Re: [PATCH] dmaengine: ioatdma: Fix error handling path in 'ioat_dma_self_test()'

[Dave Jiang]

On 07/21/2017 12:57 AM, Vinod Koul wrote:
> On Thu, Jul 20, 2017 at 09:56:45AM -0700, Dave Jiang wrote:
>>
>>
>> On 07/20/2017 12:24 AM, walter harms wrote:
>>>
>>>
>>> Am 20.07.2017 00:16, schrieb Christophe JAILLET:
>>>> If the 'memcmp' fails, free allocated resources as done in all other
>>>> error handling paths.
>>>>
>>>> Signed-off-by: Christophe JAILLET <christophe.jaillet@xxxxxxxxxx>
>>>> ---
>>>> Please review carefully, this patch looks "too obvious" to me!
>>>> ---
>>>>  drivers/dma/ioat/init.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/dma/ioat/init.c b/drivers/dma/ioat/init.c
>>>> index ed8ed1192775..948fc1f8fb5c 100644
>>>> --- a/drivers/dma/ioat/init.c
>>>> +++ b/drivers/dma/ioat/init.c
>>>> @@ -390,7 +390,7 @@ static int ioat_dma_self_test(struct ioatdma_device *ioat_dma)
>>>>  	if (memcmp(src, dest, IOAT_TEST_SIZE)) {
>>>>  		dev_err(dev, "Self-test copy failed compare, disabling\n");
>>>>  		err = -ENODEV;
>>>> -		goto free_resources;
>>>> +		goto unmap_dma;
>>>>  	}
>>>>  
>>>>  unmap_dma:
>>>
>>> ^^^^^^^^^^
>>>
>>>
>>> is the goto needed at all ?
>>
>> It's not. However, it may be better to stay there if we happen to add
>> additional code after the if block later on and guard against mistakes.
>> At least IMO.
> 
> Then lets remove it please, there is no place for dead code, if we need it
> people can add it as part of the changes they introduce..
> 

I have no strong opinion in this Christophe. I have seen mistakes and
bugs introduced because of these special optimizations. I've said my
piece and Vinod is for removing it so I'll defer to him.