[PATCH RFC 1/4] audit: Implement bpf_audit_log_*() wrappers

[Mickaël Salaün]

On Mon, Mar 16, 2026 at 06:14:40PM -0400, David Windsor wrote:
> Hi Frederick,
> 
> On Wed, Mar 11, 2026 at 04:31:17PM -0500, Frederick Lawler wrote:
> > +__bpf_kfunc int bpf_audit_log_cause(struct bpf_audit_context *ac,
> > +				    const char *cause__str)
> > +{
> > +	if (log_once(ac, BIT_ULL(LSM_AUDIT_DATA_CAUSE)))
> > +		return -EINVAL;
> > +
> > +	audit_log_format(ac->ab, " cause=");
> > +	audit_log_untrustedstring(ac->ab, cause__str);
> > +	return 0;
> > +}
> 
> Rather than putting everything in the cause field, could we perhaps
> have a separate kfunc here that appends normal stringpairs (not
> format strings) to the audit record:
> 
>   bpf_audit_log_str(ac, "result", "denied");
>   bpf_audit_log_str(ac, "op", "read");
>   bpf_audit_log_str(ac, "scontext", ctx_str);

That would mean arbitrary audit keys (and values), which would not be
acceptable (i.e. no consistency).

> 
> I know you didn't want to wrap audit_log_format(), which makes sense,
> this would be a midway point between that and stuffing everything in
> one field.