Re: [RFC, WIP, v4 06/11] media: vidtv: add wrappers for memcpy and memset

[Mauro Carvalho Chehab <mchehab+huawei@kernel.org>]

Em Sat,  2 May 2020 00:22:11 -0300
"Daniel W. S. Almeida" <dwlsalmeida@gmail.com> escreveu:

> From: "Daniel W. S. Almeida" <dwlsalmeida@gmail.com>
> 
> A lot of code in this driver is for serializing structures. This is
> error prone.
> 
> Therefore, prevent buffer overflows by wrapping memcpy and memset,
> comparing the requested length against the buffer size.
> 
> Signed-off-by: Daniel W. S. Almeida <dwlsalmeida@gmail.com>
> ---
>  drivers/media/test-drivers/vidtv/Makefile     |  3 ++
>  .../media/test-drivers/vidtv/vidtv_common.c   | 44 +++++++++++++++++++
>  .../media/test-drivers/vidtv/vidtv_common.h   | 28 ++++++++++++
>  3 files changed, 75 insertions(+)
>  create mode 100644 drivers/media/test-drivers/vidtv/vidtv_common.c
>  create mode 100644 drivers/media/test-drivers/vidtv/vidtv_common.h
> 
> diff --git a/drivers/media/test-drivers/vidtv/Makefile b/drivers/media/test-drivers/vidtv/Makefile
> index a9f1993dd5443..9ea9485d42189 100644
> --- a/drivers/media/test-drivers/vidtv/Makefile
> +++ b/drivers/media/test-drivers/vidtv/Makefile
> @@ -1,3 +1,6 @@
>  # SPDX-License-Identifier: GPL-2.0
>  
> +vidtv_demod-objs := vidtv_common.o
> +vidtv_bridge-objs := vidtv_common.o
> +
>  obj-$(CONFIG_DVB_VIDTV)	+= vidtv_tuner.o vidtv_demod.o vidtv_bridge.o
> diff --git a/drivers/media/test-drivers/vidtv/vidtv_common.c b/drivers/media/test-drivers/vidtv/vidtv_common.c
> new file mode 100644
> index 0000000000000..28f10630499a9
> --- /dev/null
> +++ b/drivers/media/test-drivers/vidtv/vidtv_common.c
> @@ -0,0 +1,44 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * The Virtual DVB test driver serves as a reference DVB driver and helps
> + * validate the existing APIs in the media subsystem. It can also aid
> + * developers working on userspace applications.
> + *
> + * Written by Daniel W. S. Almeida <dwlsalmeida@gmail.com>
> + */
> +
> +#include <linux/types.h>
> +#include <linux/string.h>
> +#include <linux/printk.h>
> +
> +u32 vidtv_memcpy(void *to,
> +		 const void *from,
> +		 size_t len,
> +		 u32 offset,
> +		 u32 buf_sz)
> +{
> +	if (buf_sz && offset + len > buf_sz) {
> +		pr_err("%s: overflow detected, skipping. Try increasing the buffer size",
> +		       __func__);
> +		return 0;

shouldn't it return an error?

> +	}
> +
> +	memcpy(to, from, len);
> +	return len;
> +}
> +
> +u32 vidtv_memset(void *to,
> +		 int c,
> +		 size_t len,
> +		 u32 offset,
> +		 u32 buf_sz)
> +{
> +	if (buf_sz && offset + len > buf_sz) {
> +		pr_err("%s: overflow detected, skipping. Try increasing the buffer size",
> +		       __func__);
> +		return 0;
> +	}
> +
> +	memset(to, c, len);
> +	return len;
> +}
> diff --git a/drivers/media/test-drivers/vidtv/vidtv_common.h b/drivers/media/test-drivers/vidtv/vidtv_common.h
> new file mode 100644
> index 0000000000000..64072c010dc66
> --- /dev/null
> +++ b/drivers/media/test-drivers/vidtv/vidtv_common.h
> @@ -0,0 +1,28 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * The Virtual DVB test driver serves as a reference DVB driver and helps
> + * validate the existing APIs in the media subsystem. It can also aid
> + * developers working on userspace applications.
> + *
> + * Written by Daniel W. S. Almeida <dwlsalmeida@gmail.com>
> + */
> +
> +#ifndef VIDTV_COMMON_H
> +#define VIDTV_COMMON_H
> +
> +#include <linux/types.h>
> +#include <media/dvb_frontend.h>
> +
> +u32 vidtv_memcpy(void *to,
> +		 const void *from,
> +		 size_t len,
> +		 u32 offset,
> +		 u32 buf_sz);
> +
> +u32 vidtv_memset(void *to,
> +		 int c,
> +		 size_t len,
> +		 u32 offset,
> +		 u32 buf_sz);
> +
> +#endif // VIDTV_COMMON_H


On a generic note, I don't like seeing functions or macros like those
re-defining existing Kernel functions like memcpy(), memset(), etc. 

This is actually a very common pattern when vendors try to submit
new drivers upstream: several of them have a generic code, and use an
OS-specific abstraction layer, with lots of defines, inline functions
and re-definitions for Kernel functions.

Before upstreaming a driver (or removing one from staging), the driver
should get rid of those.

On **this very specific case**, I see the value of having it there, as
you're not doing it as a normal Digital TV driver, but, instead, using
those in order to emulate an MPEG-TS encoding.

Yet, as this driver is meant to be a sort of "tutorial" for ones
implementing such features, please add a WARNING at both the header and
at the source code, saying that normal drivers should not do that,
explaining why, in this specific case (where you're simulating a MPEG-TS
in software) it makes sense to have such functions.

Thanks,
Mauro