This patch adds 'module' member in the 'struct vsock_transport'
in order to get/put the transport module. This prevents the
module unloading while sockets are assigned to it.

We increase the module refcnt when a socket is assigned to a
transport, and we decrease the module refcnt when the socket
is destructed.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

To allow other transports to be loaded with vmci_transport,
we register the vmci_transport as G2H or H2G only when a VMCI guest
or host is active.

To do that, this patch adds a callback registered in the vmci driver
that will be called when the host or guest becomes active.
This callback will register the vmci_transport in the VSOCK core.

Cc: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds the support of multiple transports in the
VSOCK core.

With the multi-transports support, we can use vsock with nested VMs
(using also different hypervisors) loading both guest->host and
host->guest transports at the same time.

Major changes:
- vsock core module can be loaded regardless of the transports
- vsock_core_init() and vsock_core_exit() are renamed to
  vsock_core_register() and vsock_core_unregister()
- vsock_core_register() has a feature parameter (H2G, G2H, DGRAM)
  to identify which directions the transport can handle and if it's
  support DGRAM (only vmci)
- each stream socket is assigned to a transport when the remote CID
  is set (during the connect() or when we receive a connection request
  on a listener socket).
  The remote CID is used to decide which transport to use:
  - remote CID <= VMADDR_CID_HOST will use guest->host transport;
  - remote CID == local_cid (guest->host transport) will use guest->host
    transport for loopback (host->guest transports don't support loopback);
  - remote CID > VMADDR_CID_HOST will use host->guest transport;
- listener sockets are not bound to any transports since no transport
  operations are done on it. In this way we can create a listener
  socket, also if the transports are not loaded or with VMADDR_CID_ANY
  to listen on all transports.
- DGRAM sockets are handled as before, since only the vmci_transport
  provides this feature.
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remote peer is always the host, so we set VMADDR_CID_HOST as
remote CID instead of VMADDR_CID_ANY.
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vsock_insert_unbound() was called only when 'sock' parameter of
__vsock_create() was not null. This only happened when
__vsock_create() was called by vsock_create().

In order to simplify the multi-transports support, this patch
moves vsock_insert_unbound() at the end of vsock_create().
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

All transports call __vsock_create() with the same parameters,
most of them depending on the parent socket. In order to simplify
the VSOCK core APIs exposed to the transports, this patch adds
the vsock_create_connected() callable from transports to create
a new socket when a connection request is received.
We also unexported the __vsock_create().
Suggested-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

virtio_transport and vmci_transport handle the buffer_size
sockopts in a very similar way.

In order to support multiple transports, this patch moves this
handling in the core to allow the user to change the options
also if the socket is not yet assigned to any transport.

This patch also adds the '.notify_buffer_size' callback in the
'struct virtio_transport' in order to inform the transport,
when the buffer_size is changed by the user. It is also useful
to limit the 'buffer_size' requested (e.g. virtio transports).
Acked-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since now the 'struct vsock_sock' object contains a pointer to
the transport, this patch adds a parameter to the
vsock_core_get_transport() to return the right transport
assigned to the socket.

This patch modifies also the virtio_transport_get_ops(), that
uses the vsock_core_get_transport(), adding the
'struct vsock_sock *' parameter.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We are going to add 'struct vsock_sock *' parameter to
virtio_transport_get_ops().

In some cases, like in the virtio_transport_reset_no_sock(),
we don't have any socket assigned to the packet received,
so we can't use the virtio_transport_get_ops().

In order to allow virtio_transport_reset_no_sock() to use the
'.send_pkt' callback from the 'vhost_transport' or 'virtio_transport',
we add the 'struct virtio_transport *' to it and to its caller:
virtio_transport_recv_pkt().

We moved the 'vhost_transport' and 'virtio_transport' definition,
to pass their address to the virtio_transport_recv_pkt().
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As a preparation to support multiple transports, this patch adds
the 'transport' member at the 'struct vsock_sock'.
This new field is initialized during the creation in the
__vsock_create() function.

This patch also renames the global 'transport' pointer to
'transport_single', since for now we're only supporting a single
transport registered at run-time.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This header file now only includes the "uapi/linux/vm_sockets.h".
We can include directly it when needed.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vm_sockets_get_local_cid() is only used in virtio_transport_common.c.
We can replace it calling the virtio_transport_get_ops() and
using the get_local_cid() callback registered by the transport.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The VSOCK_DEFAULT_CONNECT_TIMEOUT definition was introduced with
commit d021c344 ("VSOCK: Introduce VM Sockets"), but it is
never used in the net/vmw_vsock/vmci_transport.c.

VSOCK_DEFAULT_CONNECT_TIMEOUT is used and defined in
net/vmw_vsock/af_vsock.c

Cc: Jorgen Hansen <jhansen@vmware.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Sunil Goutham says:

====================
octeontx2-af: Debugfs support and updates to parser profile

This patchset adds debugfs support to dump various HW state machine info
which helps in debugging issues. Info includes
- Current queue context, stats, resource utilization etc
- MCAM entry utilization, miss and pkt drop counter
- CGX ingress and egress stats
- Current RVU block allocation status
- etc.

Rest patches has changes wrt
- Updated packet parsing profile for parsing more protocols.
- RSS algorithms to include inner protocols while generating hash
- Handle current version of silicon's limitations wrt shaping, coloring
  and fixed mapping of transmit limiter queue's configuration.
- Enable broadcast packet replication to PF and it's VFs.
- Support for configurable NDC cache waymask
- etc

Changes from v1:
   Removed inline keyword for newly introduced APIs in few patches.
   - Suggested by David Miller.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Traffic for a CGX mapped NIXLF can be stopped by disabling entries
in NPC MCAM or by configuring CGX and mailbox messages exist for the
two options. If traffic is stopped at CGX then VFs of that PF are
also effected hence CGX traffic should be started/stopped by
tracking all the users of it. This patch implements that CGX users
tracking. CGX is also configured along with NPC if required.

Also removed a check which mandates even number of LBK VFs.
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>

A config option is added to disable caching of dynamic entries
like SQEs and stack pages. Also locks down all HW contexts in NDC,
preventing them from being evicted.

This option is useful when the queue count is large and there are
huge NDC cache misses. It's trade off between SQ context misses and
dynamically changing entries like SQE and stack page pointers.
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Each of the NIX/NPA LFs can choose which ways of their respective
NDC caches should be used to cache their contexts. This enables
flexible configurations like disabling caching for a LF, limiting
it's context to a certain set of ways etc etc. Separate way_mask
for NIX-TX and NIX-RX is not supported.
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Ingress packet replication support has been added to 96xx B0
silicon. This patch enables using that feature to replicate
ingress broadcast packets to PF and it's VFs.

Also fixed below issues
- VFs can also install NPC MCAM entry to forward broadcast pkts.
  Otherwise, unless PF's interface is UP, VFs will not receive
  bcast packets.
- NPC MCAM entry is disabled when PF and all it's VFs are down.
- Few corner cases in installing multicast entry list.
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

CN96xx initial silicon doesn't support all features pertaining to
NIX transmit scheduling and shaping.
- It supports a fixed topology of 1:1 mapped transmit
  limiters at all levels.
- Supports DWRR only at SMQ/MDQ and TL1.
- Doesn't support shaping and coloring.

This patch adds HW capability structure by which each variant
and skew of silicon can be differentiated by their supported
features. And adds support for A0 silicon's transmit scheduler
capabilities or rather limitations.
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for few more RSS key types for flow key
algorithm to compute rss hash index.

Following flow key types have been added.
- Tunnel types like NVGRE, VXLAN, GENEVE.
- L2 offload type ETH_DMAC, Here we will consider only DMAC 6 bytes.
- And extension header IPV6_EXT (1 byte followed by IPV6 header
- Hashing inner protocol fields for inner DMAC, IPv4/v6, TCP, UDP, SCTP.
Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Writing into NPC MCAM1 and MCAM0 registers are suppressed if
they happened to form a reserved combination. Hence
clear and disable MCAM entries before update.

For HRM:
[CAM(1)]<n>=1, [CAM(0)]<n>=1: Reserved.
The reserved combination is not allowed. Hardware suppresses any
write to CAM(0) or CAM(1) that would result in the reserved combination for
any CAM bit.
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Updated NPC KPU packet parsing profile with support for following

- Fragmentation support for IPv4 IPv6 outer header
- NIX instruction header support
- QinQ with TPID of 0x8100 as non inner most vlan tag, as legacy
  network equipments still generate QinQ packets with this configuration.
- To better support RSS for tunnelled packets, udp based tunnel
  protocols such as vxlan, vxlan-gpe, geneve and gtpu are now
  captured into a separate layer E. Consequently, the inner
  packet headers are pushed one layer down to LF, LG, and LH
  accordingly.
- Support for rfc7510 mpls in udp. Up to 4 MPLS labels can be parsed
  and captured in one layer LE.
- Parser support for DSA, extended DSA and eDSA tags right after
  ethernet header by Marvell SOHO and Falcon switches. For extended
  DSA and eDSA tags, a special PKIND of 62 is used, as these tags don't
  contain a tpid field.
- Higig2 protocol header parsing support, added a NPC_LT_LA_HIGIG2_ETHER
  for a combined header of HIGIG2 and Ethernet.  Add a
  NPC_LT_LA_IH_NIX_HIGIG2_ETHER for a combined header of nix_ih,
  HIGIG2 and Ethernet on egress side. Also added 2 upper flags in LA to
  indicate the presence of nix_ih and HIGIG2.

Other changes include
- IPv4.TTL==0 IPv6.HLIM==0 check
- Per RFC 1858, mark fragment offset == 1 as error
- TCP invalid flags check
- Separate error codes for outer and inner IPv4 checksum errors.
- Fix a parser error when KPU parses incoming IPSec ESP and AH packets
- NPC vtag capture/strip hardware expect tag pointer to point to
  tpid/ethertype instead of tci. So move lb_ptr to point to tpid/ethertype.
- Fix npc parser error when parsing udp packets that don't have any payload.
- For a single MCAM entry to match on packets with one or stacked vlan tags
  combine NPC_LT_LB_STAG and NPC_LT_LB_QINQ to NPC_LT_LB_STAG_QINQ.
- NVGRE to have a separate ltype LD_NVGRE instead of combined with LD_GRE.
- Reserve top LD/LTYPEs to support custom KPU profile fields.
Signed-off-by: Hao Zheng <haoz@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For every mailbox handler added to rvu, we are adding a function
declaration in rvu header file. Cleaned this up by adding a macro
to generate these declarations automatically.
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>

If mailbox client has a bounce buffer or a intermediate buffer where
mbox messages are framed then copy them from there to HW buffer.
If 'mbase' and 'hw_mbase' are not same then assume 'mbase' points to
bounce buffer.

This patch also adds msg_size field to mbox header to copy only valid
data instead of whole buffer.
Signed-off-by: Geetha sowjanya <gakula@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Added a new mailbox API which goes through all responses
to check their IDs and response codes.

Also added logic to prevent queuing multiple works to
process the same mailbox message. This scenario happens
when AF is processing a PF's request and menawhile PF
sends ACK to AF sent UP message, then mbox_hdr->num_msgs
in the PF->AF DOWN mbox region will be nonzero and AF
will end up processing PF's request again. This is fixed
by taking a backup of num_msgs counter and clearing the
same in the mbox region before scheduling work.
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 support to display current NPC MCAM entries and counter's allocation
status ín debugfs.

cat /sys/kernel/debug/octeontx2/npc/mcam_info' will dump following info
- MCAM Rx and Tx keysize
- Total MCAM entries and counters
- Current available count
- Count of number of MCAM entries and counters allocated
  by a RVU PF/VF device.

Also, one NPC MCAM counter (last one) is reserved and mapped to
NPC RX_INTF's MISS_ACTION to count dropped packets due to no MCAM
entry match. This pkt drop counter can be checked via debugfs.
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

A CGX port is shared by a RVU PF and it's VFs. These per
CGX port level NIX Rx/Tx counters are cumilative stats of
all NIXLFs sharing this port. These stats when compared
to CGX Rx/Tx stats helps in identifying pkts dropped within
the system, if any.
Signed-off-by: Linu Cherian <lcherian@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds CGX LMAC physical interface or serdes Rx/Tx
packet stats to debugfs.

'cat cgx<idx>/lmac<idx>/stats' dumps the current interface link
status and Rx/Tx stats. Stats include pkt received/transmitted,
dropped, pause frames etc etc.
Signed-off-by: Prakash Brahmajyosyula <bprakash@marvell.com>
Signed-off-by: Linu Cherian <lcherian@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

NDC is a data cache unit which caches NPA and NIX block's
aura/pool/RQ/SQ/CQ/etc contexts to reduce number of costly
DRAM accesses.

This patch adds support to dump cache's performance stats
like cache line hit/miss counters, average cycles taken for
accessing cached and non-cached data. This will help in
checking if NPA/NIX context reads/writes are having NDC cache
misses which inturn might effect performance.

Also changed NDC enums to reflect correct NDC hardware instance.
Signed-off-by: Prakash Brahmajyosyula <bprakash@marvell.com>
Signed-off-by: Linu Cherian <lcherian@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

To aid in debugging NIX block related issues, added support to dump
NIX block LF's RQ, SQ and CQ hardware contexts in debugfs. User can
check which contexts are enabled currently and dump it's current HW
context.

Four new files 'qsize', 'rq_ctx', 'sq_ctx' and 'cq_ctx' are added to the
debugfs at 'sys/kernel/debug/octeontx2/nix/'

'echo <nixlf index> > qsize' will display current enabled CQ/SQ/RQs.
'echo <nixlf> [rq number/all] > rq_ctx',
'echo <nixlf> [sq number/all] > sq_ctx' &
'echo <nixlf> [cq number/all] > cq_ctx' will dump RQ/SQ/CQ's current
hardware context.
Signed-off-by: Prakash Brahmajyosyula <bprakash@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

To aid in debugging NPA related issues, added support to dump
NPA (pool allocator) block LF's aura and pool hardware contexts in
debugfs. User can check which contexts are enabled currently and dump
it's current HW context.

Three new files 'qsize', 'aura_ctx', 'pool_ctx' are added to the
debugfs at 'sys/kernel/debug/octeontx2/npa/'

'echo <npalf index> > qsize' will display current enabled Aura/Pools.
'echo <npalf> [aura number/all] > aura_ctx' &
'echo <npalf> [aura number/all] > pool_ctx' will dump Aura/Pool
context info.
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Signed-off-by: Prakash Brahmajyosyula <bprakash@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Added support to dump current resource provisioning status
of all resource virtualization unit (RVU) block's
(i.e NPA, NIX, SSO, SSOW, CPT, TIM) local functions attached
to a PF_FUNC into a debugfs file.

'cat /sys/kernel/debug/octeontx2/rsrc_alloc'
will show the current block LF's allocation status.
Signed-off-by: Christina Jacob <cjacob@marvell.com>
Signed-off-by: Prakash Brahmajyosyula <bprakash@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix build_skb for bm capable devices when they fall-back using swbm path
(e.g. when bm properties are configured in device tree but
CONFIG_MVNETA_BM_ENABLE is not set). In this case rx_offset_correction is
overwritten so we need to use it building skb instead of
MVNETA_SKB_HEADROOM directly

Fixes: 8dc9a088 ("net: mvneta: rely on build_skb in mvneta_rx_swbm poll routine")
Fixes: 0db51da7 ("net: mvneta: add basic XDP support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Reported-by: Andrew Lunn <andrew@lunn.ch>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

Saeed Mahameed says:

====================
mlx5-updates-2019-11-12

1) Merge mlx5-next for devlink reload and flowtable offloads dependencies
2) Devlink reload support
3) TC Flowtable offloads
4) Misc cleanup
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Use r8168d_modify_extpage() also in rtl8168f_config_eee_phy() to
simplify the code.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently, when ibmveth receive a loopback packet, it reports an
ambiguous error message "tx: h_send_logical_lan failed with rc=-4"
because the hypervisor rejects those types of packets. This fix
detects loopback packet and assures the source packet's MAC address
matches the driver's MAC address before transmitting to the
hypervisor.
Signed-off-by: Cris Forno <cforno12@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add support for the TI DP83869 Gigabit ethernet phy
device.

The DP83869 is a robust, low power, fully featured
Physical Layer transceiver with integrated PMD
sublayers to support 10BASE-T, 100BASE-TX and
1000BASE-T Ethernet protocols.
Signed-off-by: Dan Murphy <dmurphy@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add dt bindings for the TI dp83869 Gigabit ethernet phy
device.
Signed-off-by: Dan Murphy <dmurphy@ti.com>
CC: Rob Herring <robh+dt@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

When using the kernel datapath, the upcall don't
include skb hash info relatived. That will introduce
some problem, because the hash of skb is important
in kernel stack. For example, VXLAN module uses
it to select UDP src port. The tx queue selection
may also use the hash in stack.

Hash is computed in different ways. Hash is random
for a TCP socket, and hash may be computed in hardware,
or software stack. Recalculation hash is not easy.

Hash of TCP socket is computed:
tcp_v4_connect
    -> sk_set_txhash (is random)

__tcp_transmit_skb
    -> skb_set_hash_from_sk

There will be one upcall, without information of skb
hash, to ovs-vswitchd, for the first packet of a TCP
session. The rest packets will be processed in Open vSwitch
modules, hash kept. If this tcp session is forward to
VXLAN module, then the UDP src port of first tcp packet
is different from rest packets.

TCP packets may come from the host or dockers, to Open vSwitch.
To fix it, we store the hash info to upcall, and restore hash
when packets sent back.

+---------------+          +-------------------------+
|   Docker/VMs  |          |     ovs-vswitchd        |
+----+----------+          +-+--------------------+--+
     |                       ^                    |
     |                       |                    |
     |                       |  upcall            v restore packet hash (not recalculate)
     |                     +-+--------------------+--+
     |  tap netdev         |                         |   vxlan module
     +--------------->     +-->  Open vSwitch ko     +-->
       or internal type    |                         |
                           +-------------------------+

Reported-at: https://mail.openvswitch.org/pipermail/ovs-dev/2019-October/364062.htmlSigned-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>
Acked-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

MarkLee says:

====================
Rework mt762x GDM setup flow

The mt762x GDM block is mainly used to setup the HW internal
rx path from GMAC to RX DMA engine(PDMA) and the packet
switching engine(PSE) is responsed to do the data forward
following the GDM configuration.

This patch set have three goals :

1. Integrate GDM/PSE setup operations into single function "mtk_gdm_config"

2. Refine the timing of GDM/PSE setup, move it from mtk_hw_init
   to mtk_open

3. Enable GDM GDMA_DROP_ALL mode to drop all packet during the
   stop operation
====================
Signed-off-by: David S. Miller <davem@davemloft.net>