This patch adds offload command related to "ethtool -K".
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds vlan offload config commands, initializes
the rules of tx/rx vlan tag handle for hw.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch sets vlan masked, in order to avoid the received
packets being filtered.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add configuration for rss_size_max in hdev but not hardcode it.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fixes a memory leak problems in change tqps process,
the function hns3_uninit_all_ring and hns3_init_all_ring
may be called many times.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch modifies the return data of get_rxnfc, it will return
the current handle's rss_size but not the total tqp number.
because the tc_size has been change to the log2 of roundup
power of two of rss_size.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds the support to change tqps number for PF driver
by using ehtool -L command.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds the support to query tqps number for PF driver
by using ehtool -l command.
Signed-off-by: Peng Li <lipeng321@huawei.com>
Signed-off-by: Mingguang Qu <qumingguang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The 'md' is allocated from 'tun_dst = ip_tun_rx_dst' and
since we've checked 'tun_dst', 'md' will never be NULL.
The patch removes it at both ipv4 and ipv6 erspan.

Fixes: afb4c97d ("ip6_gre: fix potential memory leak in ip6erspan_rcv")
Fixes: 50670b6e ("ip_gre: fix potential memory leak in erspan_rcv")
Cc: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Dereference tp->md5sig_info in tcp_v4_destroy_sock() the same way it is
done in the adjacent call to tcp_clear_md5_list().

Resolves this sparse warning:

net/ipv4/tcp_ipv4.c:1914:17: warning: incorrect type in argument 1 (different address spaces)
net/ipv4/tcp_ipv4.c:1914:17:    expected struct callback_head *head
net/ipv4/tcp_ipv4.c:1914:17:    got struct callback_head [noderef] <asn:4>*<noident>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Acked-by: Christoph Paasch <cpaasch@apple.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Non-functional cleanups in lan9303_csr_reg_wait():
 - Change type of param 'mask' from int to u32.
 - Remove param 'value' (will probably never be used)
 - Reduced retries from 1000 to 25, consistent with lan9303_read_wait.
 - Removed comments
Signed-off-by: Egil Hjelmeland <privat@egil-hjelmeland.no>

Changes v1 -> v2:
 - Removed comments
Signed-off-by: David S. Miller <davem@davemloft.net>

If SNAT modifies the source address the resulting packet might match
an IPsec policy, reinject the packet if that's the case.

The exact same thing is already done for IPv4.
Signed-off-by: Tobias Brunner <tobias@strongswan.org>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In case of tx clean up, we set '-1' as budget. This means clean up until
wq is empty or till (1 << 32) pkts are cleaned. Under heavy load this
will run for long time and cause
"watchdog: BUG: soft lockup - CPU#25 stuck for 21s!" warning.

This patch sets wq clean up budget to 256.
Signed-off-by: Govindarajulu Varadarajan <gvaradar@cisco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ASSERT_RTNL() macro is actual open-coded variant of WARN_ONCE() with
two exceptions. First, it prints stack for multiple hits and not only
once as WARN_ONCE() does. Second, the user can disable prints of
WARN_ONCE by setting CONFIG_BUG to N.

The multiple prints of dump stack are actually not needed, because calls
without rtnl lock are programming errors and user can't do anything
about them except to complain to the mailing list after first occurrence
of such failure.

The user who disabled BUG/WARN prints did it explicitly because by default
in upstream kernel and distributions this option is enabled. It means
that user doesn't want to see prints about missing locks too.

This patch replaces open-coded variant in favor of already existing
macro and change error prints to be once only.
Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remove superfluous pin setup to get out of accessing invalid I/O pin
registers because the way for pin configuring tends to be different from
various SoCs and thus it should be better being managed and controlled by
the pinctrl driver which MT7622 already can support.
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The property "mediatek,pctl" is only required for SoCs such as MT2701 and
MT7623, so adding a few words for stating the condition.
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Lots of overlapping changes.  Also on the net-next side
the XDP state management is handled more in the generic
layers so undo the 'net' nfp fix which isn't applicable
in net-next.

Include a necessary change by Jakub Kicinski, with log message:

====================
cls_bpf no longer takes care of offload tracking.  Make sure
netdevsim performs necessary checks.  This fixes a warning
caused by TC trying to remove a filter it has not added.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller"
 "What's a holiday weekend without some networking bug fixes? [1]

   1) Fix some eBPF JIT bugs wrt. SKB pointers across helper function
      calls, from Daniel Borkmann.

   2) Fix regression from errata limiting change to marvell PHY driver,
      from Zhao Qiang.

   3) Fix u16 overflow in SCTP, from Xin Long.

   4) Fix potential memory leak during bridge newlink, from Nikolay
      Aleksandrov.

   5) Fix BPF selftest build on s390, from Hendrik Brueckner.

   6) Don't append to cfg80211 automatically generated certs file,
      always write new ones from scratch. From Thierry Reding.

   7) Fix sleep in atomic in mac80211 hwsim, from Jia-Ju Bai.

   8) Fix hang on tg3 MTU change with certain chips, from Brian King.

   9) Add stall detection to arc emac driver and reset chip when this
      happens, from Alexander Kochetkov.

  10) Fix MTU limitng in GRE tunnel drivers, from Xin Long.

  11) Fix stmmac timestamping bug due to mis-shifting of field. From
      Fredrik Hallenberg.

  12) Fix metrics match when deleting an ipv4 route. The kernel sets
      some internal metrics bits which the user isn't going to set when
      it makes the delete request. From Phil Sutter.

  13) mvneta driver loop over RX queues limits on "txq_number" :-) Fix
      from Yelena Krivosheev.

  14) Fix double free and memory corruption in get_net_ns_by_id, from
      Eric W. Biederman.

  15) Flush ipv4 FIB tables in the reverse order. Some tables can share
      their actual backing data, in particular this happens for the MAIN
      and LOCAL tables. We have to kill the LOCAL table first, because
      it uses MAIN's backing memory. Fix from Ido Schimmel.

  16) Several eBPF verifier value tracking fixes, from Edward Cree, Jann
      Horn, and Alexei Starovoitov.

  17) Make changes to ipv6 autoflowlabel sysctl really propagate to
      sockets, unless the socket has set the per-socket value
      explicitly. From Shaohua Li.

  18) Fix leaks and double callback invocations of zerocopy SKBs, from
      Willem de Bruijn"

[1] Is this a trick question? "Relaxing"? "Quiet"? "Fine"? - Linus.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (77 commits)
  skbuff: skb_copy_ubufs must release uarg even without user frags
  skbuff: orphan frags before zerocopy clone
  net: reevalulate autoflowlabel setting after sysctl setting
  openvswitch: Fix pop_vlan action for double tagged frames
  ipv6: Honor specified parameters in fibmatch lookup
  bpf: do not allow root to mangle valid pointers
  selftests/bpf: add tests for recent bugfixes
  bpf: fix integer overflows
  bpf: don't prune branches when a scalar is replaced with a pointer
  bpf: force strict alignment checks for stack pointers
  bpf: fix missing error return in check_stack_boundary()
  bpf: fix 32-bit ALU op verification
  bpf: fix incorrect tracking of register size truncation
  bpf: fix incorrect sign extension in check_alu_op()
  bpf/verifier: fix bounds calculation on BPF_RSH
  ipv4: Fix use-after-free when flushing FIB tables
  s390/qeth: fix error handling in checksum cmd callback
  tipc: remove joining group member from congested list
  selftests: net: Adding config fragment CONFIG_NUMA=y
  nfp: bpf: keep track of the offloaded program
  ...

Sven Eckelmann says:

====================
flow_dissector: Provide basic batman-adv unicast handling

we are currently starting to use batman-adv as mesh protocol on multicore
embedded devices. These usually don't have a lot of CPU power per core but
are reasonable fast when using multiple cores.

It was noticed that sending was working very well but receiving was
basically only using on CPU core per neighbor. The reason for that is
format of the (normal) incoming packet:

  +--------------------+
  | ip(v6)hdr          |
  +--------------------+
  | inner ethhdr       |
  +--------------------+
  | batadv unicast hdr |
  +--------------------+
  | outer ethhdr       |
  +--------------------+

The flow dissector will therefore stop after parsing the outer ethernet
header and will not parse the actual ipv(4|6)/... header of the packet. Our
assumption was now that it would help us to add minimal support to the flow
dissector to jump over the batman-adv unicast and inner ethernet header
(like in gre ETH_P_TEB). The patch was implemented in a slightly hacky
way [1] and the results looked quite promising.

I didn't get any feedback how the files should actually be named. So I am
now just using the names from RFC v3

The discussion of the RFC v3 can be found in the related patches of
https://patchwork.ozlabs.org/cover/849345/

The discussion of the RFC v2 can be found in the related patches of
https://patchwork.ozlabs.org/cover/844783/

Changes in v4:
==============

* added  patch to change the u8/u16 to __u8/__u16 in
  include/uapi/linux/batadv_packet.h
  - requested by Willem de Bruijn <willemdebruijn.kernel@gmail.com>

Changes in v3:
==============

* removed change of uapi/linux/batman_adv.h to uapi/linux/batadv_genl.h
  - requested by Willem de Bruijn <willemdebruijn.kernel@gmail.com>
* removed naming fixes for enums/defines in uapi/linux/batadv_genl.h
  - requested by Willem de Bruijn <willemdebruijn.kernel@gmail.com>
* renamed uapi/linux/batadv.h to uapi/linux/batadv_packet.h
* moved batadv dissector functionality in own function
  - requested by Tom Herbert <tom@herbertland.com>
* added support for flags FLOW_DISSECTOR_F_STOP_AT_ENCAP and
  FLOW_DIS_ENCAPSULATION
  - requested by Willem de Bruijn <willemdebruijn.kernel@gmail.com>

Changes in v2:
==============

* removed the batman-adv unicast packet header definition from flow_dissector.c
* moved the batman-adv packet.h/uapi headers around to provide the correct
  definitions to flow_dissector.c
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The batman-adv unicast packets contain a full layer 2 frame in encapsulated
form. The flow dissector must therefore be able to parse the batman-adv
unicast header to reach the layer 2+3 information.

  +--------------------+
  | ip(v6)hdr          |
  +--------------------+
  | inner ethhdr       |
  +--------------------+
  | batadv unicast hdr |
  +--------------------+
  | outer ethhdr       |
  +--------------------+

The obtained information from the upper layer can then be used by RPS to
schedule the processing on separate cores. This allows better distribution
of multiple flows from the same neighbor to different cores.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The header file is used by different userspace programs to inject packets
or to decode sniffed packets. It should therefore be available to them as
userspace header.

Also other components in the kernel (like the flow dissector) require
access to the packet definitions to be able to decode ETH_P_BATMAN ethernet
packets.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The uapi headers use the __u8/__u16/... version of the fixed width types
instead of u8/u16/... The use of the latter must be avoided before
packet.h is copied to include/uapi/linux/.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The BIT(x) macro is no longer available for uapi headers because it is
defined outside of it (linux/bitops.h). The use of it must therefore be
avoided and replaced by an appropriate other representation.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The headers used by packet.h should also be included by it directly. main.h
is currently dealing with it in batman-adv, but this will no longer work
when this header is moved to include/uapi/linux/.
Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Bert Kenward says:

====================
sfc: support extra stats on Medford2

X2000-series NICs add port stats for two new features: FEC (Forward Error
 Correction, used on 25G links) and CTPIO (cut-through programmed I/O).
This patch series adds support for reporting both of these sets of stats

v2: add additional Signed-off-by
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

While the Linux driver doesn't use CTPIO ('cut-through programmed I/O'),
 other drivers on the same port might, so if we're responsible for
 reporting per-port stats we need to include the CTPIO stats.
Signed-off-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There's no explicit capability bit, so we just condition them on having
 efx->num_mac_stats >= MC_CMD_MAC_NSTATS_V2.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Medford2 NICs support more than MC_CMD_MAC_NSTATS stats, and report the new
 count in a field of MC_CMD_GET_CAPABILITIES_V4.  This also means that the
 end generation count moves (it is, as before, the last 64 bits of the DMA
 buffer, but that is no longer MC_CMD_MAC_GENERATION_END).
So read num_mac_stats from the GET_CAPABILITIES response, if present;
 otherwise assume MC_CMD_MAC_NSTATS; and always use num_mac_stats - 1 rather
 than MC_CMD_MAC_GENERATION_END.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Bert Kenward <bkenward@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add device id's 0x50ac, 0x6087 for T5 and T6 cards
respectively.
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Get rid of custom hex_dump_to_buffer().

The output is slightly changed, i.e. each byte followed by white space.

Note, we don't use print_hex_dump() here since the original code uses
nedev_dbg().
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Jie Deng <jiedeng@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Attaching more than one PHY to phylink is bad news, as we store a
pointer to the PHY in a single location. Error out if more than one
PHY is attempted to be attached.
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

===================
Mellanox, mlx5 fixes 2017-12-19

The follwoing series includes some fixes for mlx5 core and etherent
driver.

Please pull and let me know if there is any problem.

This series doesn't introduce any conflict with the ongoing mlx5 for-next
submission.

For -stable:

kernels >= v4.7.y
    ("net/mlx5e: Fix possible deadlock of VXLAN lock")
    ("net/mlx5e: Add refcount to VXLAN structure")
    ("net/mlx5e: Prevent possible races in VXLAN control flow")
    ("net/mlx5e: Fix features check of IPv6 traffic")

kernels >= v4.9.y
    ("net/mlx5: Fix error flow in CREATE_QP command")
    ("net/mlx5: Fix rate limit packet pacing naming and struct")

kernels >= v4.13.y
    ("net/mlx5: FPGA, return -EINVAL if size is zero")

kernels >= v4.14.y
    ("Revert "mlx5: move affinity hints assignments to generic code")

All above patches apply and compile with no issues on corresponding -stable.
===================
Signed-off-by: David S. Miller <davem@davemloft.net>

skb_copy_ubufs creates a private copy of frags[] to release its hold
on user frags, then calls uarg->callback to notify the owner.

Call uarg->callback even when no frags exist. This edge case can
happen when zerocopy_sg_from_iter finds enough room in skb_headlen
to copy all the data.

Fixes: 3ece7826 ("sock: skb_copy_ubufs support for compound pages")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Call skb_zerocopy_clone after skb_orphan_frags, to avoid duplicate
calls to skb_uarg(skb)->callback for the same data.

skb_zerocopy_clone associates skb_shinfo(skb)->uarg from frag_skb
with each segment. This is only safe for uargs that do refcounting,
which is those that pass skb_orphan_frags without dropping their
shared frags. For others, skb_orphan_frags drops the user frags and
sets the uarg to NULL, after which sock_zerocopy_clone has no effect.

Qemu hangs were reported due to duplicate vhost_net_zerocopy_callback
calls for the same data causing the vhost_net_ubuf_ref_>refcount to
drop below zero.

Link: http://lkml.kernel.org/r/<CAF=yD-LWyCD4Y0aJ9O0e_CHLR+3JOeKicRRTEVCPxgw4XOcqGQ@mail.gmail.com>
Fixes: 1f8b977a ("sock: enable MSG_ZEROCOPY")
Reported-by: Andreas Hartmann <andihartmann@01019freenet.de>
Reported-by: David Hill <dhill@redhat.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull block fixes from Jens Axboe:
 "It's been a few weeks, so here's a small collection of fixes that
  should go into the current series.

  This contains:

   - NVMe pull request from Christoph, with a few important fixes.

   - kyber hang fix from Omar.

   - A blk-throttl fix from Shaohua, fixing a case where we double
     charge a bio.

   - Two call_single_data alignment fixes from me, fixing up some
     unfortunate changes that went into 4.14 without being properly
     reviewed on the block side (since nobody was CC'ed on the
     patch...).

   - A bounce buffer fix in two parts, one from me and one from Ming.

   - Revert bdi debug error handling patch. It's causing boot issues for
     some folks, and a week down the line, we're still no closer to a
     fix. Revert this patch for now until it's figured out, then we can
     retry for 4.16"

* 'for-linus' of git://git.kernel.dk/linux-block:
  Revert "bdi: add error handle for bdi_debug_register"
  null_blk: unalign call_single_data
  block: unalign call_single_data in struct request
  block-throttle: avoid double charge
  block: fix blk_rq_append_bio
  block: don't let passthrough IO go into .make_request_fn()
  nvme: setup streams after initializing namespace head
  nvme: check hw sectors before setting chunk sectors
  nvme: call blk_integrity_unregister after queue is cleaned up
  nvme-fc: remove double put reference if admin connect fails
  nvme: set discard_alignment to zero
  kyber: fix another domain token wait queue hang

Pull KVM fixes from Paolo Bonzini:
 "ARM fixes:
   - A bug in handling of SPE state for non-vhe systems
   - A fix for a crash on system shutdown
   - Three timer fixes, introduced by the timer optimizations for v4.15

  x86 fixes:
   - fix for a WARN that was introduced in 4.15
   - fix for SMM when guest uses PCID
   - fixes for several bugs found by syzkaller

  ... and a dozen papercut fixes for the kvm_stat tool"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (22 commits)
  tools/kvm_stat: sort '-f help' output
  kvm: x86: fix RSM when PCID is non-zero
  KVM: Fix stack-out-of-bounds read in write_mmio
  KVM: arm/arm64: Fix timer enable flow
  KVM: arm/arm64: Properly handle arch-timer IRQs after vtimer_save_state
  KVM: arm/arm64: timer: Don't set irq as forwarded if no usable GIC
  KVM: arm/arm64: Fix HYP unmapping going off limits
  arm64: kvm: Prevent restoring stale PMSCR_EL1 for vcpu
  KVM/x86: Check input paging mode when cs.l is set
  tools/kvm_stat: add line for totals
  tools/kvm_stat: stop ignoring unhandled arguments
  tools/kvm_stat: suppress usage information on command line errors
  tools/kvm_stat: handle invalid regular expressions
  tools/kvm_stat: add hint on '-f help' to man page
  tools/kvm_stat: fix child trace events accounting
  tools/kvm_stat: fix extra handling of 'help' with fields filter
  tools/kvm_stat: fix missing field update after filter change
  tools/kvm_stat: fix drilldown in events-by-guests mode
  tools/kvm_stat: fix command line option '-g'
  kvm: x86: fix WARN due to uninitialized guest FPU state
  ...

The RGMII spec allows compliance for devices that implement an internal
delay on TXC and/or RXC inside the transmitter. This patch adds the
necessary RGMII_[RX|TX]ID mode code to handle such PHYs with the
emac driver.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The ibm_emac driver predates the PHY_INTERFACE_MODE_*
enums by a few years.

And while the driver has been retrofitted to use the PHYLIB,
the old definitions have stuck around to this day.

This patch replaces all occurences of PHY_MODE_* with
the respective equivalent PHY_INTERFACE_MODE_* enum.
And finally, it purges the old macros for good.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phy_modes() in the common phy.h already defines the same phy mode
names in lower case. The deleted rgmii_mode_name() is used only
in one place and for a "notice-level" printk. Hence, it will not
be missed.
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>