Re: [PATCH] kernel: audit.c: Add __rcu notation to RCU pointer

[Amol Grover]

On Tue, Nov 26, 2019 at 09:29:25PM -0500, Joel Fernandes wrote:
> On Tue, Nov 26, 2019 at 10:57:23PM +0530, Amol Grover wrote:
> > add __rcu notation to RCU protected global pointer auditd_conn
> > 
> > Fixes multiple instances of sparse error:
> > error: incompatible types in comparison expression
> > (different address spaces)
> > 
> > Signed-off-by: Amol Grover <frextrite@xxxxxxxxx>
> > ---
> >  kernel/audit.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/kernel/audit.c b/kernel/audit.c
> > index da8dc0db5bd3..30e7fc9b8da2 100644
> > --- a/kernel/audit.c
> > +++ b/kernel/audit.c
> > @@ -102,12 +102,14 @@ struct audit_net {
> >   * This struct is RCU protected; you must either hold the RCU lock for reading
> >   * or the associated spinlock for writing.
> >   */
> > -static struct auditd_connection {
> > +struct auditd_connection {
> >  	struct pid *pid;
> >  	u32 portid;
> >  	struct net *net;
> >  	struct rcu_head rcu;
> > -} *auditd_conn = NULL;
> > +};
> > +static struct auditd_connection __rcu *auditd_conn;
> > +RCU_INIT_POINTER(auditd_conn);
> 
> Looks like this causes a build error. Always please build test your patches
> in the very least. And I also did not understand how RCU_INIT_POINTER can
> even be used outside of a function. In C, executable code cannot be outside
> functions.
> 
> Is doing the following not sufficient to fix the sparse issue?
> 
> thanks,
> 
>  - Joel
> 
> ---8<-----------------------
> 
> diff --git a/kernel/audit.c b/kernel/audit.c
> index 49b6049b26ac..c5d4b5a2dea1 100644
> --- a/kernel/audit.c
> +++ b/kernel/audit.c
> @@ -108,8 +108,8 @@ struct auditd_connection {
>  	struct net *net;
>  	struct rcu_head rcu;
>  };
> -static struct auditd_connection __rcu *auditd_conn;
> -RCU_INIT_POINTER(auditd_conn);
> +static struct auditd_connection __rcu *auditd_conn = NULL;

I ran a quick checkpatch and it gave me this error:
ERROR: do not initialise statics to NULL

So in order to fix it I decided to INIT the pointer (and failed)

Should I consider this as a false positive?

Thanks
Amol

> +
>  static DEFINE_SPINLOCK(auditd_conn_lock);
>  
>  /* If audit_rate_limit is non-zero, limit the rate of sending audit records