Received-SPF: pass (google.com: domain of linux-kernel+bounces-82482-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Date: Mon, 26 Feb 2024 16:48:38 -0800
From: Kees Cook <keescook@chromium.org>
To: Jacob Keller <jacob.e.keller@intel.com>
Cc: linux-i2c@vger.kernel.org, linux-kernel@vger.kernel.org,
	gustavoars@kernel.org, Peter Rosin <peda@axentia.se>,
	Jesse Brandeburg <jesse.brandeburg@intel.com>
Subject: Re: [PATCH v2] mux: convert mux_chip->mux to flexible array
Message-ID: <202402261647.CCCF97AF@keescook>
References: <20240226212925.3781744-1-jacob.e.keller@intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240226212925.3781744-1-jacob.e.keller@intel.com>

On Mon, Feb 26, 2024 at 01:29:25PM -0800, Jacob Keller wrote:
> The mux_chip structure size is over allocated to additionally include both
> the array of mux controllers as well as a device specific private area.
> The controllers array is then pointed to by assigning mux_chip->mux to the
> first block of extra memory, while the private area is extracted via
> mux_chip_priv() and points to the area just after the controllers.
> 
> The size of the mux_chip allocation uses direct multiplication and addition
> rather than the <linux/overflow.h> helpers. In addition, the mux_chip->mux
> struct member wastes space by having to store the pointer as part of the
> structures.
> 
> Convert struct mux_chip to use a flexible array member for the mux
> controller array. Use struct_size() and size_add() to compute the size of
> the structure while protecting against overflow.
> 
> After converting the mux pointer, notice that two 4-byte holes remain in
> the structure layout due to the alignment requirements for the dev
> sub-structure and the ops pointer.
> 
> These can be easily fixed through re-ordering the id field to the 4-byte
> hole just after the controllers member.
> 
> This changes the layout from:
> 
> struct mux_chip {
>         unsigned int               controllers;          /*     0     4 */
> 
>         /* XXX 4 bytes hole, try to pack */
> 
>         struct mux_control *       mux;                  /*     8     8 */
>         struct device              dev __attribute__((__aligned__(8))); /*    16  1488 */
> 
>         /* XXX last struct has 3 bytes of padding */
> 
>         /* --- cacheline 23 boundary (1472 bytes) was 32 bytes ago --- */
>         int                        id;                   /*  1504     4 */
> 
>         /* XXX 4 bytes hole, try to pack */
> 
>         const struct mux_control_ops  * ops;             /*  1512     8 */
> 
>         /* size: 1520, cachelines: 24, members: 5 */
>         /* sum members: 1512, holes: 2, sum holes: 8 */
>         /* paddings: 1, sum paddings: 3 */
>         /* forced alignments: 1 */
>         /* last cacheline: 48 bytes */
> } __attribute__((__aligned__(8)));
> 
> To the following:
> 
> struct mux_chip {
>         unsigned int               controllers;          /*     0     4 */
>         int                        id;                   /*     4     4 */
>         struct device              dev __attribute__((__aligned__(8))); /*     8  1488 */
> 
>         /* XXX last struct has 3 bytes of padding */
> 
>         /* --- cacheline 23 boundary (1472 bytes) was 24 bytes ago --- */
>         const struct mux_control_ops  * ops;             /*  1496     8 */
>         struct mux_control         mux[];                /*  1504     0 */
> 
>         /* size: 1504, cachelines: 24, members: 5 */
>         /* paddings: 1, sum paddings: 3 */
>         /* forced alignments: 1 */
>         /* last cacheline: 32 bytes */
> } __attribute__((__aligned__(8)));
> 
> This both removes risk of overflowing and performing an under-allocation,
> as well as saves 16 bytes of otherwise wasted space for every mux_chip.

This looks like a great clean-up, thank you!

> 
> Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
> ---
> Changes since v1:
> * Rebased and updated the commit message slightly.
> 
>  drivers/mux/core.c         |  7 +++----
>  include/linux/mux/driver.h | 10 +++++-----
>  2 files changed, 8 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/mux/core.c b/drivers/mux/core.c
> index 775816112932..9225abca7897 100644
> --- a/drivers/mux/core.c
> +++ b/drivers/mux/core.c
> @@ -98,13 +98,12 @@ struct mux_chip *mux_chip_alloc(struct device *dev,
>  	if (WARN_ON(!dev || !controllers))
>  		return ERR_PTR(-EINVAL);
>  
> -	mux_chip = kzalloc(sizeof(*mux_chip) +
> -			   controllers * sizeof(*mux_chip->mux) +
> -			   sizeof_priv, GFP_KERNEL);
> +	mux_chip = kzalloc(size_add(struct_size(mux_chip, mux, controllers),
> +				    sizeof_priv),
> +			   GFP_KERNEL);
>  	if (!mux_chip)
>  		return ERR_PTR(-ENOMEM);
>  
> -	mux_chip->mux = (struct mux_control *)(mux_chip + 1);
>  	mux_chip->dev.class = &mux_class;
>  	mux_chip->dev.type = &mux_type;
>  	mux_chip->dev.parent = dev;
> diff --git a/include/linux/mux/driver.h b/include/linux/mux/driver.h
> index 18824064f8c0..84dc0d3e79d6 100644
> --- a/include/linux/mux/driver.h
> +++ b/include/linux/mux/driver.h
> @@ -56,18 +56,18 @@ struct mux_control {
>  /**
>   * struct mux_chip -	Represents a chip holding mux controllers.
>   * @controllers:	Number of mux controllers handled by the chip.
> - * @mux:		Array of mux controllers that are handled.
> - * @dev:		Device structure.
>   * @id:			Used to identify the device internally.
> + * @dev:		Device structure.
>   * @ops:		Mux controller operations.
> + * @mux:		Flexible array of mux controllers that are handled.
>   */
>  struct mux_chip {
>  	unsigned int controllers;
> -	struct mux_control *mux;
> -	struct device dev;
>  	int id;
> -
> +	struct device dev;
>  	const struct mux_control_ops *ops;
> +
> +	struct mux_control mux[];

This can gain the __counted_by annotation as well:

	struct mux_control mux[] __counted_by(controllers);

With that:

	Reviewed-by: Kees Cook <keescook@chromium.org>

-Kees

-- 
Kees Cook