Re: [PATCH 8/8] staging: rtl8188eu: use is_broadcast_ether_addr

[Michael Straube <straube.linux@gmail.com>]

On 07/29/18 22:05, Joe Perches wrote:
> On Sun, 2018-07-29 at 20:21 +0200, Michael Straube wrote:
>> On 07/29/18 19:59, Joe Perches wrote:
>>> On Sun, 2018-07-29 at 19:42 +0200, Michael Straube wrote:
>>>> On 07/29/18 19:21, Joe Perches wrote:
>>>>> On Sun, 2018-07-29 at 19:08 +0200, Michael Straube wrote:
>>>>>> Use is_broadcast_ether_addr instead of checking each byte of the
>>>>>> address array for 0xff. Shortens the code and improves readability.
>>>>>
>>>>> You should show in the commit log that sta_addr is __aligned(2)
>>>>> as required by is_broadcast_ether_addr, otherwise you could be
>>>>> introducing runtime alignment defects.
>>>>>
>>>>
>>>> Ok, sta_addr is used from following structs.
>>>>
>>>> struct ieee_param {
>>>>            u32 cmd;
>>>>            u8 sta_addr[ETH_ALEN];
>>>>            union {
>>>>            ...
>>>>            ...
>>>>            }; u
>>>> };
>>>>
>>>> struct ieee_param_ex {
>>>> 	u32 cmd;
>>>> 	u8 sta_addr[ETH_ALEN];
>>>> 	u8 data[0];
>>>> };
>>>>
>>>> Well, looking at it now, I'm not sure about the alignment anymore
>>>> in the struct that contains the union. Is sta_addr in the first
>>>> struct __aligned(2)?
>>>>
>>>> Should I include the snippets in the commit message, or is just
>>>> writing that sta_addr is __aligned(2) enough? (if it is in the
>>>> first case...)
>>>
>>> It's enough to just state that the uses are properly aligned
>>> as long as you looked and understand that it's required.
>>>
>>
>> Ok, thank you.
>>
>> I looked at it and understand that it's required.
>> But, as mentioned, at second look I'm not sure about the union.
>>
>> I guess I need to read a little more about the alignment of unions.
>> Any hints are welcomed. :)
> 
> The union doesn't matter here.
> Only the locations of sta_addr matter.
> Both follow a u32, so they are also aligned to a u32.
> 
> That is actually not guaranteed by the c standard
> as the u8 array could have arbitrary padding bytes
> inserted between the u32, but no compiler used by
> the kernel does that.
> 

Ah ok, thank you.

Michael