Test various aspects of the FIB offload API on top of the netdevsim
implementation. Both good and bad flows are tested.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implement a set of common helpers and tests for FIB offload that can be
used by multiple drivers to check their FIB offload implementations.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Implement dummy IPv4 and IPv6 FIB "offload" in the driver by storing
currently "programmed" routes in a hash table. Each route in the hash
table is marked with "trap" indication. The indication is cleared when
the route is replaced or when the netdevsim instance is deleted.

This will later allow us to test the route offload API on top of
netdevsim.

v2:
* Convert to new fib_alias_hw_flags_set() interface
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Previous patches added support for two hardware flags for IPv4 and IPv6
routes: 'RTM_F_OFFLOAD' and 'RTM_F_TRAP'. Both indicate the presence of
the route in hardware. The first indicates that traffic is actually
offloaded from the kernel, whereas the second indicates that packets
hitting such routes are trapped to the kernel for processing (e.g., host
routes).

Use these two flags in mlxsw. The flags are modified in two places.
Firstly, whenever a route is updated in the device's table. This
includes the addition, deletion or update of a route. For example, when
a host route is promoted to perform NVE decapsulation, its action in the
device is updated, the 'RTM_F_OFFLOAD' flag set and the 'RTM_F_TRAP'
flag cleared.

Secondly, when a route is replaced and overwritten by another route, its
flags are cleared.

v2:
* Convert to new fib_alias_hw_flags_set() interface
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The driver currently uses the 'RTNH_F_OFFLOAD' flag for both routes and
nexthops, which is cumbersome and unnecessary now that we have separate
flag for the route itself.

Separate the offload indication for nexthops from routes and call it
whenever the offload state within the nexthop group changes.

Note that IPv6 (unlike IPv4) does not share the same nexthop group
between different routes, whereas mlxsw does. Therefore, whenever the
offload indication within an IPv6 nexthop group changes, all the linked
routes need to be updated.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In a similar fashion to previous patch, add "offload" and "trap"
indication to IPv6 routes.

This is done by using two unused bits in 'struct fib6_info' to hold
these indications. Capable drivers are expected to set these when
processing the various in-kernel route notifications.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Acked-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When performing L3 offload, routes and nexthops are usually programmed
into two different tables in the underlying device. Therefore, the fact
that a nexthop resides in hardware does not necessarily mean that all
the associated routes also reside in hardware and vice-versa.

While the kernel can signal to user space the presence of a nexthop in
hardware (via 'RTNH_F_OFFLOAD'), it does not have a corresponding flag
for routes. In addition, the fact that a route resides in hardware does
not necessarily mean that the traffic is offloaded. For example,
unreachable routes (i.e., 'RTN_UNREACHABLE') are programmed to trap
packets to the CPU so that the kernel will be able to generate the
appropriate ICMP error packet.

This patch adds an "offload" and "trap" indications to IPv4 routes, so
that users will have better visibility into the offload process.

'struct fib_alias' is extended with two new fields that indicate if the
route resides in hardware or not and if it is offloading traffic from
the kernel or trapping packets to it. Note that the new fields are added
in the 6 bytes hole and therefore the struct still fits in a single
cache line [1].

Capable drivers are expected to invoke fib_alias_hw_flags_set() with the
route's key in order to set the flags.

The indications are dumped to user space via a new flags (i.e.,
'RTM_F_OFFLOAD' and 'RTM_F_TRAP') in the 'rtm_flags' field in the
ancillary header.

v2:
* Make use of 'struct fib_rt_info' in fib_alias_hw_flags_set()

[1]
struct fib_alias {
        struct hlist_node  fa_list;                      /*     0    16 */
        struct fib_info *          fa_info;              /*    16     8 */
        u8                         fa_tos;               /*    24     1 */
        u8                         fa_type;              /*    25     1 */
        u8                         fa_state;             /*    26     1 */
        u8                         fa_slen;              /*    27     1 */
        u32                        tb_id;                /*    28     4 */
        s16                        fa_default;           /*    32     2 */
        u8                         offload:1;            /*    34: 0  1 */
        u8                         trap:1;               /*    34: 1  1 */
        u8                         unused:6;             /*    34: 2  1 */

        /* XXX 5 bytes hole, try to pack */

        struct callback_head rcu __attribute__((__aligned__(8))); /*    40    16 */

        /* size: 56, cachelines: 1, members: 12 */
        /* sum members: 50, holes: 1, sum holes: 5 */
        /* sum bitfield members: 8 bits (1 bytes) */
        /* forced alignments: 1, forced holes: 1, sum forced holes: 5 */
        /* last cacheline: 56 bytes */
} __attribute__((__aligned__(8)));
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

fib_dump_info() is used to prepare RTM_{NEW,DEL}ROUTE netlink messages
using the passed arguments. Currently, the function takes 11 arguments,
6 of which are attributes of the route being dumped (e.g., prefix, TOS).

The next patch will need the function to also dump to user space an
indication if the route is present in hardware or not. Instead of
passing yet another argument, change the function to take a struct
containing the different route attributes.

v2:
* Name last argument of fib_dump_info()
* Move 'struct fib_rt_info' to include/net/ip_fib.h so that it could
  later be passed to fib_alias_hw_flags_set()
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Subsequent patches will add an offload / trap indication to routes which
will signal if the route is present in hardware or not.

After programming the route to the hardware, drivers will have to ask
the IPv4 code to set the flags by passing the route's key.

In the case of route replace, the new route is notified before it is
actually inserted into the FIB alias list. This can prevent simple
drivers (e.g., netdevsim) that program the route to the hardware in the
same context it is notified in from being able to set the flag.

Solve this by first inserting the new route to the list and rollback the
operation in case the route was vetoed.
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Socionext driver can run on dma coherent and non-coherent devices.
Get rid of huge dma_sync_single_for_device in netsec_alloc_rx_data since
now the driver can let page_pool API to managed needed DMA sync
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Bjorn Andersson says:

====================
QRTR flow control improvements

In order to prevent overconsumption of resources on the remote side QRTR
implements a flow control mechanism.

Move the handling of the incoming confirm_rx to the receiving process to
ensure incoming flow is controlled. Then implement outgoing flow
control, using the recommended algorithm of counting outstanding
non-confirmed messages and blocking when hitting a limit. The last three
patches refactors the node assignment and port lookup, in order to
remove the worker in the receive path.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Rather than enqueuing messages and scheduling a worker to deliver them
to the individual sockets we can now, thanks to the previous work, move
this directly into the endpoint callback.

This saves us a context switch per incoming message and removes the
possibility of an opportunistic suspend to happen between the message is
coming from the endpoint until it ends up in the socket's receive
buffer.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The important part of qrtr_port_lookup() wrt synchronization is that the
function returns a reference counted struct qrtr_sock, or fail.

As such we need only to ensure that an decrement of the object's
refcount happens inbetween the finding of the object in the idr and
qrtr_port_lookup()'s own increment of the object.

By using RCU and putting a synchronization point after we remove the
mapping from the idr, but before it can be released we achieve this -
with the benefit of not having to hold the mutex in qrtr_port_lookup().
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Move operations on the qrtr_nodes radix tree under a separate spinlock
and make the qrtr_nodes tree GFP_ATOMIC, to allow operation from atomic
context in a subsequent patch.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

In order to prevent overconsumption of resources on the remote side QRTR
implements a flow control mechanism.

The mechanism works by the sender keeping track of the number of
outstanding unconfirmed messages that has been transmitted to a
particular node/port pair.

Upon count reaching a low watermark (L) the confirm_rx bit is set in the
outgoing message and when the count reaching a high watermark (H)
transmission will be blocked upon the reception of a resume_tx message
from the remote, that resets the counter to 0.

This guarantees that there will be at most 2H - L messages in flight.
Values chosen for L and H are 5 and 10 respectively.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

The confirm-rx bit is used to implement a per port flow control, in
order to make sure that no messages are dropped due to resource
exhaustion. Move the resume-tx transmission to recvmsg to only confirm
messages as they are consumed by the application.
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pktgen can use only one IPv6 source address from output device or src6
command setting. In pressure test we need create lots of sessions more
than 65535. So add src6_min and src6_max command to set the range.
Signed-off-by: Niu Xilei <niu_xilei@163.com>

Changes since v3:
 - function set_src_in6_addr use static instead of static inline
 - precompute min_in6_l,min_in6_h,max_in6_h,max_in6_l in setup time
Changes since v2:
 - reword subject line
Changes since v1:
 - only create IPv6 source address over least significant 64 bit range
Signed-off-by: David S. Miller <davem@davemloft.net>

the i2c_add_driver will set the .owner to THIS_MODULE
Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jason A. Donenfeld says:

====================
skb_list_walk_safe refactoring for net/*'s skb_gso_segment usage

This patchset adjusts all return values of skb_gso_segment in net/* to
use the new skb_list_walk_safe helper.

First we fix a minor bug in the helper macro that didn't come up in the
last patchset's uses. Then we adjust several cases throughout net/. The
xfrm changes were a bit hairy, but doable. Reading and thinking about
the code in mac80211 indicates a memory leak, which the commit
addresses. All the other cases were pretty trivial.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a conversion case for the new function, keeping the flow of the
existing code as intact as possible. We also switch over to using
skb_mark_not_on_list instead of a null write to skb->next.

Finally, this code appeared to have a memory leak in the case where
header building fails before the last gso segment. In that case, the
remaining segments are not freed. So this commit also adds the proper
kfree_skb_list call for the remainder of the skbs.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a straight-forward conversion case for the new function, keeping
the flow of the existing code as intact as possible.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is converts xfrm segment iteration to use the new function, keeping
the flow of the existing code as intact as possible. One case is very
straight-forward, whereas the other case has some more subtle code that
likes to peak at ->next and relink skbs. By keeping the variables the
same as before, we can upgrade this code with minimal surgery required.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This is a straight-forward conversion case for the new function,
iterating over the return value from udp_rcv_segment, which actually is
a wrapper around skb_gso_segment.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This worked before, because we made all callers name their next pointer
"next". But in trying to be more "drop-in" ready, the silliness here is
revealed. This commit fixes the problem by making the macro argument and
the member use different names.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Antoine Tenart says:

====================
net: macsec: initial support for hardware offloading

This series intends to add support for offloading MACsec transformations
to hardware enabled devices. The series adds the necessary
infrastructure for offloading MACsec configurations to hardware drivers,
in patches 1 to 5; then introduces MACsec offloading support in the
Microsemi MSCC PHY driver, in patches 6 to 10.

The series can also be found at:
https://github.com/atenart/linux/tree/net-next/macsec

IProute2 modifications can be found at:
https://github.com/atenart/iproute2/tree/macsec

MACsec hardware offloading infrastructure
-----------------------------------------

Linux has a software implementation of the MACsec standard. There are
hardware engines supporting MACsec operations, such as the Intel ixgbe
NIC and some Microsemi PHYs (the one we use in this series). This means
the MACsec offloading infrastructure should support networking PHY and
MAC drivers. Note that MAC driver preliminary support is part of this
series, but should not be merged before we actually have a provider for
this.

We do intend in this series to re-use the logic, netlink API and data
structures of the existing MACsec software implementation. This allows
not to duplicate definitions and structure storing the same information;
as well as using the same userspace tools to configure both software or
hardware offloaded MACsec flows (with `ip macsec`).

When adding a new MACsec virtual interface the existing logic is kept:
offloading is disabled by default. A user driven configuration choice is
needed to switch to offloading mode (a patch in iproute2 is needed for
this). A single MACsec interface can be offloaded for now, and some
limitations are there: no flow can be moved from one implementation to
the other so the decision needs to be done before configuring the
interface.

MACsec offloading ops are called in 2 steps: a preparation one, and a
commit one. The first step is allowed to fail and should be used to
check if a provided configuration is compatible with a given MACsec
capable hardware. The second step is not allowed to fail and should
only be used to enable a given MACsec configuration.

A limitation as of now is the counters and statistics are not reported
back from the hardware to the software MACsec implementation. This
isn't an issue when using offloaded MACsec transformations, but it
should be added in the future so that the MACsec state can be reported
to the user (which would also improve the debug).

Microsemi PHY MACsec support
----------------------------

In order to add support for the MACsec offloading feature in the
Microsemi MSCC PHY driver, the __phy_read_page and __phy_write_page
helpers had to be exported. This is because the initialization of the
PHY is done while holding the MDIO bus lock, and we need to change the
page to configure the MACsec block.

The support itself is then added in three patches. The first one adds
support for configuring the MACsec block within the PHY, so that it is
up, running and available for future configuration, but is not doing any
modification on the traffic passing through the PHY. The second patch
implements the phy_device MACsec ops in the Microsemi MSCC PHY driver,
and introduce helpers to configure MACsec transformations and flows to
match specific packets. The last one adds support for PN rollover.

Thanks!
Antoine

Since v5:
  - Fixed a compilation issue due to an inclusion from an UAPI header.
  - Added an EXPORT_SYMBOL_GPL for the PN rollover helper, to fix module
    compilation issues.
  - Added a dependency for the MSCC driver on MACSEC || MACSEC=n.
  - Removed the patches including the MAC offloading support as they are
    not to be applied for now.

Since v4:
  - Reworked the MACsec read and write functions in the MSCC PHY driver
    to remove the conditional locking.

Since v3:
  - Fixed a check when enabling offloading that was too restrictive.
  - Fixed the propagation of the changelink event to the underlying
    device drivers.

Since v2:
  - Allow selection the offloading from userspace, defaulting to the
    software implementation when adding a new MACsec interface. The
    offloading mode is now also reported through netlink.
  - Added support for letting MKA packets in and out when using MACsec
    (there are rules to let them bypass the MACsec h/w engine within the
    PHY).
  - Added support for PN rollover (following what's currently done in
    the software implementation: the flow is disabled).
  - Split patches to remove MAC offloading support for now, as there are
    no current provider for this (patches are still included).
  - Improved a few parts of the MACsec support within the MSCC PHY
    driver (e.g. default rules now block non-MACsec traffic, depending
    on the configuration).
  - Many cosmetic fixes & small improvements.

Since v1:
  - Reworked the MACsec offloading API, moving from a single helper
    called for all MACsec configuration operations, to a per-operation
    function that is provided by the underlying hardware drivers.
  - Those functions now contain a verb to describe the configuration
    action they're offloading.
  - Improved the error handling in the MACsec genl helpers to revert
    the configuration to its previous state when the offloading call
    failed.
  - Reworked the file inclusions.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for handling MACsec PN rollover in the mscc PHY
driver. When a flow rolls over, an interrupt is fired. This patch adds
the logic to check all flows and identify the one rolling over in the
handle_interrupt PHY helper, then disables the flow and report the event
to the MACsec core.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Allow to call macsec_pn_wrapped from hardware drivers to notify when a
PN rolls over. Some drivers might used an interrupt to implement this.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds MACsec offloading support to some Microsemi PHYs, to
configure flows and transformations so that matched packets can be
processed by the MACsec engine, either at egress, or at ingress.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds support for initializing the MACsec engine found within
some Microsemi PHYs. The engine is initialized in a passthrough mode and
does not modify any incoming or outgoing packet. But thanks to this it
now can be configured to perform MACsec transformations on packets,
which will be supported by a future patch.

The MACsec read and write functions are wrapped into two versions: one
called during the init phase, and the other one later on. This is
because the init functions in the Microsemi PHY driver are called while
the MDIO bus lock is taken.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MACsec offloading to underlying hardware devices is disabled by default
(the software implementation is used). This patch adds support for
changing this setting through the MACsec netlink interface. Many checks
are done when enabling offloading on a given MACsec interface as there
are limitations (it must be supported by the hardware, only a single
interface can be offloaded on a given physical device at a time, rules
can't be moved for now).
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces the MACsec hardware offloading infrastructure.

The main idea here is to re-use the logic and data structures of the
software MACsec implementation. This allows not to duplicate definitions
and structure storing the same kind of information. It also allows to
use a unified genlink interface for both MACsec implementations (so that
the same userspace tool, `ip macsec`, is used with the same arguments).
The MACsec offloading support cannot be disabled if an interface
supports it at the moment.

The MACsec configuration is passed to device drivers supporting it
through macsec_ops which are called from the MACsec genl helpers. Those
functions call the macsec ops of PHY and Ethernet drivers in two steps:
a preparation one, and a commit one. The first step is allowed to fail
and should be used to check if a provided configuration is compatible
with the features provided by a MACsec engine, while the second step is
not allowed to fail and should only be used to enable a given MACsec
configuration. Two extra calls are made: when a virtual MACsec interface
is created and when it is deleted, so that the hardware driver can stay
in sync.

The Rx and TX handlers are modified to take in account the special case
were the MACsec transformation happens in the hardware, whether in a PHY
or in a MAC, as the packets seen by the networking stack on both the
physical and MACsec virtual interface are exactly the same. This leads
to some limitations: the hardware and software implementations can't be
used on the same physical interface, as the policies would be impossible
to fulfill (such as strict validation of the frames). Also only a single
virtual MACsec interface can be offloaded to a physical port supporting
hardware offloading as it would be impossible to guess onto which
interface a given packet should go (for ingress traffic).

Another limitation as of now is that the counters and statistics are not
reported back from the hardware to the software MACsec implementation.
This isn't an issue when using offloaded MACsec transformations, but it
should be added in the future so that the MACsec state can be reported
to the user (which would also improve the debug).
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds a reference to MACsec ops in the phy_device, to allow
PHYs to support offloading MACsec operations. The phydev lock will be
held while calling those helpers.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces MACsec ops for drivers to support offloading
MACsec operations.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces the macsec_context structure. It will be used
in the kernel to exchange information between the common MACsec
implementation (macsec.c) and the MACsec hardware offloading
implementations. This structure contains pointers to MACsec specific
structures which contain the actual MACsec configuration, and to the
underlying device (phydev for now).
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch moves some structure, type and identifier definitions into a
MACsec specific header. This patch does not modify how the MACsec code
is running and only move things around. This is a preparation for the
future MACsec hardware offloading support, which will re-use those
definitions outside macsec.c.
Signed-off-by: Antoine Tenart <antoine.tenart@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Guillaume Nault says:

====================
netns: Optimise netns ID lookups

Netns ID lookups can be easily protected by RCU, rather than by holding
a spinlock.

Patch 1 prepares the code, patch 2 does the RCU conversion, and finally
patch 3 stops disabling BHs on updates (patch 2 makes that unnecessary).
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

When peernet2id() had to lock "nsid_lock" before iterating through the
nsid table, we had to disable BHs, because VXLAN can call peernet2id()
from the xmit path:
  vxlan_xmit() -> vxlan_fdb_miss() -> vxlan_fdb_notify()
    -> __vxlan_fdb_notify() -> vxlan_fdb_info() -> peernet2id().

Now that peernet2id() uses RCU protection, "nsid_lock" isn't used in BH
context anymore. Therefore, we can safely use plain
spin_lock()/spin_unlock() and let BHs run when holding "nsid_lock".
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>