Re: [PATCH] net:sched fix array-index-out-of-bounds in taprio_change

[Pavel Skripkin <paskripkin@gmail.com>]

On 8/11/21 10:44 AM, Eric Dumazet wrote:
> 
> 
> On 8/11/21 7:10 AM, tcs.kernel@gmail.com wrote:
>> From: Haimin Zhang <tcs_kernel@tencent.com>
>> 
>> syzbot report an array-index-out-of-bounds in taprio_change
>> index 16 is out of range for type '__u16 [16]'
>> that's because mqprio->num_tc is lager than TC_MAX_QUEUE,so we check
>> the return value of netdev_set_num_tc.
>> 
>> Reported-by: syzbot+2b3e5fb6c7ef285a94f6@syzkaller.appspotmail.com
>> Signed-off-by: Haimin Zhang <tcs_kernel@tencent.com>
>> ---
>>  net/sched/sch_taprio.c | 4 +++-
>>  1 file changed, 3 insertions(+), 1 deletion(-)
>> 
>> diff --git a/net/sched/sch_taprio.c b/net/sched/sch_taprio.c
>> index 9c79374..1ab2fc9 100644
>> --- a/net/sched/sch_taprio.c
>> +++ b/net/sched/sch_taprio.c
>> @@ -1513,7 +1513,9 @@ static int taprio_change(struct Qdisc *sch, struct nlattr *opt,
>>  	taprio_set_picos_per_byte(dev, q);
>>  
>>  	if (mqprio) {
>> -		netdev_set_num_tc(dev, mqprio->num_tc);
>> +		err = netdev_set_num_tc(dev, mqprio->num_tc);
>> +		if (err)
>> +			goto free_sched;
>>  		for (i = 0; i < mqprio->num_tc; i++)
>>  			netdev_set_tc_queue(dev, i,
>>  					    mqprio->count[i],
>> 
> 
> When was the bug added ?
> 
> Hint: Please provide a Fixes: tag
> 
> taprio_parse_mqprio_opt() already checks :
> 
> /* Verify num_tc is not out of max range */
> if (qopt->num_tc > TC_MAX_QUEUE) {
>      NL_SET_ERR_MSG(extack, "Number of traffic classes is outside valid range");
>      return -EINVAL;
> }
> 
> So what is happening exactly ?
> 
> 

Hi, Eric!

I've looked into this bug, but I decided to write a reproducer for it 
first. Unfortunately, I didn't finish it yesterday, but I have an idea 
about what happened:

taprio_parse_mqprio_opt() may return 0 before qopt->num_tc check:

	/* If num_tc is already set, it means that the user already
	 * configured the mqprio part
	 */
	if (dev->num_tc)
		return 0;

Then taprio_mqprio_cmp() fails here:

	if (!mqprio || mqprio->num_tc != dev->num_tc)
		return -1;

That's why we won't get shift-out-of-bound in taprio_mqprio_cmp().

And finally taprio_change() gets to buggy for with wrong mqprio->num_tc.
I don't know how to reproduce it, but I'll try to finish my reproducer 
this evening.


Does above makes any sense?



With regards,
Pavel Skripkin