[v2] exfat: fix out-of-bounds in exfat_nls_to_ucs2()

[PATCH v2] exfat: fix out-of-bounds in exfat_nls_to_ucs2()

Posted by Jeongjun Park 2 months, 1 week ago

In exfat_nls_to_ucs2(), if there is no NLS loss and the char-to-ucs2
conversion is successfully completed, the variable "i" will have the same
value as len. 

However, exfat_nls_to_ucs2() checks p_cstring[i] to determine whether nls
is lost immediately after the while loop ends, so if len is FSLABEL_MAX,
"i" will also be FSLABEL_MAX immediately after the while loop ends,
resulting in an out-of-bounds read of 1 byte from the p_cstring stack
memory.

Therefore, to prevent this and properly determine whether nls has been
lost, it should be modified to check if "i" and len are equal, rather than
dereferencing p_cstring.

Cc: <stable@vger.kernel.org>
Reported-by: syzbot+98cc76a76de46b3714d4@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=98cc76a76de46b3714d4
Fixes: 370e812b3ec1 ("exfat: add nls operations")
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
---
 fs/exfat/nls.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/exfat/nls.c b/fs/exfat/nls.c
index 8243d94ceaf4..de06abe426d7 100644
--- a/fs/exfat/nls.c
+++ b/fs/exfat/nls.c
@@ -616,7 +616,7 @@ static int exfat_nls_to_ucs2(struct super_block *sb,
 		unilen++;
 	}
 
-	if (p_cstring[i] != '\0')
+	if (i != len)
 		lossy |= NLS_NAME_OVERLEN;
 
 	*uniname = '\0';
--

Re: [PATCH v2] exfat: fix out-of-bounds in exfat_nls_to_ucs2()

Posted by Namjae Jeon 2 months, 1 week ago

On Fri, Oct 10, 2025 at 2:04 PM Jeongjun Park <aha310510@gmail.com> wrote:
>
> In exfat_nls_to_ucs2(), if there is no NLS loss and the char-to-ucs2
> conversion is successfully completed, the variable "i" will have the same
> value as len.
>
> However, exfat_nls_to_ucs2() checks p_cstring[i] to determine whether nls
> is lost immediately after the while loop ends, so if len is FSLABEL_MAX,
> "i" will also be FSLABEL_MAX immediately after the while loop ends,
> resulting in an out-of-bounds read of 1 byte from the p_cstring stack
> memory.
>
> Therefore, to prevent this and properly determine whether nls has been
> lost, it should be modified to check if "i" and len are equal, rather than
> dereferencing p_cstring.
>
> Cc: <stable@vger.kernel.org>
> Reported-by: syzbot+98cc76a76de46b3714d4@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=98cc76a76de46b3714d4
> Fixes: 370e812b3ec1 ("exfat: add nls operations")
> Signed-off-by: Jeongjun Park <aha310510@gmail.com>
> ---
>  fs/exfat/nls.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/exfat/nls.c b/fs/exfat/nls.c
> index 8243d94ceaf4..de06abe426d7 100644
> --- a/fs/exfat/nls.c
> +++ b/fs/exfat/nls.c
> @@ -616,7 +616,7 @@ static int exfat_nls_to_ucs2(struct super_block *sb,
>                 unilen++;
>         }
>
> -       if (p_cstring[i] != '\0')
> +       if (i != len)
>                 lossy |= NLS_NAME_OVERLEN;
As I said before, the core problem is passing an incorrect length in the ioctl
FS_IOC_SETFSLABEL call. While your change may fix the symptom,
it doesn't fix the root cause.
Furthermore, NLS_NAME_OVERLEN macro is unnecessary.
That will result in different errors when creating a file with trailing periods
with utf8 and the other iocharset.
Please remove it and make the same error handling for maximum length
consistent with exfat_utf8_to_utf16().

>
>         *uniname = '\0';
> --

Re: [PATCH v2] exfat: fix out-of-bounds in exfat_nls_to_ucs2()

Posted by Pali Rohár 2 months, 1 week ago

On Friday 10 October 2025 14:03:29 Jeongjun Park wrote:
> In exfat_nls_to_ucs2(), if there is no NLS loss and the char-to-ucs2
> conversion is successfully completed, the variable "i" will have the same
> value as len. 
> 
> However, exfat_nls_to_ucs2() checks p_cstring[i] to determine whether nls
> is lost immediately after the while loop ends, so if len is FSLABEL_MAX,
> "i" will also be FSLABEL_MAX immediately after the while loop ends,
> resulting in an out-of-bounds read of 1 byte from the p_cstring stack
> memory.
> 
> Therefore, to prevent this and properly determine whether nls has been
> lost, it should be modified to check if "i" and len are equal, rather than
> dereferencing p_cstring.
> 
> Cc: <stable@vger.kernel.org>
> Reported-by: syzbot+98cc76a76de46b3714d4@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=98cc76a76de46b3714d4
> Fixes: 370e812b3ec1 ("exfat: add nls operations")
> Signed-off-by: Jeongjun Park <aha310510@gmail.com>
> ---
>  fs/exfat/nls.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/exfat/nls.c b/fs/exfat/nls.c
> index 8243d94ceaf4..de06abe426d7 100644
> --- a/fs/exfat/nls.c
> +++ b/fs/exfat/nls.c
> @@ -616,7 +616,7 @@ static int exfat_nls_to_ucs2(struct super_block *sb,
>  		unilen++;
>  	}
>  
> -	if (p_cstring[i] != '\0')
> +	if (i != len)
>  		lossy |= NLS_NAME_OVERLEN;
>  
>  	*uniname = '\0';
> --

Looks good for me,

Reviewed-by: Pali Rohár <pali@kernel.org>