Re: [PATCH] staging: wilc1000: use kernel define byte order macros

From: Dan Carpenter <dan.carpenter@oracle.com>
To: Robert Perry Hooker <perry.hooker@gmail.com>
Cc: devel@driverdev.osuosl.org, gregkh@linuxfoundation.org,
	ganesh.krishna@microchip.com, linux-kernel@vger.kernel.org,
	aditya.shankar@microchip.com
Subject: Re: [PATCH] staging: wilc1000: use kernel define byte order macros
Date: Wed, 22 Mar 2017 12:24:04 +0300	[thread overview]
Message-ID: <20170322092403.GF32449@mwanda> (raw)
In-Reply-To: <1490132410.17318.6.camel@gmail.com>

On Tue, Mar 21, 2017 at 03:40:10PM -0600, Robert Perry Hooker wrote:
> Thanks for taking a look, Dan. Sorry if I missed the mark here.
> 
> Can you tell me a bit more about the bug this would introduce?
> 
> I see that ieee80211_is_action is defined like this: static inline bool ieee80211_is_action(__le16 fc)
> 
> ...and that buff[FRAME_TYPE_ID]is a u8 (since FRAME_TYPE_ID = 0).
> 
> Is there an issue with calling cpu_to_le16 on a u8 that isn't encountered by implicitly casting a u8 to __le16? Or am I
> missing something else?
> 

Oh...  Hm.  You're right.  I just was thinking that since buff was a
little endian buffer but it's only reading a u8.  It should probably
be reading a le16...  The buff likely is just a regular ieee80211_hdr
struct.

So you're fixing a bug, but probably not in the right way.  We should
instead just say "struct ieee80211_hdr *hdr = buff;" and instead of
treating it like an array of u8.  Probably it requires testing...

regards,
dan carpenter