Re: [PATCH net v4 1/1] net: sched: fix ets qdisc OOB Indexing

[Petr Machata <petrm@nvidia.com>]

Cong Wang <xiyou.wangcong@gmail.com> writes:

> On Sat, Jan 11, 2025 at 09:57:39AM -0500, Jamal Hadi Salim wrote:
>> diff --git a/net/sched/sch_ets.c b/net/sched/sch_ets.c
>> index f80bc05d4c5a..516038a44163 100644
>> --- a/net/sched/sch_ets.c
>> +++ b/net/sched/sch_ets.c
>> @@ -91,6 +91,8 @@ ets_class_from_arg(struct Qdisc *sch, unsigned long arg)
>>  {
>>  	struct ets_sched *q = qdisc_priv(sch);
>>  
>> +	if (arg == 0 || arg > q->nbands)
>> +		return NULL;
>>  	return &q->classes[arg - 1];
>>  }
>
> I must miss something here. Some callers of this function don't handle
> NULL at all, so are you sure it is safe to return NULL for all the
> callers here??
>
> For one quick example:
>
> 322 static int ets_class_dump_stats(struct Qdisc *sch, unsigned long arg,
> 323                                 struct gnet_dump *d)
> 324 {
> 325         struct ets_class *cl = ets_class_from_arg(sch, arg);
> 326         struct Qdisc *cl_q = cl->qdisc;
>
> 'cl' is not checked against NULL before dereferencing it.
>
> There are other cases too, please ensure _all_ of them handle NULL
> correctly.

Yeah, I looked through ets_class_from_arg() callers last week and I
think that besides the one call that needs patching, which already
handles NULL, in all other cases the arg passed to ets_class_from_arg()
comes from class_find, and therefore shouldn't cause the NULL return.