Re: [PATCH net-next 13/14] wireless: Use eth_<foo>_addr instead of memset

[David Miller <davem@davemloft.net>]

From: Joe Perches <joe@perches.com>
Date: Tue, 03 Mar 2015 11:03:17 -0800

> On Tue, 2015-03-03 at 13:57 -0500, David Miller wrote:
>> From: Johannes Berg <johannes@sipsolutions.net>
>> Date: Tue, 03 Mar 2015 09:16:57 +0100
>> 
>> > Other than that, I guess I'll apply this, but I really wish there was a
>> > way to distinguish more easily which of these require alignment and
>> > which don't.
>> 
>> You can't apply "this" without the dependency patch #1.
>> 
>> Therefore this should probably all go through my tree.
> 
> Hey David.
> 
> The eth_<foo>_addr functions already exist so
> 1/14 isn't a dependency.
> 
> It's just a trivial improvement on existing code
> with CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS and
> at least the arm 4.6.3 compiler.

You're right.  But I think we're skipping this series for now, see
my other reply.

If GCC can't emit as good an inline memset as we can come up with
for 6 byte constant lengths, that's a GCC bug that should be fixed.