Re: [PATCH] net: dsa: sja1105: Remove unneeded cast in sja1105_crc32()

From: Geert Uytterhoeven <geert@linux-m68k.org>
To: Vladimir Oltean <olteanv@gmail.com>
Cc: Andrew Lunn <andrew@lunn.ch>,
	Vivien Didelot <vivien.didelot@gmail.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	"David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, netdev <netdev@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] net: dsa: sja1105: Remove unneeded cast in sja1105_crc32()
Date: Thu, 25 Feb 2021 08:28:15 +0100	[thread overview]
Message-ID: <CAMuHMdVeAoPK_iB=Y73X_7zTEJnS6bFKkzCe8QyH8oyZA9OZ5A@mail.gmail.com> (raw)
In-Reply-To: <20210224224358.pysql5pu23zt7mtb@skbuf>

Hi Vladimir,

On Wed, Feb 24, 2021 at 11:44 PM Vladimir Oltean <olteanv@gmail.com> wrote:
> On Tue, Feb 23, 2021 at 12:20:03PM +0100, Geert Uytterhoeven wrote:
> > sja1105_unpack() takes a "const void *buf" as its first parameter, so
> > there is no need to cast away the "const" of the "buf" variable before
> > calling it.
> >
> > Drop the cast, as it prevents the compiler performing some checks.
> >
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > ---
>
> Reviewed-by: Vladimir Oltean <olteanv@gmail.com>

Thanks!

> By the way, your email went straight to my spam box, I just found the
> patch by mistake on patchwork.
>
>     Why is this message in spam?
>     It is in violation of Google's recommended email sender guidelines.

Yeah, sometimes Gmail can be annoying.  I recommend adding a filter
to never send emails with "PATCH" in the subject to spam.

> > Compile-tested only.
> >
> > BTW, sja1105_packing() and packing() are really bad APIs, as the input
> > pointer parameters cannot be const due to the direction depending on
> > "op".  This means the compiler cannot do const checks.  Worse, callers
> > are required to cast away constness to prevent the compiler from
> > issueing warnings.  Please don't do this!
> > ---
>
> What const checks can the compiler not do?

If you have a const and a non-const buffer, and accidentally call
packing() with the two buffer pointers exchanged (this is a common
mistake), you won't get a compiler warning.
So having separate pack() and unpack() functions would be safer.
You can rename packing() to __packing() to make it clear this function
is not to be called directly without deep consideration, and have
pack() and unpack() as wrappers just calling __packing().
Of course that means callers that do need a separate "op" parameter
still need to call __packing(), but they can provide their own safer
wrappers, too.

> Also, if you know of an existing kernel API which can replace packing(),
> I'm all ears.

No idea.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds