[Meizu 18] No vibration

[AndyCGYan]

Device: Meizu 18 (meizu_18_CN) PHH AOSP version: v309 Vendor version: R (Flyme 9)

No vibration anywhere in the OS. Logcat notably says: E servicemanager: Could not find android.hardware.vibrator.IVibrator/default in the VINTF manifest.

lshal on stock and GSI: lshal-stock.txt, lshal-gsi.txt dumpsys sensorservice: sensorservice.txt Logcat since boot on GSI: log-since-boot.txt

[phhusson]

That's uh fun, there is no vibrator HAL at all?!

K uh, got a dump of stock ROM?

[AndyCGYan]

Dumped: here

[phhusson]

Ok so vibrator events go to android.os.FlymeVibratorHelper

It directly access /sys

.field private static final BACK_STRENGTH:Ljava/lang/String; = "/sys/class/meizu/motor/set_mback"
.field private static final CONTROL_PATH_EFFECT:Ljava/lang/String; = "/sys/class/meizu/motor/on_off"
.field private static final CONTROL_PATH_RTP:Ljava/lang/String; = "/sys/class/meizu/motor/rtp"
.field private static final CONTROL_PATH_TIMEOUT:Ljava/lang/String; = "/sys/class/timed_output/vibrator/enable"
.field private static final PATH_WAVE_NUMBER:Ljava/lang/String; = "/sys/class/meizu/motor/waveform"
.field private static final PRESSURE_STRENGTH:Ljava/lang/String; = "/sys/class/meizu/motor/set_cspress"
.field private static final RTPID_PATH:Ljava/lang/String; = "/sys/class/meizu/motor/set_rtp"

vibrator on is echo 1 > /sys/class/timed_output/vibrator/enable; vibrator off is echo 0 > /sys/class/timed_output/vibrator/enable

I don't know how this works, but /sys/class/meizu/motor/set_mback contains the force of the feedback when pressing back key, it can be "2 0" for weakest, "1 0" for half, "0 0" for strongest.

Likewise, with /sys/class/meizu/motor/set_cspress for home key

There are some pre-set """musics""", which are triggered by writing their id to /sys/class/meizu/motor/rtp; Some can probably be mapped to android.hardware.vibrator@1.3::Effect. Only effect id i have is 5105 and 5001. By following flyme_phone_ringid and flyme_alarm_ringid it's possible to find more.

[AndyCGYan]

Interesting, so it's (again) implementing custom patterns alongside linear motor.

vibrator on is echo 1 > /sys/class/timed_output/vibrator/enable; vibrator off is echo 0 > /sys/class/timed_output/vibrator/enable

FYI, doing this on GSI triggers a very short vibe (like tapping a key) instead of a constant one.

[phhusson]

Well implementing custom pattern is the main usage of linear motor

Uh yeah, misread, syntax for enable is echo $durationInMs $amplitude > /sys/class/timed_output/vibrator/enable so echo 100 will do a 100ms duration, at default amplitudee I have no idea what are the possible amplituds

[AndyCGYan]

Weird, it doesn't seem to work the way you analyzed...

• echo $durationInMs > /sys/class/timed_output/vibrator/enable where durationInMs ≤ 50 triggers the same short vibe as before
• echo $durationInMs > /sys/class/timed_output/vibrator/enable where durationInMs > 50 triggers a constant but "loose" (for the lack of a better word) vibe
• Tried a few $amplitude integers, did nothing

[AndyCGYan]

Updating recent findings here for documentation sake:

• $amplitude does work but in a subtle way - the range is [1, 255], and the difference is hard to notice.
• Critical nodes like /sys/class/timed_output/vibrator/enable can't be read, states "I/O error".
• There also exists /sys/class/timed_output/vibrator/activate and /sys/class/timed_output/vibrator/duration nodes, which aren't referenced in the framework - these work to an extent, but seems like there's some other precursory node that needs to be set (by triggering a vibration from the OS) before they work as intended.
• RTP effect IDs are in the HapticFeedbackConstants class.

[AndyCGYan]

Solved by implementing a vibrator HAL - details and PR soon.

[AndyCGYan]

PR here: https://github.com/phhusson/device_phh_treble/pull/266

• /sys/class/timed_output/vibrator/enable takes duration and (optionally) amplitude as parameter, but since amplitude can't be set on its own, I set supportsAmplitudeControl() to false. Default amplitude is the maximum 255.
• /sys/class/meizu/motor/waveform takes waveform IDs corresponding to different forms of oneshot vibration. The two used most by stock are 1 (short tap) and 13 (medium strength long vibe); I tried and found out 10, 11, 12 are stronger long vibes, but 10 and 11 are so strong they might damage the device, so I chose 12.
• /sys/class/meizu/motor/on_off takes RTP effect IDs from the HapticFeedbackConstants class. @xingrz 's 1.3 implementation already laid out handily how they correspond to native effects.