We handle a corner case in the function tipc_group_update_rcv_win().
During extreme pessure it might happen that a message receiver has all
its active senders in RECLAIMING or REMITTED mode, meaning that there
is nobody to reclaim advertisements from if an additional sender tries
to go active.

Currently we just set the new sender to ACTIVE anyway, hence at least
theoretically opening up for a receiver queue overflow by exceeding the
MAX_ACTIVE limit. The correct solution to this is to instead add the
member to the pending queue, while letting the oldest member in that
queue revert to JOINED state.

In this commit we refactor the code for handling message arrival from
a JOINED member, both to make it more comprehensible and to cover the
case described above.
Acked-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

- We remove the 'reclaiming' member list in struct tipc_group, since
  it doesn't serve any purpose.

- We simplify the GRP_REMIT_MSG branch of tipc_group_protocol_rcv().
Acked-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Tariq Toukan says:

====================
ethtool ringparam upper bound

This patchset by Jenny adds sanity checks in ethtool ringparam
operation for input upper bounds, similarly to what's done in
ethtool_set_channels.

The checks are added in patch 1, using a call to get_ringparam
prior to calling set_ringparam NDO.

Patch 2 changes the function's behavior in mlx4_en, so that
it returns an error for out-of-range input, instead of rounding
it to closest valid, similar to mlx5e.

Patch 3 removes the upper bound checks in mlx5e_ethtool_set_ringparam
as it becomes redundant.

Series generated against net-next commit:
f66faae2 Merge branch 'ipv6-ipv4-nexthop-align'
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since the checks are done in upper layer ethtool code,
checks in driver are not needed any more.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In current implementation, any requested RX/TX ring size value
that is less than minimum is silently casted to nearest valid value.
Update this behavior to align with mlx5 behavior by printing warning
in dmesg and remaining the size unchanged.
Kernel is responsible for verifying against the maximum.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add a sanity check to ensure that all requested ring parameters
are within bounds, which should reduce errors in driver implementation.
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the ARRAY_SIZE macro on array seg6_action_table to determine size of
the array. Improvement suggested by coccinelle.
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the ARRAY_SIZE macro on array cmd_priv_map to determine size of the
array.  Improvement suggested by coccinelle.
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The ability to set speed and duplex for virtio_net is useful in various
scenarios as described here:

16032be5 virtio_net: add ethtool support for set and get of settings

However, it would be nice to be able to set this from the hypervisor,
such that virtio_net doesn't require custom guest ethtool commands.

Introduce a new feature flag, VIRTIO_NET_F_SPEED_DUPLEX, which allows
the hypervisor to export a linkspeed and duplex setting. The user can
subsequently overwrite it later if desired via: 'ethtool -s'.

Note that VIRTIO_NET_F_SPEED_DUPLEX is defined as bit 63, the intention
is that device feature bits are to grow down from bit 63, since the
transports are starting from bit 24 and growing up.
Signed-off-by: NJason Baron <jbaron@akamai.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: virtio-dev@lists.oasis-open.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds support for the GCM-AES-256 cipher suite as specified in
IEEE 802.1AEbn-2011. The prepared cipher suite selection mechanism is used,
with GCM-AES-128 being the default cipher suite as defined in the standard.
Signed-off-by: NFelix Walter <felix.walter@cloudandheat.com>
Cc: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jason Wang says:

====================
XDP transmission for tuntap

This series tries to implement XDP transmission (ndo_xdp_xmit) for
tuntap. Pointer ring was used for queuing both XDP buffers and
sk_buff, this is done by encoding the type into lowest bit of the
pointer and storin XDP metadata in the headroom of XDP buff.

Tests gets 3.05 Mpps when doing xdp_redirect_map from ixgbe to VM
(testpmd + virtio-net in guest). This gives us ~20% improvments
compared to use skb during redirect.

Please review.

Changes from V1:

- slient warnings
- fix typos
- add skb mode number in the commit log
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch implements XDP transmission for TAP. Since we can't create
new queues for TAP during XDP set, exist ptr_ring was reused for
queuing XDP buffers. To differ xdp_buff from sk_buff, TUN_XDP_FLAG
(0x1UL) was encoded into lowest bit of xpd_buff pointer during
ptr_ring_produce, and was decoded during consuming. XDP metadata was
stored in the headroom of the packet which should work in most of
cases since driver usually reserve enough headroom. Very minor changes
were done for vhost_net: it just need to peek the length depends on
the type of pointer.

Tests were done on two Intel E5-2630 2.40GHz machines connected back
to back through two 82599ES. Traffic were generated/received through
MoonGen/testpmd(rxonly). It reports ~20% improvements when
xdp_redirect_map is doing redirection from ixgbe to TAP (from 2.50Mpps
to 3.05Mpps)

Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: NJason Wang <jasowang@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch switches to use ptr_ring instead of skb_array. This will be
used to enqueue different types of pointers by encoding type into
lower bits.
Signed-off-by: NJason Wang <jasowang@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller:

 1) Frag and UDP handling fixes in i40e driver, from Amritha Nambiar and
    Alexander Duyck.

 2) Undo unintentional UAPI change in netfilter conntrack, from Florian
    Westphal.

 3) Revert a change to how error codes are returned from
    dev_get_valid_name(), it broke some apps.

 4) Cannot cache routes for ipv6 tunnels in the tunnel is ipv4/ipv6
    dual-stack. From Eli Cooper.

 5) Fix missed PMTU updates in geneve, from Xin Long.

 6) Cure double free in macvlan, from Gao Feng.

 7) Fix heap out-of-bounds write in rds_message_alloc_sgs(), from
    Mohamed Ghannam.

 8) FEC bug fixes from FUgang Duan (mis-accounting of dev_id, missed
    deferral of probe when the regulator is not ready yet).

 9) Missing DMA mapping error checks in 3c59x, from Neil Horman.

10) Turn off Broadcom tags for some b53 switches, from Florian Fainelli.

11) Fix OOPS when get_target_net() is passed an SKB whose NETLINK_CB()
    isn't initialized. From Andrei Vagin.

12) Fix crashes in fib6_add(), from Wei Wang.

13) PMTU bug fixes in SCTP from Marcelo Ricardo Leitner.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (56 commits)
  sh_eth: fix TXALCR1 offsets
  mdio-sun4i: Fix a memory leak
  phylink: mark expected switch fall-throughs in phylink_mii_ioctl
  sctp: fix the handling of ICMP Frag Needed for too small MTUs
  sctp: do not retransmit upon FragNeeded if PMTU discovery is disabled
  xen-netfront: enable device after manual module load
  bnxt_en: Fix the 'Invalid VF' id check in bnxt_vf_ndo_prep routine.
  bnxt_en: Fix population of flow_type in bnxt_hwrm_cfa_flow_alloc()
  sh_eth: fix SH7757 GEther initialization
  net: fec: free/restore resource in related probe error pathes
  uapi/if_ether.h: prevent redefinition of struct ethhdr
  ipv6: fix general protection fault in fib6_add()
  RDS: null pointer dereference in rds_atomic_free_op
  sh_eth: fix TSU resource handling
  net: stmmac: enable EEE in MII, GMII or RGMII only
  rtnetlink: give a user socket to get_target_net()
  MAINTAINERS: Update my email address.
  can: ems_usb: improve error reporting for error warning and error passive
  can: flex_can: Correct the checking for frame length in flexcan_start_xmit()
  can: gs_usb: fix return value of the "set_bittiming" callback
  ...

Preempt counter APIs have been split out, currently, hardirq.h just
includes irq_enter/exit APIs which are not used by TIPC at all.

So, remove the unused hardirq.h.
Signed-off-by: NYang Shi <yang.s@alibaba-inc.com>
Acked-by: NYing Xue <ying.xue@windriver.com>
Tested-by: NYing Xue <ying.xue@windriver.com>
Cc: Jon Maloy <jon.maloy@ericsson.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Preempt counter APIs have been split out, currently, hardirq.h just
includes irq_enter/exit APIs which are not used by openvswitch at all.

So, remove the unused hardirq.h.
Signed-off-by: NYang Shi <yang.s@alibaba-inc.com>
Acked-by: NPravin B Shelar <pshelar@ovn.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: dev@openvswitch.org
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Preempt counter APIs have been split out, currently, hardirq.h just
includes irq_enter/exit APIs which are not used by caif at all.

So, remove the unused hardirq.h.
Signed-off-by: NYang Shi <yang.s@alibaba-inc.com>
Cc: Dmitry Tarnyagin <dmitry.tarnyagin@lockless.no>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Gal Pressman says:

====================
Replace WARN_ONCE usages with netdev_WARN_ONCE

This series will fix an issue in netdev_WARN_ONCE, improve its formatting and
replace drivers' usage of WARN_ONCE to netdev_WARN_ONCE.

Driver specific patches were compilation tested, in addition, functional tested
on Mellanox NIC.

v1->v2:
- Addressed commit message comments in patch #1
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the more appropriate netdev_WARN_ONCE instead of WARN_ONCE macro.
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the more appropriate netdev_WARN_ONCE instead of WARN_ONCE macro.
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Cc: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the more appropriate netdev_WARN_ONCE instead of WARN_ONCE macro.
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the more appropriate netdev_WARN_ONCE instead of WARN_ONCE macro.
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove the unnecessary line break between the netdev name and reg state
to the actual message that should be printed.

For example, this:
[86730.307236] ------------[ cut here ]------------
[86730.313496] netdevice: enp27s0f0
Message from the driver
[...]

Will be replaced with:
[86770.259289] ------------[ cut here ]------------
[86770.265191] netdevice: enp27s0f0: Message from the driver
[...]
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

netdev_WARN_ONCE is broken (whoops..), this fix will remove the
unnecessary "condition" parameter, add the missing comma and change
"arg" to "args".

Fixes: 375ef2b1 ("net: Introduce netdev_*_once functions")
Signed-off-by: NGal Pressman <galp@mellanox.com>
Reviewed-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pablo Neira Ayuso says:

====================
Netfilter/IPVS updates for net-next

The following patchset contains Netfilter/IPVS updates for your
net-next tree:

1) Free hooks via call_rcu to speed up netns release path, from
   Florian Westphal.

2) Reduce memory footprint of hook arrays, skip allocation if family is
   not present - useful in case decnet support is not compiled built-in.
   Patches from Florian Westphal.

3) Remove defensive check for malformed IPv4 - including ihl field - and
   IPv6 headers in x_tables and nf_tables.

4) Add generic flow table offload infrastructure for nf_tables, this
   includes the netlink control plane and support for IPv4, IPv6 and
   mixed IPv4/IPv6 dataplanes. This comes with NAT support too. This
   patchset adds the IPS_OFFLOAD conntrack status bit to indicate that
   this flow has been offloaded.

5) Add secpath matching support for nf_tables, from Florian.

6) Save some code bytes in the fast path for the nf_tables netdev,
   bridge and inet families.

7) Allow one single NAT hook per point and do not allow to register NAT
   hooks in nf_tables before the conntrack hook, patches from Florian.

8) Seven patches to remove the struct nf_af_info abstraction, instead
   we perform direct calls for IPv4 which is faster. IPv6 indirections
   are still needed to avoid dependencies with the 'ipv6' module, but
   these now reside in struct nf_ipv6_ops.

9) Seven patches to handle NFPROTO_INET from the Netfilter core,
   hence we can remove specific code in nf_tables to handle this
   pseudofamily.

10) No need for synchronize_net() call for nf_queue after conversion
    to hook arrays. Also from Florian.

11) Call cond_resched_rcu() when dumping large sets in ipset to avoid
    softlockup. Again from Florian.

12) Pass lockdep_nfnl_is_held() to rcu_dereference_protected(), patch
    from Florian Westphal.

13) Fix matching of counters in ipset, from Jozsef Kadlecsik.

14) Missing nfnl lock protection in the ip_set_net_exit path, also
    from Jozsef.

15) Move connlimit code that we can reuse from nf_tables into
    nf_conncount, from Florian Westhal.

And asorted cleanups:

16) Get rid of nft_dereference(), it only has one single caller.

17) Add nft_set_is_anonymous() helper function.

18) Remove NF_ARP_FORWARD leftover chain definition in nf_tables_arp.

19) Remove unnecessary comments in nf_conntrack_h323_asn1.c
    From Varsha Rao.

20) Remove useless parameters in frag_safe_skb_hp(), from Gao Feng.

21) Constify layer 4 conntrack protocol definitions, function
    parameters to register/unregister these protocol trackers, and
    timeouts. Patches from Florian Westphal.

22) Remove nlattr_size indirection, from Florian Westphal.

23) Add fall-through comments as -Wimplicit-fallthrough needs this,
    from Gustavo A. R. Silva.

24) Use swap() macro to exchange values in ipset, patch from
    Gustavo A. R. Silva.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull rdma fixes from Doug Ledford:

 - One line fix to mlx4 error flow (same as mlx5 fix in last pull
   request, just in the mlx4 driver)

 - Fix a race condition in the IPoIB driver. This patch is larger than
   just a one line fix, but resolves a race condition in a fairly
   straight forward manner

 - Fix a locking issue in the RDMA netlink code. This patch is also
   larger than I would like for a late -rc. It has, however, had a week
   to bake in the rdma tree prior to this pull request

 - One line fix to fix granting remote machine access to memory that
   they don't need and shouldn't have

 - One line fix to correct the fact that our sgid/dgid pair is swapped
   from what you would expect when receiving an incoming connection
   request

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  IB/srpt: Fix ACL lookup during login
  IB/srpt: Disable RDMA access by the initiator
  RDMA/netlink: Fix locking around __ib_get_device_by_index
  IB/ipoib: Fix race condition in neigh creation
  IB/mlx4: Fix mlx4_ib_alloc_mr error flow

Pull x86 platform driver fix from Darren Hart:
 "Address a wmi initcall ordering race resulting in a difficult to
  reproduce boot failure"

* tag 'platform-drivers-x86-v4.15-4' of git://git.infradead.org/linux-platform-drivers-x86:
  platform/x86: wmi: Call acpi_wmi_init() later

As of now, there're two sk_family are traced with sock:inet_sock_set_state,
which are AF_INET and AF_INET6.
So the sk_family are exposed as well.
Then we can conveniently use it to do the filter.

Both sk_family and sk_protocol are showed in the printk message, so we need
not expose them as tracepoint arguments.
Suggested-by: NBrendan Gregg <brendan.d.gregg@gmail.com>
Suggested-by: NSong Liu <songliubraving@fb.com>
Signed-off-by: NYafang Shao <laoar.shao@gmail.com>
Reviewed-by: NSong Liu <songliubraving@fb.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The  TXALCR1 offsets are incorrect in the register offset tables, most
probably due to copy&paste error.  Luckily, the driver never uses this
register. :-)

Fixes: 4a55530f ("net: sh_eth: modify the definitions of register")
Signed-off-by: NSergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If the probing of the regulator is deferred, the memory allocated by
'mdiobus_alloc_size()' will be leaking.
It should be freed before the next call to 'sun4i_mdio_probe()' which will
reallocate it.

Fixes: 4bdcb1dd ("net: Add MDIO bus driver for the Allwinner EMAC")
Signed-off-by: NChristophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The "offset" option has been removed by
commit 900631ee ("l2tp: remove configurable payload offset").
Signed-off-by: NGuillaume Nault <g.nault@alphalink.fr>
Acked-by: NJames Chapman <jchapman@katalix.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In preparation to enabling -Wimplicit-fallthrough, mark switch cases
where we are expecting to fall through.

Addresses-Coverity-ID: 1463447 ("Missing break in switch")
Signed-off-by: NGustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Marcelo Ricardo Leitner says:

====================
SCTP PMTU discovery fixes

This patchset fixes 2 issues with PMTU discovery that can lead to flood
of retransmissions.
The first patch fixes the issue for when PMTUD is disabled by the
application, while the second fixes it for when its enabled.

Please consider these to stable.
====================
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The mask argument for phy_modify() in several locations was inverted.

Fixes: fea23fb5 ("net: phy: convert read-modify-write to phy_modify()")
Reported-by: NHeiner Kallweit <hkallweit1@gmail.com>
Tested-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[  246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512

That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).

As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.

The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.

Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.

Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad3919 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
  resulted in a change or not
Changes from v2:
none

See-also: https://lkml.org/lkml/2017/12/22/811Reported-by: Nsyzbot <syzkaller@googlegroups.com>
Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, if PMTU discovery is disabled on a given transport, but the
configured value is higher than the actual PMTU, it is likely that we
will get some icmp Frag Needed. The issue is, if PMTU discovery is
disabled, we won't update the information and will issue a
retransmission immediately, which may very well trigger another ICMP,
and another retransmission, leading to a loop.

The fix is to simply not trigger immediate retransmissions if PMTU
discovery is disabled on the given transport.

Changes from v2:
- updated stale comment, noticed by Xin Long
Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When loading the module after unloading it, the network interface would
not be enabled and thus wouldn't have a backend counterpart and unable
to be used by the guest.

The guest would face errors like:

  [root@guest ~]# ethtool -i eth0
  Cannot get driver information: No such device

  [root@guest ~]# ifconfig eth0
  eth0: error fetching interface information: Device not found

This patch initializes the state of the netfront device whenever it is
loaded manually, this state would communicate the netback to create its
device and establish the connection between them.
Signed-off-by: NEduardo Otubo <otubo@redhat.com>
Reviewed-by: NBoris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since both first_rx and rx_ring are the head of rx ring, it not
necessary to use two structure members to statically indicate
the head of rx ring. So first_rx is removed.

CC: Srinivas Eeda <srinivas.eeda@oracle.com>
CC: Joe Jin <joe.jin@oracle.com>
CC: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: NZhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Michael Chan says:

====================
bnxt_en: 2 small bug fixes.

The first one fixes the TC Flower flow parameter passed to firmware.  The
2nd one fixes the VF index range checking for iproute2 SRIOV related commands.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>