Re: [PATCH 1/1] powerpc/ps3: Fix error return code in ps3_register_devices()

["Leizhen (ThunderTown)" <thunder.leizhen@huawei.com>]

On 2021/5/24 4:15, Geoff Levand wrote:
> Hi,
> 
> On 5/20/21 5:20 AM, Michael Ellerman wrote:
>> Zhen Lei <thunder.leizhen@huawei.com> writes:
>>> When call ps3_start_probe_thread() failed, further initialization should
>>> be stopped and the returned error code should be propagated.
> ...
>>> --- a/arch/powerpc/platforms/ps3/device-init.c
>>>  
>>>  	result = ps3_start_probe_thread(PS3_BUS_TYPE_STORAGE);
>>> +	if (result < 0)
>>> +		return result;
>>
>> If you bail out here you skip:
>>
>>>  	ps3_register_vuart_devices();
>>
>> Which I suspect means there will be no console output?
>>
>> Presumably the system won't boot if the probe thread fails, but it might
>> at least print an oops, whereas if we return we might get nothing at
>> all. Though I'm just guessing, I don't know this code that well.
> 
> That probe is for the storage devices (PS3_BUS_TYPE_STORAGE).
> 
> There are cases where the system is usable even if the storage
> devices are not available, for example, when using an NFS root
> filesystem.
> 
> ps3_start_probe_thread was made to be quite verbose on error
> to make up for it's return value not being checked.

So should we delete the local variable 'result' and ignore the return value?
- result = ps3_start_probe_thread(PS3_BUS_TYPE_STORAGE);
- (void)ps3_start_probe_thread(PS3_BUS_TYPE_STORAGE);

> 
>> Anyway please leave this code alone unless you're willing to test your
>> changes, or at least provide a more thorough justification for them.
> 
> Agreed, this change should not be merged.
> 
> -Geoff
> 
> .
>