Re: [PATCH v1] net: ipv6: fix racey clock check in route cache aging logic

From: Eric Dumazet
Date: Thu Oct 25 2018 - 18:15:09 EST




On 10/25/2018 02:46 PM, Brendan Higgins wrote:
> On Thu, Oct 25, 2018 at 2:40 PM Eric Dumazet <eric.dumazet@xxxxxxxxx> wrote:
>> On 10/25/2018 02:13 PM, Brendan Higgins wrote:
> <snip>
>>>
>>> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
>>> index 2a7423c394560..54d28b91fd840 100644
>>> --- a/net/ipv6/route.c
>>> +++ b/net/ipv6/route.c
>>> @@ -1734,7 +1734,7 @@ static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
>>> rt6_remove_exception(bucket, rt6_ex);
>>> return;
>>> }
>>> - } else if (time_after(jiffies, rt->dst.expires)) {
>>> + } else if (time_after(now, rt->dst.expires)) {
>>> RT6_TRACE("purging expired route %p\n", rt);
>>> rt6_remove_exception(bucket, rt6_ex);
>>> return;
>>>
>>
>>
>> I do not think there is a bug here ?
>>
>> As a matter of fact, using the latest value of jiffies is probably better,
>> since in some cases the @now variable could be quite in the past.
>
> Then why do we pass the `now` parameter in at all and use it at all,
> like here: https://elixir.bootlin.com/linux/latest/source/net/ipv6/route.c#L1764
> ?
>
> I am still skeptical that we should check jiffies in each check, but
> we should at least be consistent.

Well, this is a case where we do not really care.

When a bug is fixed (you added a Fixes: tag which is good), we want
to understand the real problem that needs to be fixed on stable kernels.

Since this does not seem to be a real issue, I would suggest you send a cleanup
patch when net-next is open (few days after linux-4.20-rc1 is release)