[7.1-rc1 BUG] System hang when reading clock controller sysfs 'registers' files

[7.1-rc1 BUG] System hang when reading clock controller sysfs 'registers' files

[Daniel J Blueman]

On a Qualcomm X1 Hamoa-based Lenovo Slim7x with upstream 7.1-rc1, when
reading any of the /sys/kernel/debug/regmap/*clock*/registers files
[1] (which could happen inadvertently with grep), we see a complete
and immediate system hang.

If this relates to clock/power gating, it could be avoided by
inhibiting that prior to the read. It also reproduces when booted with
"clk_ignore_unused pd_ignore_unused".

Thanks,
  Dan

-- [1]

/sys/kernel/debug/regmap/100000.clock-controller/registers
/sys/kernel/debug/regmap/1fc0000.clock-controller/registers
/sys/kernel/debug/regmap/3d90000.clock-controller/registers
/sys/kernel/debug/regmap/6b6c000.clock-controller-lpass-audio-csr/registers
/sys/kernel/debug/regmap/6ea0000.clock-controller-lpass-tcsr/registers
/sys/kernel/debug/regmap/aaf0000.clock-controller/registers
/sys/kernel/debug/regmap/af00000.clock-controller/registers
/sys/kernel/debug/regmap/dummy-clock-controller@0x0000000001fc0000/registers
--
Daniel J Blueman

Re: [7.1-rc1 BUG] System hang when reading clock controller sysfs 'registers' files

[Konrad Dybcio]

On 5/2/26 8:40 AM, Daniel J Blueman wrote:
> On a Qualcomm X1 Hamoa-based Lenovo Slim7x with upstream 7.1-rc1, when
> reading any of the /sys/kernel/debug/regmap/*clock*/registers files
> [1] (which could happen inadvertently with grep), we see a complete
> and immediate system hang.
> 
> If this relates to clock/power gating, it could be avoided by
> inhibiting that prior to the read. It also reproduces when booted with
> "clk_ignore_unused pd_ignore_unused".

Does this attempt to read the entire regmap-ped address space of the
device?

If so, that's probably ""intended"", since there are TZ-only (and other
forms of secure) registers sprinkled throughout the range

Konrad

> 
> Thanks,
>   Dan
> 
> -- [1]
> 
> /sys/kernel/debug/regmap/100000.clock-controller/registers
> /sys/kernel/debug/regmap/1fc0000.clock-controller/registers
> /sys/kernel/debug/regmap/3d90000.clock-controller/registers
> /sys/kernel/debug/regmap/6b6c000.clock-controller-lpass-audio-csr/registers
> /sys/kernel/debug/regmap/6ea0000.clock-controller-lpass-tcsr/registers
> /sys/kernel/debug/regmap/aaf0000.clock-controller/registers
> /sys/kernel/debug/regmap/af00000.clock-controller/registers
> /sys/kernel/debug/regmap/dummy-clock-controller@0x0000000001fc0000/registers
> --
> Daniel J Blueman

Re: [7.1-rc1 BUG] System hang when reading clock controller sysfs 'registers' files

[Krzysztof Kozlowski]

On 04/05/2026 09:49, Konrad Dybcio wrote:
> On 5/2/26 8:40 AM, Daniel J Blueman wrote:
>> On a Qualcomm X1 Hamoa-based Lenovo Slim7x with upstream 7.1-rc1, when
>> reading any of the /sys/kernel/debug/regmap/*clock*/registers files
>> [1] (which could happen inadvertently with grep), we see a complete
>> and immediate system hang.
>>
>> If this relates to clock/power gating, it could be avoided by
>> inhibiting that prior to the read. It also reproduces when booted with
>> "clk_ignore_unused pd_ignore_unused".
> 
> Does this attempt to read the entire regmap-ped address space of the
> device?
> 
> If so, that's probably ""intended"", since there are TZ-only (and other
> forms of secure) registers sprinkled throughout the range

Yep. See also commit d81901a5406e ("clk: qcom: dispcc-sm8750: Allow
dumping regmap").

We might need something similar for all other devices.

Best regards,
Krzysztof