Re: [RFC 2/4] lib/strncpy_from_user: Remove redundant user space pointer range check

From: Andrey Konovalov
Date: Wed Jan 15 2020 - 09:42:56 EST


On Tue, Jan 14, 2020 at 9:08 PM Vineet Gupta <Vineet.Gupta1@xxxxxxxxxxxx> wrote:
>
> This came up when switching ARC to word-at-a-time interface and using
> generic/optimized strncpy_from_user
>
> It seems the existing code checks for user buffer/string range multiple
> times and one of tem cn be avoided.
>
> There's an open-coded range check which computes @max off of user_addr_max()
> and thus typically way larger than the kernel buffer @count and subsequently
> discarded in do_strncpy_from_user()
>
> if (max > count)
> max = count;
>
> The canonical user_access_begin() => access_ok() follow anyways and even
> with @count it should suffice for an intial range check as is true for
> any copy_{to,from}_user()
>
> And in case actual user space buffer is smaller than kernel dest pointer
> (i.e. @max < @count) the usual string copy, null byte detection would
> abort the process early anyways
>
> Signed-off-by: Vineet Gupta <vgupta@xxxxxxxxxxxx>
> ---
> lib/strncpy_from_user.c | 36 +++++++++++-------------------------
> lib/strnlen_user.c | 28 +++++++---------------------
> 2 files changed, 18 insertions(+), 46 deletions(-)
>
> diff --git a/lib/strncpy_from_user.c b/lib/strncpy_from_user.c
> index dccb95af6003..a1622d71f037 100644
> --- a/lib/strncpy_from_user.c
> +++ b/lib/strncpy_from_user.c
> @@ -21,22 +21,15 @@
> /*
> * Do a strncpy, return length of string without final '\0'.
> * 'count' is the user-supplied count (return 'count' if we
> - * hit it), 'max' is the address space maximum (and we return
> - * -EFAULT if we hit it).
> + * hit it). If access fails, return -EFAULT.
> */
> static inline long do_strncpy_from_user(char *dst, const char __user *src,
> - unsigned long count, unsigned long max)
> + unsigned long count)
> {
> const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS;
> + unsigned long max = count;
> unsigned long res = 0;
>
> - /*
> - * Truncate 'max' to the user-specified limit, so that
> - * we only have one limit we need to check in the loop
> - */
> - if (max > count)
> - max = count;
> -
> if (IS_UNALIGNED(src, dst))
> goto byte_at_a_time;
>
> @@ -72,7 +65,7 @@ static inline long do_strncpy_from_user(char *dst, const char __user *src,
> * Uhhuh. We hit 'max'. But was that the user-specified maximum
> * too? If so, that's ok - we got as much as the user asked for.
> */
> - if (res >= count)
> + if (res == count)
> return res;
>
> /*
> @@ -103,25 +96,18 @@ static inline long do_strncpy_from_user(char *dst, const char __user *src,
> */
> long strncpy_from_user(char *dst, const char __user *src, long count)
> {
> - unsigned long max_addr, src_addr;
> -
> if (unlikely(count <= 0))
> return 0;
>
> - max_addr = user_addr_max();
> - src_addr = (unsigned long)untagged_addr(src);

If you end up changing this code, you need to keep the untagged_addr()
logic, otherwise this breaks arm64 tagged address ABI [1].

[1] https://www.kernel.org/doc/html/latest/arm64/tagged-address-abi.html

> - if (likely(src_addr < max_addr)) {
> - unsigned long max = max_addr - src_addr;
> + kasan_check_write(dst, count);
> + check_object_size(dst, count, false);
> + if (user_access_begin(src, count)) {
> long retval;
> -
> - kasan_check_write(dst, count);
> - check_object_size(dst, count, false);
> - if (user_access_begin(src, max)) {
> - retval = do_strncpy_from_user(dst, src, count, max);
> - user_access_end();
> - return retval;
> - }
> + retval = do_strncpy_from_user(dst, src, count);
> + user_access_end();
> + return retval;
> }
> +
> return -EFAULT;
> }
> EXPORT_SYMBOL(strncpy_from_user);
> diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
> index 6c0005d5dd5c..5ce61f303d6e 100644
> --- a/lib/strnlen_user.c
> +++ b/lib/strnlen_user.c
> @@ -20,19 +20,13 @@
> * if it fits in a aligned 'long'. The caller needs to check
> * the return value against "> max".
> */
> -static inline long do_strnlen_user(const char __user *src, unsigned long count, unsigned long max)
> +static inline long do_strnlen_user(const char __user *src, unsigned long count)
> {
> const struct word_at_a_time constants = WORD_AT_A_TIME_CONSTANTS;
> unsigned long align, res = 0;
> + unsigned long max = count;
> unsigned long c;
>
> - /*
> - * Truncate 'max' to the user-specified limit, so that
> - * we only have one limit we need to check in the loop
> - */
> - if (max > count)
> - max = count;
> -
> /*
> * Do everything aligned. But that means that we
> * need to also expand the maximum..
> @@ -64,7 +58,7 @@ static inline long do_strnlen_user(const char __user *src, unsigned long count,
> * Uhhuh. We hit 'max'. But was that the user-specified maximum
> * too? If so, return the marker for "too long".
> */
> - if (res >= count)
> + if (res == count)
> return count+1;
>
> /*
> @@ -98,22 +92,14 @@ static inline long do_strnlen_user(const char __user *src, unsigned long count,
> */
> long strnlen_user(const char __user *str, long count)
> {
> - unsigned long max_addr, src_addr;
> -
> if (unlikely(count <= 0))
> return 0;
>
> - max_addr = user_addr_max();
> - src_addr = (unsigned long)untagged_addr(str);
> - if (likely(src_addr < max_addr)) {
> - unsigned long max = max_addr - src_addr;
> + if (user_access_begin(str, count)) {
> long retval;
> -
> - if (user_access_begin(str, max)) {
> - retval = do_strnlen_user(str, count, max);
> - user_access_end();
> - return retval;
> - }
> + retval = do_strnlen_user(str, count);
> + user_access_end();
> + return retval;
> }
> return 0;
> }
> --
> 2.20.1
>