The current definition of dsaf_drv_tbl_tcam_key is only suitable for
Little-endian. If data is stored in Big-endian, this may lead to
error in data use. Shift operation can make it work normally in both
Big-endian and Little-endian.
Signed-off-by: NQianqian Xie <xieqianqian@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Hardware ring buffer data is stored in Little-endian. Thus cpu data
should be modified to Little-endian.
Signed-off-by: NQianqian Xie <xieqianqian@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In current scenario, when the interface is disabled we reset the XGMAC
RX/TX functionality. This operation does not affects the PHY layer/SFP
and which appears UP to the remote end(this behaviour is unlike GMAC).
The result is remote end keeps on sending the packets which gets partly
processed by XMAC and dropped. Since these are partly processed these
appears as errored packets in the packet counter statistics.

This patch fixes this behaviour and adds local-fault and remote-fault
functionality which can be used to intimate the remote peer whenever
the state of the interface changes. This patch also removes the
existing hns_dsaf_xge_core_srst_by_port function which was being used
to reset the RX/TX functionality at XGE Core.
Reported-by: NJun He <hjat2005@huawei.com>
Signed-off-by: NDaode Huang <huangdaode@hisilicon.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch modify the gmac_rx_filt_pkt and gmac_rx_octets_total_filt
statistics value. The two statistics is inconsistent with register,
and just the opposite.
Signed-off-by: NQianqian Xie <xieqianqian@huawei.com>
Signed-off-by: NJun He <hjat2005@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch deletes redundant macro definitions in hns drivers.
And change the .h file containing relation to make the layers
more clearly
Signed-off-by: NQianqian Xie <xieqianqian@huawei.com>
Signed-off-by: NWeiwei Deng <dengweiwei@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When set auto-negotiation off and duplex half, if run "ethtool -r ethX"
on port with phy, then the port will be failed to work. It should
forbid to start auto-negotiation when auto-negotiate is off. This
patch add the limited condition.
Reported-by: NJinchuang Tian <tianjinchuang1@huawei.com>
Signed-off-by: NDaode Huang <huangdaode@hisilicon.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Reviewed-by: Nlipeng <lipeng321@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The default mac pause time set to 0xff which is too short for pausing,
this patch change it to the max value 0xffff.
Signed-off-by: NDaode Huang <huangdaode@hisilicon.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Reviewed-by: Nlipeng <lipeng321@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If set promisc mode when there is some traffic, The service nic will
cause system halted. We reserve the last 6 tcam entry for the 6 ports.
If promisc mode is enabled, we can config the relative tcam as fuzzy
matching and set to be valid, or set the tcam to be invalid
Signed-off-by: NKejian Yan <yankejian@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since there is not enough tcam table entries for vlan and multicast
address, HNSv2 needs to add support of fuzzy matching of TCAM tables.
To add fuzzy match of TCAM, we Add the property to mask the bits to
be fuzzy matched
Signed-off-by: NKejian Yan <yankejian@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since there is not enough tcam table entries for every vlan and multicast
address, HNS needs to add support of fuzzy matching of TCAM tables. Adding
the property to mask the bits to be fuzzy matched, so update the bindings
document
Signed-off-by: NKejian Yan <yankejian@huawei.com>
Reviewed-by: NYisen Zhuang <yisen.zhuang@huawei.com>
Signed-off-by: NSalil Mehta <salil.mehta@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After commit 6ed46d12 ("sock_diag: align nlattr properly when
needed"), tcp_get_info() gets 64bit aligned memory, so we can avoid
the unaligned helpers.
Suggested-by: NDavid Miller <davem@davemloft.net>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-by: NSoheil Hassas Yeganeh <soheil@google.com>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Simon Wunderlich says:

====================
pull request for net-next: batman-adv 2016-11-08 v2

This feature and cleanup patchset includes the following changes:

 - netlink and code cleanups by Sven Eckelmann (3 patches)

 - Cleanup and minor fixes by Linus Luessing (3 patches)

 - Speed up multicast update intervals, by Linus Luessing

 - Avoid (re)broadcast in meshes for some easy cases,
   by Linus Luessing

 - Clean up tx return state handling, by Sven Eckelmann (6 patches)

 - Fix some special mac address handling cases, by Sven Eckelmann
   (3 patches)
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Richard Cochran says:

====================
PHC frequency fine tuning

This series expands the PTP Hardware Clock subsystem by adding a
method that passes the frequency tuning word to the the drivers
without dropping the low order bits.  Keeping those bits is useful for
drivers whose frequency resolution is higher than 1 ppb.

The appended script (below) runs a simple demonstration of the
improvement.  This test needs two Intel i210 PCIe cards installed in
the same PC, with their SDP0 pins connected by copper wire.  Measuring
the estimated offset (from the ptp4l servo) and the true offset (from
the PPS) over one hour yields the following statistics.

|        |   Est. Before |    Est. After |   True Before |    True After |
|--------+---------------+---------------+---------------+---------------|
| min    | -5.200000e+01 | -1.600000e+01 | -3.100000e+01 | -1.000000e+00 |
| max    | +5.700000e+01 | +2.500000e+01 | +8.500000e+01 | +4.000000e+01 |
| pk-pk: | +1.090000e+02 | +4.100000e+01 | +1.160000e+02 | +4.100000e+01 |
| mean   | +6.472222e-02 | +1.277778e-02 | +2.422083e+01 | +1.826083e+01 |
| stddev | +1.158006e+01 | +4.581982e+00 | +1.207708e+01 | +4.981435e+00 |

Here the numbers in units of nanoseconds, and the ~20 nanosecond PPS
offset is due to input/output delays on the i210's external interface
logic.

With the series applied, both the peak to peak error and the standard
deviation improve by a factor of more than two.  These two graphs show
the improvement nicely.

  http://linuxptp.sourceforge.net/fine-tuning/fine-est.png

  http://linuxptp.sourceforge.net/fine-tuning/fine-tru.png
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The dp83640 has a frequency resolution of about 0.029 ppb.
This patch lets users of the device benefit from the
increased frequency resolution when tuning the clock.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The 82580 and related devices offer a frequency resolution of about
0.029 ppb.  This patch lets users of the device benefit from the
increased frequency resolution when tuning the clock.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The internal PTP Hardware Clock (PHC) interface limits the resolution for
frequency adjustments to one part per billion.  However, some hardware
devices allow finer adjustment, and making use of the increased resolution
improves synchronization measurably on such devices.

This patch adds an alternative method that allows finer frequency tuning
by passing the scaled ppm value to PHC drivers.  This value comes from
user space, and it has a resolution of about 0.015 ppb.  We also deprecate
the older method, anticipating its removal once existing drivers have been
converted over.
Signed-off-by: NRichard Cochran <richardcochran@gmail.com>
Suggested-by: NUlrik De Bie <ulrik.debie-os@e2big.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are no more users except from net/core/dev.c
napi_hash_add() can now be static.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This is automatically done from netif_napi_add(), and we want to not
export napi_hash_add() anymore in the following patch.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Michael Chan <michael.chan@broadcom.com>
Acked-by: NMichael Chan <michael.chan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove the unused but set variables min_set and max_set in
adjust_reg_min_max_vals to fix the following warning when building with
'W=1':

  kernel/bpf/verifier.c:1483:7: warning: variable ‘min_set’ set but not used [-Wunused-but-set-variable]

There is no warning about max_set being unused, but since it is only
used in the assignment of min_set it can be removed as well.

They were introduced in commit 48461135 ("bpf: allow access into map
value arrays") but seem to have never been used.

Cc: Josef Bacik <jbacik@fb.com>
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

tc_act macro addressed a non existing field, and was not used in the
kernel source.
Signed-off-by: NYotam Gigi <yotamg@mellanox.com>
Reviewed-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

David Lebrun says:

====================
net: add support for IPv6 Segment Routing

v5:
 - Check SRH validity when adding a new route with lwtunnels and
   when setting an IPV6_RTHDR socket option.
 - Check that hdr->segments_left is not out of bounds when processing
   an SR-enabled packet.
 - Add __ro_after_init attribute to seg6_genl_policy structure.
 - Add CONFIG_IPV6_SEG6_INLINE option to enable or disable
   direct header insertion.

v4:
 - Change @cleanup in ipv6_srh_rcv() from int to bool
 - Move checksum helper functions into header file
 - Add common definition for SR TLVs
 - Add comments for HMAC computation algorithm
 - Use rhashtable to store HMAC infos instead of linked list
 - Remove packed attribute for struct sr6_tlv_hmac
 - Use dst cache only if CONFIG_DST_CACHE is enabled

v3:
 - Fix compilation for CONFIG_IPV6={n,m}

v2:
 - Remove packed attribute from sr6 struct and replaced unaligned
   16-bit flags with two 8-bit flags.
 - SR code now included by default. Option CONFIG_IPV6_SEG6_HMAC
   exists for HMAC support (which requires crypto dependencies).
 - Replace "hidden" calls to mutex_{un,}lock to direct calls.
 - Fix reverse xmas tree coding style.
 - Fix cast-from-void*'s.
 - Update skb->csum to account for SR modifications.
 - Add dst_cache in seg6_output.

Segment Routing (SR) is a source routing paradigm, architecturally
defined in draft-ietf-spring-segment-routing-09 [1]. The IPv6 flavor of
SR is defined in draft-ietf-6man-segment-routing-header-02 [2].

The main idea is that an SR-enabled packet contains a list of segments,
which represent mandatory waypoints. Each waypoint is called a segment
endpoint. The SR-enabled packet is routed normally (e.g. shortest path)
between the segment endpoints. A node that inserts an SRH into a packet
is called an ingress node, and a node that is the last segment endpoint
is called an egress node.

From an IPv6 viewpoint, an SR-enabled packet contains an IPv6 extension
header, which is a Routing Header type 4, defined as follows:

struct ipv6_sr_hdr {
        __u8    nexthdr;
        __u8    hdrlen;
        __u8    type;
        __u8    segments_left;
        __u8    first_segment;
        __u8    flag_1;
        __u8    flag_2;
        __u8    reserved;

        struct in6_addr segments[0];
};

The first 4 bytes of the SRH is consistent with the Routing Header
definition in RFC 2460. The type is set to `4' (SRH).

Each segment is encoded as an IPv6 address. The segments are encoded in
reverse order: segments[0] is the last segment of the path, and
segments[first_segment] is the first segment of the path.

segments[segments_left] points to the currently active segment and
segments_left is decremented at each segment endpoint.

There exist two ways for a packet to receive an SRH, we call them
encap mode and inline mode. In the encap mode, the packet is encapsulated
in an outer IPv6 header that contains the SRH. The inner (original) packet
is not modified. A virtual tunnel is thus created between the ingress node
(the node that encapsulates) and the egress node (the last segment of the path).
Once an encapsulated SR packet reaches the egress node, the node decapsulates
the packet and performs a routing decision on the inner packet. This kind of
SRH insertion is intended to use for routers that encapsulates in-transit
packet.

The second SRH insertion method, the inline mode, acts by directly inserting
the SRH right after the IPv6 header of the original packet. For this method,
if a particular flag (SR6_FLAG_CLEANUP) is set, then the penultimate segment
endpoint must strip the SRH from the packet before forwarding it to the last
segment endpoint. This insertion method is intended to use for endhosts,
however it is also used for in-transit packets by some industry actors.
Note that directly inserting extension headers may break several mechanisms
such as Path MTU Discovery, IPSec AH, etc. For this reason, this insertion
method is only available if CONFIG_IPV6_SEG6_INLINE is enabled.

Finally, the SRH may contain TLVs after the segments list. Several types of
TLVs are defined, but we currently consider only the HMAC TLV. This TLV is
an answer to the deprecation of the RH0 and enables to ensure the authenticity
and integrity of the SRH. The HMAC text contains the flags, the first_segment
index, the full list of segments, and the source address of the packet. While
SR is intended to use mostly within a single administrative domain, the HMAC
TLV allows to verify SR packets coming from an untrusted source.

This patches series implements support for the IPv6 flavor of SR and is
logically divided into the following components:

        (1) Data plane support (patch 01). This patch adds a function
            in net/ipv6/exthdrs.c to handle the Routing Header type 4.
            It enables the kernel to act as a segment endpoint, by supporting
            the following operations: decrementation of the segments_left field,
            cleanup flag support (removal of the SRH if we are the penultimate
            segment endpoint) and decapsulation of the inner packet as an egress
            node.

        (2) Control plane support (patches 02..03 and 07..09). These patches enables
            to insert SRH on locally emitted and/or forwarded packets, both with
            encap mode and with inline mode. The SRH insertion is controlled through
            the lightweight tunnels mechanism. Furthermore, patch 08 enables the
            applications to insert an SRH on a per-socket basis, through the
            setsockopt() system call. The mechanism to specify a per-socket
            Routing Header was already defined for RH0 and no special modification
            was performed on this side. However, the code to actually push the RH
            onto the packets had to be adapted for the SRH specifications.

        (3) HMAC support (patches 04..06). These patches adds the support of the
            HMAC TLV verification for the dataplane part, and generation for
            the control plane part. Two hashing algorithms are supported
            (SHA-1 as legacy and SHA-256 as required by the IETF draft), but
            additional algorithms can be easily supported by simply adding an
            entry into an array.

[1] https://tools.ietf.org/html/draft-ietf-spring-segment-routing-09
[2] https://tools.ietf.org/html/draft-ietf-6man-segment-routing-header-02
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds documentation for some SR-related per-interface
sysctls.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for per-socket SRH injection with the setsockopt
system call through the IPPROTO_IPV6, IPV6_RTHDR options.
The SRH is pushed through the ipv6_push_nfrag_opts function.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch prepares for insertion of SRH through setsockopt().
The new source address argument is used when an HMAC field is
present in the SRH, which must be filled. The HMAC signature
process requires the source address as input text.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch enables the verification of the HMAC signature for transiting
SR-enabled packets, and its insertion on encapsulated/injected SRH.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch provides an implementation of the genetlink commands
to associate a given HMAC key identifier with an hashing algorithm
and a secret.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds the necessary functions to compute and check the HMAC signature
of an SR-enabled packet. Two HMAC algorithms are supported: hmac(sha1) and
hmac(sha256).

In order to avoid dynamic memory allocation for each HMAC computation,
a per-cpu ring buffer is allocated for this purpose.

A new per-interface sysctl called seg6_require_hmac is added, allowing a
user-defined policy for processing HMAC-signed SR-enabled packets.
A value of -1 means that the HMAC field will always be ignored.
A value of 0 means that if an HMAC field is present, its validity will
be enforced (the packet is dropped is the signature is incorrect).
Finally, a value of 1 means that any SR-enabled packet that does not
contain an HMAC signature or whose signature is incorrect will be dropped.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch creates a new type of interfaceless lightweight tunnel (SEG6),
enabling the encapsulation and injection of SRH within locally emitted
packets and forwarded packets.

>From a configuration viewpoint, a seg6 tunnel would be configured as follows:

  ip -6 ro ad fc00::1/128 encap seg6 mode encap segs fc42::1,fc42::2,fc42::3 dev eth0

Any packet whose destination address is fc00::1 would thus be encapsulated
within an outer IPv6 header containing the SRH with three segments, and would
actually be routed to the first segment of the list. If `mode inline' was
specified instead of `mode encap', then the SRH would be directly inserted
after the IPv6 header without outer encapsulation.

The inline mode is only available if CONFIG_IPV6_SEG6_INLINE is enabled. This
feature was made configurable because direct header insertion may break
several mechanisms such as PMTUD or IPSec AH.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds the necessary hooks and structures to provide support
for SR-IPv6 control plane, essentially the Generic Netlink commands
that will be used for userspace control over the Segment Routing
kernel structures.

The genetlink commands provide control over two different structures:
tunnel source and HMAC data. The tunnel source is the source address
that will be used by default when encapsulating packets into an
outer IPv6 header + SRH. If the tunnel source is set to :: then an
address of the outgoing interface will be selected as the source.

The HMAC commands currently just return ENOTSUPP and will be implemented
in a future patch.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement minimal support for processing of SR-enabled packets
as described in
https://tools.ietf.org/html/draft-ietf-6man-segment-routing-header-02.

This patch implements the following operations:
- Intermediate segment endpoint: incrementation of active segment and rerouting.
- Egress for SR-encapsulated packets: decapsulation of outer IPv6 header + SRH
  and routing of inner packet.
- Cleanup flag support for SR-inlined packets: removal of SRH if we are the
  penultimate segment endpoint.

A per-interface sysctl seg6_enabled is provided, to accept/deny SR-enabled
packets. Default is deny.

This patch does not provide support for HMAC-signed packets.
Signed-off-by: NDavid Lebrun <david.lebrun@uclouvain.be>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The newly introduced mii_ethtool_get_link_ksettings function sets
lp_advertising to an uninitialized value when BMCR_ANENABLE is not
set:

drivers/net/mii.c: In function 'mii_ethtool_get_link_ksettings':
drivers/net/mii.c:224:2: error: 'lp_advertising' may be used uninitialized in this function [-Werror=maybe-uninitialized]

As documented in include/uapi/linux/ethtool.h, the value is
expected to be zero when we don't know it, so let's initialize
it to that.

Fixes: bc8ee596 ("net: mii: add generic function to support ksetting support")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For single items being collected this should be preferred as being more
typesafe (as the compiler can check format string and to-be-written-to
variable match) and more efficient (requiring one less parameter to be
passed).
Signed-off-by: NJan Beulich <jbeulich@suse.com>
Reviewed-by: NPaul Durrant <paul.durrant@citrix.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is some difference between force_igmp_version and force_mld_version.
Add document to make users aware of this.
Signed-off-by: NHangbin Liu <liuhangbin@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

recv_seq, send_seq and lns_mode mode are all defined as
unsigned int foo:1;
Signed-off-by: NAsbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

These assignments follow this pattern:

	unsigned int foo:1;
	struct nlattr *nla = info->attrs[bar];

	if (nla)
		foo = nla_get_flag(nla); /* expands to: foo = !!nla */

This could be simplified to: if (nla) foo = 1;
but lets just remove the condition and use the macro,

	foo = nla_get_flag(nla);
Signed-off-by: NAsbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch causes the proper attribute flags to be set,
in the case that IPv6 UDP checksums are disabled, so that
userspace ie. `ip l2tp show tunnel` knows about it.
Signed-off-by: NAsbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Only set L2TP_ATTR_UDP_CSUM in l2tp_nl_tunnel_send()
when it's running over IPv4.

This prepares the code to also have IPv6 specific attributes.
Signed-off-by: NAsbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The attributes L2TP_ATTR_UDP_ZERO_CSUM6_RX and
L2TP_ATTR_UDP_ZERO_CSUM6_TX are used as flags,
but is defined as a u8 in a comment.

This patch redocuments them as flags.

Adding nla_policy entries would break API, so not doing that.

CC: Tom Herbert <therbert@google.com>
Signed-off-by: NAsbjoern Sloth Toennesen <asbjorn@asbjorn.st>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Receiving a GSO packet in dev_gro_receive() is not uncommon
in stacked devices, or devices partially implementing LRO/GRO
like bnx2x. GRO is implementing the aggregation the device
was not able to do itself.

Current code causes reorders, like in following case :

For a given flow where sender sent 3 packets P1,P2,P3,P4

Receiver might receive P1 as a single packet, stored in GRO engine.

Then P2-P4 are received as a single GSO packet, immediately given to
upper stack, while P1 is held in GRO engine.

This patch will make sure P1 is given to upper stack, then P2-P4
immediately after.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Edward Cree says:

====================
sfc: enable 4-tuple UDP RSS hashing

EF10 based NICs have configurable RSS hash fields, and can be made to take the
ports into the hash on UDP (they already do so for TCP).  This patch series
enables this, in order to improve spreading of UDP traffic.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>