[bug report] UBI: Fastmap: Do not add vol if it already exists

[Sheng Yong]

On 10/26/2016 4:25 PM, Dan Carpenter wrote:
> On Wed, Oct 26, 2016 at 09:46:51AM +0800, Sheng Yong wrote:
>> Hi Dan,
>>
>> On 10/26/2016 4:46 AM, Dan Carpenter wrote:
>>> Hello shengyong,
>>>
>>> The patch e96a8a3bb671: "UBI: Fastmap: Do not add vol if it already
>>> exists" from May 26, 2015, leads to the following static checker
>>> warning:
>>>
>>> 	drivers/mtd/ubi/fastmap.c:712 ubi_attach_fastmap()
>>> 	warn: PTR_ERR(av) is never (-22)
>>>
>>> drivers/mtd/ubi/fastmap.c
>>>    703  
>>>    704                  av = add_vol(ai, be32_to_cpu(fmvhdr->vol_id),
>>>    705                               be32_to_cpu(fmvhdr->used_ebs),
>>>    706                               be32_to_cpu(fmvhdr->data_pad),
>>>    707                               fmvhdr->vol_type,
>>>    708                               be32_to_cpu(fmvhdr->last_eb_bytes));
>>>    709  
>>>    710                  if (!av)
>>>    711                          goto fail_bad;
>>>    712                  if (PTR_ERR(av) == -EINVAL) {
>>>
>>> av is either -EEXIST or -ENOMEM.  It's never -EINVAL.
>> The commit e96a8a3bb671 ("UBI: Fastmap: Do not add vol if it already exists")
>> adds a "return ERR_PTR(-EINVAL);" to add_vol(). So I think av could be -EINVAL.
>> You mean add_vol should return -EEXIST instead of -EINVAL?
>>
> 
> Oh, ah.  I'm on linux-next.  Commit de4c455b3e9f ("UBI: factorize code
> used to manipulate volumes at attach time") removed the -EINVAL.
Sorry, my bad. I didn't check linux-next. You're right, the caller of add_vol should
have checked -EEXIST. And Boris's patch could fix it. Thanks Dan for pointing that
out.

thanks,
Sheng
> 
> It's really Boris to blame for this warning.  I'm not sure what the fix
> is.
> 
> regards,
> dan carpenter
> 
> 
> .
>