Re: [PATCH net-next v5 02/10] net: sched: act_sample: add action cookie to sample

[Eelco Chaudron]

On 26 Jun 2024, at 22:06, Adrián Moreno wrote:

> On Wed, Jun 26, 2024 at 04:28:01PM GMT, Eelco Chaudron wrote:
>>
>>
>> On 25 Jun 2024, at 22:51, Adrian Moreno wrote:
>>
>>> If the action has a user_cookie, pass it along to the sample so it can
>>> be easily identified.
>>>
>>> Signed-off-by: Adrian Moreno <amorenoz@xxxxxxxxxx>
>>> ---
>>>  net/sched/act_sample.c | 12 ++++++++++++
>>>  1 file changed, 12 insertions(+)
>>>
>>> diff --git a/net/sched/act_sample.c b/net/sched/act_sample.c
>>> index a69b53d54039..2ceb4d141b71 100644
>>> --- a/net/sched/act_sample.c
>>> +++ b/net/sched/act_sample.c
>>> @@ -167,7 +167,9 @@ TC_INDIRECT_SCOPE int tcf_sample_act(struct sk_buff *skb,
>>>  {
>>>  	struct tcf_sample *s = to_sample(a);
>>>  	struct psample_group *psample_group;
>>> +	u8 cookie_data[TC_COOKIE_MAX_SIZE];
>>>  	struct psample_metadata md = {};
>>> +	struct tc_cookie *user_cookie;
>>>  	int retval;
>>>
>>>  	tcf_lastuse_update(&s->tcf_tm);
>>> @@ -189,6 +191,16 @@ TC_INDIRECT_SCOPE int tcf_sample_act(struct sk_buff *skb,
>>>  		if (skb_at_tc_ingress(skb) && tcf_sample_dev_ok_push(skb->dev))
>>>  			skb_push(skb, skb->mac_len);
>>>
>>> +		rcu_read_lock();
>>> +		user_cookie = rcu_dereference(a->user_cookie);
>>> +		if (user_cookie) {
>>> +			memcpy(cookie_data, user_cookie->data,
>>> +			       user_cookie->len);
>>
>> Maybe I’m over paranoid, but can we assume user_cookie->len, will not be larger than TC_COOKIE_MAX_SIZE?
>> Or should we do something like min(user_cookie->len, sizeof(cookie_data))
>>
>
> I think it's good to be paranoid with this kind of things. I do,
> however, think it should be safe to use. The cookie is extracted from
> the netlink attribute directly and its length is verified with the
> nla_policy [1]. So nothing that comes into the kernel should be larger
> than TC_COOKIE_MAX_SIZE.

ACK, confirmed that [1] seems to be the only way to set the cookie. So this patch seems fine to me too.

Acked-by: Eelco Chaudron <echaudro@xxxxxxxxxx>

> I guess if there is some previous bug that allows for the size to get
> corrupted, then this might happen but doing those kind of checks in the
> fast path seems a bit excessive. For example, Ilya argued in v2 [2] that
> we should avoid zeroing "u8 cookie_data[TC_COOKIE_MAX_SIZE]" to safe the
> unneeded cycles.
>
> [1] https://github.com/torvalds/linux/blob/55027e689933ba2e64f3d245fb1ff185b3e7fc81/net/sched/act_api.c#L1299
> [2] https://patchwork.kernel.org/project/netdevbpf/patch/20240603185647.2310748-3-amorenoz@xxxxxxxxxx/
>
> Thanks.
> Adrián
>
>>> +			md.user_cookie = cookie_data;
>>> +			md.user_cookie_len = user_cookie->len;
>>> +		}
>>> +		rcu_read_unlock();
>>> +
>>>  		md.trunc_size = s->truncate ? s->trunc_size : skb->len;
>>>  		psample_sample_packet(psample_group, skb, s->rate, &md);
>>>
>>> --
>>> 2.45.1
>>