MT76xx - WLAN leds doesn't work with the testing kernel

[e9hack]

In some Makefiles, EXTRA_CFLAGS += DCONFIG_MT76_LEDS must be replaced by ccflags-y += DCONFIG_MT76_LEDS.

[rany2]

Already fixed in OpenWrt: https://github.com/openwrt/openwrt/blob/main/package/kernel/mt76/patches/003-pass-LED-define-via-ccflags-y.patch

[e9hack]

Nothing is fixed:

mt76/mt7915/Makefile

Line 3 in 018f603

EXTRA_CFLAGS += -DCONFIG_MT76_LEDS

To get the leds working, I had to modify this file. It exist more Makefiles with EXTRA_CFLAGS += DCONFIG_MT76_LEDS

[rany2]

In my case that patch was enough to fix this issue for mt7915. I think those are redundant as it's already in the parent Makefile.

[mbc07]

Just chiming in to say the current nightly for D-Link DIR-882 (dual MT7615 radios) still doesn't expose the LEDs...

[zectaiga]

I tested this on an ELECOM WRC-2533GST2, which uses MT7621A + 2x MT7615 PCIe.

With only the top-level mt76 Makefile changed from EXTRA_CFLAGS to ccflags-y, the mt76 core side got CONFIG_MT76_LEDS, but the MT7615 PCI driver side still did not behave correctly.

In particular, with mt7615/Makefile still using:

EXTRA_CFLAGS += -DCONFIG_MT76_LEDS

the LED callback setup in mt7615/pci_init.o was not built with CONFIG_MT76_LEDS, so the MT7615 LEDs were not registered.

After changing mt7615/Makefile as well:

ccflags-y += -DCONFIG_MT76_LEDS

the mt76-* LED class devices appeared on the device.

So at least for MT7615 PCIe devices, changing only the top-level mt76 Makefile is not sufficient. The driver subdirectory Makefile also needs to use ccflags-y.