With tracepoints support present in the mailbox
code this patch adds tracepoints in PF and VF drivers
at places where mailbox messages are allocated,
sent and at message interrupts.
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Added tracepoints in mailbox code so that
the mailbox operations like message allocation,
sending message and message interrupts are traced.
Also the mailbox errors occurred like timeout
or wrong responses are traced.
These will help in debugging mailbox issues.

Here's an example output showing one of the mailbox
messages sent by PF to AF and AF responding to it:

~# mount -t tracefs none /sys/kernel/tracing/
~# echo 1 > /sys/kernel/tracing/events/rvu/enable
~# ifconfig eth0 up
~# cat /sys/kernel/tracing/trace

~# cat /sys/kernel/tracing/trace
 tracer: nop

		      _-----=> irqs-off
		     / _----=> need-resched
		    | / _---=> hardirq/softirq
		    || / _--=> preempt-depth
		    ||| /     delay
   TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
      | |       |   ||||       |         |
ifconfig-2382  [002] ....   756.161892: otx2_msg_alloc: [0002:02:00.0] msg:(0x400) size:40

ifconfig-2382  [002] ...1   756.161895: otx2_msg_send: [0002:02:00.0] sent 1 msg(s) of size:48

 <idle>-0     [000] d.h1   756.161902: otx2_msg_interrupt: [0002:01:00.0] mbox interrupt PF(s) to AF (0x2)

kworker/u49:0-1165  [000] ....   756.162049: otx2_msg_process: [0002:01:00.0] msg:(0x400) error:0

kworker/u49:0-1165  [000] ...1   756.162051: otx2_msg_send: [0002:01:00.0] sent 1 msg(s) of size:32

kworker/u49:0-1165  [000] d.h.   756.162056: otx2_msg_interrupt: [0002:02:00.0] mbox interrupt AF to PF (0x1)
Signed-off-by: Subbaraya Sundeep <sbhatta@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The comment "holders of db->lock must always block IRQs" and related
code to do irqsave and irqrestore don't make sense since we are in a
IRQ-disabled hardIRQ context.

Cc: Maxime Ripard <mripard@kernel.org>
Cc: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

A number of static variables were not modified. Make them const to allow
the compiler to put them in read-only memory. In order to do so,
constify a couple of input pointers as well as some local pointers.
This moves about 35Kb to read-only memory as seen by the output of the
size command.

Before:
   text    data     bss     dec     hex filename
 404938  111534     640  517112   7e3f8 drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge.ko

After:
   text    data     bss     dec     hex filename
 439499   76974     640  517113   7e3f9 drivers/net/ethernet/hisilicon/hns3/hns3pf/hclge.ko
Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Nikolay Aleksandrov says:

====================
net: bridge: mcast: IGMPv3/MLDv2 fast-path (part 2)

This is the second part of the IGMPv3/MLDv2 support which adds support
for the fast-path. In order to be able to handle source entries we add
mdb support for S,G entries (i.e. we add source address support to
br_ip), that requires to extend the current mdb netlink API, fortunately
we just add another attribute which will contain nested future mdb
attributes, then we use it to add support for S,G user- add, del and
dump. The lookup sequence is simple: when IGMPv3/MLDv2 are enabled do
the S,G lookup first and if it fails fallback to *,G. The more complex
part is when we begin handling source lists and auto-installing S,G entries
and *,G filter mode transitions. We have the following cases:
 1) *,G INCLUDE -> EXCLUDE transition: we need to install the port in
    all of *,G's installed S,G entries for proper replication (except
    the ones explicitly blocked), this is also necessary when adding a
    new *,G EXCLUDE port group

 2) *,G EXCLUDE -> INCLUDE transition: we need to remove the port from
    all of *,G's installed S,G entries, this is also necessary when
    removing a *,G port group

 3) New S,G port entry: we need to install all current *,G EXCLUDE ports

 4) Remove S,G port entry: if all other port groups were auto-installed we
    can safely remove them and delete the whole S,G entry

Currently we compute these operations from the available ports, their
source lists and their filter mode. In the future we can extend the port
group structure and reduce the running time of these ops. Also one
current limitation is that host-joined S,G entries are not supported.
I.e. one cannot add "dev bridge port bridge" mdb S,G entries. The host
join is currently considered an EXCLUDE {} join, so it's reflected in
all of *,G's installed S,G entries. If an S,G,port entry is added as
temporary then the kernel can take it over if a source shows up from a
report, permanent entries are skipped. In order to properly handle
blocked sources we add a new port group blocked flag to avoid forwarding
to that port group in the S,G. Finally when forwarding we use the port
group filter mode (if it's INCLUDE and the port group is from a *,G then
don't replicate to it, respectively if it's EXCLUDE then forward) and the
blocked flag (obviously if it's set - skip that port unless it's a
router port) to decide if the port should be skipped. Another limitation
is that we can't do some of the above transitions without small traffic
drop while installing/removing entries. That will be taken care of when
we add atomic swap of port replication lists later.

Patch break down:
 patches 1-3: prepare the mdb code for better extack support which is
              used in future patches to return a more meaningful error
 patches 4-6: add the source address field to struct br_ip, and do minor
              cleanups around it
 patches 7-8: extend the mdb netlink API so we can send new mdb
              attributes and uses the new API for S,G entry add/del/dump
              support
 patch     9: takes care of S,G entries when doing a lookup (first S,G
              then *,G lookup)
 patch    10: adds a new port group field and attribute for origin protocol
              we use the already available RTPROT_ definitions,
              currently user-space entries are added as RTPROT_STATIC and
              kernel entries are added as RTPROT_KERNEL, we may allow
              user-space to set custom values later (e.g. for FRR, clag)
 patch    11: adds an internal S,G,port rhashtable to speed up filter
              mode transitions
 patch    12: initial automatic install of S,G entries based on port
              groups' source lists
 patch    13: handles port group modes on transitions or when new
              port group entries are added
 patch    14: self-explanatory - adds support for blocked port group
              entries needed to stop forwarding to particular S,G,port
              entries
 patch    15: handles host-join/leave state changes, treats host-joins
              as EXCLUDE {} groups (reflected in all *,G's S,G entries)
 patch    16: finally adds the fast-path filter mode and block flag
              support

Here're the sets that will come next (in order):
 - iproute2 support for IGMPv3/MLDv2
 - selftests for all mode transitions and group flags
 - explicit host tracking for proper fast-leave support
 - atomic port replication lists (these are also needed for broadcast
   forwarding optimizations)
 - mode transition optimization and removal of open-coded sorted lists

Not implemented yet:
 - Host IGMPv3/MLDv2 filter support (currently we handle only join/leave
   as before)
 - Proper other querier source timer and value updates
 - IGMPv3/v2 MLDv2/v1 compat (I have a few rough patches for this one)

v2: fix build with CONFIG_BATMAN_ADV_MCAST in patch 6
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

We need to avoid forwarding to ports in MCAST_INCLUDE filter mode when the
mdst entry is a *,G or when the port has the blocked flag.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since host joins are considered as EXCLUDE {} joins we need to reflect
that in all of *,G ports' S,G entries. Since the S,Gs can have
host_joined == true only set automatically we can safely set it to false
when removing all automatically added entries upon S,G delete.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When excluding S,G entries we need a way to block a particular S,G,port.
The new port group flag is managed based on the source's timer as per
RFCs 3376 and 3810. When a source expires and its port group is in
EXCLUDE mode, it will be blocked.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We need to handle group filter mode transitions and initial state.
To change a port group's INCLUDE -> EXCLUDE mode (or when we have added
a new port group in EXCLUDE mode) we need to add that port to all of
*,G ports' S,G entries for proper replication. When the EXCLUDE state is
changed from IGMPv3 report, br_multicast_fwd_filter_exclude() must be
called after the source list processing because the assumption is that
all of the group's S,G entries will be created before transitioning to
EXCLUDE mode, i.e. most importantly its blocked entries will already be
added so it will not get automatically added to them.
The transition EXCLUDE -> INCLUDE happens only when a port group timer
expires, it requires us to remove that port from all of *,G ports' S,G
entries where it was automatically added previously.
Finally when we are adding a new S,G entry we must add all of *,G's
EXCLUDE ports to it.
In order to distinguish automatically added *,G EXCLUDE ports we have a
new port group flag - MDB_PG_FLAGS_STAR_EXCL.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for automatic install of S,G mdb entries based
on the port group's source list and the source entry's timer.
Once installed the S,G will be used when forwarding packets if the
approprate multicast/mld versions are set. A new source flag called
BR_SGRP_F_INSTALLED denotes if the source has a forwarding mdb entry
installed.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

To speedup S,G forward handling we need to be able to quickly find out
if a port is a member of an S,G group. To do that add a global S,G port
rhashtable with key: source addr, group addr, protocol, vid (all br_ip
fields) and port pointer.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We need to be able to differentiate between pg entries created by
user-space and the kernel when we start generating S,G entries for
IGMPv3/MLDv2's fast path. User-space entries are created by default as
RTPROT_STATIC and the kernel entries are RTPROT_KERNEL. Later we can
allow user-space to provide the entry rt_protocol so we can
differentiate between who added the entries specifically (e.g. clag,
admin, frr etc).
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If (S,G) entries are enabled (igmpv3/mldv2) then look them up first. If
there isn't a present (S,G) entry then try to find (*,G).
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add new mdb attributes (MDBE_ATTR_SOURCE for setting,
MDBA_MDB_EATTR_SOURCE for dumping) to allow add/del and dump of mdb
entries with a source address (S,G). New S,G entries are created with
filter mode of MCAST_INCLUDE. The same attributes are used for IPv4 and
IPv6, they're validated and parsed based on their protocol.
S,G host joined entries which are added by user are not allowed yet.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since the MDB add/del code expects an exact struct br_mdb_entry we can't
really add any extensions, thus add a new nested attribute at the level of
MDBA_SET_ENTRY called MDBA_SET_ENTRY_ATTRS which will be used to pass
all new options via netlink attributes. This patch doesn't change
anything functionally since the new attribute is not used yet, only
parsed.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since now we have src in br_ip, u no longer makes sense so rename
it to dst. No functional changes.

v2: fix build with CONFIG_BATMAN_ADV_MCAST

CC: Marek Lindner <mareklindner@neomailbox.ch>
CC: Simon Wunderlich <sw@simonwunderlich.de>
CC: Antonio Quartulli <a@unstable.cc>
CC: Sven Eckelmann <sven@narfation.org>
CC: b.a.t.m.a.n@lists.open-mesh.org
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Now that we have src and dst in br_ip it is logical to use the src field
for the cases where we need to work with a source address such as
querier source address and group source address.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add a new src field to struct br_ip which will be used to lookup S, G
entries. When SSM option is added we will enable full br_ip lookups.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pass and use extack all the way down to br_mdb_add_group().
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

To avoid doing duplicate device checks and searches (the same were done
in br_mdb_add and __br_mdb_add) pass the already found port to __br_mdb_add
and pull the bridge's netif_running and enabled multicast checks to
br_mdb_add. This would also simplify the future extack errors.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can drop the pr_info() calls and just use extack to return a
meaningful error to user-space when br_mdb_parse() fails.
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

drivers/net/ethernet/realtek/8139cp.c: In function cp_tx_timeout:
drivers/net/ethernet/realtek/8139cp.c:1242:6: warning: variable ‘rc’ set but not used [-Wunused-but-set-variable]

`rc` is never used, so remove it.
Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix the warnings about function header comments when building hinic
driver with "W=1" option.
Signed-off-by: Luo bin <luobin9@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Alexei Starovoitov says:

====================
pull-request: bpf-next 2020-09-23

The following pull-request contains BPF updates for your *net-next* tree.

We've added 95 non-merge commits during the last 22 day(s) which contain
a total of 124 files changed, 4211 insertions(+), 2040 deletions(-).

The main changes are:

1) Full multi function support in libbpf, from Andrii.

2) Refactoring of function argument checks, from Lorenz.

3) Make bpf_tail_call compatible with functions (subprograms), from Maciej.

4) Program metadata support, from YiFei.

5) bpf iterator optimizations, from Yonghong.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Seth reported problem with cross builds, that fail
on resolve_btfids build, because we are trying to
build it on cross build arch.

Fixing this by always forcing the host arch.
Reported-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200923185735.3048198-2-jolsa@kernel.org

Currently all the resolve_btfids 'users' are under CONFIG_BPF
code, so if we have CONFIG_BPF disabled, resolve_btfids will
fail, because there's no data to resolve.

Disabling resolve_btfids if there's CONFIG_BPF disabled,
so we won't fail such builds.
Suggested-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200923185735.3048198-1-jolsa@kernel.org

Merge tag 'linux-can-next-for-5.10-20200923' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next

Marc Kleine-Budde says:

====================
pull-request: can-next 2020-09-23

this is a pull request of 20 patches for net-next.

The complete series target the flexcan driver and is created by Joakim
Zhang and me.

The first six patches are cleanup (sort include files alphabetically,
remove stray empty line, get rid of long lines) and adding more
registers and documentation (registers and wakeup interrupt).

Then in two patches the transceiver regulator is made optional, and a
check for maximum transceiver bitrate is added.

Then the ECC support for HW thats supports this is added.

The next three patches improve suspend and low power mode handling.

Followed by six patches that add CAN-FD support and CAN-FD related
features.

The last two patches add support for the flexcan IP core on the imx8qm
and lx2160ar1.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Julian Wiedmann says:

====================
s390/qeth: updates 2020-09-23

please apply the following patch series for qeth to netdev's net-next tree.

This brings all sorts of cleanups. Highlights are more code sharing in
the init/teardown paths, and more fine-grained rollback on errors during
initialization (instead of a full-blown teardown).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Shuffle some code around (primarily all the discipline-related stuff) to
get rid of all the unnecessary forward declarations.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Clarify which discipline-specific steps are needed to roll back after
error in qeth_l?_set_online(), and which are common to roll back
from qeth_hardsetup_card().

Some steps (cancelling the RX modeset, draining the TX queues) are only
necessary if the netdev was potentially UP before, so move them to the
common qeth_set_offline().
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move duplicated code from the disciplines into the core path.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Originators of cmd IO typically hold the rtnl or conf_mutex to protect
against a concurrent teardown.
Since qeth_set_offline() already holds the conf_mutex, the main reason
why we still care about cancelling pending cmds is so that they release
the rtnl when we need it ourselves.

So move this step a little earlier into the teardown sequence.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The programming of ucast IPs via qeth_l3_modify_ip() is driven
independently from any of our typical locking mechanisms (eg. detaching
the netdevice, or holding the conf_mutex).
So when we inspect the card state to check whether the required cmd IO
should be deferred, there is no protection against concurrent state
changes.

But by slightly re-ordering the teardown sequence, we can rely on the
ip_lock to sufficiently serialize things:

1. when running concurrently to qeth_l3_set_online(), any instance of
   qeth_l3_modify_ip() that aquires the ip_lock _after_
   qeth_l3_recover_ip() will observe the state as CARD_STATE_SOFTSETUP
   and not defer the IO.
2. when running concurrently to qeth_l3_set_offline(), any instance of
   qeth_l3_modify_ip() that aquires the ip_lock _after_
   qeth_l3_clear_ip_htable() will observe the state as CARD_STATE_DOWN
   and defer the IO.

These guarantees in mind, we can now drop the conf_mutex from the
qeth_l3_modify_rxip_vipa() wrapper.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Convert the remaining occurences in sysfs code to kstrtouint().

While at it move some input parsing out of locked sections, replace an
open-coded clamp() and remove some unnecessary run-time checks for
ipatoe->mask_bits that are already enforced when creating the object.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Indicate the max number of to-be-parsed characters, and avoid copying
the address sub-string.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

card->ipato is currently protected by the conf_mutex. But most users
also hold the ip_lock - in particular qeth_l3_add_ip().

So slightly expand the sections under ip_lock in a few places (to
effectively cover a few error & no-op cases), and then drop the
conf_mutex where it's no longer needed.
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mcast IP objects are allocated within qeth_l3_add_mcast_rtnl(),
with .ref_counter already set to 1 via qeth_l3_init_ipaddr().
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Arrays with designated initializers have an implicit length of the highest
initialized value plus one. I used this to ensure that newly added entries
in enum bpf_reg_type get a NULL entry in compatible_reg_types.

This is difficult to understand since it requires knowledge of the
peculiarities of designated initializers. Use __BPF_ARG_TYPE_MAX to size
the array instead.
Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200923160156.80814-1-lmb@cloudflare.com

drivers/net/ethernet/microchip/lan743x_main.c: In function lan743x_pm_suspend:

`ret` is set but not used. In fact, `pci_prepare_to_sleep` function value should
be the right value of `lan743x_pm_suspend` function, therefore, fix it.
Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
mlx5-updates-2020-09-21

Multi packet TX descriptor support for SKBs.

This series introduces some refactoring of the regular TX data path in
mlx5 and adds the Enhanced TX MPWQE feature support. MPWQE stands for
multi-packet work queue element, and it can serve multiple packets,
reducing the PCI bandwidth spent on control traffic. It should improve
performance in scenarios where PCI is the bottleneck, and xmit_more is
signaled by the kernel. The refactoring done in this series also
improves the packet rate on its own.

MPWQE is already implemented in the XDP tx path, this series adds the
support of MPWQE for regular kernel SKB tx path.

MPWQE is supported from ConnectX-5 and onward, for legacy devices we need
to keep backward compatibility for regular (Single packet) WQE descriptor.

MPWQE is not compatible with certain offloads and features, such as TLS
offload, TSO, nonlinear SKBs. If such incompatible features are in use,
the driver gracefully falls back to non-MPWQE per SKB.

Prior to the final patch "net/mlx5e: Enhanced TX MPWQE for SKBs" that adds
the actual support, Maxim did some refactoring to the tx data path to
split it into stages and smaller helper functions that can be utilized and
reused for both legacy and new MPWQE feature.

Performance testing:

UDP performance is improved in a single stream pktgen test:
  Packet rate: 16.86 Mpps (±0.15 Mpps) -> 20.94 Mpps (±0.33 Mpps)
  Instructions per packet: 434 -> 329
  Cycles per packet: 158 -> 123
  Instructions per cycle: 2.75 -> 2.67

TCP and XDP_TX single stream tests show no performance difference.

MPWQE can reduce PCI bandwidth:
  PCI Gen2, pktgen at fixed rate of 36864000 pps on 24 CPU cores:
    Inbound PCI utilization with MPWQE off: 80.3%
    Inbound PCI utilization with MPWQE on: 59.0%
  PCI Gen3, pktgen at fixed rate of 56064000 pps on 24 CPU cores:
    Inbound PCI utilization with MPWQE off: 65.4%
    Inbound PCI utilization with MPWQE on: 49.3%

MPWQE can also reduce CPU load, increasing the packet rate in case of
CPU bottleneck:
  PCI Gen2, pktgen at full rate on 24 CPU cores:
    Packet rate with MPWQE off: 37.5 Mpps
    Packet rate with MPWQE on: 49.0 Mpps
  PCI Gen3, pktgen at full rate on 24 CPU cores:
    Packet rate with MPWQE off: 57.0 Mpps
    Packet rate with MPWQE on: 66.8 Mpps

Burst size in all pktgen tests is 32.

CPU: Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz (x86_64)
NIC: Mellanox ConnectX-6 Dx
GCC 10.2.0
====================
Signed-off-by: David S. Miller <davem@davemloft.net>