lkml.org 
[lkml]   [2024]   [Apr]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH net v1 2/2] net: gro: add p_off param in *_gro_complete
Richard Gobert wrote:
> Willem de Bruijn wrote:
> > Richard Gobert wrote:
> >> Commits a602456 ("udp: Add GRO functions to UDP socket") and 57c67ff ("udp:
> >> additional GRO support") introduce incorrect usage of {ip,ipv6}_hdr in the
> >> complete phase of gro. The functions always return skb->network_header,
> >> which in the case of encapsulated packets at the gro complete phase, is
> >> always set to the innermost L3 of the packet. That means that calling
> >> {ip,ipv6}_hdr for skbs which completed the GRO receive phase (both in
> >> gro_list and *_gro_complete) when parsing an encapsulated packet's _outer_
> >> L3/L4 may return an unexpected value.
> >>
> >> This incorrect usage leads to a bug in GRO's UDP socket lookup.
> >> udp{4,6}_lib_lookup_skb functions use ip_hdr/ipv6_hdr respectively. These
> >> *_hdr functions return network_header which will point to the innermost L3,
> >> resulting in the wrong offset being used in __udp{4,6}_lib_lookup with
> >> encapsulated packets.
> >>
> >> To fix this issue p_off param is used in *_gro_complete to pass off the
> >> offset of the previous layer.
> >>
> >> Reproduction example:
> >>
> >> Endpoint configuration example (fou + local address bind)
> >>
> >> # ip fou add port 6666 ipproto 4
> >> # ip link add name tun1 type ipip remote 2.2.2.1 local 2.2.2.2 encap fou encap-dport 5555 encap-sport 6666 mode ipip
> >> # ip link set tun1 up
> >> # ip a add 1.1.1.2/24 dev tun1
> >>
> >> Netperf TCP_STREAM result on net-next before patch is applied:
> >>
> >> net-next main, GRO enabled:
> >> $ netperf -H 1.1.1.2 -t TCP_STREAM -l 5
> >> Recv Send Send
> >> Socket Socket Message Elapsed
> >> Size Size Size Time Throughput
> >> bytes bytes bytes secs. 10^6bits/sec
> >>
> >> 131072 16384 16384 5.28 2.37
> >>
> >> net-next main, GRO disabled:
> >> $ netperf -H 1.1.1.2 -t TCP_STREAM -l 5
> >> Recv Send Send
> >> Socket Socket Message Elapsed
> >> Size Size Size Time Throughput
> >> bytes bytes bytes secs. 10^6bits/sec
> >>
> >> 131072 16384 16384 5.01 2745.06
> >>
> >> patch applied, GRO enabled:
> >> $ netperf -H 1.1.1.2 -t TCP_STREAM -l 5
> >> Recv Send Send
> >> Socket Socket Message Elapsed
> >> Size Size Size Time Throughput
> >> bytes bytes bytes secs. 10^6bits/sec
> >>
> >> 131072 16384 16384 5.01 2877.38
> >>
> >> Fixes: 57c67ff4bd92 ("udp: additional GRO support")
> >> Suggested-by: Eric Dumazet <edumazet@google.com>
> >> Signed-off-by: Richard Gobert <richardbgobert@gmail.com>
> >
> >> diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
> >> index 163f94a5a58f..9c18a39b0d0c 100644
> >> --- a/drivers/net/geneve.c
> >> +++ b/drivers/net/geneve.c
> >> @@ -555,7 +555,7 @@ static struct sk_buff *geneve_gro_receive(struct sock *sk,
> >> }
> >>
> >> static int geneve_gro_complete(struct sock *sk, struct sk_buff *skb,
> >> - int nhoff)
> >> + int p_off, int nhoff)
> >> {
> >> struct genevehdr *gh;
> >> struct packet_offload *ptype;
> >> @@ -569,11 +569,12 @@ static int geneve_gro_complete(struct sock *sk, struct sk_buff *skb,
> >>
> >> /* since skb->encapsulation is set, eth_gro_complete() sets the inner mac header */
> >> if (likely(type == htons(ETH_P_TEB)))
> >> - return eth_gro_complete(skb, nhoff + gh_len);
> >> + return eth_gro_complete(skb, p_off, nhoff + gh_len);
> >
> > Since the new field to the callback is only used between IP and
> > transport layer callback implementations, I think the others should
> > just return zero, to make it clear that the value is unused.
> >
>
> Got it. I'll push it in v2.
>
> > I still think that if the only issue is with udp, we can just special
> > case those and pass the nhoff instead of thoff in the existing one
> > available offset field, and compute the transport offset in the udp
> > function. For much less code churn. But unless anyone else agrees you
> > can ignore that suggestion.
> >
> >> -int inet_gro_complete(struct sk_buff *skb, int nhoff)
> >> +int inet_gro_complete(struct sk_buff *skb, int prior_off, int nhoff)
> >> {
> >> struct iphdr *iph = (struct iphdr *)(skb->data + nhoff);
> >> const struct net_offload *ops;
> >> @@ -1667,17 +1667,17 @@ int inet_gro_complete(struct sk_buff *skb, int nhoff)
> >> */
> >> err = INDIRECT_CALL_2(ops->callbacks.gro_complete,
> >> tcp4_gro_complete, udp4_gro_complete,
> >> - skb, nhoff + sizeof(*iph));
> >> + skb, nhoff, nhoff + sizeof(*iph));
> >
> > Identation change
> >
>
> Will fix in v2 as well, thanks!
>
> >> struct sock *udp4_lib_lookup_skb(const struct sk_buff *skb,
> >> + int nhoff,
> >> __be16 sport, __be16 dport)
> >> {
> >> - const struct iphdr *iph = ip_hdr(skb);
> >> + const struct iphdr *iph = (const struct iphdr *)(skb->data + nhoff);
> >
> > How about instead just pass the saddr and daddr and leave the iph
> > pointer to the caller (which also computes the udph pointer).
>
> Here's a snippet I wrote - could you make sure this is what you mean?
>
> BTW I couldn't find a union type which contains both in_addr and in6_addr
> and is generic enough (I wrote udp_offload_addr which is similar to
> tcp_ao_addr, is there a more generic one? Should it be created?)

Thanks for the sketch. That is a bit more complex than I had in mind.

>
> diff --git a/include/net/gro.h b/include/net/gro.h
> index ebead1d642b4..56e5e21feb00 100644
> --- a/include/net/gro.h
> +++ b/include/net/gro.h
> @@ -405,7 +405,8 @@ INDIRECT_CALLABLE_DECLARE(int udp6_gro_complete(struct sk_buff *, int, int));
>
> struct sk_buff *udp_gro_receive(struct list_head *head, struct sk_buff *skb,
> struct udphdr *uh, struct sock *sk);
> -int udp_gro_complete(struct sk_buff *skb, int nhoff, int thoff,
> +int udp_gro_complete(struct sk_buff *skb, const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr, int thoff,
> udp_lookup_t lookup);
>
> static inline struct udphdr *udp_gro_udphdr(struct sk_buff *skb)
> diff --git a/include/net/udp.h b/include/net/udp.h
> index 601d1c3b677a..5f7224e6eb1e 100644
> --- a/include/net/udp.h
> +++ b/include/net/udp.h
> @@ -62,6 +62,11 @@ struct udp_hslot {
> spinlock_t lock;
> } __attribute__((aligned(2 * sizeof(long))));
>
> +union udp_offload_addr {
> + __be32 ipaddr;
> + struct in6_addr ip6addr;
> +};
> +

No need for such a union if the code just either calls
udp4_lib_lookup_skb or udp6_lib_lookup_skb
> /**
> * struct udp_table - UDP table
> *
> @@ -166,7 +171,9 @@ static inline void udp_csum_pull_header(struct sk_buff *skb)
> UDP_SKB_CB(skb)->cscov -= sizeof(struct udphdr);
> }
>
> -typedef struct sock *(*udp_lookup_t)(const struct sk_buff *skb, int nhoff,
> +typedef struct sock *(*udp_lookup_t)(const struct sk_buff *skb,
> + const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr,
> __be16 sport, __be16 dport);
>
> void udp_v6_early_demux(struct sk_buff *skb);
> @@ -301,7 +308,9 @@ struct sock *udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
> struct sock *__udp4_lib_lookup(struct net *net, __be32 saddr, __be16 sport,
> __be32 daddr, __be16 dport, int dif, int sdif,
> struct udp_table *tbl, struct sk_buff *skb);
> -struct sock *udp4_lib_lookup_skb(const struct sk_buff *skb, int nhoff,
> +struct sock *udp4_lib_lookup_skb(const struct sk_buff *skb,
> + const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr,
> __be16 sport, __be16 dport);
> struct sock *udp6_lib_lookup(struct net *net,
> const struct in6_addr *saddr, __be16 sport,
> @@ -312,7 +321,9 @@ struct sock *__udp6_lib_lookup(struct net *net,
> const struct in6_addr *daddr, __be16 dport,
> int dif, int sdif, struct udp_table *tbl,
> struct sk_buff *skb);
> -struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb, int nhoff,
> +struct sock *udp6_lib_lookup_skb(const struct sk_buff *skb,
> + const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr,
> __be16 sport, __be16 dport);
> int udp_read_skb(struct sock *sk, skb_read_actor_t recv_actor);
>
> diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> index 306d2a78fefa..151c3adecc21 100644
> --- a/net/ipv4/udp.c
> +++ b/net/ipv4/udp.c
> @@ -529,17 +529,18 @@ static inline struct sock *__udp4_lib_lookup_skb(struct sk_buff *skb,
> inet_sdif(skb), udptable, skb);
> }
>
> -struct sock *udp4_lib_lookup_skb(const struct sk_buff *skb, int nhoff,
> +struct sock *udp4_lib_lookup_skb(const struct sk_buff *skb,
> + const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr,
> __be16 sport, __be16 dport)
> {
> - const struct iphdr *iph = (const struct iphdr *)(skb->data + nhoff);
> struct net *net = dev_net(skb->dev);
> int iif, sdif;
>
> inet_get_iif_sdif(skb, &iif, &sdif);
>
> - return __udp4_lib_lookup(net, iph->saddr, sport,
> - iph->daddr, dport, iif,
> + return __udp4_lib_lookup(net, saddr->ipaddr, sport,
> + daddr->ipaddr, dport, iif,
> sdif, net->ipv4.udp_table, NULL);
> }
>
> diff --git a/net/ipv4/udp_offload.c b/net/ipv4/udp_offload.c
> index dcf8124b1a6a..0f05c7ed05d3 100644
> --- a/net/ipv4/udp_offload.c
> +++ b/net/ipv4/udp_offload.c
> @@ -689,7 +689,8 @@ static int udp_gro_complete_segment(struct sk_buff *skb)
> return 0;
> }
>
> -int udp_gro_complete(struct sk_buff *skb, int nhoff, int thoff,
> +int udp_gro_complete(struct sk_buff *skb, const union udp_offload_addr *saddr,
> + const union udp_offload_addr *daddr, int thoff,
> udp_lookup_t lookup)
> {
> struct udphdr *uh = (struct udphdr *)(skb->data + thoff);
> @@ -700,8 +701,8 @@ int udp_gro_complete(struct sk_buff *skb, int nhoff, int thoff,
> uh->len = newlen;
>
> sk = INDIRECT_CALL_INET(lookup, udp6_lib_lookup_skb,
> - udp4_lib_lookup_skb, skb, nhoff, uh->source,
> - uh->dest);
> + udp4_lib_lookup_skb, skb, saddr, daddr,
> + uh->source, uh->dest);

Since this INDIRECT_CALL_INET needs to know about the definitions of
udp4_lib_lookup_skb and udp6_lib_lookup_skb anyway, we can just get
rid of the whole udp_lookup_t type and function pointer passing?

Or move the entire lookup to udp4_gro_complete/udp6_gro_complete and
pass the sk to udp_gro_complete.

\
 
 \ /
  Last update: 2024-04-17 21:40    [W:0.075 / U:0.576 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site