL_DESIGN to stop returning the hardcoded 10 Ah value. Use the battery profile/configured capacity data when available; otherwise remove or reject these unsupported properties rather than reporting a fabricated capacity.
777-797: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winMake SoC notification state per device.
static int socis shared by every CW221x instance and survives device removal. One gauge can suppresspower_supply_changed()notifications for another gauge.Store the previous SoC in
struct cw_batteryand initialize it per instance.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch` around lines 777 - 797, Update cw_bat_work to use a previous-SoC field on struct cw_battery instead of the shared static soc variable, and initialize that field when each battery instance is created so its first comparison is well-defined. Keep power_supply_changed notification behavior based on changes to that per-device state.
858-865: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winReturn the read error from power-supply get_property.
cw_get_voltage()andcw_get_current()already return negative values on I2C read failures, butcw_battery_get_property()discards them and returns success with stale cached values forPOWER_SUPPLY_PROP_VOLTAGE_NOWandPOWER_SUPPLY_PROP_CURRENT_NOW. Store the return value and return it for these cases.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch` around lines 858 - 865, Update cw_battery_get_property() so the POWER_SUPPLY_PROP_VOLTAGE_NOW and POWER_SUPPLY_PROP_CURRENT_NOW cases capture the return values from cw_get_voltage() and cw_get_current(), respectively, and immediately return any negative read error instead of using stale cached values; retain the existing value conversion and successful return behavior.
561-574: 🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy liftSerialize gauge I/O with cached property reads.
cw_battery_get_property()readscw_batfields directly, and properties likePOWER_SUPPLY_PROP_VOLTAGE_NOWandPOWER_SUPPLY_PROP_CURRENT_NOWalso call refresh I/O whilecw_update_data()does the same from delayed work. Add one per-device mutex that protects bothcw_update_data()I/O/cache writes and property refresh reads, so the battery cannot expose a mixed partial state.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch` around lines 561 - 574, Update the cw_battery synchronization by adding a per-device mutex to struct cw_battery and using it to serialize all cw_update_data() gauge I/O and cached-field writes with cw_battery_get_property() refreshes and reads, including VOLTAGE_NOW and CURRENT_NOW. Initialize and destroy the mutex with the device lifecycle, and ensure every relevant access path locks and unlocks it without exposing a mixed partial state.
969-985: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winWait for battery delayed work before suspend completes.
cw_bat_work()callscw_update_data()and then requeues itself withqueue_delayed_work().cancel_delayed_work()does not wait for the current work execution, so I2C reads can continue after suspend starts. Usecancel_delayed_work_sync()and add a suspended state check incw_bat_work()so it cannot requeue while suspended.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-power-supply-add-cw221x.patch` around lines 969 - 985, Update cw_bat_suspend to use cancel_delayed_work_sync so any running battery_delay_work completes before suspend returns, and add a suspended-state guard in cw_bat_work that prevents requeueing while suspended. Ensure the suspend/resume state is set consistently so cw_bat_resume can restart the delayed work without allowing it to reschedule during suspension.patch/kernel/archive/rockchip64-6.18/general-disable-mtu-validation.patch (1)
21-46: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winPreserve the 16 KiB MTU ceiling.
This patch removes the independent
BUF_SIZE_16KiBlimit, sostmmac_change_mtu()can accept MTUs above the stmmac DMA hardware limit. Keep that upper bound in place, or make all affected kernel versions enforce the same ceiling elsewhere.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/general-disable-mtu-validation.patch` around lines 21 - 46, Update stmmac_change_mtu() to retain the BUF_SIZE_16KiB upper-bound validation while removing or bypassing only the FIFO-size check. Ensure MTUs above the 16 KiB DMA limit still return -EINVAL, consistently across all affected kernel versions.patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-04-leds.patch (1)
217-224: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winHandle suspend transfer failures.
spi_sync_transfer()can return a negative error code, and this path ignores it aftermcu_joypad_encode_leds()marks the LED update as dirty. A failed suspend transfer can prevent the MCU from entering the LED-off suspend state, with no return value, retry, or diagnosis. Log or return the error, and keepled_dirtyuntil the MCU update completes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-input-add-spi-mcu-joypad-04-leds.patch` around lines 217 - 224, Update the suspend LED-off flow around mcu_joypad_encode_leds and spi_sync_transfer to capture and handle the transfer’s negative return value. Preserve led_dirty on failure so the update can be retried, and provide appropriate error reporting since this path has no return value; clear the dirty state only after a successful MCU update.patch/kernel/archive/rockchip64-6.18/general-drm-panel-add-yixian-yx0345-panel.patch (1)
196-208: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winReturn accumulated DSI initialization failures.
mipi_dsi_dcs_write_buffer_multi()writes failures intomctx.accum_err, soyixian_yx0345_prepare()can fail while the panel enable path is left active. Returnmctx.accum_errat the end of the sequence.Proposed fix
- return 0; + return mctx.accum_err;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/general-drm-panel-add-yixian-yx0345-panel.patch` around lines 196 - 208, Update yixian_yx0345_prepare() to return mctx.accum_err after completing the initialization command loop and final delay, instead of unconditionally returning 0. Preserve the existing command execution and delay behavior.patch/kernel/archive/rockchip64-7.1/rk3399-usbc-Revert-usb-typec-tcpm-unregister-existing-source-cap.patch (1)
16-36: 🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick winKeep the partner source capability un-registration.
tcpm_register_source_caps()can run when source PDOs or negotiated revisions change. Keep the existingcap = port->partner_source_capslogic, unregister it withusb_power_delivery_unregister_capabilities(cap)when it is present, clearport->partner_source_caps, then assign the replacement fromusb_power_delivery_register_capabilities(). The 6.18 and 7.2 archive patches keep the same lifecycle.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3399-usbc-Revert-usb-typec-tcpm-unregister-existing-source-cap.patch` around lines 16 - 36, Restore the existing partner source capability lifecycle in tcpm_register_source_caps(): initialize cap from port->partner_source_caps, unregister it when present, and clear port->partner_source_caps before registering the replacement. Keep assigning the result of usb_power_delivery_register_capabilities() to cap and preserve its error handling.Source: MCP tools
patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch (6)
1564-1568: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winBalance the devfreq event enable.
devfreq_event_enable_edev(data->edev)runs beforerk3328_dmc_init()andrk3328_devfreq_init(). Later probe failures return directly, andrk3328_dmcfreq_remove()only callsrk3328_devfreq_fini(), so the event can remain enabled.Disable
data->edevon failed probe paths and inrk3328_dmcfreq_remove()to match the enable.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 1564 - 1568, Balance the enable in the probe flow by disabling data->edev on every later failure path after devfreq_event_enable_edev succeeds, including failures from rk3328_dmc_init() and rk3328_devfreq_init(). Also update rk3328_dmcfreq_remove() to call devfreq_event_disable_edev(data->edev) alongside rk3328_devfreq_fini(), while preserving the existing return handling.
355-363: 🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick winEnable the DMC timing phandle before adding board-side status.
Add
ddr_timing = <&ddr_timing>;here too, or every enabled board override.of_get_rk3328_timings()reads that phandle and continues with firmware initialization even when it is missing or incomplete, so the generic DMC node can leave timing data unavailable for rockchip64-7.1 boards that enable DMC in DTS.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 355 - 363, Add the ddr_timing phandle to the rk3328 DMC device node alongside operating-points-v2, referencing the existing ddr_timing node so of_get_rk3328_timings() receives timing data when boards enable the generic DMC node.
1560-1563: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winReturn the actual
devfreq_event_get_edev_by_phandle()error.
devfreq_event_get_edev_by_phandle()can fail with permanent errors such as-EINVALor-ENODEV, but this probe currently treats every failure as-EPROBE_DEFER. ReturnPTR_ERR(data->edev)and defer only when the error is-EPROBE_DEFER.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 1560 - 1563, Update the error handling after devfreq_event_get_edev_by_phandle() in the probe path to capture PTR_ERR(data->edev), return -EPROBE_DEFER only when that error matches -EPROBE_DEFER, and otherwise return the original error.
1500-1509: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick winInitialize
rdev->ratewith the selected OPP frequency.
rk3328_devfreq_init()selectscur_freqfrom the OPP table, butdmcfreq->rateis only assigned from*freqinrk3328_dmcfreq_target(). Sinceget_cur_freq()andget_dev_status()return that cached rate, expose the selected frequency before devfreq registration so active DMC frequency queries do not return zero.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 1500 - 1509, Update rk3328_devfreq_init() after devfreq_recommended_opp() successfully selects cur_freq to initialize the DMC device's cached rate (rdev->rate, or the corresponding dmcfreq rate field) with cur_freq before devfreq registration. Preserve the existing OPP release and error handling, and ensure get_cur_freq() and get_dev_status() observe the selected initial frequency instead of zero.
461-468: 🩺 Stability & Availability | 🔴 Critical | ⚡ Quick winValidate every
ioremap()result before use.
rockchip_ddrclk_data_init()setsinited_flageven ifioremap()returnsNULL, so the v2 clock callbacks can dereferenceddr_data.share_memorybefore the mapping exists.rk3328_dmc_init()also passesddr_psci_paramtoof_get_rk3328_timings()immediately after mapping. Check each mapping, return-ENOMEMon failure, and unmap it on later initialization failure or inrk3328_dmcfreq_remove().🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 461 - 468, Validate the ioremap result in rockchip_ddrclk_data_init before assigning ddr_data.inited_flag, returning -ENOMEM when the mapping fails. In rk3328_dmc_init, validate ddr_psci_param before passing it to of_get_rk3328_timings, and unmap any successful mappings when later initialization fails. Also release the mappings in rk3328_dmcfreq_remove.
659-668: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winKeep unsupported DDR types as errors.
rockchip_dfi_ddrtype_to_ctrl()is shared by the existing Rockchip DFI path. Its default case now converts every unsupporteddfi->ddr_typeto controller register value 0 and returns success. Keep the upstream behavior and return-EOPNOTSUPP, or apply a fallback only in the RK3228/RK3328-specific init path.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/rk3328-add-dmc-driver.patch` around lines 659 - 668, Update rockchip_dfi_ddrtype_to_ctrl() so its default case continues logging unsupported memory types and returning -EOPNOTSUPP instead of assigning *ctrl = 0 and succeeding. If RK3228/RK3328 requires a fallback, implement it only in that platform-specific initialization path.Source: MCP tools
patch/kernel/archive/rockchip64-6.18/rk3399-usbc-phy-rockchip-naneng-Add-fallback-for-old-DTs.patch (2)
193-217: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winPreserve the detected host role.
At Line 214, the non-quirk branch assigns
DWC3_GCTL_PRTCAP_DEVICEafter the code has already selectedDWC3_GCTL_PRTCAP_HOST. Whenusb3_phy_reset_quirkis false, every host notification is forced into device mode.Set device mode only when
modeis not alreadyDWC3_GCTL_PRTCAP_HOST.Proposed fix
- } else { + } else if (mode != DWC3_GCTL_PRTCAP_HOST) { mode = DWC3_GCTL_PRTCAP_DEVICE; }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/rk3399-usbc-phy-rockchip-naneng-Add-fallback-for-old-DTs.patch` around lines 193 - 217, Update the non-quirk branch in the extcon handling around dwc3_set_mode so it assigns DWC3_GCTL_PRTCAP_DEVICE only when mode is not already DWC3_GCTL_PRTCAP_HOST. Preserve the host mode selected from EXTCON_USB_HOST and keep the existing quirk-specific USB state handling unchanged.
70-75: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winHandle USB3 PHY power-on failures before continuing.
At Line 72,
phy_power_on()failures are stored only inusb3_phy_powered; the mode transition continues. A failed power-on can leave DWC3 configured for a role with an unpowered PHY.With multiple ports, the final assignment at Line 74 also hides earlier failures. The paired
rk3399-usbc-usb-dwc3-Track-the-power-state-of-usb3_generic_phy.patchuses this flag to decide whether to power off all USB3 PHYs. Roll back partial activation, abort the transition on failure, and set the flag only after all ports succeed.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/rk3399-usbc-phy-rockchip-naneng-Add-fallback-for-old-DTs.patch` around lines 70 - 75, The USB3 PHY activation loop in the usb3_phy_reset_quirk path must handle partial failures instead of continuing with an unpowered PHY. Update the phy_power_on flow to abort immediately on any failure, power off previously activated ports, leave usb3_phy_powered unset, and set it only after every port powers on successfully.patch/kernel/archive/rockchip64-6.18/temporary-workaround-dma-reset.patch (1)
20-28: 🩺 Stability & Availability | 🟠 Major | 🏗️ Heavy liftDo not convert DMA-reset failure into success.
Both patches continue DMA initialization after
stmmac_reset()fails. A failed reset does not establish that the DMA engine can safely process descriptors.
patch/kernel/archive/rockchip64-6.18/temporary-workaround-dma-reset.patch#L20-L28: Restore the error return, or add a tested, Rockchip-specific false-failure condition before continuing.patch/kernel/archive/rockchip64-7.1/temporary-workaround-dma-reset.patch#L20-L28: Apply the same constrained error handling.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-6.18/temporary-workaround-dma-reset.patch` around lines 20 - 28, In both patch/kernel/archive/rockchip64-6.18/temporary-workaround-dma-reset.patch lines 20-28 and patch/kernel/archive/rockchip64-7.1/temporary-workaround-dma-reset.patch lines 20-28, update stmmac_init_dma_engine so stmmac_reset failure is not converted to success: restore returning ret, or continue only under a tested Rockchip-specific false-failure condition with constrained handling.patch/kernel/archive/rockchip64-7.1/general-drm-panel-add-rocknix-generic-dsi.patch (1)
556-616: 🎯 Functional Correctness | 🟠 Major | ⚡ Quick winReturn the actual mode count from
generic_panel_get_modes.The function always returns 1 at line 615, regardless of how many modes were added by
drm_mode_probed_add()in the preceding loop.drm_panel_funcs.get_modesmust return the number of modes added.If
ctx->modesis empty, this reports a mode when none exists. If multiple modes were parsed, this undercounts them. Track and return the actual count.🐛 Proposed fix to return the actual mode count
struct generic_panel *ctx = panel_to_generic_panel(panel); struct drm_display_mode mode_tmp; struct drm_display_mode *mode; struct generic_panel_mode *genmode = ctx->modes; + int count = 0; while (genmode) { ... drm_mode_probed_add(connector, mode); + count++; genmode = genmode->prev; } connector->display_info.width_mm = ctx->size.width; connector->display_info.height_mm = ctx->size.height; drm_connector_set_panel_orientation(connector, ctx->orientation); - return 1; + return count;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/rockchip64-7.1/general-drm-panel-add-rocknix-generic-dsi.patch` around lines 556 - 616, Update generic_panel_get_modes to initialize a mode counter, increment it only after each successful drm_mode_probed_add call, and return that counter instead of the unconditional 1. Preserve the existing -ENOMEM error path so failed mode duplication does not increase the count, and return zero when ctx->modes is empty.
🧹 Nitpick comments (1)
patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch (1)
52-52: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winRemove the unconditional GPIO IRQ trace from both patch variants.
printk()runs during normal GPIO IRQ mapping and adds default-level log noise.
patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch#L52-L52: replaceprintk()withdev_dbg()or remove it.patch/kernel/archive/meson64-7.1/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch#L52-L52: apply the same logging change.Proposed change
- printk("gpio irq setup: hwirq: 0x%X irqfirst: 0x%X irqlast: 0x%X pin[%d]\n", hwirq, bank->irq_first, bank->irq_last, gpio); + dev_dbg(pc->dev, "gpio irq setup: hwirq: 0x%X irqfirst: 0x%X irqlast: 0x%X pin[%d]\n", hwirq, bank->irq_first, bank->irq_last, gpio);🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch` at line 52, Remove the unconditional GPIO IRQ printk trace at both affected sites: patch/kernel/archive/meson64-6.18/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch line 52 and patch/kernel/archive/meson64-7.1/jethome-0001-Fix-meson64-add-gpio-irq-patch-from-https-lkml.org-l.patch line 52. Either delete the trace or replace it with debug-level logging such as dev_dbg(), preserving the surrounding GPIO IRQ mapping logic.
1 participant
Description
rewrite various patch sets to bring them back in shape
How Has This Been Tested?
Checklist:
Summary by CodeRabbit
New Features
Bug Fixes