We should check the parameter mlen before using 'mlen - 1'
expression for the 'mesg' array index.

Signed-off-by: Ma Ke <make24@iscas.ac.cn>
---
 net/sunrpc/svcauth_unix.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/sunrpc/svcauth_unix.c b/net/sunrpc/svcauth_unix.c
index 04b45588ae6f..816bf56597dd 100644
--- a/net/sunrpc/svcauth_unix.c
+++ b/net/sunrpc/svcauth_unix.c
@@ -196,7 +196,7 @@ static int ip_map_parse(struct cache_detail *cd,
 	struct auth_domain *dom;
 	time64_t expiry;
 
-	if (mesg[mlen-1] != '\n')
+	if (mlen && mesg[mlen - 1] != '\n')
 		return -EINVAL;
 	mesg[mlen-1] = 0;
 
-- 
2.25.1

On Mon, 24 Jun 2024, Ma Ke wrote:
> We should check the parameter mlen before using 'mlen - 1'
> expression for the 'mesg' array index.

There is no need.  This function is only called from cache_do_downcall()
and that function already checks for zero.

That function is only called from cache_downcall() which checks the
size_t count is not >= 32768 so the fact that it is cast to an int for
the ->cache_parse function cannot cause and overflow.

I wouldn't object to ->cache_parse() and qword_get() and maybe others
having their len parameter changed from int to size_t.  
But adding this extra test on mlen add no value.

Thanks,
NeilBrown


> 
> Signed-off-by: Ma Ke <make24@iscas.ac.cn>
> ---
>  net/sunrpc/svcauth_unix.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/net/sunrpc/svcauth_unix.c b/net/sunrpc/svcauth_unix.c
> index 04b45588ae6f..816bf56597dd 100644
> --- a/net/sunrpc/svcauth_unix.c
> +++ b/net/sunrpc/svcauth_unix.c
> @@ -196,7 +196,7 @@ static int ip_map_parse(struct cache_detail *cd,
>  	struct auth_domain *dom;
>  	time64_t expiry;
>  
> -	if (mesg[mlen-1] != '\n')
> +	if (mlen && mesg[mlen - 1] != '\n')
>  		return -EINVAL;
>  	mesg[mlen-1] = 0;
>  
> -- 
> 2.25.1
> 
>