Re: drivers/dma/pl330.c:2981:9: warning: Identical condition 'ret', second condition is always false

[Marek Szyprowski]

Hi Vinod,

On 08.08.2020 14:59, Vinod Koul wrote:
> On 08-08-20, 10:53, kernel test robot wrote:
>> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>> head:   30185b69a2d533c4ba6ca926b8390ce7de495e29
>> commit: a39cddc9e3775100100a4272feed64faac831be9 dmaengine: pl330: Drop boilerplate code for suspend/resume
>> date:   8 months ago
>> compiler: aarch64-linux-gcc (GCC) 9.3.0
>>
>> If you fix the issue, kindly add following tag as appropriate
>> Reported-by: kernel test robot <lkp@xxxxxxxxx>
>>
>>
>> cppcheck warnings: (new ones prefixed by >>)
>>
>>>> drivers/dma/pl330.c:2981:9: warning: Identical condition 'ret', second condition is always false [identicalConditionAfterEarlyExit]
>>      return ret;
>>             ^
>>     drivers/dma/pl330.c:2976:6: note: first condition
>>      if (ret)
>>          ^
>>     drivers/dma/pl330.c:2981:9: note: second condition
>>      return ret;
> This one could be fixed by making this as return 0, but is harmless

Right, with CONFIG_PM disabled, pm_runtime_force_resume() is noop, what 
causes the above warning.

>>             ^
>>>> drivers/dma/pl330.c:2798:23: warning: Variable 'desc->rqcfg.brst_len' is reassigned a value before the old one has been used. [redundantAssignment]
>>      desc->rqcfg.brst_len = get_burst_len(desc, len);
>>                           ^
>>     drivers/dma/pl330.c:2796:24: note: Variable 'desc->rqcfg.brst_len' is reassigned a value before the old one has been used.
>>       desc->rqcfg.brst_len = 1;
>>                            ^
>>     drivers/dma/pl330.c:2798:23: note: Variable 'desc->rqcfg.brst_len' is reassigned a value before the old one has been used.
>>      desc->rqcfg.brst_len = get_burst_len(desc, len);
> This one actually seems like a bug. Reading the code I think
> get_burst_len() should be called first before checking if burst size is
> smaller and setting to 1 in that case
>
> Sugar Zhang, Marek Szyprowski can you folks check this?

Indeed. It look that the commit 137bd11090d89b added 
desc->rqcfg.brst_len = 1 assignment before the desc->rqcfg.brst_len is 
set. Maybe this was a result of the broken rebase or so. No idea. It 
makes sense to switch the order and call desc->rqcfg.brst_len = 
get_burst_len(desc, len) first. I can send a patch if you want.

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland