When the bound device of a tunnel device is down, encapsulated packets
are not egressed anymore, but tunnel decap still works. Extend
mlxsw_sp_nexthop_rif_update() to take IFF_UP into consideration when
deciding whether a given next hop should be offloaded.

Because the new logic was added to mlxsw_sp_nexthop_rif_update(), this
fixes the case where a newly-added tunnel has a down bound device, which
would previously be fully offloaded. Now the down state of the bound
device is noted and next hops forwarding to such tunnel are not
offloaded.

In addition to that, notice NETDEV_UP and NETDEV_DOWN of a bound device
to force refresh of tunnel encap route offloads.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a bound device of an IP-in-IP tunnel changes, such as through
'ip tunnel change name $name dev $dev', the loopback backing the tunnel
needs to be recreated.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Changes to L3 tunnel netdevices (through `ip tunnel change' as well as
`ip link set') lead to NETDEV_CHANGE being generated on the tunnel
device. Because what is relevant for the tunnel in question depends on
the tunnel type, handling of the event is dispatched to the IPIP module
through a newly-added interface mlxsw_sp_ipip_ops.ol_netdev_change().

IPIP tunnels now remember the last set of tunnel parameters in struct
mlxsw_sp_ipip_entry.parms, and use it to figure out what exactly has
changed.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a bound device of a tunnel netdevice changes VRF, the loopback RIF
that backs the tunnel needs to be updated and existing encapsulating
routes need to be refreshed.

Note that several tunnels can share the same bound device, in which case
all the impacted tunnels need to be updated.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The approach for offloading IP tunnels implemented currently by mlxsw
doesn't allow two tunnels that have the same local IP address in the
same (underlay) VRF. Previously, offloads were introduced on demand as
encap routes were formed. When such a route was created that would cause
offload of a conflicting tunnel, mlxsw_sp_ipip_entry_create() would
detect it and return -EEXIST, which would propagate up and cause FIB
abort.

Now however IPIP entries are created as soon as an offloadable netdevice
is created, and the failure prevents creation of such device.
Furthermore, if the driver is installed at the point where such
conflicting tunnels exist, the failure actually prevents successful
modprobe.

Furthermore, follow-up patches implement handling of NETDEV_CHANGE due
to the local address change. However, NETDEV_CHANGE can't be vetoed. The
failure merely means that the offloads weren't updated, but the change
in Linux configuration is not rolled back. It is thus desirable to have
a robust way of handling these conflicts, which can later be reused for
handling NETDEV_CHANGE as well.

To fix this, when a conflicting tunnel is created, instead of failing,
simply pull the old tunnel to slow path and reject offloading the
new one.

Introduce two functions: mlxsw_sp_ipip_entry_demote_tunnel() and
mlxsw_sp_ipip_demote_tunnel_by_saddr() to handle this. Make them both
public, because they will be useful later on in this patchset.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When trying to determine whether there are other offloaded tunnels with
the same local address, mlxsw_sp_ipip_entry_create() should look for a
tunnel with matching UL protocol, matching saddr, in the same VRF.
However instead of taking into account the UL protocol of the tunnel
netdevice (which mlxsw_sp_ipip_entry_saddr_matches() then compares to
the UL protocol of inspected IPIP entry), it deduces the UL protocol
from the inspected IPIP entry (and that's compared to itself).

This is currently immaterial, because only one tunnel type is offloaded,
and therefore the UL protocol always matches, but introducing support
for a tunnel with IPv6 underlay would uncover this error.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The work that needs to be done to update HW configuration in response to
changes is similar to what __mlxsw_sp_ipip_entry_update_tunnel() already
does, but with a number of twists: each change requires a different
subset of things to happen. Extend the function to support all these
uses, and allow finely-grained configuration of what should happen at
each call through a suite of function arguments.

Publish the updated function to allow use from the spectrum_ipip module.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The work that's done by mlxsw_sp_netdevice_ipip_ol_vrf_event() is a good
basis for a more versatile function that would take care of all sorts of
tunnel updates requests: __mlxsw_sp_ipip_entry_update_tunnel(). Extract
that function. Factor out a helper mlxsw_sp_ipip_entry_ol_lb_update() as
well.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function mlxsw_sp_rif_create() takes an extack parameter. So far,
for creation of loopback interfaces, NULL was passed. For some events
however the extack can be extracted and passed along. So do that for
NETDEV_CHANGEUPPER handler.

Use the opportunity to update the type of info argument that
mlxsw_sp_netdevice_ipip_ol_event() takes. Follow-up patches will
introduce handling of more changes, and some of them carry an extack as
well, but in an info structure of a different type. Though not strictly
erroneous (the pointer could be cast whichever way), it makes no sense
to pretend the value is always of a certain type, when in fact it isn't.
So change the prototype of the above-mentioned function as well.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The piece of logic to promote decap route, if any, is useful for generic
tunnel updates, not just for handling of NETDEV_UP events on tunnel
interfaces. Extract it to a separate function.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This function only ever returns 0, so don't pretend it returns anything
useful and just make it void.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To implement NETDEV_CHANGE notifications on IP-in-IP tunnels, the
handler needs to figure out what actually changed, to understand how
exactly to update the offloads. It will do so by storing struct
ip_tunnel_parm with previous configuration, and comparing that to the
new version.

To facilitate these comparisons, extract the code that operates on
struct ip_tunnel_parm from the existing accessor functions, and make
those a thin wrapper that extracts tunnel parameters and dispatches.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

These functions ideologically belong to the IPIP module, and some
follow-up work will benefit from their presence there.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some of the code down the road needs this logic as well.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To distinguish between events related to tunnel device itself and its
bound device, rename a number of functions related to handling tunneling
netdevice events to include _ol_ (for "overlay") in the name. That
leaves room in the namespace for underlay-related functions, which would
have _ul_ in the name.
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make sure the device and the kernel are performing the multipath hash
according to the same parameters by updating the device whenever the
relevant netevent is generated.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NPetr Machata <petrm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Up until now we used the hardware's defaults for multipath hash
computation. This patch aligns the hardware's multipath parameters with
the kernel's.

For IPv4 packets, the parameters are determined according to the
'fib_multipath_hash_policy' sysctl during module initialization. In case
L3-mode is requested, only the source and destination IP addresses are
used. There is no special handling of ICMP error packets.

In case L4-mode is requested, a 5-tuple is used: source and destination
IP addresses, source and destination ports and IP protocol. Note that
the layer 4 fields are not considered for fragmented packets.

For IPv6 packets, the source and destination IP addresses are used, as
well as the flow label and the next header fields.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NPetr Machata <petrm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The RECRv2 register is used for setting up the router's ECMP hash
configuration.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NPetr Machata <petrm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The struct containing the work item queued from the netevent handler is
named after the only event it is currently used for, which is neighbour
updates.

Use a more appropriate name for the struct, as we are going to use it
for more events.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NPetr Machata <petrm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We are going to need to respond to netevents notifying us about
multipath hash updates by configuring the device's hash parameters.

Embed the netevent notifier in the router struct so that we could
retrieve it upon notifications and use it to configure the device.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Reviewed-by: NPetr Machata <petrm@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This restores the original behaviour before the block callbacks were
introduced. Allow the drivers to do binding of block always, no matter
if the NETIF_F_HW_TC feature is on or off. Move the check to the block
callback which is called for rule insertion.
Reported-by: NAlexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It fixes a problem for the last chunk where 'chunk_size' is smaller than
MLXSW_I2C_BLK_MAX and data is copied to the wrong offset, overriding
previous data.

Fixes: 6882b0ae ("mlxsw: Introduce support for I2C bus")
Signed-off-by: NVadim Pasternak <vadimp@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The ASIC has the ability to generate events whenever a sensor indicates
the temperature goes above or below its high or low thresholds,
respectively.

In new firmware versions the firmware enforces a minimum of 5
degrees Celsius difference between both thresholds. Make the driver
conform to this requirement.

Note that this is required even when the events are disabled, as in
certain systems interrupts are generated via GPIO based on these
thresholds.

Fixes: 85926f87 ("mlxsw: reg: Add definition of temperature management registers")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Adding a FIB rule on a spectrum platform silently aborts FIB offload:
    $ ip ru add pref 99 from all to 192.168.1.1 table 10
    $ dmesg -c
    [  623.144736] mlxsw_spectrum 0000:03:00.0: FIB abort triggered. Note that FIB entries are no longer being offloaded to this device.

This patch reworks FIB rule handling to return a message to the user:
    $ ip ru add pref 99 from all to 8.8.8.8 table 11
    Error: spectrum: FIB rules not supported. Aborting offload.

spectrum currently only checks whether the fib rule is a default rule or
an l3mdev rule, both of which it knows how to handle. Any other it aborts
FIB offload. Move the processing to check the rule type inline with the
user request. If the rule is an unsupported one, then a work queue entry
is used to abort the offload. Change the rule delete handling to just
return since it does nothing at the moment.
Signed-off-by: NDavid Ahern <dsahern@gmail.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace recurring magic number in PPCNT register with a define.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Change the HW stats cache to be local. Rename it for better clarity.
It holds the results of the last result of HW stats that are being read
periodically, in order to have answer for stats request immediately.
Signed-off-by: NNogah Frankel <nogahf@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Make the spectrum_mr_tcam.c include the spectrum_mr_tcam.h header file.

Cleans up sparse warning:
symbol 'mlxsw_sp_mr_tcam_ops' was not declared. Should it be static?

Fixes: 0e14c777 ("mlxsw: spectrum: Add the multicast routing hardware logic")
Signed-off-by: NYotam Gigi <yotamg@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The function is only used internally in spectrum_mr.c and is not declared
in the header file, thus make it static.

Cleans up sparse warning:
symbol 'mlxsw_sp_mr_dev_vif_lookup' was not declared. Should it be static?

Fixes: c011ec1b ("mlxsw: spectrum: Add the multicast routing offloading logic")
Signed-off-by: NYotam Gigi <yotamg@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix various endianness issues in comparisons and assignments. The fix is
entirely cosmetic as all the values fixed are endianness-agnostic.

Cleans up sparse warnings:
spectrum_mr.c:156:49: warning: restricted __be32 degrades to integer
spectrum_mr.c:206:26: warning: restricted __be32 degrades to integer
spectrum_mr.c:212:31: warning: incorrect type in assignment (different
  base types)
spectrum_mr.c:212:31:    expected restricted __be32 [usertype] addr4
spectrum_mr.c:212:31:    got unsigned int
spectrum_mr.c:214:32: warning: incorrect type in assignment (different
  base types)
spectrum_mr.c:214:32:    expected restricted __be32 [usertype] addr4
spectrum_mr.c:214:32:    got unsigned int
spectrum_mr.c:461:16: warning: restricted __be32 degrades to integer
spectrum_mr.c:461:49: warning: restricted __be32 degrades to integer

Fixes: c011ec1b ("mlxsw: spectrum: Add the multicast routing offloading logic")
Signed-off-by: NYotam Gigi <yotamg@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

During the dump the per netlink packet entry counter should be zeroed out
when new packet is created.

Fixes: 190d38a5 ("mlxsw: spectrum_dpipe: Add support for adjacency table dump")
Signed-off-by: NArkadi Sharshevsky <arkadis@mellanox.com>
Reported-by: NDavid Ahern <dsahern@gmail.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The KVD linear is currently partitioned into two partitions. One for
single entries and another for groups of 32 entries.

Add another partition consisting of groups of 512 entries which will
allow us to more accurately represent the nexthop weights in non-equal
cost multi-path routing.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory region where adjacency entries (nexthops) are stored is
called the KVD linear and is configured during initialization with a
size of 64K.

Extend this area with 32K more entries, that will be partitioned into 64
groups of 0.5K entries, thereby allowing us to support weighted nexthops
with high accuracy.

Change the ratio between both types of hash entries, so as to prevent
reduction in the number of double hash entries, which are used for IPv6
neighbours and routes with a prefix length greater than 64.

Note that the user will be able to control all these sizes once the
devlink resource manager is introduced.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Up until now the driver assumed all the nexthops have an equal weight
and wrote each to a single adjacency entry.

This patch takes the `weight` parameter into account and populates the
adjacency group according to the relative weight of each nexthop.

Specifically, the weights of all the nexthops that should be offloaded
are first normalized and then used to calculate the upper adjacency
index of each nexthop. This is done according to the hash-threshold
algorithm used by the kernel for IPv4 multi-path routing.

Adjacency groups are currently limited to 32 entries which limits the
weights that can be used, but follow-up patches will introduce groups of
512 entries.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The device has certain restrictions regarding the size of an adjacency
group.

Have the router determine the size of the adjacency group according to
available KVDL allocation sizes and these restrictions.

This was not needed until now since only allocations of up 32 entries
were supported and these are all valid sizes for an adjacency group.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

As the first step towards non-equal-cost multi-path support, store each
nexthop's weight.

For IPv6 nexthops always set the weight to 1, as it only supports ECMP.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The current KVDL allocation API allows the user to specify the requested
number of entries, but the user has no way of knowing how many entries
were actually allocated.

This works because existing users (e.g., router) request the exact
number they end up using. With the introduction of large adjacency
groups, this will change, as the router will have the ability to choose
from several allocation sizes, where larger allocations provide higher
accuracy with respect to requested weights and better resilience against
nexthop failures.

One option is to have the router try several allocations of descending
size until one succeeds, but a better way is to simply allow it to query
the actual allocation size and then size its request accordingly.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The KVD linear (KVDL) allocator currently consists of a very large
bitmap that reflects the KVDL's usage. The boundaries of each partition
as well as their allocation size are represented using defines.

This representation requires us to patch all the functions that act on a
partition whenever the partitioning scheme is changed. In addition, it
does not enable the dynamic configuration of the KVDL using the
up-coming resource manager.

Add objects to represent these partitions as well as the accompanying
code that acts on them to perform allocations and de-allocations.

In the following patches, this will allow us to easily add another
partition as well as new operations to act on these partitions.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The adjacency group size is part of the match on the adjacency group and
should therefore be exposed using dpipe.

When non-equal-cost multi-path support will be introduced, the group's
size will help users understand the exact number of adjacency entries
each nexthop occupies, as a nexthop will no longer correspond to a
single entry.

The output for a multi-path route with two nexthops, one with weight 255
and the second 1 will be:

Example:

$ devlink dpipe table dump pci/0000:01:00.0 name mlxsw_adj
pci/0000:01:00.0:
  index 0
  match_value:
    type field_exact header mlxsw_meta field adj_index value 65536
    type field_exact header mlxsw_meta field adj_size value 512
    type field_exact header mlxsw_meta field adj_hash_index value 0
  action_value:
    type field_modify header ethernet field destination mac value e4:1d:2d:a5:f3:64
    type field_modify header mlxsw_meta field erif_port mapping ifindex mapping_value 3 value 1

  index 1
  match_value:
    type field_exact header mlxsw_meta field adj_index value 65536
    type field_exact header mlxsw_meta field adj_size value 512
    type field_exact header mlxsw_meta field adj_hash_index value 510
  action_value:
    type field_modify header ethernet field destination mac value e4:1d:2d:a5:f3:65
    type field_modify header mlxsw_meta field erif_port mapping ifindex mapping_value 4 value 2

Thus, the first nexthop occupies 510 adjacency entries and the second 2,
which leads to a ratio of 255 to 1.
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Spectrum tunnels do not default to ttl of "inherit" like the Linux ones
do. Configure TIGCR on router init so that the TTL of tunnel packets is
copied from the overlay packets.

Fixes: ee954d1a ("mlxsw: spectrum_router: Support GRE tunnels")
Signed-off-by: NPetr Machata <petrm@mellanox.com>
Reviewed-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>