svn commit: r273552 - head/sys/kern

[Mateusz Guzik]

On Thu, Oct 23, 2014 at 04:19:05PM -0700, Xin Li wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA512
> 
> On 10/23/14 15:47, Dag-Erling Smørgrav wrote:
> > Xin LI <delphij at FreeBSD.org> writes:
> >> Log: Test if 'env' is NULL before doing memset() and strlen(), 
> >> the caller may pass NULL to freeenv().
> > 
> > If this is in response to a panic in early boot, the real bug is 
> > elsewhere (see r273564).  Adding a NULL check here only hides it.
> 
> Yes that would fix it.  Does this look good to you?
> 
> Index: sys/kern/kern_environment.c
> ===================================================================
> - --- sys/kern/kern_environment.c	(revision 273564)
> +++ sys/kern/kern_environment.c	(working copy)
> @@ -262,7 +262,8 @@ void
>  freeenv(char *env)
>  {
> 
> - -	if (dynamic_kenv && env != NULL) {
> +	MPASS(env != NULL);
> +	if (dynamic_kenv) {
>  		memset(env, 0, strlen(env));
>  		free(env, M_KENV);
>  	}
> 

There are at least 80 consumers of this function. Unless someone is up
to reviewing them all, can we go with a warning + backtrace for the time
being?

-- 
Mateusz Guzik <mjguzik gmail.com>