linux-kernel - Re: [PATCH] ARM: dts: qcom: msm8960: expressatt Add PWM vibrator

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <052f8def-8c6b-4644-a3a8-f7706189d76e@oss.qualcomm.com>
Date: Thu, 29 Jan 2026 11:31:01 +0100
From: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
To: Rudraksha Gupta <guptarud@...il.com>,
        Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>, Rob Herring <robh@...nel.org>,
        Krzysztof Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>
Cc: linux-arm-msm@...r.kernel.org, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] ARM: dts: qcom: msm8960: expressatt Add PWM vibrator

On 1/29/26 8:55 AM, Rudraksha Gupta wrote:
> Hello Konrad,
> 
> 
> On 12/16/25 06:22, Konrad Dybcio wrote:
>> On 12/11/25 11:23 AM, Rudraksha Gupta via B4 Relay wrote:
>>> From: Rudraksha Gupta <guptarud@...il.com>
>>>
>>> Add a pwm-vibrator to expressatt. Currently this vibrates only at 100%
>>>
>>> Signed-off-by: Rudraksha Gupta <guptarud@...il.com>
>>> ---
>>> Add a pwm-vibrator to expressatt. Currently this vibrates only at 100%
>>>
>>> Link:
>>> - https://github.com/LineageOS/android_kernel_samsung_d2/blob/stable/cm-12.0-YNG4N/arch/arm/mach-msm/board-express.c#L1767
>>> - https://github.com/LineageOS/android_kernel_samsung_d2/blob/stable/cm-12.0-YNG4N/drivers/motor/Makefile#L5
>>>
>>> Test:
>>> =====================
>>> samsung-expressatt:~$ dmesg | grep vibra
>>> [   79.892226] input: pwm-vibrator as /devices/platform/vibrator/input/input4
>>> samsung-expressatt:~$ fftest /dev/input/event4
>>> Force feedback test program.
>>> HOLD FIRMLY YOUR WHEEL OR JOYSTICK TO PREVENT DAMAGES
>>>
>>> Device /dev/input/event4 opened
>>> Features:
>>>    * Absolute axes:
>>>      [00 00 00 00 00 00 00 00 ]
>>>    * Relative axes:
>>>      [00 00 ]
>>>    * Force feedback effects types: Periodic, Rumble, Gain,
>>>      Force feedback periodic effects: Square, Triangle, Sine,
>>>      [00 00 00 00 00 00 00 00 00 00 03 07 01 00 00 00 ]
>>>    * Number of simultaneous effects: 16
>>>
>>> Setting master gain to 75% ... OK
>>> Uploading effect #0 (Periodic sinusoidal) ... OK (id 0)
>>> Uploading effect #1 (Constant) ... Error: Invalid argument
>>> Uploading effect #2 (Spring) ... Error: Invalid argument
>>> Uploading effect #3 (Damper) ... Error: Invalid argument
>> Looks like this is because GP1_CLK does not implement .set_duty_cycle..
>>
>> The downstream driver you linked to does so in a terribly hacky way
>> (from the vibrator driver and not the clock driver):
>>
>> https://github.com/LineageOS/android_kernel_samsung_d2/blob/cm-14.1/drivers/motor/immvibespi.c#L53
>>
>> Upstream, we have an implementation for clk_rcg*2*_ops, whereas 8960
>> uses clk_rcg_ops.
>>
>> They look very similar though, perhaps you can hack it up..
>>
>> One thing to note is that you're passing GP1_CLK to the clk-pwm (as
>> you should), but we need to do clk_set_duty_cycle on its parent,
>> GP1_CLK*_SRC*. The framework will take care of this:
>>
>> ```
>> --- drivers/clk/clk.c
>>     if (!core->ops->get_duty_cycle)
>>         return clk_core_update_duty_cycle_parent_nolock(core);
>> ```
>>
>> so long as you add CLK_DUTY_CYCLE_PARENT to the child
>>
>> [...]
> 
> 
> Sorry for the delay, but I'm struggling to get this implemented. I have my in progress work here:
> 
> https://codeberg.org/LogicalErzor/linux/commit/4e7f94d9fdd1e4e1d688b2eb518494d710f157fb

The changes to drivers/clk/clk-pwm.c are unnecessary, you're using
clk-pwm (PWM implemented using a clock reference) and not pwm-clock
(clock implemented using a PWM reference)

> I was able to get rid of the -16 error that was getting spammed when using fftest, but it is still showing
> 
> Uploading effect #1 (Constant) ... Error: Invalid argument
> Uploading effect #2 (Spring) ... Error: Invalid argument
> Uploading effect #3 (Damper) ... Error: Invalid argument


Is there a chance you can pr_err the entire call stack from
pwm_vibrator_start() downwards and see where it fails?

As a guess, I'd check if you're not hitting the EINVAL return path in
clk_core_update_duty_cycle_nolock()


Konrad