Martin KaFai Lau says:

====================
bpf: LRU performance and test-program improvements

The first 4 patches make a few improvements to the LRU tests.

Patch 5/6 is to improve the performance of BPF_F_NO_COMMON_LRU map.

Patch 6/6 adds an example in using LRU map with map-in-map.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds a map-in-map LRU example.
If we know only a subset of cores will use the
LRU, we can allocate a common LRU list per targeting core
and store it into an array-of-hashs.

It allows using the common LRU map with map-update performance
comparable to the BPF_F_NO_COMMON_LRU map but without wasting memory
on the unused cores that we know they will never access the LRU map.

BPF_F_NO_COMMON_LRU:
> map_perf_test 32 8 10000000 10000000 | awk '{sum += $3}END{print sum}'
9234314 (9.23M/s)

map-in-map LRU:
> map_perf_test 512 8 1260000 80000000 | awk '{sum += $3}END{print sum}'
9962743 (9.96M/s)

Notes that the max_entries for the map-in-map LRU test is 1260000 which
is the max_entries for each inner LRU map.  8 processes have been
started, so 8 * 1260000 = 10080000 (~10M) which is close to what is
used in the BPF_F_NO_COMMON_LRU test.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After doing map_perf_test with a much bigger
BPF_F_NO_COMMON_LRU map, the perf report shows a
lot of time spent in rotating the inactive list (i.e.
__bpf_lru_list_rotate_inactive):
> map_perf_test 32 8 10000 1000000 | awk '{sum += $3}END{print sum}'
19644783 (19M/s)
> map_perf_test 32 8 10000000 10000000 |  awk '{sum += $3}END{print sum}'
6283930 (6.28M/s)

By inactive, it usually means the element is not in cache.  Hence,
there is a need to tune the PERCPU_NR_SCANS value.

This patch finds a better number of elements to
scan during each list rotation.  The PERCPU_NR_SCANS (which
is defined the same as PERCPU_FREE_TARGET) decreases
from 16 elements to 4 elements.  This change only
affects the BPF_F_NO_COMMON_LRU map.

The test_lru_dist does not show meaningful difference
between 16 and 4.  Our production L4 load balancer which uses
the LRU map for conntrack-ing also shows little change in cache
hit rate.  Since both benchmark and production data show no
cache-hit difference, PERCPU_NR_SCANS is lowered from 16 to 4.
We can consider making it configurable if we find a usecase
later that shows another value works better and/or use
a different rotation strategy.

After this change:
> map_perf_test 32 8 10000000 10000000 |  awk '{sum += $3}END{print sum}'
9240324 (9.2M/s)

i.e. 6.28M/s -> 9.2M/s

The test_lru_dist has not shown meaningful difference:
> test_lru_dist zipf.100k.a1_01.out 4000 1:
nr_misses: 31575 (Before) vs 31566 (After)

> test_lru_dist zipf.100k.a0_01.out 40000 1
nr_misses: 67036 (Before) vs 67031 (After)
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The current bpf_map_def is statically defined during compile
time.  This patch allows the *_user.c program to change it during
runtime.  It is done by adding load_bpf_file_fixup_map() which
takes a callback.  The callback will be called before creating
each map so that it has a chance to modify the bpf_map_def.

The current usecase is to change max_entries in map_perf_test.
It is interesting to test with a much bigger map size in
some cases (e.g. the following patch on bpf_lru_map.c).
However,  it is hard to find one size to fit all testing
environment.  Hence, it is handy to take the max_entries
as a cmdline arg and then configure the bpf_map_def during
runtime.

This patch adds two cmdline args.  One is to configure
the map's max_entries.  Another is to configure the max_cnt
which controls how many times a syscall is called.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

One more LRU test will be added later in this patch series.
In this patch, we first move all existing LRU map tests into
a single syscall (connect) first so that the future new
LRU test can be added without hunting another syscall.

One of the map name is also changed from percpu_lru_hash_map
to nocommon_lru_hash_map to avoid the confusion with percpu_hash_map.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch does the following cleanup on test_lru_map.c
1) Fix indentation (Replace spaces by tabs)
2) Remove redundant BPF_F_NO_COMMON_LRU test
3) Simplify some comments
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

test_lru_sanity3 is not applicable to BPF_F_NO_COMMON_LRU.
It just happens to work when PERCPU_FREE_TARGET == 16.

This patch:
1) Disable test_lru_sanity3 for BPF_F_NO_COMMON_LRU
2) Add test_lru_sanity6 to test list rotation for
   the BPF_F_NO_COMMON_LRU map.
Signed-off-by: NMartin KaFai Lau <kafai@fb.com>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Recently, suspend/resume and WOL support are added into mvneta driver.
If we enable WOL, then we get some error as below on Marvell BG4CT
platforms during suspend:

[  184.149723] dpm_run_callback(): mdio_bus_suspend+0x0/0x50 returns -16
[  184.149727] PM: Device f7b62004.mdio-mi:00 failed to suspend: error -16

-16 means -EBUSY, phy_suspend() will return -EBUSY if it finds the
device has WOL enabled.

We fix this issue by properly setting the netdev's power.can_wakeup
and power.wakeup, i.e

1. in mvneta_mdio_probe(), call device_set_wakeup_capable() to set
power.can_wakeup if the phy support WOL.

2. in mvneta_ethtool_set_wol(), call device_set_wakeup_enable() to
set power.wakeup if WOL has been successfully enabled in phy.
Signed-off-by: NJisheng Zhang <jszhang@marvell.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Recently we added support for SW fdbs to take over HW ones, but that
results in changing a user-visible fdb flag thus we need to send a
notification, also it's consistent with how HW takes over SW entries.
Signed-off-by: NNikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

struct kcm_clone only contains fd, and kcm_clone() only
writes this struct, so there is no need to copy it from user.

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The section is not specific only to "TC classifiers", but applies to the
whole TC subsystem. Also, add couple of forgotten headers.
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Acked-by: NJamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Simplify the loop in phy_supported_speeds().
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

phylib has some undesirable behaviour when forcing a link mode through
ethtool.  phylib uses this code:

	idx = phy_find_valid(phy_find_setting(phydev->speed, phydev->duplex),
			features);

to find an index in the settings table.  phy_find_setting() starts at
index 0, and scans upwards looking for an exact speed and duplex match.
When it doesn't find it, it returns MAX_NUM_SETTINGS - 1, which is
10baseT-Half duplex.

phy_find_valid() then scans from the point (and effectively only checks
one entry) before bailing out, returning MAX_NUM_SETTINGS - 1.

phy_sanitize_settings() then sets ->speed to SPEED_10 and ->duplex to
DUPLEX_HALF whether or not 10baseT-Half is supported or not.  This goes
against all the comments against these functions, and 10baseT-Half may
not even be supported by the hardware.

Rework these functions, introducing a new method of scanning the table.
There are two modes of lookup that phylib wants: exact, and inexact.

- in exact mode, we return either an exact match or failure
- in inexact mode, we return an exact match if it exists, a match at
  the highest speed that is not greater than the requested speed
  (ignoring duplex), or failing that, the lowest supported speed, or
  failure.

The biggest difference is that we always check whether the entry is
supported before further consideration, so all unsupported entries are
not considered as candidates.

This results in arguably saner behaviour, better matches the comments,
and is probably what users would expect.

This becomes important as ethernet speeds increase, PHYs exist which do
not support the 10Mbit speeds, and half-duplex is likely to become
obsolete - it's already not even an option on 10Gbit and faster links.
Signed-off-by: NRussell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since 3.12 it has been possible to configure the default queuing
discipline via sysctl. This patch adds ability to configure the
default queue discipline in kernel configuration. This is useful for
environments where configuring the value from userspace is difficult
to manage.

The default is still the same as before (pfifo_fast) and it is
possible to change after kernel init with sysctl. This is similar
to how TCP congestion control works.
Signed-off-by: NStephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Manish Chopra says:

====================
qed/qede: aRFS support

This series adds support for Accelerated Flow Steering
in qede driver for TCP/UDP over IPv4/IPv6 protocols.

Please consider applying this series to "net-next"
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds support for aRFS for TCP and UDP
protocols with IPv4/IPv6.
Signed-off-by: NManish Chopra <manish.chopra@cavium.com>
Signed-off-by: NYuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch adds necessary APIs to interface with
qede aRFS support in successive patch.

It also reserves separate PTT entry for aRFS,
[as being in fastpath flow] for hardware access instead of
trying to acquire it at run time from the ptt pool.
Signed-off-by: NManish Chopra <manish.chopra@cavium.com>
Signed-off-by: NYuval Mintz <yuval.mintz@cavium.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This chip is used by a lot of embedded devices and also by the Raspberry
Pi 1, 2 & 3 which were created to promote the study of computer
sciences. Students wanting to learn kernel / network device driver
programming through those devices can only rely on the Linux kernel
driver source to make their own.

This commit adds a lot of comments to the registers definition to expand
the register names.

Cc: Steve Glendinning <steve.glendinning@shawell.net>
Cc: Microchip Linux Driver Support <UNGLinuxDriver@microchip.com>
CC: David Miller <davem@davemloft.net>
Signed-off-by: NMartin Wetterwald <martin@wetterwald.eu>
Reviewed-by: NAndrew Lunn <andrew@lunn.ch>
Acked-by: NSteve Glendinning <steve.glendinning@shawell.net>
Acked-by: NWoojung Huh <Woojung.Huh@microchip.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The MTU overhead calculation in L2TP device set-up
merged via commit b784e7eb
needs to be adjusted to lock the tunnel socket while
referencing the sub-data structures to derive the
socket's IP overhead.
Reported-by: NGuillaume Nault <g.nault@alphalink.fr>
Tested-by: NGuillaume Nault <g.nault@alphalink.fr>
Signed-off-by: NR. Parameswaran <rparames@brocade.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ndisc_notify is the ipv6 equivalent to arp_notify. When arp_notify is
set to 1, gratuitous arp requests are sent when the device is brought up.
The same is expected when ndisc_notify is set to 1 (per ndisc_notify in
Documentation/networking/ip-sysctl.txt). The NA is not sent on NETDEV_UP
event; add it.

Fixes: 5cb04436 ("ipv6: add knob to send unsolicited ND on link-layer address change")
Signed-off-by: NDavid Ahern <dsa@cumulusnetworks.com>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
Mellanox, mlx5 RDMA net device support

This series provides the lower level mlx5 support of RDMA netdevice
creation API [1] suggested and introduced by Intel's HFI OPA VNIC
netdevice driver [2], to enable IPoIB mlx5 RDMA netdevice creation.

mlx5 IPoIB RDMA netdev will serve as an acceleration netdevice for the current
IPoIB ULP generic netdevice, providing:
	- mlx5 RSS support.
	- mlx5 HW RX,TX offloads (checksum, TSO, LRO, etc ..).
	- Full mlx5 HW features transparent to the ULP itself.

The idea here is to reuse and benefit from the already implemented mlx5e netdevice
management and channels API for both etherent and RDMA netdevices, since both IPoIB
and Ethernet netdevices share same common mlx5 HW resources (with some small
exceptions) and share most of the control/data path logic, it is more natural to
have them share the same code.

The differences between IPoIB and Ethernet netdevices can be summarized to:

Steering:
In mlx5, IPoIB traffic is sent and received from an underlay special QP, and in Ethernet
the traffic is handled by vports and vport steering is managed by e-switch or FW.

For IPoIB traffic to get steered correctly the only thing we need to do is to create RSS
HW contexts for RX and TX HW contexts for TX (similar to mlx5e) with the underlay QP attached to
them (underlay QP will be 0 in case of Ethernet).

RX,TX:
Since IPoIB traffic is different, slightly modified RX and TX handlers are required,
still we do some code reuse in data path via common helper functions.

All of the other generic netdevice and mlx5 aspects will be shared between mlx5 Ethernet
and IPoIB netdevices, e.g.
	- Channels creation and handling (RQs,SQs,CQs, NAPI, interrupt moderation, etc..)
	- Offloads, checksum, GRO, LRO, TSO, and more.
        - netdevice logic and non Ethernet specific ndos (open/close, etc..)

In order to achieve what we want:

In patchet 1 to 3, Erez added the supported for underlay QP in mlx5_ifc and refactored
the mlx5 steering code to accept the underlay QP as a parameter for creating steering
objects and enabled flow steering for IB link.

Then we are going to use the mlx5e netdevice profile, which is already used to separate between
NIC and VF representors netdevices, to create new type of IPoIB netdevice profile.

For that, one small refactoring is required to make mlx5e netdevice profile management
more genetic and agnostic to link type which is done in patch #4.

In patch #5, we introduce ipoib.c to host all of mlx5 IPoIB (mlx5i) specific logic and a
skeleton for the IPoIB mlx5 netdevice profile, and we will start filling it in next patches,
using mlx5e already existing APIs.

Patch #6 and #7, Implement init/cleanup RX mlx5i netdev profile handlers to create mlx5 RSS
resources, same as mlx5e but without vlan and L2 steering tables.

Patch #8, Implement init/cleanup TX mlx5i netdev profile handlers, to create TX resources
same as mlx5e but with one TC (tc = 0) support.

Patch #9, Implement mlx5i open/close ndos, where we reuese the mlx5e channels API, to start/stop TX/RX channels.

Patch #10, Create the underlay QP and attach it to mlx5i RSS and TX HW contexts.

Patch #11 and #12, Break down the mlx5e xmit flow into smaller helper function and implement the
mlx5i IPoIB xmit routine.

Patch #13 and #14, Have an RX handler per netdevice profile. We already do this before this series
in a non clean way to separate between NIC netdev and VF representor RX handlers, in patch 13 we make
the RX handler generic and bound to a profile and in patch 14 we implement the IPoIB RX handlers.

Patch #15, Small cleanup to avoid e-switch with IPoIB netdev.

In order to enable mlx5 IPoIB, a merge between the IPoIB RDMA netdev offolad support [3]
- which was alread submitted to the rdma mailing list - and this series is required
plus an extra small patch [4] which will connect between both sides and actually enables the offload.

Once both patch-sets are merged into linux we will have to submit the extra small patch [4], to enable
the feature.

Thanks,
Saeed.

[1] https://patchwork.kernel.org/patch/9676637/

[2] https://lwn.net/Articles/715453/
    https://patchwork.kernel.org/patch/9587815/

[3] https://patchwork.kernel.org/patch/9672069/
[4] https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux.git/commit/?id=0141db6a686e32294dee015b7d07706162ba48d8
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add check for bit IB_QP_CREATE_NETIF_QP while creating QP.
Signed-off-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently the driver support only ethernet eswitch, and we want to
protect downstream IPoIB netdev from trying to access it in IB link.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement IPoIB RX SKB handler.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In order to have different RX handler per profile, fix and refactor the
current code to take the rx handler directly from the netdevice profile
rather than computing it on runtime as it was done with the switchdev
mode representor rx handler.

This will also remove the current wrong assumption in mlx5e_alloc_rq
code that mlx5e_priv->ppriv is of the type vport_rep.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement mlx5e's IPoIB SKB transmit using the helper functions provided
by mlx5e ethernet tx flow, the only difference in the code between
mlx5e_xmit and mlx5i_xmit is that IPoIB has some extra fields to fill
(UD datagram segment) in the TX descriptor (WQE) and it doesn't need to
have any vlan handling.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Break current mlx5e xmit flow into smaller blocks (helper functions)
in order to reuse them for IPoIB SKB transmission.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Create IPoIB underlay QP needed by the IPoIB netdevice profile for RSS
and TX HW context to perform on IPoIB traffic.

Reset the underlay QP on dev_uninit ndo to stop IPoIB traffic going
through this QP when the ULP IPoIB decides to cleanup.

Implement attach/detach mcast RDMA netdev callbacks for later RDMA
netdev use.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement open/close of IPoIB netdevice ndos using mlx5e's
channels API to manage data path resources (RQs/SQs/CQs).

Set IPoIB netdev address on dev_init ndo.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Modify mlx5e tis creation function to accept underlay qp number, which
will be needed by IPoIB.

Implement mlx5i (IPoIB) tx init/cleanup netdevice profile flows to
create one TIS with the IPoIB underlay qp, for IPoIB TX SQs.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Like the mlx5e ethernet mode, on IPoIB mode we need to create RX steering
tables, but IPoIB do not require MAC and VLAN steering tables so the
only tables we create in here are:
1. TTC Table (Traffic Type Classifier table for RSS steering)
2. ARFS Table (for accelerated RFS support)

Creation of those tables is identical to mlx5e ethernet mode, hence the
use of mlx5e_create_ttc_table and mlx5e_arfs_create_tables.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Implement IPoIB RX RSS (RQTs and TIRs) HW objects creation,
All we do here is simply reuse the mlx5e implementation to create
direct and indirect (RSS) steering HW objects.

For that we just expose
mlx5e_{create,destroy}_{direct,indirect}_{rqt,tir} functions into en.h
and call them from ipoib.c in init/cleanup_rx IPoIB netdevice profile
callbacks.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Create mlx5e IPoIB netdevice profile skeleton in the new ipoib.c
file with empty implementation.

Downstream patches will provide the full mlx5 rdma netdevice acceleration
support for IPoIB into this new file, by using the mlx5e netdevice
profile and new mlx5_channels APIs and infrastructures.
Same as already done in mlx5e NIC netdevice and switchdev mode VF
representors.
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In preparation for mlx5e RDMA net_device support, here we generalize
mlx5e_attach/detach in a way that those functions will be agnostic
to link type.  For that we move ethernet specific NIC net device logic out
of those functions into {nic,rep}_{enable/disable} mlx5e NIC and
representor profiles callbacks.

Also some of the logic was moved only to NIC profile since it is not right
to have this logic for representor net device (e.g. set port MTU).
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Reviewed-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Get the relevant capabilities if supports ipoib_enhanced_offloads and
init the flow steering table accordingly.
Signed-off-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

IB flow tables need the underlay qp to perform flow steering.
Here we change the API of the flow tables creation to accept the
underlay QP number as a parameter in order to support IB (IPoIB) flow
steering.
Signed-off-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

New capability bit: ipoib_enhanced_offloads, indicates new ability for UD
QP to do RSS and enhanced IPoIB offloads and acceleration.

Add underlay_qpn to the TIS and flow_table objects In order to support
SET_ROOT command, to connect between IPoIB QPs and flow steering tables.
Signed-off-by: NErez Shitrit <erezsh@mellanox.com>
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Azure hosts are not supporting non-TCP port numbers in vRSS hashing for
now. For example, UDP packet loss rate will be high if port numbers are
also included in vRSS hash.

So, we created this patch to use only IP numbers for hashing in non-TCP
traffic.
Signed-off-by: NHaiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: NStephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If the outgoing skb has a RX queue mapping available, we use the queue
number directly, other than put it through Send Indirection Table.
Signed-off-by: NHaiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: NStephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch moves as is the legacy DSA code from dsa.c to legacy.c,
except the few shared symbols which remain in dsa.c.
Signed-off-by: NVivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>