Recently I got a Caldigit TS3 Thunderbolt 3 dock, and noticed that upon
hotplugging my kernel would immediately crash due to igb:

[  680.825801] kernel BUG at drivers/pci/msi.c:352!
[  680.828388] invalid opcode: 0000 [#1] SMP
[  680.829194] Modules linked in: igb(O) thunderbolt i2c_algo_bit joydev vfat fat btusb btrtl btbcm btintel bluetooth ecdh_generic hp_wmi sparse_keymap rfkill wmi_bmof iTCO_wdt intel_rapl x86_pkg_temp_thermal coretemp crc32_pclmul snd_pcm rtsx_pci_ms mei_me snd_timer memstick snd pcspkr mei soundcore i2c_i801 tpm_tis psmouse shpchp wmi tpm_tis_core tpm video hp_wireless acpi_pad rtsx_pci_sdmmc mmc_core crc32c_intel serio_raw rtsx_pci mfd_core xhci_pci xhci_hcd i2c_hid i2c_core [last unloaded: igb]
[  680.831085] CPU: 1 PID: 78 Comm: kworker/u16:1 Tainted: G           O     4.15.0-rc3Lyude-Test+ #6
[  680.831596] Hardware name: HP HP ZBook Studio G4/826B, BIOS P71 Ver. 01.03 06/09/2017
[  680.832168] Workqueue: kacpi_hotplug acpi_hotplug_work_fn
[  680.832687] RIP: 0010:free_msi_irqs+0x180/0x1b0
[  680.833271] RSP: 0018:ffffc9000030fbf0 EFLAGS: 00010286
[  680.833761] RAX: ffff8803405f9c00 RBX: ffff88033e3d2e40 RCX: 000000000000002c
[  680.834278] RDX: 0000000000000000 RSI: 00000000000000ac RDI: ffff880340be2178
[  680.834832] RBP: 0000000000000000 R08: ffff880340be1ff0 R09: ffff8803405f9c00
[  680.835342] R10: 0000000000000000 R11: 0000000000000040 R12: ffff88033d63a298
[  680.835822] R13: ffff88033d63a000 R14: 0000000000000060 R15: ffff880341959000
[  680.836332] FS:  0000000000000000(0000) GS:ffff88034f440000(0000) knlGS:0000000000000000
[  680.836817] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  680.837360] CR2: 000055e64044afdf CR3: 0000000001c09002 CR4: 00000000003606e0
[  680.837954] Call Trace:
[  680.838853]  pci_disable_msix+0xce/0xf0
[  680.839616]  igb_reset_interrupt_capability+0x5d/0x60 [igb]
[  680.840278]  igb_remove+0x9d/0x110 [igb]
[  680.840764]  pci_device_remove+0x36/0xb0
[  680.841279]  device_release_driver_internal+0x157/0x220
[  680.841739]  pci_stop_bus_device+0x7d/0xa0
[  680.842255]  pci_stop_bus_device+0x2b/0xa0
[  680.842722]  pci_stop_bus_device+0x3d/0xa0
[  680.843189]  pci_stop_and_remove_bus_device+0xe/0x20
[  680.843627]  trim_stale_devices+0xf3/0x140
[  680.844086]  trim_stale_devices+0x94/0x140
[  680.844532]  trim_stale_devices+0xa6/0x140
[  680.845031]  ? get_slot_status+0x90/0xc0
[  680.845536]  acpiphp_check_bridge.part.5+0xfe/0x140
[  680.846021]  acpiphp_hotplug_notify+0x175/0x200
[  680.846581]  ? free_bridge+0x100/0x100
[  680.847113]  acpi_device_hotplug+0x8a/0x490
[  680.847535]  acpi_hotplug_work_fn+0x1a/0x30
[  680.848076]  process_one_work+0x182/0x3a0
[  680.848543]  worker_thread+0x2e/0x380
[  680.848963]  ? process_one_work+0x3a0/0x3a0
[  680.849373]  kthread+0x111/0x130
[  680.849776]  ? kthread_create_worker_on_cpu+0x50/0x50
[  680.850188]  ret_from_fork+0x1f/0x30
[  680.850601] Code: 43 14 85 c0 0f 84 d5 fe ff ff 31 ed eb 0f 83 c5 01 39 6b 14 0f 86 c5 fe ff ff 8b 7b 10 01 ef e8 b7 e4 d2 ff 48 83 78 70 00 74 e3 <0f> 0b 49 8d b5 a0 00 00 00 e8 62 6f d3 ff e9 c7 fe ff ff 48 8b
[  680.851497] RIP: free_msi_irqs+0x180/0x1b0 RSP: ffffc9000030fbf0

As it turns out, normally the freeing of IRQs that would fix this is called
inside of the scope of __igb_close(). However, since the device is
already gone by the point we try to unregister the netdevice from the
driver due to a hotplug we end up seeing that the netif isn't present
and thus, forget to free any of the device IRQs.

So: make sure that if we're in the process of dismantling the netdev, we
always allow __igb_close() to be called so that IRQs may be freed
normally. Additionally, only allow igb_close() to be called from
__igb_close() if it hasn't already been called for the given adapter.
Signed-off-by: NLyude Paul <lyude@redhat.com>
Fixes: 9474933c ("igb: close/suspend race in netif_device_detach")
Cc: Todd Fujinaka <todd.fujinaka@intel.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: stable@vger.kernel.org
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

I personally spent a long time trying to decypher why my CPU would not
reach deeper C-states. Let's just tell the next user what's going on.
Signed-off-by: NMatt Turner <matt.turner@intel.com>
Acked-by: NShannon Nelson <shannon.nelson@oracle.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

By design, the idleslope increments are restricted to 16.384kbps steps.
Add a comment to igb_main.c making that explicit and add one example
that illustrates the impact of that.
Signed-off-by: NJesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

According to section 12.0.3.4.13 "Receive Descriptor Control - RXDCTL"
of the Intel® 82579 Gigabit Ethernet PHY Datasheet v2.1:

    "HTHRESH should be given a non zero value when ever PTHRESH is
     used."

In RXDCTL(0), PTHRESH lives at bits 5:0, and HTHREST lives at bits 13:8.
Set only bit 8 of HTHREST as is done in e1000_flush_rx_ring(). Found by
inspection.
Signed-off-by: NMatt Turner <matt.turner@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch adds a new function igb_get_max_rss_queues() to get maximum
RSS queues, this will reduce duplicate code and facilitate future
maintenance.
Signed-off-by: NZhang Shengju <zhangshengju@cmss.chinamobile.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Before libvirt modifies the MAC address and vlan tag for an SRIOV VF
for use by a virtual machine (either using vfio device assignment or
macvtap passthru mode), it saves the current MAC address and vlan tag
so that it can reset them to their original value when the guest is
done.  Libvirt can't leave the VF MAC set to the value used by the
now-defunct guest since it may be started again later using a
different VF, but it certainly shouldn't just pick any random value,
either. So it saves the state of everything prior to using the VF, and
resets it to that.

The igb driver initializes the MAC addresses of all VFs to
00:00:00:00:00:00, and reports that when asked (via an RTM_GETLINK
netlink message, also visible in the list of VFs in the output of "ip
link show"). But when libvirt attempts to restore the MAC address back
to 00:00:00:00:00:00 (using an RTM_SETLINK netlink message) the kernel
responds with "Invalid argument".

Forbidding a reset back to the original value leaves the VF MAC at the
value set for the now-defunct virtual machine. Especially on a system
with NetworkManager enabled, this has very bad consequences, since
NetworkManager forces all interfacess to be IFF_UP all the time - if
the same virtual machine is restarted using a different VF (or even on
a different host), there will be multiple interfaces watching for
traffic with the same MAC address.

To allow libvirt to revert to the original state, we need a way to
remove the administrative set MAC on a VF, to allow normal host
operation again, and to reset/overwrite the VF MAC via VF netdev.

This patch implements the outlined scenario by allowing to set the
VF MAC to 00:00:00:00:00:00 via RTM_SETLINK on the PF.
igb_ndo_set_vf_mac resets the IGB_VF_FLAG_PF_SET_MAC flag to 0,
so it's possible to reset the VF MAC back to the original value via
the VF netdev.

Note: Recent patches to libvirt allow for a workaround if the NIC
isn't capable of resetting the administrative MAC back to all 0, but
in theory the NIC should allow resetting the MAC in the first place.
Signed-off-by: NCorinna Vinschen <vinschen@redhat.com>
Tested-by: NAaron Brown <arron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Dmitry Safonov says:

====================
pktgen: Behavior flags fixes

v2:
  o fixed a nitpick from David Miller

There are a bunch of fixes/cleanups/Documentations.
Diffstat says for itself, regardless added docs and missed flag
parameters.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Don't use error-prone-brute-force way.
Signed-off-by: NDmitry Safonov <dima@arista.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add macro generated pkt_flag_names array, with a little help of which
the flags can be printed by using an index.
Signed-off-by: NDmitry Safonov <dima@arista.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

PKT_FALGS macro will be used to add package behavior names definitions
to simplify the code that prints/reads pkg flags.
Sorted the array in order of printing the flags in pktgen_if_show()
Note: Renamed IPSEC_ON => IPSEC for simplicity.

No visible behavior change expected.
Signed-off-by: NDmitry Safonov <dima@arista.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o FLOW_SEQ now can be disabled with pgset "flag !FLOW_SEQ"
o FLOW_SEQ and FLOW_RND are antonyms, as it's shown by pktgen_if_show()
o IPSEC now may be disabled

Note, that IPV6 is enabled with dst6/src6 parameters, not with
a flag parameter.
Signed-off-by: NDmitry Safonov <dima@arista.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

o Change process name in ps output: looks like, these days the process
  is named kpktgend_<cpu>, rather than pktgen/<cpu>.
o Use pg_ctrl for start/stop as it can work well with pgset without
  changes to $(PGDEV) variable.
o Clarify a bit needed $(PGDEV) definition for sample scripts and that
  one needs to `source functions.sh`.
o Document how-to unset a behaviour flag, note about history expansion.
o Fix pgset spi parameter value.

Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-doc@vger.kernel.org
Signed-off-by: NDmitry Safonov <dima@arista.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rahul Lakkireddy says:

====================
cxgb4: fix build error

Patch 1 fixes build error with compiling cudbg_zlib.c when
CONFIG_ZLIB_DEFLATE macro is not defined.

Patch 2 fixes following sparse warning:
"Using plain integer as NULL pointer"
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

memset variables to 0 to fix sparse warnings:

drivers/net/ethernet/chelsio/cxgb4/cxgb4_cudbg.c:409:42: sparse: Using
plain integer as NULL pointer

drivers/net/ethernet/chelsio/cxgb4/cudbg_zlib.c:43:47: sparse: Using
plain integer as NULL pointer

Fixes: ad75b7d3 ("cxgb4: implement ethtool dump data operations")
Fixes: 91c1953d ("cxgb4: use zlib deflate to compress firmware dump")
Signed-off-by: NRahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fixes:
drivers/net/ethernet/chelsio/cxgb4/cudbg_zlib.c:39:5: error:
redefinition of 'cudbg_compress_buff'
    int cudbg_compress_buff(struct cudbg_init *pdbg_init,
        ^~~~~~~~~~~~~~~~~~~
   In file included from
drivers/net/ethernet/chelsio/cxgb4/cudbg_zlib.c:23:0:
   drivers/net/ethernet/chelsio/cxgb4/cudbg_zlib.h:45:19: note: previous
definition of 'cudbg_compress_buff' was here
    static inline int cudbg_compress_buff(struct cudbg_init *pdbg_init,
                      ^~~~~~~~~~~~~~~~~~~

Fixes: 91c1953d ("cxgb4: use zlib deflate to compress firmware dump")
Signed-off-by: NRahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Ursula Braun says:

====================
net/smc: socket closing improvements

while the first 2 patches are just small cleanups, the remaing
patches affect socket closing.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If the peer sends a shutdown WRITE, this should not affect sending
in general, and waiting for send buffer space in particular.
Stop waiting of the local socket for send buffer space only, if peer
signals closing, but not if peer signals just shutdown WRITE.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a socket is closed or shutdown, smc waits for data being transmitted
in certain states. If the state changes during this wait, the close
switch depending on state should be reentered.
In addition, state change is avoided if sending of close or shutdown fails.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Work requests are needed for every ib_post_send(), among them the
ib_post_send() to signal closing. If an smc socket program is cancelled,
the smc connections should be cleaned up, and require sending of closing
signals to the peer. This may fail, if a wait for
a free work request is needed, but is cancelled immediately due to the
cancel interrupt. To guarantee notification of the peer, the wait for
a work request is changed to uninterruptible.

And the area to receive work request completion info with
ib_poll_cq() is cleared first.
And _tx_ variable names are used in the _tx_routines for the
demultiplexing common type in the header.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is no need to wait for confirmation of pending tx requests
for a closing connection, since pending tx slots are dismissed
when finishing a connection.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Cleanup to avoid duplicate code in smc_clcsock_accept().
No functional change.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Cleanup to consistently exploit the local struct sock definitions.
No functional change.
Signed-off-by: NUrsula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

implement ndo_set_vf_vlan for mgmt netdevice to configure
the PCIe VF.

Original work by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: NGanesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jakub Kicinski says:

====================
bpf and netdevsim test updates

A number of test improvements (delayed by merges).  Quentin provides
patches for checking printing to the verifier log from the drivers
and validating extack messages are propagated.  There is also a test
for replacing TC filters to avoid adding back the bug Daniel recently
fixed in net and stable.
====================
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Daniel discovered recently I broke TC filter replace (and fixed
it in commit ad9294db ("bpf: fix cls_bpf on filter replace")).
Add a test to make sure it never happens again.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make netdevsim print a message to the BPF verifier log buffer when a
program is offloaded.

Then use this message in hardware offload selftests to make sure that
using this buffer actually prints the message to the console for
eBPF hardware offload.

The message is appended after the last instruction is processed with the
verifying function from netdevsim. Output looks like the following:

    $ tc filter add dev foo ingress bpf obj sample_ret0.o \
        sec .text verbose skip_sw

    Prog section '.text' loaded (5)!
     - Type:         3
     - Instructions: 2 (0 over limit)
     - License:

    Verifier analysis:

    0: (b7) r0 = 0
    1: (95) exit
    [netdevsim] Hello from netdevsim!
    processed 2 insns, stack depth 0

"verbose" flag is required to see it in the console since netdevsim does
not throw an error after printing the message.
Signed-off-by: NQuentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We should not compile netdevsim/bpf.c if BPF syscall is not
enabled.  Otherwise bpf core would have to provide wrappers
for all functions offload drivers may call, even though
system will never see a BPF object.
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: NQuentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add checks to test that netlink extack messages are correctly displayed
in some expected error cases for eBPF offload to netdevsim with TC and
XDP.

iproute2 may be built without libmnl support, in which case the extack
messages will not be reported.  Try to detect this condition, and when
enountered print a mild warning to the user and skip the extack validation.
Signed-off-by: NQuentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the recently added extack support for TC eBPF filters in netdevsim.
Signed-off-by: NQuentin Monnet <quentin.monnet@netronome.com>
Reviewed-by: NJakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
40GbE Intel Wired LAN Driver Updates 2018-01-23

This series contains updates to i40e and i40evf only.

Pawel enables FlatNVM support on x722 devices by allowing nvmupdate tool
to configure the preservation flags in the AdminQ command.

Mitch fixes a potential divide by zero error when DCB is enabled and
the firmware fails to configure the VSI, so check for this state.
Fixed a bug where the driver could fail to adhere to ETS bandwidth
allocations if 8 traffic classes were configured on the switch.

Sudheer fixes a potential deadlock by avoiding to call
flush_schedule_work() in i40evf_remove(), since cancel_work_sync()
and cancel_delayed_work_sync() already cleans up necessary work items.
Fixed an issue with the problematic detection and recovery from
hung queues in the PF which was causing lost interrupts.  This is done
by triggering a software interrupt so that interrupts are forced on
and if we are already in napi_poll and an interrupt fires, napi_poll
will not be rescheduled and the interrupt is lost.

Avinash fixes an issue in the VF where is was possible to issue a
reset_task while the device is currently being removed.

Michal fixes an issue occurring while calling i40e_led_set() with
the blink parameter set to true, which was causing the activity LED
instead of the link LED to blink for port identification.

Shiraz changes the client interface to not call client close/open on
netdev down/up events, since this causes a lot of thrash that is
not needed.  Instead, disable the PE TCP-ENA flag during a netdev
down event and re-enable on a netdev up event, since this blocks all
TCP traffic to the RDMA protocol engine.

Alan fixes an issue which was causing a potential transmit hang by
ignoring the PF link up message if the VF state is not yet in the
RUNNING state.

Amritha fixes the channel VSI recreation during the reset flow to
reconfigure the transmit rings and the queue context associated with
the channel VSI.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Davide Caratti says:

====================
net/sched: remove spinlock from 'csum' action

Similarly to what has been done earlier with other actions [1][2], this
series tries to improve the performance of 'csum' tc action, removing a
spinlock in the data path. Patch 1 lets act_csum use per-CPU counters;
patch 2 removes spin_{,un}lock_bh() calls from the act() method.

test procedure (using pktgen from https://github.com/netoptimizer):

 # ip link add name eth1 type dummy
 # ip link set dev eth1 up
 # tc qdisc add dev eth1 root handle 1: prio
 # for a in pass drop; do
 > tc filter del dev eth1 parent 1: pref 10 matchall action csum udp
 > tc filter add dev eth1 parent 1: pref 10 matchall action csum udp $a
 > for n in 2 4; do
 > ./pktgen_bench_xmit_mode_queue_xmit.sh -v -s 64 -t $n -n 1000000 -i eth1
 > done
 > done

test results:

      |    |  before patch   |   after patch
  $a  | $n | avg. pps/thread | avg. pps/thread
 -----+----+-----------------+----------------
 pass |  2 |    1671463 ± 4% |    1920789 ± 3%
 pass |  4 |     648797 ± 1% |     738190 ± 1%
 drop |  2 |    3212692 ± 2% |    3719811 ± 2%
 drop |  4 |    1078824 ± 1% |    1328099 ± 1%

references:

[1] https://www.spinics.net/lists/netdev/msg334760.html
[2] https://www.spinics.net/lists/netdev/msg465862.html

v3 changes:
 - use rtnl_dereference() in place of rcu_dereference() in tcf_csum_dump()

v2 changes:
 - add 'drop' test, it produces more contentions
 - use RCU-protected struct to store 'action' and 'update_flags', to avoid
   reading the values from subsequent configurations
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

use RCU instead of spin_{,unlock}_bh() to protect concurrent read/write on
act_csum configuration, to reduce the effects of contention in the data
path when multiple readers are present.
Signed-off-by: NDavide Caratti <dcaratti@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

use per-CPU counters, like other TC actions do, instead of maintaining one
set of stats across all cores. This allows updating act_csum stats without
the need of protecting them using spin_{,un}lock_bh() invocations.
Signed-off-by: NDavide Caratti <dcaratti@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It takes 1sec for bond link down notification to hit user-space
when all slaves of the bond go down. 1sec is too long for
protocol daemons in user-space relying on bond notification
to recover (eg: multichassis lag implementations in user-space).
Since the link event code already marks team device port link events
 as urgent, this patch moves the code to cover all lag ports and master.
Signed-off-by: NRoopa Prabhu <roopa@cumulusnetworks.com>
Acked-by: NJiri Pirko <jiri@mellanox.com>
Reviewed-by: NNikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
10GbE Intel Wired LAN Driver Updates 2018-01-23

This series contains updates to ixgbe only.

Shannon Nelson provides an implementation of the ipsec hardware offload
feature for the ixgbe driver for these devices: x540, x550, 82599.

The ixgbe NICs support ipsec offload for 1024 Rx and 1024 Tx Security
Associations (SAs), using up to 128 inbound IP addresses, and using the
rfc4106(gcm(aes)) encryption.  This code does not yet support checksum
offload, or TSO in conjunction with the ipsec offload - those will be
added in the future.

This code shows improvements in both packet throughput and CPU utilization.
For example, here are some quicky numbers that show the magnitude of the
performance gain on a single run of "iperf -c <dest>" with the ipsec
offload on both ends of a point-to-point connection:

	9.4 Gbps - normal case
	7.6 Gbps - ipsec with offload
	343 Mbps - ipsec no offload
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Sebastian Reichel says:

====================
GEHC Bx50 Switch Support

This adds support for the internal switch found in GE Healthcare
B450v3, B650v3 and B850v3. All devices use a GPIO bitbanged MDIO
bus to communicate with the switch and a PCIe based network card
for exchanging network data. The cpu network data link requires,
that the switch's internal phy interface is enabled, so support
for that is added by the first patch in this series.

The patch series is based on v4.15-rc8.

Changes since PATCHv4:
 * Introduce dsa_port_link_(un)register_of and mark the fixed
   variant static.
 * Update patch description to describe the phy<->phy connection
   from i210 to the Marvell switch
Changes since PATCHv3:
 * Enable the phy in dsa_port_setup() instead of abusing the
   fixed link setup function
Changes since PATCHv2:
 * Add phy nodes to switch in bx50.dtsi and reference them
   from switch ports
 * Enable cpu-port's phy based on 'phy-handle' instead of 'phy-mode'
Changes since PATCHv1:
 * Use 'marvell,mv88e6085' instead of introducing compatible
   string for mv88e6240.
 * Fix indention of DT nodes
 * Only enable 'cpu' phy, if explicitly set to "internal".
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds support for the Marvell switch and names the network
ports according to the labels, that can be found next to the
connectors. The switch is connected to the host system using a
PCI based network card.

The PCI bus configuration has been written using the following
information:

root@b450v3# lspci -tv
-[0000:00]---00.0-[01]----00.0  Intel Corporation I210 Gigabit Network Connection
root@b450v3# lspci -nn
00:00.0 PCI bridge [0604]: Synopsys, Inc. Device [16c3:abcd] (rev 01)
01:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)
Signed-off-by: NSebastian Reichel <sebastian.reichel@collabora.co.uk>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds support for the Marvell switch and names the network
ports according to the labels, that can be found next to the
connectors. The switch is connected to the host system using a
PCI based network card.

The PCI bus configuration has been written using the following
information:

root@b650v3# lspci -tv
-[0000:00]---00.0-[01]----00.0  Intel Corporation I210 Gigabit Network Connection
root@b650v3# lspci -nn
00:00.0 PCI bridge [0604]: Synopsys, Inc. Device [16c3:abcd] (rev 01)
01:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)
Signed-off-by: NSebastian Reichel <sebastian.reichel@collabora.co.uk>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This adds support for the Marvell switch and names the network
ports according to the labels, that can be found next to the
connectors ("ID", "IX", "ePort 1", "ePort 2"). The switch is
connected to the host system using a PCI based network card.

The PCI bus configuration has been written using the following
information:

root@b850v3# lspci -tv
-[0000:00]---00.0-[01]----00.0-[02-05]--+-01.0-[03]----00.0  Intel Corporation I210 Gigabit Network Connection
                                        +-02.0-[04]----00.0  Intel Corporation I210 Gigabit Network Connection
                                        \-03.0-[05]--
root@b850v3# lspci -nn
00:00.0 PCI bridge [0604]: Synopsys, Inc. Device [16c3:abcd] (rev 01)
01:00.0 PCI bridge [0604]: PLX Technology, Inc. PEX 8605 PCI Express 4-port Gen2 Switch [10b5:8605] (rev ab)
02:01.0 PCI bridge [0604]: PLX Technology, Inc. PEX 8605 PCI Express 4-port Gen2 Switch [10b5:8605] (rev ab)
02:02.0 PCI bridge [0604]: PLX Technology, Inc. PEX 8605 PCI Express 4-port Gen2 Switch [10b5:8605] (rev ab)
02:03.0 PCI bridge [0604]: PLX Technology, Inc. PEX 8605 PCI Express 4-port Gen2 Switch [10b5:8605] (rev ab)
03:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)
04:00.0 Ethernet controller [0200]: Intel Corporation I210 Gigabit Network Connection [8086:1533] (rev 03)
Signed-off-by: NSebastian Reichel <sebastian.reichel@collabora.co.uk>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

B850v3, B650v3 and B450v3 all have a GPIO bit banged MDIO bus to
communicate with a Marvell switch. On all devices the switch is
connected to a PCI based network card, which needs to be referenced
by DT, so this also adds the common PCI root node.
Signed-off-by: NSebastian Reichel <sebastian.reichel@collabora.co.uk>
Reviewed-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>