Some 1000BASE-T PHY modules take a while for the PHY to wake up.
Retry the probe a number of times before deciding that the module has
no PHY.

Tested with:
 Sourcephotonics SPGBTXCNFC - PHY takes less than 50ms to respond.
 Champion One 1000SFPT - PHY takes about 200ms to respond.
 Mikrotik S-RJ01 - no PHY
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename sm_retries as sm_fault_retries, as this is what this member is
tracking.
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Russell King says:

====================
Add support for SFP+ copper modules

This series adds support for Copper SFP+ modules with Clause 45 PHYs.
Specifically the patches:

1. drop support for the probably never tested 100BASE-*X modules.
2. drop EEPROM ID from sfp_select_interface()
3. add more compliance code definitions from SFF-8024, renaming the
   existing definitions.
4. add module start/stop methods so phylink knows when a module is
   about to become active. The module start method is called after
   we have probed for a PHY on the module.
5. move start/stop of module PHY down into phylink using the new
   module start/stop methods.
6. add support for Clause 45 I2C accesses, tested with Methode DM7052.
   Other modules appear to use the same protocol, but slight
   differences, but I do not have those modules to test with.
   (if someone does, please holler!)
7. rearrange how we attach to PHYs so that we can support Clause 45
   PHYs with indeterminant interface modes.  (Clause 45 PHYs appear
   to like to change their PHY interface mode depending on the
   negotiated speed.)
8. add support for phylink to connect to a clause 45 PHY on a SFP
   module.
9. split the link_an_mode between the configured value and the
   currently selected mode value; some clause 45 PHYs have no
   capability to provide in-band negotiation.
10. split the link configuration on SFP module insertion in phylink
    so we can use it in other code paths.
11. delay MAC configuration for copper modules without a PHY to the
    module start method - after any module PHY has been probed.  If
    the module has a PHY, then we setup the MAC when the PHY is
    detected.
12. the Broadcom 84881 PHY does not support in-band negotiation even
    though it uses SGMII and 2500BASE-X.  Having the MAC operating
    with in-band negotiation enabled, even with AN bypass enabled,
    results in no link - Broadcom say that the host MAC must always
    be forced.
13. add support for the Broadcom 84881 PHY found on the Methode
    DM7052 module.
14. add support to SFP to probe for a Clause 45 PHY on copper SFP+
    modules.

v3: now bisectable!
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some SFP+ modules have a Clause 45 PHY onboard, which is accessible via
the normal I2C address.  Detect 10G BASE-T PHYs which may have an
accessible PHY and probe for it.
Reviewed-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>

Add a rudimentary Clause 45 driver for the BCM84881 PHY, found on
Methode DM7052 SFPs.
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The Broadcom BCM84881 does not appear to send the SGMII control word
when operating in SGMII mode, which causes network adapters to fail
to link with the PHY, or decide to operate at fixed 1G speed, even if
the PHY negotiated 100M.

Work around this by detecting the Broadcom BCM84881 and switch to phy
mode rather than inband mode.
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Knowing whether we need to delay the MAC configuration because a module
may have a PHY is useful to phylink to allow NBASE-T modules to work on
systems supporting no more than 2.5G speeds.

This commit allows us to delay such configuration until after the PHY
has been probed by recording the parsed capabilities, and if the module
may have a PHY, doing no more until the module_start() notification is
called.  At that point, we either have a PHY, or we don't.

We move the PHY-based setup a little later, and use the PHYs support
capabilities rather than the EEPROM parsed capabilities to determine
whether we can support the PHY.
Reviewed-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>

Split out the configuration step from phylink_sfp_module_insert() so
we can re-use this later.
Reviewed-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>

Split link_an_mode between the configured setting and the current
operating setting.  This is an important distinction to make when we
need to configure PHY mode for a plugged SFP+ module that does not
use in-band signalling.
Reviewed-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>

Some SFP+ modules have Clause 45 PHYs embedded on them, which need a
little more handling in order to ensure that they are correctly setup,
as they switch the PHY link mode according to the negotiated speed.

With Clause 22 PHYs, we assumed that they would operate in SGMII mode,
but this assumption is now false.  Adapt phylink to support Clause 45
PHYs on SFP+ modules.
Reviewed-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>

In order to support Clause 45 PHYs on SFP+ modules, which have an
indeterminant phy interface mode, we need to be able to call
phylink_bringup_phy() with a different interface mode to that used when
binding the PHY. Reduce __phylink_connect_phy() to an attach operation,
and move the call to phylink_bringup_phy() to its call sites.
Reviewed-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>

Some SFP+ modules have PHYs on them just like SFP modules do, except
they are Clause 45 PHYs.  The I2C protocol used to access them is
modified slightly in order to send the device address and 16-bit
register index.
Reviewed-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>

Move phy_start() and phy_stop() into the module_start and module_stop
notifications in phylink, rather than having them in the SFP code.
This gives phylink responsibility for controlling the PHY, rather
than having SFP start and stop the PHY state machine.
Reviewed-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>

When dealing with some copper modules, we can't positively know the
module capabilities are until we have probed the PHY. Without the full
capabilities, we may end up failing a module that we could otherwise
drive with a restricted set of capabilities.

An example of this would be a module with a NBASE-T PHY plugged into
a host that supports phy interface modes 2500BASE-X and SGMII. The
PHY supports 10GBASE-R, 5000BASE-X, 2500BASE-X, SGMII interface modes,
which means a subset of the capabilities are compatible with the host.

However, reading the module EEPROM leads us to believe that the module
only supports ethtool link mode 10GBASE-T, which is incompatible with
the host - and thus results in the module being rejected.

This patch adds an extra notification which are triggered after the
SFP module's PHY probe, and a corresponding notification just before
the PHY is removed.
Reviewed-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>

SFF-8024 is used to define various constants re-used in several SFF
SFP-related specifications.  Split these constants from the enum, and
rename them to indicate that they're defined by SFF-8024.

Add and use updated SFF-8024 extended compliance code definitions for
10GBASE-T, 5GBASE-T and 2.5GBASE-T modules.
Reviewed-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>

We don't need the EEPROM ID to derive the phy interface mode as we can
derive it merely from the ethtool link modes.  Remove the EEPROM ID
argument to sfp_select_interface().
Reviewed-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>

The 100BASE-FX and 100BASE-LX support assumes a PHY is present; this
is probably an incorrect assumption. In any case, sfp_parse_support()
will fail such a module. Let's stop pretending we support these
modules.
Reviewed-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>

T6 has a separate region known as high priority filter region
that allows classifying packets going through ULD path. So,
query firmware for HPFILTER resources and enable the high
priority offload filter support when it is available.
Signed-off-by: NShahjada Abul Husain <shahjada@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fixes gcc '-Wunused-but-set-variable' warning:

drivers/net/ethernet/freescale/enetc/enetc_qos.c: In function enetc_setup_tc_cbs:
drivers/net/ethernet/freescale/enetc/enetc_qos.c:195:6: warning: variable tc_max_sized_frame set but not used [-Wunused-but-set-variable]

Fixes: c431047c ("enetc: add support Credit Based Shaper(CBS) for hardware offload")
Signed-off-by: NChen Wandun <chenwandun@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Device stats are currently hard coded in the PCI BAR0 layout.
Add a ability to read them from the TLV area instead.
Names for the stats are maintained by the driver, and their
meaning documented. This allows us to more easily add and
remove device stats.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a TCP socket is created, sk->sk_state is initialized twice as
TCP_CLOSE in sock_init_data() and tcp_init_sock(). The tcp_init_sock() is
always called after the sock_init_data(), so it is not necessary to update
sk->sk_state in the tcp_init_sock().

Before v2.1.8, the code of the two functions was in the inet_create(). In
the patch of v2.1.8, the tcp_v4/v6_init_sock() were added and the code of
initialization of sk->state was duplicated.
Signed-off-by: NKuniyuki Iwashima <kuni1840@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Provide a software TX timestamp and add it to the ethtool query
interface.

skb_tx_timestamp() is also needed if one would like to use PHY
timestamping.
Signed-off-by: NMichael Walle <michael@walle.cc>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jon Maloy says:

====================
tipc: introduce variable window congestion control

We improve thoughput greatly by introducing a variety of the Reno
congestion control algorithm at the link level.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We introduce a simple variable window congestion control for links.
The algorithm is inspired by the Reno algorithm, covering both 'slow
start', 'congestion avoidance', and 'fast recovery' modes.

- We introduce hard lower and upper window limits per link, still
  different and configurable per bearer type.

- We introduce a 'slow start theshold' variable, initially set to
  the maximum window size.

- We let a link start at the minimum congestion window, i.e. in slow
  start mode, and then let is grow rapidly (+1 per rceived ACK) until
  it reaches the slow start threshold and enters congestion avoidance
  mode.

- In congestion avoidance mode we increment the congestion window for
  each window-size number of acked packets, up to a possible maximum
  equal to the configured maximum window.

- For each non-duplicate NACK received, we drop back to fast recovery
  mode, by setting the both the slow start threshold to and the
  congestion window to (current_congestion_window / 2).

- If the timeout handler finds that the transmit queue has not moved
  since the previous timeout, it drops the link back to slow start
  and forces a probe containing the last sent sequence number to the
  sent to the peer, so that this can discover the stale situation.

This change does in reality have effect only on unicast ethernet
transport, as we have seen that there is no room whatsoever for
increasing the window max size for the UDP bearer.
For now, we also choose to keep the limits for the broadcast link
unchanged and equal.

This algorithm seems to give a 50-100% throughput improvement for
messages larger than MTU.
Suggested-by: NXin Long <lucien.xin@gmail.com>
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>

When we increase the link tranmsit window we often observe the following
scenario:

1) A STATE message bypasses a sequence of traffic packets and arrives
   far ahead of those to the receiver. STATE messages contain a
   'peers_nxt_snt' field to indicate which was the last packet sent
   from the peer. This mechanism is intended as a last resort for the
   receiver to detect missing packets, e.g., during very low traffic
   when there is no packet flow to help early loss detection.
3) The receiving link compares the 'peer_nxt_snt' field to its own
   'rcv_nxt', finds that there is a gap, and immediately sends a
   NACK message back to the peer.
4) When this NACKs arrives at the sender, all the requested
   retransmissions are performed, since it is a first-time request.

Just like in the scenario described in the previous commit this leads
to many redundant retransmissions, with decreased throughput as a
consequence.

We fix this by adding two more conditions before we send a NACK in
this sitution. First, the deferred queue must be empty, so we cannot
assume that the potential packet loss has already been detected by
other means. Second, we check the 'peers_snd_nxt' field only in probe/
probe_reply messages, thus turning this into a true mechanism of last
resort as it was really meant to be.
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>

When we increase the link send window we sometimes observe the
following scenario:

1) A packet #N arrives out of order far ahead of a sequence of older
   packets which are still under way. The packet is added to the
   deferred queue.
2) The missing packets arrive in sequence, and for each 16th of them
   an ACK is sent back to the receiver, as it should be.
3) When building those ACK messages, it is checked if there is a gap
   between the link's 'rcv_nxt' and the first packet in the deferred
   queue. This is always the case until packet number #N-1 arrives, and
   a 'gap' indicator is added, effectively turning them into NACK
   messages.
4) When those NACKs arrive at the sender, all the requested
   retransmissions are done, since it is a first-time request.

This sometimes leads to a huge amount of redundant retransmissions,
causing a drop in max throughput. This problem gets worse when we
in a later commit introduce variable window congestion control,
since it drops the link back to 'fast recovery' much more often
than necessary.

We now fix this by not sending any 'gap' indicator in regular ACK
messages. We already have a mechanism for sending explicit NACKs
in place, and this is sufficient to keep up the packet flow.
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>

Clang warns:

../drivers/net/ppp/ppp_async.c:877:6: warning: misleading indentation;
statement is not part of the previous 'if' [-Wmisleading-indentation]
                                ap->rpkt = skb;
                                ^
../drivers/net/ppp/ppp_async.c:875:5: note: previous statement is here
                                if (!skb)
                                ^
1 warning generated.

This warning occurs because there is a space before the tab on this
line. Clean up this entire block's indentation so that it is consistent
with the Linux kernel coding style and clang no longer warns.

Fixes: 6722e78c ("[PPP]: handle misaligned accesses")
Link: https://github.com/ClangBuiltLinux/linux/issues/800Signed-off-by: NNathan Chancellor <natechancellor@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Clang warns:

../drivers/net/ethernet/smsc/smc911x.c:939:3: warning: misleading
indentation; statement is not part of the previous 'if'
[-Wmisleading-indentation]
         if (!lp->ctl_rfduplx)
         ^
../drivers/net/ethernet/smsc/smc911x.c:936:2: note: previous statement
is here
        if (lp->ctl_rspeed != 100)
        ^
1 warning generated.

This warning occurs because there is a space after the tab on this line.
Remove it so that the indentation is consistent with the Linux kernel
coding style and clang no longer warns.

Fixes: 0a0c72c9 ("[PATCH] RE: [PATCH 1/1] net driver: Add support for SMSC LAN911x line of ethernet chips")
Link: https://github.com/ClangBuiltLinux/linux/issues/796Signed-off-by: NNathan Chancellor <natechancellor@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Clang warns:

../drivers/net/ethernet/dec/tulip/uli526x.c:1812:3: warning: misleading
indentation; statement is not part of the previous 'if'
[-Wmisleading-indentation]
        switch (mode) {
        ^
../drivers/net/ethernet/dec/tulip/uli526x.c:1809:2: note: previous
statement is here
        if (cr6set)
        ^
1 warning generated.

../drivers/net/ethernet/dec/tulip/dmfe.c:2217:3: warning: misleading
indentation; statement is not part of the previous 'if'
[-Wmisleading-indentation]
        switch(mode) {
        ^
../drivers/net/ethernet/dec/tulip/dmfe.c:2214:2: note: previous
statement is here
        if (cr6set)
        ^
1 warning generated.

This warning occurs because there is a space before the tab on these
lines. Remove them so that the indentation is consistent with the Linux
kernel coding style and clang no longer warns.

While we are here, adjust the default block in dmfe_init_module to have
a proper break between the label and assignment and add a space between
the switch and opening parentheses to avoid a checkpatch warning.

Fixes: e1c3e501 ("[PATCH] initialisation cleanup for ULI526x-net-driver")
Link: https://github.com/ClangBuiltLinux/linux/issues/795Signed-off-by: NNathan Chancellor <natechancellor@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dan Murphy says:

====================
Fix Tx/Rx FIFO depth for DP83867

The DP83867 supports both the RGMII and SGMII modes.  The Tx and Rx FIFO depths
are configurable in these modes but may not applicable for both modes.

When the device is configured for RGMII mode the Tx FIFO depth is applicable
and for SGMII mode both Tx and Rx FIFO depth settings are applicable.  When
the driver was originally written only the RGMII device was available and there
were no standard fifo-depth DT properties.

The patchset converts the special ti,fifo-depth property to the standard
tx-fifo-depth property while still allowing the ti,fifo-depth property to be
set as to maintain backward compatibility.

In addition to this change the rx-fifo-depth property support was added and only
written when the device is configured for SGMII mode.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This code changes the TI specific ti,fifo-depth to the common
tx-fifo-depth property.  The tx depth is applicable for both RGMII and
SGMII modes of operation.

rx-fifo-depth was added as well but this is only applicable for SGMII
mode.

So in summary
if RGMII mode write tx fifo depth only
if SGMII mode write both rx and tx fifo depths

If the property is not populated in the device tree then set the value
to the default values.
Signed-off-by: NDan Murphy <dmurphy@ti.com>
Reported-by: NAdrian Bunk <bunk@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Convert the ti,fifo-depth from a TI specific property to the common
tx-fifo-depth property.  Also add support for the rx-fifo-depth.

These are optional properties for this device and if these are not
available then the fifo depths are set to device default values.
Signed-off-by: NDan Murphy <dmurphy@ti.com>
Reported-by: NAdrian Bunk <bunk@kernel.org>
CC: Rob Herring <robh@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch introduces a sysctl knob "net.ipv4.tcp_no_ssthresh_metrics_save"
that disables TCP ssthresh metrics cache by default. Other parts of TCP
metrics cache, e.g. rtt, cwnd, remain unchanged.

As modern networks becoming more and more dynamic, TCP metrics cache
today often causes more harm than benefits. For example, the same IP
address is often shared by different subscribers behind NAT in residential
networks. Even if the IP address is not shared by different users,
caching the slow-start threshold of a previous short flow using loss-based
congestion control (e.g. cubic) often causes the future longer flows of
the same network path to exit slow-start prematurely with abysmal
throughput.

Caching ssthresh is very risky and can lead to terrible performance.
Therefore it makes sense to make disabling ssthresh caching by
default and opt-in for specific networks by the administrators.
This practice also has worked well for several years of deployment with
CUBIC congestion control at Google.
Acked-by: NEric Dumazet <edumazet@google.com>
Acked-by: NNeal Cardwell <ncardwell@google.com>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NKevin(Yudong) Yang <yyd@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 31243461 ("sctp: cache netns in sctp_ep_common") set netns
in asoc and ep base since they're created, and it will never change.
It's a better way to get netns from asoc and ep base, comparing to
calling sock_net().

This patch is to replace them.

v1->v2:
  - no change.
Suggested-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: NXin Long <lucien.xin@gmail.com>
Acked-by: NNeil Horman <nhorman@tuxdriver.com>
Acked-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The Nokia GPON module can hold tx-fault active while it is initialising
which can take up to 60s. Avoid this causing the module to be declared
faulty after the SFP MSA defined non-cooled module timeout.
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The variable rc is assigned with a value that is never read and
it is re-assigned a new value later on.  The assignment is redundant
and can be removed.  Clean up multiple occurrances of this pattern.

Addresses-Coverity: ("Unused value")
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Convert cpu_to_le16(le16_to_cpu(frame->datalen) + len) to
use le16_add_cpu(), which is more concise and does the same thing.
Reported-by: NHulk Robot <hulkci@huawei.com>
Signed-off-by: NMao Wenan <maowenan@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2019-12-09

Here's the first bluetooth-next pull request for 5.6:

 - Devicetree bindings updates for Broadcom controllers
 - Add support for PCM configuration for Broadcom controllers
 - btusb: Fixes for Realtek devices
 - butsb: A few other smaller fixes (mem leak & non-atomic allocation issue)

Please let me know if there are any issues pulling. Thanks.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>