Alexei Starovoitov says:

====================
bpf: introduce bpf_skb_vlan_push/pop() helpers

Let TC+eBPF programs call skb_vlan_push/pop via helpers.

v1->v2:
- reworded commit log to better explain correctness of re-caching
  and fixed comparison of mixed endiannes (suggested by Eric)
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

improve accuracy of timing in test_bpf and add two stress tests:
- {skb->data[0], get_smp_processor_id} repeated 2k times
- {skb->data[0], vlan_push} x 68 followed by {skb->data[0], vlan_pop} x 68

1st test is useful to test performance of JIT implementation of BPF_LD_ABS
together with BPF_CALL instructions.
2nd test is stressing skb_vlan_push/pop logic together with skb->data access
via BPF_LD_ABS insn which checks that re-caching of skb->data is done correctly.

In order to call bpf_skb_vlan_push() from test_bpf.ko have to add
three export_symbol_gpl.
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Allow eBPF programs attached to TC qdiscs call skb_vlan_push/pop via
helper functions. These functions may change skb->data/hlen which are
cached by some JITs to improve performance of ld_abs/ld_ind instructions.
Therefore JITs need to recognize bpf_skb_vlan_push/pop() calls,
re-compute header len and re-cache skb->data/hlen back into cpu registers.
Note, skb->data/hlen are not directly accessible from the programs,
so any changes to skb->data done either by these helpers or by other
TC actions are safe.

eBPF JIT supported by three architectures:
- arm64 JIT is using bpf_load_pointer() without caching, so it's ok as-is.
- x64 JIT re-caches skb->data/hlen unconditionally after vlan_push/pop calls
  (experiments showed that conditional re-caching is slower).
- s390 JIT falls back to interpreter for now when bpf_skb_vlan_push() is present
  in the program (re-caching is tbd).

These helpers allow more scalable handling of vlan from the programs.
Instead of creating thousands of vlan netdevs on top of eth0 and attaching
TC+ingress+bpf to all of them, the program can be attached to eth0 directly
and manipulate vlans as necessary.
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2015-07-17

This series contains updates to igb, ixgbe, ixgbevf, i40e, bnx2x,
freescale, siena and dp83640.

Jacob provides several patches to clarify the intended way to implement
both SIOCSHWTSTAMP and ethtool's get_ts_info().  It is okay to support
the specific filters in SIOCSHWTSTAMP by upscaling them to the generic
filters.

Alex Duyck provides a igb patch to pull the time stamp from the fragment
before it gets added to the skb, to avoid a possible issue in which the
fragment can possibly be less than IGB_RX_HDR_LEN due to the time stamp
being pulled after the copybreak check.  Also provides a ixgbevf patch to
fold the ixgbevf_pull_tail() call into ixgbevf_add_rx_frag(), which gives
the advantage that the fragment does not have to be modified after it is
added to the skb.

Fan provides patches for ixgbe/ixgbevf to set the receive hash type
based on receive descriptor RSS type.

Todd provides a fix for igb where on check for link on any media other
than copper was not being detected since it was looking on the incorrect
PHY page (due to the page being used gets switched before the function
to check link gets executed).
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Florian Fainelli says:

====================
net: bcmgenet: PHY initialization rework

This patch series reworks how we perform PHY initialization and resets in the
GENET driver. Although this contains mostly fixes, some of the changes are a
bit too intrusive to be backported to 'net' at the moment.

Some of the motivations behind these changes were to reduce the time spent in how
performing MDIO transactions, since it is better to perform then when we have
interrupts enabled. This reduces the bring-up time of GENET from ~600 msecs down
to ~8 msecs, and about the same time for suspend/resume.

Since I do not currently have a system which is not DT-aware, can you (Petri,
Jaedon) give this a try and confirm things keep working as expected?
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now that we have reworked the way we perform the PHY initialization, we
no longer need to differentiate between init time vs. non-init time
calls, just use a dev_info_once() print to print the PHY type.
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We are currently doing a full PHY initialization and even starting the
pHY state machine during bcmgenet_mii_init() which is executed in the
driver's probe function. This is convenient to determine whether we can
attach to a proper PHY device but comes at the expense of spending up to
10ms per MDIO transactions (to reach the waitqueue timeout), which slows
things down.

This also creates a sitaution where we end-up attaching twice to the
PHY, which is not quite correct either.

Fix this by moving bcmgenet_mii_probe() into bcmgenet_open() and update
its error path accordingly.

Avoid printing the message "attached PHY at address 1 [...]" every time
we bring up/down the interface and remove this print since it duplicates
what the PHY driver already does for us.

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Our internal GPHY might be powered off before we attempt scanning the
MDIO bus and bind a driver to it. The way we are currently determining
whether a PHY is internal or not is done *after* we have successfully
matched its driver. If the PHY is powered down, it will not respond to
the MDIO bus, so we will not be able to bind a driver to it.

Our Device Tree for GENET interfaces specifies a "phy-mode" value:
"internal" which tells if this internal uses an internal PHY or not.

If of_get_phy_mode() fails to parse the 'phy-mode' property, do an
additional manual lookup, and if we find "internal" set the
corresponding internal variable accordingly.

Replace all uses of phy_is_internal() with a check against
priv->internal_phy to avoid having to rely on whether or not
priv->phydev is set correctly.

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We are currently disabling the GPHY interface during bcmgenet_close(),
and attempting to power it back on during bcmgenet_open(). This works
fine for the first time, because we called bcmgenet_mii_config() which
took care of enabling the interface, however, bcmgenet_power_up() really
needs to power on the GPHY for correctness.

This will be particularly important as we want to move
bcmgenet_mii_probe() down to bcmgenet_open() to avoid seeing the "PHY
already attached" message.

Fixes: a642c4f7 ("net: bcmgenet: power up and down integrated GPHY when unused")
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bcmgenet_open()'s error path call free_irq() with a dev_id argument
different from the one we used to call request_irq() with, this will
make us trip over the warning in kernel/irq/manage.c:__free_irq()

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We are currently issuing multiple PHY resets during a suspend/resume,
first during bcmgenet_power_up() which does a hardware reset, then a
software reset by calling bcmgenet_mii_reset(). This is both unnecessary
and can take as long as 10ms per MDIO transactions while we re-apply
workarounds because we do not yet have MDIO interrupts enabled.

phy_resume() takes care of re-apply our workarounds in case we need any,
and bcmgenet_power_up() does a PHY hardware reset, all of this is more
than enough to guarantee that the PHY operates correctly.

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: NFlorian Fainelli <f.fainelli@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Joachim Eastwood says:

====================
stmmac clean up for 4.3 part1

This patch set continues the conversion of the dwmac glue layers
to more proper platform drivers. The first part of the patch set
cleans up stmmac_platform a bit. Refactors code from the common
probe function and exports two functions that will be used in
the dwmac-* drivers.

Second part converts two simple dwmac-* drivers to have their
own probe function and use the exported functions. This brings
us closer to point where stmmac_platform is only a library of
common functions for the dwmac-* drivers to use.

The plan next is:
 * add probe functions to the rest of the dwmac-* drivers
 * move probe function in stmmac_platform to dwmac-generic
 * remove struct stmmac_of_data and let those drivers
   that actually need match data handle it themselves
 * clean up include/linux/stmmac.h

Note that this patch set has only been tested on lpc18xx so
testing on other platforms is greatly appreciated.

Previous parts can be found here:
http://www.spinics.net/lists/netdev/msg328997.html
http://www.spinics.net/lists/netdev/msg329932.html
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Both of these fields are unused and has been unused since they
were added 3 and 5 years ago. Drop them since they are clearly
not very useful.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

By using a few functions from stmmac_platform we can now create
a proper probe function in this driver. By doing so we can drop
the OF match data and simplify the overall driver.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

By using a few functions from stmmac_platform we can now create
a proper probe function in this driver. By doing so we can drop
the OF match data and simplify the overall driver.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Export stmmac_probe_config_dt() and stmmac_get_platform_resources()
so they can be used in the dwmac-* drivers themselves. This will
allow us to build more flexible and standalone drivers which just
use stmmac_platform as a library for setup functions.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since stmmac_probe_config_dt() allocates the platform data structure
it is cleaner if it just returned this structure directly. This
function will later be used in the probe function in dwmac-* drivers.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Refactor all code that deals with platform resources into it's
own get function. This function will later be used in the probe
function in dwmac-* drivers.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Refactor code to clearly separate probing non-dt versus dt. In the
non-dt case platform data must be supplied to probe successfully.
For dt the platform data structure is created and match data is
copied into it. Note that support for supplying platform data in
dt from AUXDATA is dropped as no users in mainline does this.

This change will allow dt dwmac-* drivers to call the config_dt()
function from probe to create the needed platform data struct and
retrieve common dt properties.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

By using of_device_get_match_data() the code that retrieve
match data can be simplified quite a bit.
Signed-off-by: NJoachim Eastwood <manabian@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Jon Maloy says:

====================
tipc: separate link and link aggregation layer

This is the first batch of a longer series that has two main objectives:

o Finer lock granularity during message sending and reception,
  especially regarding usage of the node spinlock.

o Better separation between the link layer implementation and the link
  aggregation layer, represented by node.c::struct tipc_node.

Hopefully these changes also make this part of code somewhat easier
to comprehend and maintain.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We convert packet/message reception according to the same principle
we have been using for message sending and timeout handling:

We move the function tipc_rcv() to node.c, hence handling the initial
packet reception at the link aggregation level. The function grabs
the node lock, selects the receiving link, and accesses it via a new
call tipc_link_rcv(). This function appends buffers to the input
queue for delivery upwards, but it may also append outgoing packets
to the xmit queue, just as we do during regular message sending. The
latter will happen when buffers are forwarded from the link backlog,
or when retransmission is requested.

Upon return of this function, and after having released the node lock,
tipc_rcv() delivers/tranmsits the contents of those queues, but it may
also perform actions such as link activation or reset, as indicated by
the return flags from the link.

This reduces the number of cpu cycles spent inside the node spinlock,
and reduces contention on that lock.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The logics for determining when a node is permitted to establish
and maintain contact with its peer node becomes non-trivial in the
presence of multiple parallel links that may come and go independently.

A known failure scenario is that one endpoint registers both its links
to the peer lost, cleans up it binding table, and prepares for a table
update once contact is re-establihed, while the other endpoint may
see its links reset and re-established one by one, hence seeing
no need to re-synchronize the binding table. To avoid this, a node
must not allow re-establishing contact until it has confirmation that
even the peer has lost both links.

Currently, the mechanism for handling this consists of setting and
resetting two state flags from different locations in the code. This
solution is hard to understand and maintain. A closer analysis even
reveals that it is not completely safe.

In this commit we do instead introduce an FSM that keeps track of
the conditions for when the node can establish and maintain links.
It has six states and four events, and is strictly based on explicit
knowledge about the own node's and the peer node's contact states.
Only events leading to state change are shown as edges in the figure
below.

                             +--------------+
                             | SELF_UP/     |
           +---------------->| PEER_COMING  |-----------------+
    SELF_  |                 +--------------+                 |PEER_
    ESTBL_ |                        |                         |ESTBL_
    CONTACT|      SELF_LOST_CONTACT |                         |CONTACT
           |                        v                         |
           |                 +--------------+                 |
           |      PEER_      | SELF_DOWN/   |     SELF_       |
           |      LOST_   +--| PEER_LEAVING |<--+ LOST_       v
+-------------+   CONTACT |  +--------------+   | CONTACT  +-----------+
| SELF_DOWN/  |<----------+                     +----------| SELF_UP/  |
| PEER_DOWN   |<----------+                     +----------| PEER_UP   |
+-------------+   SELF_   |  +--------------+   | PEER_    +-----------+
           |      LOST_   +--| SELF_LEAVING/|<--+ LOST_       A
           |      CONTACT    | PEER_DOWN    |     CONTACT     |
           |                 +--------------+                 |
           |                         A                        |
    PEER_  |       PEER_LOST_CONTACT |                        |SELF_
    ESTBL_ |                         |                        |ESTBL_
    CONTACT|                 +--------------+                 |CONTACT
           +---------------->| PEER_UP/     |-----------------+
                             | SELF_COMING  |
                             +--------------+
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In our effort to move control of the links to the link aggregation
layer, we move the perodic link supervision timer to struct tipc_node.
The new timer is shared between all links belonging to the node, thus
saving resources, while still kicking the FSM on both its pertaining
links at each expiration.

The current link timer and corresponding functions are removed.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We create a second, simpler, link timer function, tipc_link_timeout().
The new function  makes use of the new FSM function introduced in the
previous commit, and just like it, takes a buffer queue as parameter.
It returns an event bit field and potentially a link protocol packet
to the caller.

The existing timer function, link_timeout(), is still needed for a
while, so we redesign it to become a wrapper around the new function.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The link FSM implementation is currently unnecessarily complex.
It sometimes checks for conditional state outside the FSM data
before deciding next state, and often performs actions directly
inside the FSM logics.

In this commit, we create a second, simpler FSM implementation,
that as far as possible acts only on states and events that it is
strictly defined for, and postpone any actions until it is finished
with its decisions. It also returns an event flag field and an a
buffer queue which may potentially contain a protocol message to
be sent by the caller.

Unfortunately, we cannot yet make the FSM "clean", in the sense
that its decisions are only based on FSM state and event, and that
state changes happen only here. That will have to wait until the
activate/reset logics has been cleaned up in a future commit.

We also rename the link states as follows:

WORKING_WORKING -> TIPC_LINK_WORKING
WORKING_UNKNOWN -> TIPC_LINK_PROBING
RESET_UNKNOWN   -> TIPC_LINK_RESETTING
RESET_RESET     -> TIPC_LINK_ESTABLISHING

The existing FSM function, link_state_event(), is still needed for
a while, so we redesign it to make use of the new function.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As a preparation for later changes, we introduce a new function
tipc_link_build_proto_msg(). Instead of actually sending the created
protocol message, it only creates it and adds it to the head of a
skb queue provided by the caller.

Since we still need the existing function tipc_link_protocol_xmit()
for a while, we redesign it to make use of the new function.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The status flag LINK_STOPPED is not needed any more, since the
mechanism for delayed deletion of links has been removed.
Likewise, LINK_STARTED and LINK_START_EVT are unnecessary,
because we can just as well start the link timer directly from
inside tipc_link_create().

We eliminate these flags in this commit.

Instead of the above flags, we now introduce three new link modes,
TIPC_LINK_OPEN, TIPC_LINK_BLOCKED and TIPC_LINK_TUNNEL. The values
indicate whether, and in the case of TIPC_LINK_TUNNEL, which, messages
the link is allowed to receive in this state. TIPC_LINK_BLOCKED also
blocks timer-driven protocol messages to be sent out, and any change
to the link FSM. Since the modes are mutually exclusive, we convert
them to state values, and rename the 'flags' field in struct tipc_link
to 'exec_mode'.

Finally, we move the #defines for link FSM states and events from link.h
into enums inside the file link.c, which is the real usage scope of
these definitions.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently, message sending is performed through a deep call chain,
where the node spinlock is grabbed and held during a significant
part of the transmission time. This is clearly detrimental to
overall throughput performance; it would be better if we could send
the message after the spinlock has been released.

In this commit, we do instead let the call revert on the stack after
the buffer chain has been added to the transmission queue, whereafter
clones of the buffers are transmitted to the device layer outside the
spinlock scope.

As a further step in our effort to separate the roles of the node
and link entities we also move the function tipc_link_xmit() to
node.c, and rename it to tipc_node_xmit().
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the function tipc_link_xmit() is given a buffer list for
transmission, it currently consumes the list both when transmission
is successful and when it fails, except for the special case when
it encounters link congestion.

This behavior is inconsistent, and needs to be corrected if we want
to avoid problems in later commits in this series.

In this commit, we change this to let the function consume the list
only when transmission is successful, and leave the list with the
sender in all other cases. We also modifiy the socket code so that
it adapts to this change, i.e., purges the list when a non-congestion
error code is returned.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

struct tipc_node currently holds two arrays of link pointers; one,
indexed by bearer identity, which contains all links irrespective of
current state, and one two-slot array for the currently active link
or links. The latter array contains direct pointers into the elements
of the former. This has the effect that we cannot know the bearer id of
a link when accessing it via the "active_links[]" array without actually
dereferencing the pointer, something we want to avoid in some cases.

In this commit, we do instead store the bearer identity in the
"active_links" array, and use this as an index to find the right element
in the overall link entry array. This change should be seen as a
preparation for the later commits in this series.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

At present, the link input queue and the name distributor receive
queues are fields aggregated in struct tipc_link. This is a hazard,
because a link might be deleted while a receiving socket still keeps
reference to one of the queues.

This commit fixes this bug. However, rather than adding yet another
reference counter to the critical data path, we move the two queues
to safe ground inside struct tipc_node, which is already protected, and
let the link code only handle references to the queues. This is also
in line with planned later changes in this area.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As a step towards turning links into node internal entities, we move the
creation of links from the neighbor discovery logics to the node's link
control logics.

We also create an additional entry for the link's media address in the
newly introduced struct tipc_link_entry, since this is where it is
needed in the upcoming commits. The current copy in struct tipc_link
is kept for now, but will be removed later.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

struct 'tipc_node' currently contains two arrays for link attributes,
one for the link pointers, and one for the usable link MTUs.

We now group those into a new struct 'tipc_link_entry', and intoduce
one single array consisting of such enties. Apart from being a cosmetic
improvement, this is a starting point for the strict master-slave
relation between node and link that we will introduce in the following
commits.
Reviewed-by: NYing Xue <ying.xue@windriver.com>
Signed-off-by: NJon Maloy <jon.maloy@ericsson.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's not used anywhere.
Signed-off-by: NJiri Benc <jbenc@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Scott Feldman says:

====================
switchdev: avoid duplicate packet forwarding

v3:

 - Per Nicolas Dichtel review: remove errant empty union.

v2:

 - Per davem review: in sk_buff, union fwd_mark with secmark to save space
   since features appear to be mutually exclusive.
 - Per Simon Horman review:
   - fix grammar in switchdev.txt wrt fwd_mark
   - remove some unrelated changes that snuck in

v1:

This patchset was previously submitted as RFC.  No changes from the last
version (v2) sent under RFC.  Including RFC version history here for reference.

RFC v2:

 - s/fwd_mark/offload_fwd_mark
 - use consume_skb rather than kfree_skb when dropping pkt on egress.
 - Use Jiri's suggestion to use ifindex of one of the ports in a group
   as the mark for all the ports in the group.  This can be done with
   no additional storage (no hashtable from v1).  To pull it off, we
   need some simple recursive routines to walk the netdev tree ensuring
   all leaves in the tree (ports) in the same group (e.g. bridge)
   belonging to the same switch device will have the same offload fwd mark.
   Maybe someone sees a better design for the recusive routines?  They're
   not too bad, and should cover the stacked driver cases.

RFC v1:

With switchdev support for offloading L2/L3 forwarding data path to a
switch device, we have a general problem where both the device and the
kernel may forward the packet, resulting in duplicate packets on the wire.
Anytime a packet is forwarded by the device and a copy is sent to the CPU,
there is potential for duplicate forwarding, as the kernel may also do a
forwarding lookup and send the packet on the wire.

The specific problem this patch series is interested in solving is avoiding
duplicate packets on bridged ports.  There was a previous RFC from Roopa
(http://marc.info/?l=linux-netdev&m=142687073314252&w=2) to address this
problem, but didn't solve the problem of mixed ports in the bridge from
different devices; there was no way to exclude some ports from forwarding
and include others.  This RFC solves that problem by tagging the ingressing
packet with a unique mark, and then comparing the packet mark with the
egress port mark, and skip forwarding when there is a match.  For the mixed
ports bridge case, only those ports with matching marks are skipped.

The switchdev port driver must do two things:

1) Generate a fwd_mark for each switch port, using some unique key of the
   switch device (and optionally port).  This is done when the port netdev
   is registered or if the port's group membership changes (joins/leaves
   a bridge, for example).

2) On packet ingress from port, mark the skb with the ingress port's
   fwd_mark.  If the device supports it, it's useful to only mark skbs
   which were already forwarded by the device.  If the device does not
   support such indication, all skbs can be marked, even if they're
   local dst.

Two new 32-bit fields are added to struct sk_buff and struct netdevice to
hold the fwd_mark.  I've wrapped these with CONFIG_NET_SWITCHDEV for now. I
tried using skb->mark for this purpose, but ebtables can overwrite the
skb->mark before the bridge gets it, so that will not work.

In general, this fwd_mark can be used for any case where a packet is
forwarded by the device and a copy is sent to the CPU, to avoid the kernel
re-forwarding the packet.  sFlow is another use-case that comes to mind,
but I haven't explored the details.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Acked-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

If device flags ingress packet as "fwd offload", mark the
skb->offlaod_fwd_mark using the ingress port's dev->offlaod_fwd_mark.  This
will be the hint to the kernel that this packet has already been forwarded
by device to egress ports matching skb->offlaod_fwd_mark.

For rocker, derive port dev->offlaod_fwd_mark based on device switch ID and
port ifindex.  If port is bridged, use the bridge ifindex rather than the
port ifindex.
Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Acked-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

skb->offload_fwd_mark and dev->offload_fwd_mark are 32-bit and should be
unique for device and may even be unique for a sub-set of ports within
device, so add switchdev helper function to generate unique marks based on
port's switch ID and group_ifindex.  group_ifindex would typically be the
container dev's ifindex, such as the bridge's ifindex.

The generator uses a global hash table to store offload_fwd_marks hashed by
{switch ID, group_ifindex} key.
Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Acked-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NScott Feldman <sfeldma@gmail.com>
Acked-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>