These patches are meant to try and address the fact the VTI tunnels are
currently overwriting the skb->mark value. I am generally happy with the
first two patches, however the third patch still modifies the skb->mark,
though it undoes after the fact.
The main problem I am trying to address is the fact that currently if I use
an v6 over v6 VTI tunnel I cannot receive any traffic on the interface as
the skb->mark is bleeding through and causing the traffic to be dropped.
---
Alexander Duyck (3):
ip_vti/ip6_vti: Do not touch skb->mark on xmit
xfrm: Override skb->mark with tunnel->parm.i_key in xfrm_input
ip_vti/ip6_vti: Preserve skb->mark after rcv_cb call
net/ipv4/ip_vti.c | 14 ++++++++++----
net/ipv6/ip6_vti.c | 13 ++++++++++---
net/xfrm/xfrm_input.c | 17 ++++++++++++++++-
3 files changed, 36 insertions(+), 8 deletions(-)
--
Instead of modifying skb->mark we can simply modify the flowi_mark that is
generated as a result of the xfrm_decode_session. By doing this we don't
need to actually touch the skb->mark and it can be preserved as it passes
out through the tunnel.
Signed-off-by: Alexander Duyck <[email protected]>
---
net/ipv4/ip_vti.c | 5 +++--
net/ipv6/ip6_vti.c | 4 +++-
2 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/ip_vti.c b/net/ipv4/ip_vti.c
index 9f7269f3c54a..4c318e1c13c8 100644
--- a/net/ipv4/ip_vti.c
+++ b/net/ipv4/ip_vti.c
@@ -216,8 +216,6 @@ static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
memset(&fl, 0, sizeof(fl));
- skb->mark = be32_to_cpu(tunnel->parms.o_key);
-
switch (skb->protocol) {
case htons(ETH_P_IP):
xfrm_decode_session(skb, &fl, AF_INET);
@@ -233,6 +231,9 @@ static netdev_tx_t vti_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
return NETDEV_TX_OK;
}
+ /* override mark with tunnel output key */
+ fl.flowi_mark = be32_to_cpu(tunnel->parms.o_key);
+
return vti_xmit(skb, dev, &fl);
}
diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
index ed9d681207fa..104de4da3ff3 100644
--- a/net/ipv6/ip6_vti.c
+++ b/net/ipv6/ip6_vti.c
@@ -495,7 +495,6 @@ vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
int ret;
memset(&fl, 0, sizeof(fl));
- skb->mark = be32_to_cpu(t->parms.o_key);
switch (skb->protocol) {
case htons(ETH_P_IPV6):
@@ -516,6 +515,9 @@ vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
goto tx_err;
}
+ /* override mark with tunnel output key */
+ fl.flowi_mark = be32_to_cpu(t->parms.o_key);
+
ret = vti6_xmit(skb, dev, &fl);
if (ret < 0)
goto tx_err;
This change makes it so that if a tunnel is defined we just use the mark
from the tunnel instead of the mark from the skb header. By doing this we
can avoid the need to set skb->mark inside of the tunnel receive functions.
Signed-off-by: Alexander Duyck <[email protected]>
---
net/xfrm/xfrm_input.c | 17 ++++++++++++++++-
1 file changed, 16 insertions(+), 1 deletion(-)
diff --git a/net/xfrm/xfrm_input.c b/net/xfrm/xfrm_input.c
index 526c4feb3b50..b58286ecd156 100644
--- a/net/xfrm/xfrm_input.c
+++ b/net/xfrm/xfrm_input.c
@@ -13,6 +13,8 @@
#include <net/dst.h>
#include <net/ip.h>
#include <net/xfrm.h>
+#include <net/ip_tunnels.h>
+#include <net/ip6_tunnel.h>
static struct kmem_cache *secpath_cachep __read_mostly;
@@ -186,6 +188,7 @@ int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
struct xfrm_state *x = NULL;
xfrm_address_t *daddr;
struct xfrm_mode *inner_mode;
+ u32 mark = skb->mark;
unsigned int family;
int decaps = 0;
int async = 0;
@@ -203,6 +206,18 @@ int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
XFRM_SPI_SKB_CB(skb)->daddroff);
family = XFRM_SPI_SKB_CB(skb)->family;
+ /* if tunnel is present override skb->mark value with tunnel i_key */
+ if (XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4) {
+ switch (family) {
+ case AF_INET:
+ mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4->parms.i_key);
+ break;
+ case AF_INET6:
+ mark = be32_to_cpu(XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6->parms.i_key);
+ break;
+ }
+ }
+
/* Allocate new secpath or COW existing one. */
if (!skb->sp || atomic_read(&skb->sp->refcnt) != 1) {
struct sec_path *sp;
@@ -229,7 +244,7 @@ int xfrm_input(struct sk_buff *skb, int nexthdr, __be32 spi, int encap_type)
goto drop;
}
- x = xfrm_state_lookup(net, skb->mark, daddr, spi, nexthdr, family);
+ x = xfrm_state_lookup(net, mark, daddr, spi, nexthdr, family);
if (x == NULL) {
XFRM_INC_STATS(net, LINUX_MIB_XFRMINNOSTATES);
xfrm_audit_state_notfound(skb, family, spi, seq);
The vti6_rcv_cb and vti_rcv_cb calls were leaving the skb->mark modified
after completing the function. This resulted in the original skb->mark
value being lost. Since we only need skb->mark to be set for
xfrm_policy_check we can pull the assignment into the rcv_cb calls and then
just restore the original mark after xfrm_policy_check has been completed.
Signed-off-by: Alexander Duyck <[email protected]>
---
net/ipv4/ip_vti.c | 9 +++++++--
net/ipv6/ip6_vti.c | 9 +++++++--
2 files changed, 14 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/ip_vti.c b/net/ipv4/ip_vti.c
index 4c318e1c13c8..0c152087ca15 100644
--- a/net/ipv4/ip_vti.c
+++ b/net/ipv4/ip_vti.c
@@ -65,7 +65,6 @@ static int vti_input(struct sk_buff *skb, int nexthdr, __be32 spi,
goto drop;
XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4 = tunnel;
- skb->mark = be32_to_cpu(tunnel->parms.i_key);
return xfrm_input(skb, nexthdr, spi, encap_type);
}
@@ -91,6 +90,8 @@ static int vti_rcv_cb(struct sk_buff *skb, int err)
struct pcpu_sw_netstats *tstats;
struct xfrm_state *x;
struct ip_tunnel *tunnel = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip4;
+ u32 orig_mark = skb->mark;
+ int ret;
if (!tunnel)
return 1;
@@ -107,7 +108,11 @@ static int vti_rcv_cb(struct sk_buff *skb, int err)
x = xfrm_input_state(skb);
family = x->inner_mode->afinfo->family;
- if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
+ skb->mark = be32_to_cpu(tunnel->parms.i_key);
+ ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
+ skb->mark = orig_mark;
+
+ if (!ret)
return -EPERM;
skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(skb->dev)));
diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
index 104de4da3ff3..ff3bd863fa03 100644
--- a/net/ipv6/ip6_vti.c
+++ b/net/ipv6/ip6_vti.c
@@ -322,7 +322,6 @@ static int vti6_rcv(struct sk_buff *skb)
}
XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
- skb->mark = be32_to_cpu(t->parms.i_key);
rcu_read_unlock();
@@ -342,6 +341,8 @@ static int vti6_rcv_cb(struct sk_buff *skb, int err)
struct pcpu_sw_netstats *tstats;
struct xfrm_state *x;
struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
+ u32 orig_mark = skb->mark;
+ int ret;
if (!t)
return 1;
@@ -358,7 +359,11 @@ static int vti6_rcv_cb(struct sk_buff *skb, int err)
x = xfrm_input_state(skb);
family = x->inner_mode->afinfo->family;
- if (!xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family))
+ skb->mark = be32_to_cpu(t->parms.i_key);
+ ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
+ skb->mark = orig_mark;
+
+ if (!ret)
return -EPERM;
skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
On Wed, May 27, 2015 at 07:16:37AM -0700, Alexander Duyck wrote:
> These patches are meant to try and address the fact the VTI tunnels are
> currently overwriting the skb->mark value. I am generally happy with the
> first two patches, however the third patch still modifies the skb->mark,
> though it undoes after the fact.
>
> The main problem I am trying to address is the fact that currently if I use
> an v6 over v6 VTI tunnel I cannot receive any traffic on the interface as
> the skb->mark is bleeding through and causing the traffic to be dropped.
>
> ---
>
> Alexander Duyck (3):
> ip_vti/ip6_vti: Do not touch skb->mark on xmit
> xfrm: Override skb->mark with tunnel->parm.i_key in xfrm_input
> ip_vti/ip6_vti: Preserve skb->mark after rcv_cb call
All applied to the ipsec tree, thanks a lot Alexander!