Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 03 Aug 2023 22:20:20 -0400
Message-ID: <059e2bd127d8a6d59290d606a707a999.paul@paul-moore.com>
From:   Paul Moore <paul@paul-moore.com>
To:     =?UTF-8?q?Christian=20G=C3=B6ttsche?= <cgzones@googlemail.com>,
        selinux@vger.kernel.org
Cc:     Stephen Smalley <stephen.smalley.work@gmail.com>,
        Eric Paris <eparis@parisplace.org>,
        Ondrej Mosnacek <omosnace@redhat.com>,
        "GONG, Ruiqi" <gongruiqi1@huawei.com>, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 5/9] selinux: services: update type for number of class  permissions
References: <20230728155501.39632-4-cgzones@googlemail.com>
In-Reply-To: <20230728155501.39632-4-cgzones@googlemail.com>
Precedence: bulk

On Jul 28, 2023 =?UTF-8?q?Christian=20G=C3=B6ttsche?= <cgzones@googlemail.com> wrote:
> 
> Security classes have only up to 32 permissions, hence using an u16 is
> sufficient (while improving padding in struct selinux_mapping).
> 
> Also use a fixed sized cast in a bit shift to avoid (well defined)
> overflows on architectures where sizeof(unsigned int) != sizeof(u32)
> resulting in no bits set.
> 
> Signed-off-by: Christian Göttsche <cgzones@googlemail.com>
> ---
> v2:
>    update commit description:
>      - mention struct selinux_mapping  in the padding argument
>        (currently between the first and second member there are 2 bytes
>         padding)
>      - mention overflow in the cast argument and the result of setting
>        no bits due to it
> ---
>  security/selinux/ss/services.c | 6 +++---
>  security/selinux/ss/services.h | 2 +-
>  2 files changed, 4 insertions(+), 4 deletions(-)

This looks good, I would just like to request one small change
(see below).

> diff --git a/security/selinux/ss/services.c b/security/selinux/ss/services.c
> index 2c5be06fbada..cf4b87ec4a0e 100644
> --- a/security/selinux/ss/services.c
> +++ b/security/selinux/ss/services.c
> @@ -97,7 +97,6 @@ static int selinux_set_mapping(struct policydb *pol,
>  			       struct selinux_map *out_map)
>  {
>  	u16 i, j;
> -	unsigned k;
>  	bool print_unknown_handle = false;
>  
>  	/* Find number of classes in the input mapping */
> @@ -117,6 +116,7 @@ static int selinux_set_mapping(struct policydb *pol,
>  	while (map[j].name) {
>  		const struct security_class_mapping *p_in = map + (j++);
>  		struct selinux_mapping *p_out = out_map->mapping + j;
> +		u16 k;
>  
>  		/* An empty class string skips ahead */
>  		if (!strcmp(p_in->name, "")) {
> @@ -202,7 +202,7 @@ static void map_decision(struct selinux_map *map,
>  {
>  	if (tclass < map->size) {
>  		struct selinux_mapping *mapping = &map->mapping[tclass];
> -		unsigned int i, n = mapping->num_perms;
> +		u16 i, n = mapping->num_perms;
>  		u32 result;
>  
>  		for (i = 0, result = 0; i < n; i++) {
> @@ -230,7 +230,7 @@ static void map_decision(struct selinux_map *map,
>  		 * should audit that denial
>  		 */
>  		for (; i < (sizeof(u32)*8); i++)
> -			result |= 1<<i;
> +			result |= 1<<((u32)i);

Given that the for-loop bounds the value of 'i' to a maximum of 32
(31 within the valid portion of the loop), this cast seems
unnecessary and potentially problematic in the future.  Please drop
this casting.

>  		avd->auditdeny = result;
>  	}
>  }

--
paul-moore.com