Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sat, 13 Jul 2019 07:02:56 -0300
From:   Mauro Carvalho Chehab <mchehab+samsung@kernel.org>
To:     Marc Gonzalez <marc.w.gonzalez@free.fr>
Cc:     Brad Love <brad@nextdimension.cc>, Antti Palosaari <crope@iki.fi>,
        Jonathan =?UTF-8?B?TmV1c2Now6RmZXI=?= <j.neuschaefer@gmx.net>,
        linux-media <linux-media@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        MSM <linux-arm-msm@vger.kernel.org>,
        Bjorn Andersson <bjorn.andersson@linaro.org>
Subject: Re: [PATCH v3] media: si2168: Refactor command setup code
Message-ID: <20190713070256.3495de51@coco.lan>
In-Reply-To: <10f064c5-1634-c9f9-fcc9-6ab51b7f8f0b@free.fr>
References: <544859b5-108a-1909-d612-64f67a02aeec@free.fr>
        <bde6e367-61a4-7501-2459-eecad5db1d1b@nextdimension.cc>
        <20190712144537.2bad2482@coco.lan>
        <10f064c5-1634-c9f9-fcc9-6ab51b7f8f0b@free.fr>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Em Sat, 13 Jul 2019 00:11:12 +0200
Marc Gonzalez <marc.w.gonzalez@free.fr> escreveu:

> On 12/07/2019 19:45, Mauro Carvalho Chehab wrote:
> 
> > Brad Love <brad@nextdimension.cc> escreveu:
> >   
> >> On 04/07/2019 05.33, Marc Gonzalez wrote:
> >>  
> >>> +#define CMD_SETUP(cmd, args, rlen) \
> >>> +	cmd_setup(cmd, args, sizeof(args) - 1, rlen)
> >>> +    
> >>
> >> This is only a valid helper if args is a null terminated string. It just
> >> so happens that every instance in this driver is, but that could be a
> >> silent pitfall if someone used a u8 array with this macro.  
> > 
> > Actually, it is uglier than that. If one writes something like:
> > 
> > 	char buf[20];
> > 
> > 	buf[0] = 0x20;
> > 	buf[1] = 0x03;
> > 
> > 	CMD_SETUP(cmd, buf, 0);
> > 
> > 	// some other init, up to 5 values, then another CMD_SETUP()  
> 
> I'm not sure what you mean in the // comment.
> What kind of init? Why up to 5 values? Why another CMD_SETUP?

I mean that the same buffer could be re-used to do something like:

 	char buf[20];
 
 	buf[0] = 0x20;
 	buf[1] = 0x03;
 
 	CMD_SETUP(cmd, buf, 0);   // write size here should be 2

	buf[2] = 0x04
	buf[3] = 0x00
	buf[4] = 0x05

	CMD_SETUP(cmd, buf, 0); // write size here should be 5

This kind of pattern happens on other drivers and someone may
end needing something like that at this driver on some future.

> > sizeof() will evaluate to 20, and not to 2, with would be the
> > expected buffer size, and it will pass 18 random values.
> > 
> > IMHO, using sizeof() here is a very bad idea.  
> 
> You may have a point...
> (Though I'm not proposing a kernel API function, merely code
> refactoring for a single file that's unlikely to change going
> forward.)

Yes, I know, but we had already some bugs due to the usage of
sizeof() on similar macros at drivers in the past.

> It's also bad form to repeat the cmd size (twice) when the compiler
> can figure it out automatically for string literals (which is 95%
> of the use-cases).
> 
> I can drop the macro, and just use the helper...

The helper function sounds fine.

> 
> Or maybe there's a GCC extension to test that an argument is a
> string literal...

If this could be evaluated by some advanced macro logic that
would work not only with gcc but also with clang, then a
macro that does what you proposed could be useful.

There are some ways to check the type of a macro argument, but I'm
not sure if are there any way for it to distinguish between a
string constant from a char * array.

Thanks,
Mauro