Re: [PATCH] platform/x86: Fix NULL deref in amd_isp_probe

[Mario Limonciello <mario.limonciello@amd.com>]

On 8/10/26 16:52, Nirujogi, Pratap wrote:
> 
> 
> On 8/9/2026 11:24 PM, Xueqin Luo wrote:
>> [You don't often get email from luoxueqin@kylinos.cn. Learn why this 
>> is important at https://aka.ms/LearnAboutSenderIdentification ]
>>
>> Caution: This message originated from an External Source. Use proper 
>> caution when opening attachments, clicking links, or responding.
>>
>>
>> amd_isp_probe() uses ACPI_COMPANION() without NULL check. Add it to
>> prevent crash when driver_override matches a device without ACPI
>> companion.
>>
>> Fixes: e1af6f0146d6 ("platform/x86: Update swnode graph for amd isp4")
>> Signed-off-by: Xueqin Luo <luoxueqin@kylinos.cn>
>> ---
>>   drivers/platform/x86/amd/amd_isp4.c | 7 +++++++
>>   1 file changed, 7 insertions(+)
>>
>> diff --git a/drivers/platform/x86/amd/amd_isp4.c b/drivers/platform/ 
>> x86/amd/amd_isp4.c
>> index 0d494899502c..0026d6663c5c 100644
>> --- a/drivers/platform/x86/amd/amd_isp4.c
>> +++ b/drivers/platform/x86/amd/amd_isp4.c
>> @@ -379,6 +379,11 @@ static int amd_isp_probe(struct platform_device 
>> *pdev)
>>                  goto error_unregister_sw_node;
>>
>>          adev = ACPI_COMPANION(&pdev->dev);
>> +       if (!adev) {
>> +               ret = -ENODEV;
>> +               goto error_unregister_notifier;
>> +       }
>> +
>>          /* initialize root amd_camera_node */
>>          adev->driver_data = (void *)pinfo->swnodes[0];
>>
>> @@ -388,6 +393,8 @@ static int amd_isp_probe(struct platform_device 
>> *pdev)
>>          platform_set_drvdata(pdev, isp4_platform);
>>          return 0;
>>
>> +error_unregister_notifier:
>> +       bus_unregister_notifier(&i2c_bus_type, &isp4_platform->i2c_nb);
>>   error_unregister_sw_node:
>>          software_node_unregister_node_group(isp4_platform->pinfo- 
>> >swnodes);
>>          return ret;
> LGTM. I agree that adding the NULL check is good practice. That said, I 
> believe it's not strictly needed as the device_get_match_data() in 
> line#366 is expected to fail if there is no ACPI companion.

It sounds like it's essentially dead code then.

> 
> https://github.com/torvalds/linux/blob/master/drivers/platform/x86/amd/ 
> amd_isp4.c#L366
> 
> Reviewed-by: Pratap Nirujogi <pratap.nirujogi@amd.com>
> 
>> -- 
>> 2.43.0
>>
>