RE: [PATCH] include/uapi/linux/swab.h: add __u16 cast to __swab16 conditional

From: David Laight <David.Laight@ACULAB.COM>
To: 'Nick Desaulniers' <ndesaulniers@google.com>,
	Al Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Justin Stitt <jstitt007@gmail.com>,
	Nathan Chancellor <nathan@kernel.org>, Tom Rix <trix@redhat.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"llvm@lists.linux.dev" <llvm@lists.linux.dev>,
	Richard Smith <richardsmith@google.com>
Subject: RE: [PATCH] include/uapi/linux/swab.h: add __u16 cast to __swab16 conditional
Date: Fri, 10 Jun 2022 07:19:38 +0000	[thread overview]
Message-ID: <ef8069e60f7a4e23a26dcbedd0bc77d7@AcuMS.aculab.com> (raw)
In-Reply-To: <CAKwvOdnSSY0jexXioDTZOWSTi0fkaudZbgSjigPr5uzTRmA_Rg@mail.gmail.com>

From: Nick Desaulniers
> Sent: 08 June 2022 20:35
....
> The issue we're facing is more so that `ntohs` is being used in
> printf-like expressions; clang's -Wformat warns about default argument
> promotion so we need to clean up cases where smaller-than-int format
> flags are being used for promoted-to-int params.  While looking at
> that, Nathan noticed that __swab16 will return either a __u16 or an
> int based on whether __HAVE_BUILTIN_BSWAP16__ is defined, which
> depends on BOTH the compiler being used and target architecture.  This
> patch from Justin just cleans that up.

The 'problem' is that the (__u16) cast is likely to add an
extra '& 0xffff' instruction that is almost certainly not
required.

OTOH the lack of this masking has always been a difference between
htons() on BE and LE systems.

But clang is also being over-enthusiastic with its warnings.
IIRC varargs parameters always get integer promotion.
So if %hd ever makes sense for printf then it implies a
mask inside printf.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)