[PATCH 07/28] gss_krb5: Use random value to initialize confounder

[Chuck Lever]

On Mar 31, 2008, at 10:31 AM, Kevin Coffman wrote:
> Initialize the value used for the confounder to a random value
> rather than starting from zero.
> Allow for confounders of length 8 or 16 (which will be needed for  
> AES).
>
> Signed-off-by: Kevin Coffman <kwc at citi.umich.edu>
> ---
>
>  net/sunrpc/auth_gss/gss_krb5_wrap.c |   17 ++++++++++++++---
>  1 files changed, 14 insertions(+), 3 deletions(-)
>
> diff --git a/net/sunrpc/auth_gss/gss_krb5_wrap.c b/net/sunrpc/ 
> auth_gss/gss_krb5_wrap.c
> index 3cd99a7..89338a4 100644
> --- a/net/sunrpc/auth_gss/gss_krb5_wrap.c
> +++ b/net/sunrpc/auth_gss/gss_krb5_wrap.c
> @@ -88,7 +88,7 @@ out:
>  }
>
>  static inline void
> -make_confounder(char *p, int blocksize)
> +make_confounder(char *p, int conflen)

I would use an unsigned type for conflen... "size_t" is probably  
better than "int".

It might also be helpful to strip off the "inline" directive and let  
the compiler decide whether inlining is a useful optimization.

>  {
>  	static u64 i = 0;
>  	u64 *q = (u64 *)p;
> @@ -102,8 +102,19 @@ make_confounder(char *p, int blocksize)
>  	 * uniqueness would mean worrying about atomicity and rollover,  
> and I
>  	 * don't care enough. */
>
> -	BUG_ON(blocksize != 8);
> -	*q = i++;
> +	/* initialize to random value */
> +	if (i == 0) {
> +		i = random32();
> +		i = (i << 32) | random32();
> +	}

Nit: The RPC layer usually uses net_random() instead of calling  
random32() directly.

> +
> +	if (conflen == 8)
> +		*q = i++;
> +	else if (conflen == 16) {
> +		*q++ = i++;
> +		*q++ = i++;
> +	} else
> +		BUG();

Nit: a "switch" would be more clear and slightly easier to maintain.

>  }
>
>  /* Assumptions: the head and tail of inbuf are ours to play with.

--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com