Re: [RFC PATCH 2/2] x86: Allow paranoid __{get,put}_user

[Kees Cook]

On Fri, Nov 3, 2017 at 4:04 PM, Laura Abbott <labbott@xxxxxxxxxx> wrote:
> __{get,put}_user calls are designed to be fast and have no checks,
> relying on the caller to have made the appropriate calls previously.
> It's very easy to forget a check though, leaving the kernel vulnerable
> to exploits. Add an option to do the checks and kill the kernel if it
> catches something bad.
>
> Signed-off-by: Laura Abbott <labbott@xxxxxxxxxx>
> ---
> This is the actual implemtation for __{get,put}_user on x86 based on
> Mark Rutland's work for arm66
> lkml.kernel.org/r/<20171026090942.7041-1-mark.rutland@xxxxxxx>
>
> x86 turns out to be easier since the safe and unsafe paths are mostly
> disjoint so we don't have to worry about gcc optimizing out access_ok.
> I tweaked the Kconfig to someting a bit more generic.
>
> The size increase was ~8K in text with a config I tested.

Specifically, this feature would have caught the waitid() bug in 4.13
immediately.

> ---
>  arch/x86/Kconfig               |  3 +++
>  arch/x86/include/asm/uaccess.h | 11 ++++++++++-
>  security/Kconfig               | 11 +++++++++++
>  3 files changed, 24 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 2fdb23313dd5..10c6e150a91e 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -261,6 +261,9 @@ config RWSEM_XCHGADD_ALGORITHM
>  config GENERIC_CALIBRATE_DELAY
>         def_bool y
>
> +config ARCH_HAS_PARANOID_UACCESS
> +       def_bool y
> +
>  config ARCH_HAS_CPU_RELAX
>         def_bool y
>
> diff --git a/arch/x86/include/asm/uaccess.h b/arch/x86/include/asm/uaccess.h
> index d23fb5844404..767febe1c720 100644
> --- a/arch/x86/include/asm/uaccess.h
> +++ b/arch/x86/include/asm/uaccess.h
> @@ -132,6 +132,13 @@ extern int __get_user_bad(void);
>  #define __inttype(x) \
>  __typeof__(__builtin_choose_expr(sizeof(x) > sizeof(0UL), 0ULL, 0UL))
>
> +
> +#define verify_uaccess(dir, ptr)                                        \
> +({                                                                      \
> +        if (IS_ENABLED(CONFIG_PARANOID_UACCESS))                  \
> +                BUG_ON(!access_ok(dir, (ptr), sizeof(*(ptr))));         \
> +})
> +
>  /**
>   * get_user: - Get a simple variable from user space.
>   * @x:   Variable to store result.
> @@ -278,6 +285,7 @@ do {                                                                        \
>         typeof(ptr) __pu_ptr = (ptr);                                   \
>         retval = 0;                                                     \
>         __chk_user_ptr(__pu_ptr);                                       \
> +       verify_uaccess(VERIFY_WRITE, __pu_ptr);                         \
>         switch (size) {                                                 \
>         case 1:                                                         \
>                 __put_user_asm(x, __pu_ptr, retval, "b", "b", "iq",     \
> @@ -293,7 +301,7 @@ do {                                                                        \
>                 break;                                                  \
>         case 8:                                                         \
>                 __put_user_asm_u64((__typeof__(*ptr))(x), __pu_ptr,     \
> -                               retval, \ errret);                      \
> +                               retval, errret);                        \
>                 break;                                                  \
>         default:                                                        \
>                 __put_user_bad();                                       \

Which tree is this against? I don't see the weird line break in my tree?

> @@ -359,6 +367,7 @@ do {                                                                        \
>         typeof(ptr) __gu_ptr = (ptr);                                   \
>         retval = 0;                                                     \
>         __chk_user_ptr(__gu_ptr);                                       \
> +       verify_uaccess(VERIFY_READ, __gu_ptr);                          \
>         switch (size) {                                                 \
>         case 1:                                                         \
>                 __get_user_asm(x, __gu_ptr, retval, "b", "b", "=q",     \

Does __put/get_user_size_ex() need additions too? (And does
put/get_user_ex() lack access_ok() checks as-is? Looks like the users
are have access_ok() checks, but that naming really shouldn't be
aliased to "put/get_user_ex" -- otherwise it gives the impression it's
doing access_ok() checks...)

> diff --git a/security/Kconfig b/security/Kconfig
> index e8e449444e65..0a9ec1a4e86b 100644
> --- a/security/Kconfig
> +++ b/security/Kconfig
> @@ -205,6 +205,17 @@ config STATIC_USERMODEHELPER_PATH
>           If you wish for all usermode helper programs to be disabled,
>           specify an empty string here (i.e. "").
>
> +config PARANOID_UACCESS
> +       bool "Use paranoid uaccess primitives"
> +       depends on ARCH_HAS_PARANOID_UACCESS
> +       help
> +         Forces access_ok() checks in __get_user(), __put_user(), and other
> +         low-level uaccess primitives which usually do not have checks. This
> +         can limit the effect of missing access_ok() checks in higher-level
> +         primitives, with a runtime performance overhead in some cases and a
> +         small code size overhead.
> +
> +
>  source security/selinux/Kconfig
>  source security/smack/Kconfig
>  source security/tomoyo/Kconfig
> --
> 2.13.5
>

-Kees

-- 
Kees Cook
Pixel Security