Re: [syzbot] WARNING: ODEBUG bug in route4_destroy

[Zhen Chen <chenzhen126@huawei.com>]

On 22/6/14 22:46, Dan Carpenter wrote:
> On Tue, Jun 14, 2022 at 10:35:44PM +0800, 'Zhen Chen' via syzkaller-bugs wrote:
>>
>> This looks like  route4_destroy is deleting the 'fold' which has been
>> freed by tcf_queue_work in route4_change. It means 'fold' is still in
>> the table.
>> I have tested this patch on syzbot and it works well, but I am not
>> sure whether it will introduce other issues...
>>
>> diff --git a/net/sched/cls_route.c b/net/sched/cls_route.c
>> index a35ab8c27866..758c21f9d628 100644
>> --- a/net/sched/cls_route.c
>> +++ b/net/sched/cls_route.c
>> @@ -526,7 +526,7 @@ static int route4_change(struct net *net, struct sk_buff *in_skb,
>>  	rcu_assign_pointer(f->next, f1);
>>  	rcu_assign_pointer(*fp, f);
>>  
>> -	if (fold && fold->handle && f->handle != fold->handle) {
>> +	if (fold && f->handle != fold->handle) {
>                                  ^^^^^^^^^^^^
> There is still a dereference here so your patch doesn't make sense. :/
> 
> regards,
> dan carpenter

Thanks for your reply but I think the dereference may not be the point.
If fold->handle equals 0, it will not be removed from the hash table, but afterwards the old filter will be freed because it only checks the pointer 'fold' is null or not.

if (fold) {
	tcf_unbind_filter(tp, &fold->res);
	tcf_exts_get_net(&fold->exts);
	tcf_queue_work(&fold->rwork, route4_delete_filter_work);
}

So my patch simply eliminates the handle judgement and it seems to work fine on syzbot.
If I misunderstood anything, pleaese let me know :)   Thanks!