Re: [PATCH v3 1/4] soc: qcom: ice: Fix race between qcom_ice_probe() and of_qcom_ice_get()

[Neeraj Soni]

On 2/24/2026 10:16 AM, Manivannan Sadhasivam wrote:
> + Neeraj
> 
> On Mon, Feb 23, 2026 at 02:35:04PM -0600, Bjorn Andersson wrote:
>> On Mon, Feb 23, 2026 at 01:32:52PM +0530, Manivannan Sadhasivam via B4 Relay wrote:
>>> From: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxxxxxxxx>
>>>
>>> The current platform driver design causes probe ordering races with
>>> consumers (UFS, eMMC) due to ICE's dependency on SCM firmware calls. If ICE
>>> probe fails (missing ICE SCM or DT registers), devm_of_qcom_ice_get() loops
>>> with -EPROBE_DEFER, leaving consumers non-functional even when ICE should
>>> be gracefully disabled. devm_of_qcom_ice_get() doesn't know if the ICE
>>> driver probe has failed due to above reasons or it is waiting for the SCM
>>> driver.
>>>
>>> Moreover, there is no devlink dependency between ICE and consumer drivers
>>> as 'qcom,ice' is not considered as a DT 'supplier'. So the consumer drivers
>>> have no idea of when the ICE driver is going to probe.
>>>
>>> To address these issues, introduce a global ice_handle to store the valid
>>> ICE handle pointer, and set during successful ICE driver probe. On probe
>>> failure, set it to an error pointer and propagate the error from
>>> of_qcom_ice_get().
>>>
>>> Additionally, add a global ice_mutex to synchronize qcom_ice_probe() and
>>> of_qcom_ice_get().
>>>
>>> Note that this change only fixes the standalone ICE DT node bindings and
>>> not the ones with 'ice' range embedded in the consumer nodes, where there
>>> is no issue.
>>>
>>> Cc: <stable@xxxxxxxxxxxxxxx> # 6.4
>>> Fixes: 2afbf43a4aec ("soc: qcom: Make the Qualcomm UFS/SDCC ICE a dedicated driver")
>>> Reported-by: Sumit Garg <sumit.garg@xxxxxxxxxxxxxxxx>
>>> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxxxxxxxx>
>>> ---
>>>  drivers/soc/qcom/ice.c | 44 +++++++++++++++++++++++++++-----------------
>>>  1 file changed, 27 insertions(+), 17 deletions(-)
>>>
>>> diff --git a/drivers/soc/qcom/ice.c b/drivers/soc/qcom/ice.c
>>> index b203bc685cad..3c3c189e24f9 100644
>>> --- a/drivers/soc/qcom/ice.c
>>> +++ b/drivers/soc/qcom/ice.c
>>> @@ -113,6 +113,9 @@ struct qcom_ice {
>>>  	u8 hwkm_version;
>>>  };
>>>  
>>> +static DEFINE_MUTEX(ice_mutex);
>>> +static struct qcom_ice *ice_handle;
>>
>> Did we get confirmation that in the UFS + SDCC case, there's only a
>> single ICE instance per SoC?
>>
> 
> Right now there is only a single instance per SoC. But Neeraj told me that
> upcoming SoCs are going to have multiple instances. But I don't want to spend

Yes and patches for same are under review here:
https://lore.kernel.org/all/20260217052526.2335759-1-neeraj.soni@xxxxxxxxxxxxxxxx/

> too much time on *upcoming* support, but rather fix the current
> implementations.
> 
> Extending this to multiple instances would just require storing the ice_handle
> with node name/address pair in xarray or in some other data structures.
> 
> - Mani
>