Add support for remote checksum offload in VXLAN. This uses a
reserved bit to indicate that RCO is being done, and uses the low order
reserved eight bits of the VNI to hold the start and offset values in a
compressed manner.
Start is encoded in the low order seven bits of VNI. This is start >> 1
so that the checksum start offset is 0-254 using even values only.
Checksum offset (transport checksum field) is indicated in the high
order bit in the low order byte of the VNI. If the bit is set, the
checksum field is for UDP (so offset = start + 6), else checksum
field is for TCP (so offset = start + 16). Only TCP and UDP are
supported in this implementation.
Remote checksum offload for VXLAN is described in:
https://tools.ietf.org/html/draft-herbert-vxlan-rco-00
Tested by running 200 TCP_STREAM connections with VXLAN (over IPv4).
With UDP checksums and Remote Checksum Offload
IPv4
Client
11.84% CPU utilization
Server
12.96% CPU utilization
9197 Mbps
IPv6
Client
12.46% CPU utilization
Server
14.48% CPU utilization
8963 Mbps
With UDP checksums, no remote checksum offload
IPv4
Client
15.67% CPU utilization
Server
14.83% CPU utilization
9094 Mbps
IPv6
Client
16.21% CPU utilization
Server
14.32% CPU utilization
9058 Mbps
No UDP checksums
IPv4
Client
15.03% CPU utilization
Server
23.09% CPU utilization
9089 Mbps
IPv6
Client
16.18% CPU utilization
Server
26.57% CPU utilization
8954 Mbps
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces udp_offload_callbacks which has the same
GRO functions (but not a GSO function) as offload_callbacks,
except there is an argument to a udp_offload struct passed to
gro_receive and gro_complete functions. This additional argument
can be used to retrieve the per port structure of the encapsulation
for use in gro processing (mostly by doing container_of on the
structure).
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The same macros are used for rx as well. So rename it.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch cleans up the header flags of VXLAN in anticipation of
defining some new ones:
- Move header related definitions from vxlan.c to vxlan.h
- Change VXLAN_FLAGS to be VXLAN_HF_VNI (only currently defined flag)
- Move check for unknown flags to after we find vxlan_sock, this
assumes that some flags may be processed based on tunnel
configuration
- Add a comment about why the stack treating unknown set flags as an
error instead of ignoring them
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the only tunnel protocol that supports GRO with encapsulated
Ethernet is VXLAN. This pulls out the Ethernet code into a proper layer
so that it can be used by other tunnel protocols such as GRE and Geneve.
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of error vxlan_xmit_one() can free already freed skb.
Also fixes memory leak of dst-entry.
Fixes: acbf74a763 ("vxlan: Refactor vxlan driver to make use
of the common UDP tunnel functions").
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, when trying to reuse a socket, vxlan_sock_add will grab
vn->sock_lock, locate a reusable socket, inc refcount and release
vn->sock_lock.
But vxlan_sock_release() will first decrement refcount, and then grab
that lock. refcnt operations are atomic but as currently we have
deferred works which hold vs->refcnt each, this might happen, leading to
a use after free (specially after vxlan_igmp_leave):
CPU 1 CPU 2
deferred work vxlan_sock_add
... ...
spin_lock(&vn->sock_lock)
vs = vxlan_find_sock();
vxlan_sock_release
dec vs->refcnt, reaches 0
spin_lock(&vn->sock_lock)
vxlan_sock_hold(vs), refcnt=1
spin_unlock(&vn->sock_lock)
hlist_del_rcu(&vs->hlist);
vxlan_notify_del_rx_port(vs)
spin_unlock(&vn->sock_lock)
So when we look for a reusable socket, we check if it wasn't freed
already before reusing it.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Fixes: 7c47cedf43 ("vxlan: move IGMP join/leave to work queue")
Signed-off-by: David S. Miller <davem@davemloft.net>
Do the work of parsing NDA_VLAN directly in rtnetlink code, pass simple
u16 vid to drivers from there.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Andy Gospodarek <gospo@cumulusnetworks.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In "vxlan: Call udp_sock_create" there was a logic error that resulted in
the default for IPv6 VXLAN tunnels going from using checksums to not using
checksums. Since there is currently no support in iproute2 for setting
these values it means that a kernel after the change cannot talk over a IPv6
VXLAN tunnel to a kernel prior the change.
Fixes: 3ee64f3 ("vxlan: Call udp_sock_create")
Cc: Tom Herbert <therbert@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Acked-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ieee802154/fakehard.c
A bug fix went into 'net' for ieee802154/fakehard.c, which is removed
in 'net-next'.
Add build fix into the merge from Stephen Rothwell in openvswitch, the
logging macros take a new initial 'log' argument, a new call was added
in 'net' so when we merge that in here we have to explicitly add the
new 'log' arg to it else the build fails.
Signed-off-by: David S. Miller <davem@davemloft.net>
Use them to push skb->vlan_tci into the payload and avoid code
duplication.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Name fits better. Plus there's going to be introduced
__vlan_insert_tag later on.
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Suggested-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most NICs that report NETIF_F_GSO_UDP_TUNNEL support VXLAN, and not
other UDP-based encapsulation protocols where the format and size of the
header differs. This patch implements a generic ndo_gso_check() for
VXLAN which will only advertise GSO support when the skb looks like it
contains VXLAN (or no UDP tunnelling at all).
Implementation shamelessly stolen from Tom Herbert:
http://thread.gmane.org/gmane.linux.network/332428/focus=333111
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/chelsio/cxgb4vf/sge.c
drivers/net/ethernet/intel/ixgbe/ixgbe_phy.c
sge.c was overlapping two changes, one to use the new
__dev_alloc_page() in net-next, and one to use s->fl_pg_order in net.
ixgbe_phy.c was a set of overlapping whitespace changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, we only match against local port number in order to reuse
socket. But if this new vxlan wants an IPv6 socket and a IPv4 one bound
to that port, vxlan will reuse an IPv4 socket as IPv6 and a panic will
follow. The following steps reproduce it:
# ip link add vxlan6 type vxlan id 42 group 229.10.10.10 \
srcport 5000 6000 dev eth0
# ip link add vxlan7 type vxlan id 43 group ff0e::110 \
srcport 5000 6000 dev eth0
# ip link set vxlan6 up
# ip link set vxlan7 up
<panic>
[ 4.187481] BUG: unable to handle kernel NULL pointer dereference at 0000000000000058
...
[ 4.188076] Call Trace:
[ 4.188085] [<ffffffff81667c4a>] ? ipv6_sock_mc_join+0x3a/0x630
[ 4.188098] [<ffffffffa05a6ad6>] vxlan_igmp_join+0x66/0xd0 [vxlan]
[ 4.188113] [<ffffffff810a3430>] process_one_work+0x220/0x710
[ 4.188125] [<ffffffff810a33c4>] ? process_one_work+0x1b4/0x710
[ 4.188138] [<ffffffff810a3a3b>] worker_thread+0x11b/0x3a0
[ 4.188149] [<ffffffff810a3920>] ? process_one_work+0x710/0x710
So address family must also match in order to reuse a socket.
Reported-by: Jean-Tsung Hsiao <jhsiao@redhat.com>
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When doing GRO processing for UDP tunnels, we never add
SKB_GSO_UDP_TUNNEL to gso_type - only the type of the inner protocol
is added (such as SKB_GSO_TCPV4). The result is that if the packet is
later resegmented we will do GSO but not treat it as a tunnel. This
results in UDP fragmentation of the outer header instead of (i.e.) TCP
segmentation of the inner header as was originally on the wire.
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add definition to vxlan nla_policy for UDP checksum. This is necessary
to enable UDP checksums on VXLAN.
In some instances, enabling UDP checksums can improve performance on
receive for devices that return legacy checksum-unnecessary for UDP/IP.
Also, UDP checksum provides some protection against VNI corruption.
Testing:
Ran 200 instances of TCP_STREAM and TCP_RR on bnx2x.
TCP_STREAM
IPv4, without UDP checksums
14.41% TX CPU utilization
25.71% RX CPU utilization
9083.4 Mbps
IPv4, with UDP checksums
13.99% TX CPU utilization
13.40% RX CPU utilization
9095.65 Mbps
TCP_RR
IPv4, without UDP checksums
94.08% TX CPU utilization
156/248/462 90/95/99% latencies
1.12743e+06 tps
IPv4, with UDP checksums
94.43% TX CPU utilization
158/250/462 90/95/99% latencies
1.13345e+06 tps
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
pskb_may_pull maybe change skb->data and make eth pointer oboslete,
so eth needs to reload
Fixes: 91269e390d ("vxlan: using pskb_may_pull as early as possible")
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
pskb_may_pull should be used to check if skb->data has enough space,
skb->len can not ensure that.
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
when netif_rx() is done, the netif_rx handled skb maybe be freed,
and should not be used.
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Testing xmit_more support with netperf and connected UDP sockets,
I found strange dst refcount false sharing.
Current handling of IFF_XMIT_DST_RELEASE is not optimal.
Dropping dst in validate_xmit_skb() is certainly too late in case
packet was queued by cpu X but dequeued by cpu Y
The logical point to take care of drop/force is in __dev_queue_xmit()
before even taking qdisc lock.
As Julian Anastasov pointed out, need for skb_dst() might come from some
packet schedulers or classifiers.
This patch adds new helper to cleanly express needs of various drivers
or qdiscs/classifiers.
Drivers that need skb_dst() in their ndo_start_xmit() should call
following helper in their setup instead of the prior :
dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
->
netif_keep_dst(dev);
Instead of using a single bit, we use two bits, one being
eventually rebuilt in bonding/team drivers.
The other one, is permanent and blocks IFF_XMIT_DST_RELEASE being
rebuilt in bonding/team. Eventually, we could add something
smarter later.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
Call skb_set_inner_protocol to set inner Ethernet protocol to
ETH_P_TEB before transmit. This is needed for GSO with UDP tunnels.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit acbf74a763 ("vxlan: Refactor vxlan driver to make use of the common UDP tunnel functions." introduced a bug in vxlan_xmit_one()
function, causing it to transmit Vxlan packets without proper
Vxlan header inserted. The change was not needed in the first
place. Revert it.
Reported-by: Tom Herbert <therbert@google.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simplify vxlan implementation using common UDP tunnel APIs.
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch:
- Clarifies the specific requirements of devices returning
CHECKSUM_UNNECESSARY (comments in skbuff.h).
- Adds csum_level field to skbuff. This is used to express how
many checksums are covered by CHECKSUM_UNNECESSARY (stores n - 1).
This replaces the overloading of skb->encapsulation, that field is
is now only used to indicate inner headers are valid.
- Change __skb_checksum_validate_needed to "consume" each checksum
as indicated by csum_level as layers of the the packet are parsed.
- Remove skb_pop_rcv_encapsulation, no longer needed in the new
csum_level model.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The first initializer in the following
union vxlan_addr ipa = {
.sin.sin_addr.s_addr = tip,
.sa.sa_family = AF_INET,
};
is optimised away by the compiler, due to the second initializer,
therefore initialising .sin.sin_addr.s_addr always to 0.
This results in netlink messages indicating a L3 miss never contain the
missed IP address. This was observed with GCC 4.8 and 4.9. I do not know about previous versions.
The problem affects user space programs relying on an IP address being
sent as part of a netlink message indicating a L3 miss.
Changing
.sa.sa_family = AF_INET,
to
.sin.sin_family = AF_INET,
fixes the problem.
Signed-off-by: Gerhard Stenzel <gerhard.stenzel@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ndm_type means L3 address type, in neighbour proxy and vxlan, it's RTN_UNICAST.
NDA_DST is for netlink TLV type, hence it's not right value in this context.
Signed-off-by: Jun Zhao <mypopydev@gmail.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In vxlan driver call common function udp_sock_create to create the
listener UDP port.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dumping a bridge fdb dumps every fdb entry
held. With this change we are going to filter
on selected bridge port.
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In vxlan and OVS vport-vxlan call common function to get source port
for a UDP tunnel. Removed vxlan_src_port since the functionality is
now in udp_flow_src_port.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Call skb_pop_rcv_encapsulation and postpull_rcsum for the Ethernet
header to work properly with checksum complete.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we mirror packets from a vxlan tunnel to other device,
the mirror device should see the same packets (that is, without
outer header). Because vxlan tunnel sets dev->hard_header_len,
tcf_mirred() resets mac header back to outer mac, the mirror device
actually sees packets with outer headers
Vxlan tunnel should set dev->needed_headroom instead of
dev->hard_header_len, like what other ip tunnels do. This fixes
the above problem.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: stephen hemminger <stephen@networkplumber.org>
Cc: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Need to gro_postpull_rcsum for GRO to work with checksum complete.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added VXLAN link configuration for sending UDP checksums, and allowing
TX and RX of UDP6 checksums.
Also, call common iptunnel_handle_offloads and added GSO support for
checksums.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net: get rid of SET_ETHTOOL_OPS
Dave Miller mentioned he'd like to see SET_ETHTOOL_OPS gone.
This does that.
Mostly done via coccinelle script:
@@
struct ethtool_ops *ops;
struct net_device *dev;
@@
- SET_ETHTOOL_OPS(dev, ops);
+ dev->ethtool_ops = ops;
Compile tested only, but I'd seriously wonder if this broke anything.
Suggested-by: Dave Miller <davem@davemloft.net>
Signed-off-by: Wilfried Klaebe <w-lkml@lebenslange-mailadresse.de>
Acked-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch allows to switch the netns when packet is encapsulated or
decapsulated.
The vxlan socket is openned into the i/o netns, ie into the netns where
encapsulated packets are received. The socket lookup is done into this netns to
find the corresponding vxlan tunnel. After decapsulation, the packet is
injecting into the corresponding interface which may stand to another netns.
When one of the two netns is removed, the tunnel is destroyed.
Configuration example:
ip netns add netns1
ip netns exec netns1 ip link set lo up
ip link add vxlan10 type vxlan id 10 group 239.0.0.10 dev eth0 dstport 0
ip link set vxlan10 netns netns1
ip netns exec netns1 ip addr add 192.168.0.249/24 broadcast 192.168.0.255 dev vxlan10
ip netns exec netns1 ip link set vxlan10 up
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The goal of this patch is to fix rtnelink notification. The main problem was
about notification for fdb entry with more than one remote. Before the patch,
when a remote was added to an existing fdb entry, the kernel advertised the
first remote instead of the added one. Also when a remote was removed from a fdb
entry with several remotes, the deleted remote was not advertised.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the dst->output() path for ipv4, the code assumes the skb it has to
transmit is attached to an inet socket, specifically via
ip_mc_output() : The sk_mc_loop() test triggers a WARN_ON() when the
provider of the packet is an AF_PACKET socket.
The dst->output() method gets an additional 'struct sock *sk'
parameter. This needs a cascade of changes so that this parameter can
be propagated from vxlan to final consumer.
Fixes: 8f646c922d ("vxlan: keep original skb ownership")
Reported-by: lucien xin <lucien.xin@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
Documentation/devicetree/bindings/net/micrel-ks8851.txt
net/core/netpoll.c
The net/core/netpoll.c conflict is a bug fix in 'net' happening
to code which is completely removed in 'net-next'.
In micrel-ks8851.txt we simply have overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
The VXLAN neigh_reduce() code is completely non-functional since
check-in. Specific errors:
1) The original code drops all packets with a multicast destination address,
even though neighbor solicitations are sent to the solicited-node
address, a multicast address. The code after this check was never run.
2) The neighbor table lookup used the IPv6 header destination, which is the
solicited node address, rather than the target address from the
neighbor solicitation. So neighbor lookups would always fail if it
got this far. Also for L3MISSes.
3) The code calls ndisc_send_na(), which does a send on the tunnel device.
The context for neigh_reduce() is the transmit path, vxlan_xmit(),
where the host or a bridge-attached neighbor is trying to transmit
a neighbor solicitation. To respond to it, the tunnel endpoint needs
to do a *receive* of the appropriate neighbor advertisement. Doing a
send, would only try to send the advertisement, encapsulated, to the
remote destinations in the fdb -- hosts that definitely did not do the
corresponding solicitation.
4) The code uses the tunnel endpoint IPv6 forwarding flag to determine the
isrouter flag in the advertisement. This has nothing to do with whether
or not the target is a router, and generally won't be set since the
tunnel endpoint is bridging, not routing, traffic.
The patch below creates a proxy neighbor advertisement to respond to
neighbor solicitions as intended, providing proper IPv6 support for neighbor
reduction.
Signed-off-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a NULL pointer dereference in the event of an
skb allocation failure in arp_reduce().
Signed-Off-By: David L Stevens <dlstevens@us.ibm.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are many drivers calling alloc_percpu() to allocate pcpu stats
and then initializing ->syncp. So just introduce a helper function for them.
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The loop in vxlan_gro_receive() over the current set of candidates for
coalescing was wrongly aborted once a match was found. In rare cases,
this can cause a false-positives matching in the next layer GRO checks.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tom Herbert