netdev - Re: [PATCH net-next] net-gro: Prepare GRO stack for the upcoming tunneling support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 12 Dec 2013 02:07:25 +0800
From:	Jerry Chu <hkchu@...gle.com>
To:	Ben Hutchings <bhutchings@...arflare.com>
Cc:	Eric Dumazet <edumazet@...gle.com>,
	Herbert Xu <herbert@...dor.apana.org.au>,
	Or Gerlitz <ogerlitz@...lanox.com>,
	David Miller <davem@...emloft.net>,
	netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next] net-gro: Prepare GRO stack for the upcoming
 tunneling support

On Wed, Dec 11, 2013 at 3:19 AM, Ben Hutchings
<bhutchings@...arflare.com> wrote:
> On Sun, 2013-12-08 at 08:04 -0800, H.K. Jerry Chu wrote:
>> From: Jerry Chu <hkchu@...gle.com>
>>
>> This patch modifies the GRO stack to remove the assumption that
>> only one IP hdr is present in the encapsulation chain. It avoids
>> the use of ip_hdr()/ipv6_hdr() macro in IP's *_gro_receive()/
>> *_gro_complete() functions because there may be more than one IP
>> hdr present in the encapsulation chain when various flavors of IP
>> tunneling support are added. By doing so it also allows multiple
>> level, not just a single level (i.e., with only two IP hdrs like
>> IP-in-IP) of encapsulation to be supported in the future.
>>
>> With this patch, the GRO stack traversing now is mostly based on
>> skb_gro_offset rather than special hdr offsets saved in skb (e.g.,
>> skb->network_header, skb->transport_header,...) As a consequence
>> all but the top layer (which is likely to be the transport layer)
>> must have hdrs of the same length in order for a pkt to be
>> considered for aggregation. Therefore when adding a new layer
>> (e.g., for tunneling), one must check and skip flows (e.g., setting
>> NAPI_GRO_CB(p)->same_flow to 0) that have a different hdr length.
>>
>> Signed-off-by: H.K. Jerry Chu <hkchu@...gle.com>
>> Suggested-by: Eric Dumazet <edumazet@...gle.com>
>> Reviewed-by: Eric Dumazet <edumazet@...gle.com>
> [...]
>> --- a/net/ipv4/tcp_offload.c
>> +++ b/net/ipv4/tcp_offload.c
> [...]
>> @@ -303,16 +308,16 @@ skip_csum:
>>       return tcp_gro_receive(head, skb);
>>  }
>>
>> -static int tcp4_gro_complete(struct sk_buff *skb)
>> +static int tcp4_gro_complete(struct sk_buff *skb, int nhoff)
>
> The offset should definitely be called thoff in this function.

Ok.

>
> [...]
>> --- a/net/ipv6/ip6_offload.c
>> +++ b/net/ipv6/ip6_offload.c
> [...]
>> @@ -198,25 +228,28 @@ static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
>>
>>               iph = ipv6_hdr(skb);
>>       }
>> -
>>       NAPI_GRO_CB(skb)->proto = proto;
>>
>>       flush--;
>> -     nlen = skb_network_header_len(skb);
>> +     nlen = next_header - skb->network_header;
>>
>>       for (p = *head; p; p = p->next) {
>> -             const struct ipv6hdr *iph2;
>> +             struct ipv6hdr *iph2;
>
> Why remove const here?

Done to make the compiler happy but now the call to ipv6_exthdrs_len() may
not be necessary so i'll put the const back.

>
> [...]
>> -static int ipv6_gro_complete(struct sk_buff *skb)
>> +static int ipv6_gro_complete(struct sk_buff *skb, int nhoff)
>>  {
>>       const struct net_offload *ops;
>> -     struct ipv6hdr *iph = ipv6_hdr(skb);
>> +     struct ipv6hdr *iph = (struct ipv6hdr *)(skb->data + nhoff);
>>       int err = -ENOSYS;
>>
>> -     iph->payload_len = htons(skb->len - skb_network_offset(skb) -
>> -                              sizeof(*iph));
>> +     iph->payload_len = htons(skb->len - nhoff - sizeof(*iph));
>>
>>       rcu_read_lock();
>> -     ops = rcu_dereference(inet6_offloads[NAPI_GRO_CB(skb)->proto]);
>> +
>> +     nhoff += sizeof(*iph) + ipv6_exthdrs_len(iph, &ops);
>
> Would be clearer as:
>         int thoff;
>         ...
>         thoff = nhoff + ...

Perhaps...

>
>>       if (WARN_ON(!ops || !ops->callbacks.gro_complete))
>>               goto out_unlock;
>>
>> -     err = ops->callbacks.gro_complete(skb);
>> +     err = ops->callbacks.gro_complete(skb, nhoff);
>>
>>  out_unlock:
>>       rcu_read_unlock();
>> diff --git a/net/ipv6/tcpv6_offload.c b/net/ipv6/tcpv6_offload.c
>> index 6d18157..a107ad6 100644
>> --- a/net/ipv6/tcpv6_offload.c
>> +++ b/net/ipv6/tcpv6_offload.c
>> @@ -66,16 +66,16 @@ skip_csum:
>>       return tcp_gro_receive(head, skb);
>>  }
>>
>> -static int tcp6_gro_complete(struct sk_buff *skb)
>> +static int tcp6_gro_complete(struct sk_buff *skb, int nhoff)
> [...]
>
> Same comment as for tcp4_gro_complete().

Ok.

Jerry

>
> Ben.
>
> --
> Ben Hutchings, Staff Engineer, Solarflare
> Not speaking for my employer; that's the marketing department's job.
> They asked us to note that Solarflare product names are trademarked.
>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html