platform/kernel/linux-rpi.git
8 years agonet: bpf: reject invalid shifts
Rabin Vincent [Tue, 12 Jan 2016 19:17:08 +0000 (20:17 +0100)]
net: bpf: reject invalid shifts

On ARM64, a BUG() is triggered in the eBPF JIT if a filter with a
constant shift that can't be encoded in the immediate field of the
UBFM/SBFM instructions is passed to the JIT.  Since these shifts
amounts, which are negative or >= regsize, are invalid, reject them in
the eBPF verifier and the classic BPF filter checker, for all
architectures.

Signed-off-by: Rabin Vincent <rabin@rab.in>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophonet: properly unshare skbs in phonet_rcv()
Eric Dumazet [Tue, 12 Jan 2016 16:58:00 +0000 (08:58 -0800)]
phonet: properly unshare skbs in phonet_rcv()

Ivaylo Dimitrov reported a regression caused by commit 7866a621043f
("dev: add per net_device packet type chains").

skb->dev becomes NULL and we crash in __netif_receive_skb_core().

Before above commit, different kind of bugs or corruptions could happen
without major crash.

But the root cause is that phonet_rcv() can queue skb without checking
if skb is shared or not.

Many thanks to Ivaylo Dimitrov for his help, diagnosis and tests.

Reported-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com>
Tested-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Remi Denis-Courmont <courmisch@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodwc_eth_qos: Fix dma address for multi-fragment skbs
Lars Persson [Tue, 12 Jan 2016 14:28:13 +0000 (15:28 +0100)]
dwc_eth_qos: Fix dma address for multi-fragment skbs

The offset inside the fragment was not used for the dma address and
silent data corruption resulted because TSO makes the checksum match.

Fixes: 077742dac2c7 ("dwc_eth_qos: Add support for Synopsys DWC Ethernet QoS")
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: remove an unneeded condition
Dan Carpenter [Tue, 12 Jan 2016 09:36:21 +0000 (12:36 +0300)]
phy: remove an unneeded condition

It used to be that bus->irq was a pointer but after e7f4dc3536a4
('mdio: Move allocation of interrupts into core') it's an array inside
the mdio struct, so it can never be NULL.  Let's remove the check.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomdio: remove an unneed condition
Dan Carpenter [Tue, 12 Jan 2016 09:35:34 +0000 (12:35 +0300)]
mdio: remove an unneed condition

It used to be that mdio->irq was a pointer but after e7f4dc3536a4
('mdio: Move allocation of interrupts into core') it's an array inside
the mdio struct so it can never be NULL.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomdio_bus: NULL dereference on allocation error
Dan Carpenter [Tue, 12 Jan 2016 09:34:36 +0000 (12:34 +0300)]
mdio_bus: NULL dereference on allocation error

If bus = kzalloc() fails then we end up dereferencing bus when we do
"bus->irq[i] = PHY_POLL;".  The code is a little simpler if we reverse
the NULL check and return directly on failure.

Fixes: e7f4dc3536a4 ('mdio: Move allocation of interrupts into core')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Tue, 12 Jan 2016 04:55:43 +0000 (23:55 -0500)]
Merge git://git./linux/kernel/git/davem/net

Conflicts:
drivers/net/bonding/bond_main.c
drivers/net/ethernet/mellanox/mlxsw/spectrum.h
drivers/net/ethernet/mellanox/mlxsw/spectrum_switchdev.c

The bond_main.c and mellanox switch conflicts were cases of
overlapping changes.

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: Fix typo in netdev_intersect_features
Tom Herbert [Mon, 11 Jan 2016 18:19:10 +0000 (10:19 -0800)]
net: Fix typo in netdev_intersect_features

Obviously need to 'or in NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM.

Fixes: c8cd0989bd151f ("net: Eliminate NETIF_F_GEN_CSUM and NETIF_F_V[46]_CSUM")
Reported-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mdio-build-failures'
David S. Miller [Tue, 12 Jan 2016 04:27:26 +0000 (23:27 -0500)]
Merge branch 'mdio-build-failures'

Andrew Lunn says:

====================
More mdio device build failure fixes

These patches fix two build errors reported by Guenter Roeck
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: freescale: mac-fec: Fix build error from phy_device API change
Andrew Lunn [Tue, 12 Jan 2016 02:24:31 +0000 (03:24 +0100)]
net: freescale: mac-fec: Fix build error from phy_device API change

dev has moved inside the new mdio structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: freescale: ucc_geth: Fix build error from phy_device API change
Andrew Lunn [Tue, 12 Jan 2016 02:24:30 +0000 (03:24 +0100)]
net: freescale: ucc_geth: Fix build error from phy_device API change

dev has moved inside the new mdio structure.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobonding: Prevent IPv6 link local address on enslaved devices
Karl Heiss [Mon, 11 Jan 2016 13:28:43 +0000 (08:28 -0500)]
bonding: Prevent IPv6 link local address on enslaved devices

Commit 1f718f0f4f97 ("bonding: populate neighbour's private on enslave")
undoes the fix provided by commit c2edacf80e15 ("bonding / ipv6: no addrconf
for slaves separately from master") by effectively setting the slave flag
after the slave has been opened.  If the slave comes up quickly enough, it
will go through the IPv6 addrconf before the slave flag has been set and
will get a link local IPv6 address.

In order to ensure that addrconf knows to ignore the slave devices on state
change, set IFF_SLAVE before dev_open() during bonding enslavement.

Fixes: 1f718f0f4f97 ("bonding: populate neighbour's private on enslave")
Signed-off-by: Karl Heiss <kheiss@gmail.com>
Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Reviewed-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: Andy Gospodarek <gospo@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx5-enhanced-flow-steering'
David S. Miller [Mon, 11 Jan 2016 22:48:54 +0000 (17:48 -0500)]
Merge branch 'mlx5-enhanced-flow-steering'

Or Gerlitz says:

====================
net/mlx5_core: Enhance flow steering support

v0 --> v1 changes:
  - fixed improperly formatted comments.
  - compare value of ib_spec->eth.mask.ether_type in network byte order
     in ('IB/mlx5: Add flow steering utilities').

v1 --> v2 changes:
  - made sure that service functions added in the IB driver are only static-fied
    on the last commit, to make sure bisection with -Werror works fine.

v2 --> v3 changes:
   - squashed patches 11 and 12 into one patch, s.t Dave's comment
     on unused static functions gcc complaints during bisection is
     correctly addressed.

v3 has been generated against net-next commit c9c9931 "Merge tag
'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge"

The series is signed by Matan who was revently assigned to a maintainer for
the mlx5_core and IB drivers (this is a 4.5-rc1 change to the maintainers file coming
from the rdma tree) -- as such I didn't see a neeed to add my signature (Or).

This series adds three new functionalists to the driver flow-steering
infrastructure: auto-grouped flow tables, chaining of flow tables and
updates for the root flow table.

1. Auto-grouped flow tables - Flow table with auto grouping management.
When a flow table is created, hints regarding the number of rule types
and the number of rules are given in advance. Thus, a flow table is
divided into #NUM_TYPES+1 groups each contains
(#NUM_RULES)/(#NUM_TYPES+1) rules. The first #NUM_TYPES parts are groups
which are filled if the added rule matches the group specification or
the group is empty. The last part is filled by rules that can't fit
any of the former groups.

2. Chaining flow tables - Flow tables from different priorities are chained
together, if there is no match in flow table of priority i we continue
searching for a match in priority i+1. This is both true if priorities
i and i+1 belongs to the same namespace or not.

3. Updating the root flow table - the root flow table is the flow table
with the lowest level. The hardware start searching for a match in the
root flow table and continue according to the matches it find along
the way.

The first usage for the new functionality is flow steering for user-space
ConnectX-4 offloaded HW Eth RX queues done through the mlx5 IB driver.

When the mlx5 core driver is loaded, it opens three flow namespaces:
1. By-pass namespace (used by mlx5 IB driver).
2. Kernel namespace (used in order to get packets to the networking stack
through mlx5 EN driver).
3. Leftovers namespace (used by mlx5 IB and future sniffer)

The series is built as follows:

Patch #1 introduces auto-grouped flow tables support.

Patch #2 add utility functions for finding the next and the previous
flow tables in different priorities. This is used in order to chain
the flow tables in a downstream patch.

Patch #3 introduces a firmware command for updating the root flow table.

Patch #4 introduces modify flow table firmware command, this command is used
when we want to change the next flow table of an existing flow table.
This is used for chaining flow tables as well.

Patch #5 connect/disconnect flow tables. This is actually the chaining
process when we want to link flow tables. This means that if we couldn't
find a match in the first flow table, we'll continue in the chained
flow table.

Patch #6 updates priority's attributes that is required for flow table
level allocation. We update both the max_fts (the number of allowed FTs
in the sub-tree of this priority) and the start_level (which is the first
level we'll assign to the flow-tables created inside the priority).

Patch #7 adds checking of required device capabilities. Some namespaces
could be only created if the hardware supports certain attributes.
This is especially true for the Bypass and leftovers namespaces. This
adds a generic mechanism to check these required attributes.

Patch #8 creates two additional namespaces:
a. Bypass flow rules(has nine priorities)
b. Leftovers packets(have one priority) - for unmatched packets.

Patch #9 re-factors ipv4/ipv6 match fields in the mlx5 firmware interface
header to be more clear.

Patch #10 exports the flow steering API for mlx5_ib usage

Patch #11 implements the required support in mlx5_ib in order
to support the RDMA flow steering verbs.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoIB/mlx5: Add flow steering support
Maor Gottlieb [Mon, 11 Jan 2016 08:26:07 +0000 (10:26 +0200)]
IB/mlx5: Add flow steering support

Adding flow steering support by creating a flow-table per
priority (if rules exist in the priority). mlx5_ib uses
autogrouping and thus only creates the required destinations.

Also includes adding of these flow steering utilities

1. Parsing verbs flow attributes hardware steering specs.

2. Check if flow is multicast - this is required in order to decide
to which flow table will we add the steering rule.

3. Set outer headers in flow match criteria to zeros.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Export flow steering API
Maor Gottlieb [Mon, 11 Jan 2016 08:26:06 +0000 (10:26 +0200)]
net/mlx5_core: Export flow steering API

Add exports to flow steering API for mlx5_ib usage.
The following functions are exported:

1. mlx5_create_auto_grouped_flow_table - used to create flow
table with auto flow grouping management (create and destroy
flow groups). In auto-grouped flow tables, we create groups
automatically if needed (if we don't find an existing
flow group with same match criteria when we add new rule).

2. mlx5_destroy_flow_table - used to destroy  a flow table.

3. mlx5_add_flow_rule - used to add flow rule into a flow table.

4. mlx5_del_flow_rule - used to delete flow rule from its flow table.

5. mlx5_get_flow_namespace - used to get a handle to the required
namespace sub-tree.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Make ipv4/ipv6 location more clear
Maor Gottlieb [Mon, 11 Jan 2016 08:26:05 +0000 (10:26 +0200)]
net/mlx5_core: Make ipv4/ipv6 location more clear

Change the mlx5 firmware interface header to make it
more clear which bytes should be used by IPv4 or
IPv6 addresses.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Enable flow steering support for the IB driver
Maor Gottlieb [Mon, 11 Jan 2016 08:26:04 +0000 (10:26 +0200)]
net/mlx5_core: Enable flow steering support for the IB driver

When the driver is loaded, we create flow steering namespace
for kernel bypass with nine priorities and another namespace
for leftovers(in order to catch packets that weren't matched).
Verbs applications will use these priorities.
we found nine as a number that balances the requirements from the
user and retains performance.

The bypass namespace is used by verbs applications that want to bypass
the kernel networking stack. The leftovers namespace is used by verbs
applications and the sniffer in order to catch packets that weren't
handled by any preceding rules.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Initialize namespaces only when supported by device
Maor Gottlieb [Mon, 11 Jan 2016 08:26:03 +0000 (10:26 +0200)]
net/mlx5_core: Initialize namespaces only when supported by device

Before we create the sub tree of a steering namespaces(kernel, bypass,
leftovers) we check that the device has the required capabilities
in order to create this subtree.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Set priority attributes
Maor Gottlieb [Mon, 11 Jan 2016 08:26:02 +0000 (10:26 +0200)]
net/mlx5_core: Set priority attributes

Each priority has two attributes:
1. max_ft - maximum allowed flow tables under this priority.
2. start_level - start level range of the flow tables
in the priority.

These attributes are set by traversing the tree nodes by
DFS and set start level and max flow tables to each priority.
Start level depends on the max flow tables of the prior priorities
in the tree.

The leaves of the trees have max_ft set in them. Each node accumulates
the max_ft of its children and set it accordingly.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Connect flow tables
Maor Gottlieb [Mon, 11 Jan 2016 08:26:01 +0000 (10:26 +0200)]
net/mlx5_core: Connect flow tables

Flow tables from different priorities should be chained together.
When a packet arrives we search for a match in the
by-pass flow tables (first we search for a match in priority 0
and if we don't find a match we move to the next priority).
If we can't find a match in any of the bypass flow-tables, we continue
searching in the flow-tables of the next priority, which are the
kernel's flow tables.

Setting the miss flow table in a new flow table to be the next one in
the list is performed via create flow table API. If we want to change an
existing flow table, for example in order to point from an
existing flow table to the new next-in-list flow table, we use the
modify flow table API.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Introduce modify flow table command
Maor Gottlieb [Mon, 11 Jan 2016 08:26:00 +0000 (10:26 +0200)]
net/mlx5_core: Introduce modify flow table command

Introduce the modify flow table command. This command is used when
we want to change the next flow table of an existing flow table.
The next flow table is defined as the table we search (in order
to find a match), if we couldn't find a match in any of the flow table
entries in the current flow table.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Managing root flow table
Maor Gottlieb [Mon, 11 Jan 2016 08:25:59 +0000 (10:25 +0200)]
net/mlx5_core: Managing root flow table

The root Flow Table for each Flow Table Type is defined,
by default, as the Flow Table with level 0.

In order not to use an empty flow tables and introduce new hops,
but still preserve space for flow-tables that have a priority
greater(lower number) than the current flow table, we introduce this
new set root flow table command.
This command tells the HW to start matching packets from the
assigned root flow table.
This command is used when we create new flow table with level lower than the
current lowest flow table or it is the first flow table.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Add utilities to find next and prev flow-tables
Maor Gottlieb [Mon, 11 Jan 2016 08:25:58 +0000 (10:25 +0200)]
net/mlx5_core: Add utilities to find next and prev flow-tables

Add two utility functions for find next and prev flow table.
Find next flow table function gets priority and return the
first flow table of the next priority in the tree.
Find prev flow table return the last flow table of
the previous priority in the tree.

These utility functions are used for chaining flow table from different
priorities.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Introduce flow steering autogrouped flow table
Maor Gottlieb [Mon, 11 Jan 2016 08:25:57 +0000 (10:25 +0200)]
net/mlx5_core: Introduce flow steering autogrouped flow table

When user add rule to autogrouped flow table, we search
for flow group with the same match criteria, if we don't
find such group then we create new flow group with the
required match criteria and insert the rule to this group.

We divide the flow table into required_groups + 1,
in order to reserve a part of the flow table for rules
which don't match any existing group.

Signed-off-by: Maor Gottlieb <maorg@mellanox.com>
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoudp: disallow UFO for sockets with SO_NO_CHECK option
Michal Kubeček [Mon, 11 Jan 2016 06:50:30 +0000 (07:50 +0100)]
udp: disallow UFO for sockets with SO_NO_CHECK option

Commit acf8dd0a9d0b ("udp: only allow UFO for packets from SOCK_DGRAM
sockets") disallows UFO for packets sent from raw sockets. We need to do
the same also for SOCK_DGRAM sockets with SO_NO_CHECK options, even if
for a bit different reason: while such socket would override the
CHECKSUM_PARTIAL set by ip_ufo_append_data(), gso_size is still set and
bad offloading flags warning is triggered in __skb_gso_segment().

In the IPv6 case, SO_NO_CHECK option is ignored but we need to disallow
UFO for packets sent by sockets with UDP_NO_CHECK6_TX option.

Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Tested-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: pktgen: fix null ptr deref in skb allocation
John Fastabend [Mon, 11 Jan 2016 05:38:44 +0000 (21:38 -0800)]
net: pktgen: fix null ptr deref in skb allocation

Fix possible null pointer dereference that may occur when calling
skb_reserve() on a null skb.

Fixes: 879c7220e82 ("net: pktgen: Observe needed_headroom of the device")
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bpf-next'
David S. Miller [Mon, 11 Jan 2016 22:32:56 +0000 (17:32 -0500)]
Merge branch 'bpf-next'

Daniel Borkmann says:

====================
BPF update

This set adds IPv6 support for bpf_skb_{set,get}_tunnel_key() helper.
It also exports flags to user space that are being used in helpers and
weren't exported thus far. For more details, please see the individual
patches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: support ipv6 for bpf_skb_{set,get}_tunnel_key
Daniel Borkmann [Mon, 11 Jan 2016 00:16:39 +0000 (01:16 +0100)]
bpf: support ipv6 for bpf_skb_{set,get}_tunnel_key

After IPv6 support has recently been added to metadata dst and related
encaps, add support for populating/reading it from an eBPF program.

Commit d3aa45ce6b ("bpf: add helpers to access tunnel metadata") started
with initial IPv4-only support back then (due to IPv6 metadata support
not being available yet).

To stay compatible with older programs, we need to test for the passed
structure size. Also TOS and TTL support from the ip_tunnel_info key has
been added. Tested with vxlan devs in collect meta data mode with IPv4,
IPv6 and in compat mode over different network namespaces.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: export helper function flags and reject invalid ones
Daniel Borkmann [Mon, 11 Jan 2016 00:16:38 +0000 (01:16 +0100)]
bpf: export helper function flags and reject invalid ones

Export flags used by eBPF helper functions through UAPI, so they can be
used by programs (instead of them redefining all flags each time or just
using the hard-coded values). It also gives a better overview what flags
are used where and we can further get rid of the extra macros defined in
filter.c. Moreover, reject invalid flags.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'renesas-eth-fixes'
David S. Miller [Mon, 11 Jan 2016 22:31:10 +0000 (17:31 -0500)]
Merge branch 'renesas-eth-fixes'

Sergei Shtylyov says:

====================
Fix some dubious code in the Renesas Ethernet drivers

   Here's a set of 2 patches against DaveM's 'net.git' repo. While initializing
EMAC the code tries to respect the duplex mode both programmed into ECMR and
stored in its own private data -- this just can't be right.

[1/2] ravb: stop reading ECMR in ravb_emac_init()
[2/2] sh_eth: stop reading ECMR in sh_eth_dev_init()
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosh_eth: stop reading ECMR in sh_eth_dev_init()
Sergei Shtylyov [Sun, 10 Jan 2016 21:28:14 +0000 (00:28 +0300)]
sh_eth: stop reading ECMR in sh_eth_dev_init()

The code in sh_eth_dev_init()  twiddling the ECMR bits always looked a bit
strange to me:  if one intends to respect 'mdp->duplex', why save old value
of the ECMR.DM bit? As all the other bits are zeroed anyway, we don't really
need to read ECMR before writing to it.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoravb: stop reading ECMR in ravb_emac_init()
Sergei Shtylyov [Sun, 10 Jan 2016 21:27:38 +0000 (00:27 +0300)]
ravb: stop reading ECMR in ravb_emac_init()

The code in ravb_emac_init() twiddling the ECMR bits always looked a bit
strange to me: if one intends to respect 'priv->duplex', why save old value
of the ECMR.DM bit?   As all the other bits are zeroed anyway, we don't
really need to read ECMR before writing to it.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosched,cls_flower: set key address type when present
Jamal Hadi Salim [Sun, 10 Jan 2016 16:47:01 +0000 (11:47 -0500)]
sched,cls_flower: set key address type when present

only when user space passes the addresses should we consider their
presence

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotcp_yeah: don't set ssthresh below 2
Neal Cardwell [Mon, 11 Jan 2016 18:42:43 +0000 (13:42 -0500)]
tcp_yeah: don't set ssthresh below 2

For tcp_yeah, use an ssthresh floor of 2, the same floor used by Reno
and CUBIC, per RFC 5681 (equation 4).

tcp_yeah_ssthresh() was sometimes returning a 0 or negative ssthresh
value if the intended reduction is as big or bigger than the current
cwnd. Congestion control modules should never return a zero or
negative ssthresh. A zero ssthresh generally results in a zero cwnd,
causing the connection to stall. A negative ssthresh value will be
interpreted as a u32 and will set a target cwnd for PRR near 4
billion.

Oleksandr Natalenko reported that a system using tcp_yeah with ECN
could see a warning about a prior_cwnd of 0 in
tcp_cwnd_reduction(). Testing verified that this was due to
tcp_yeah_ssthresh() misbehaving in this way.

Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobonding: make mii_status sysfs node consistent
Jarod Wilson [Sat, 9 Jan 2016 01:35:36 +0000 (20:35 -0500)]
bonding: make mii_status sysfs node consistent

The spew in /proc/net/bonding/bond0 uses netif_carrier_ok() to determine
mii_status, while /sys/class/net/bond0/bonding/mii_status looks at
curr_active_slave, which doesn't actually seem to be set sometimes when
the bond actually is up. A mode 4 bond configured via ifcfg-foo files on a
Red Hat Enterprise Linux system, after boot, comes up clean and
functional, but the sysfs node shows mii_status of down, while proc shows
up. A simple enough fix here seems to be to use the same method for
determining up or down in both places, and I'd opt for the one that seems
to match reality.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <gospo@cumulusnetworks.com>
CC: netdev@vger.kernel.org
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosctp: fix use-after-free in pr_debug statement
Marcelo Ricardo Leitner [Fri, 8 Jan 2016 13:00:54 +0000 (11:00 -0200)]
sctp: fix use-after-free in pr_debug statement

Dmitry Vyukov reported a use-after-free in the code expanded by the
macro debug_post_sfx, which is caused by the use of the asoc pointer
after it was freed within sctp_side_effect() scope.

This patch fixes it by allowing sctp_side_effect to clear that asoc
pointer when the TCB is freed.

As Vlad explained, we also have to cover the SCTP_DISPOSITION_ABORT case
because it will trigger DELETE_TCB too on that same loop.

Also, there were places issuing SCTP_CMD_INIT_FAILED and ASSOC_FAILED
but returning SCTP_DISPOSITION_CONSUME, which would fool the scheme
above. Fix it by returning SCTP_DISPOSITION_ABORT instead.

The macro is already prepared to handle such NULL pointer.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/rtnetlink: remove unused sz_idx variable
Alexander Kuleshov [Sun, 10 Jan 2016 15:26:57 +0000 (21:26 +0600)]
net/rtnetlink: remove unused sz_idx variable

The sz_idx variable is defined in the rtnetlink_rcv_msg(), but
not used anywhere. Let's remove it.

Signed-off-by: Alexander Kuleshov <kuleshovmail@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Add FDB lock to prevent session interleaving
Ido Schimmel [Sun, 10 Jan 2016 08:32:16 +0000 (09:32 +0100)]
mlxsw: spectrum: Add FDB lock to prevent session interleaving

Dumping the FDB (invoked with a process context) or handling FDB
notifications (polled periodicly in delayed work) might each entail
multiple EMAD transcations due to the number of entries.

While we only allow one EMAD transaction at a time, there is nothing
stopping the dump and notification processing sessions from
interleaving. However, this is forbidden by the hardware, so we need to
make sure only one of these sessions can run at a time.

Solve this by adding a mutex ('fdb_lock'), as both kernel threads can
sleep while waiting for the response EMAD.

Fixes: 56ade8fe3f ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agounix: properly account for FDs passed over unix sockets
willy tarreau [Sun, 10 Jan 2016 06:54:56 +0000 (07:54 +0100)]
unix: properly account for FDs passed over unix sockets

It is possible for a process to allocate and accumulate far more FDs than
the process' limit by sending them over a unix socket then closing them
to keep the process' fd count low.

This change addresses this problem by keeping track of the number of FDs
in flight per user and preventing non-privileged processes from having
more FDs in flight than their configured FD limit.

Reported-by: socketpair@gmail.com
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Mitigates: CVE-2013-4312 (Linux 2.0+)
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: bfin_mac: Use phy_find_first() instead of open-coding it
Guenter Roeck [Mon, 11 Jan 2016 04:43:49 +0000 (20:43 -0800)]
net: bfin_mac: Use phy_find_first() instead of open-coding it

Use phy_find_first() to find the first phy device instead of
open-coding it.

Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'ovs-cleanups'
David S. Miller [Mon, 11 Jan 2016 04:49:21 +0000 (23:49 -0500)]
Merge branch 'ovs-cleanups'

Jean Sacren says:

====================
Trivial fix-ups for openvswitch

This series does trivial fix-ups for openvswitch as follows:

1) Clean up the leftover of the unused function.

2) Fix up the twisted struct geneve_port member name.

3) Update the kernel doc to reflect the changes in struct vport.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoopenvswitch: update kernel doc for struct vport
Jean Sacren [Sat, 9 Jan 2016 23:07:11 +0000 (16:07 -0700)]
openvswitch: update kernel doc for struct vport

commit be4ace6e6b1b ("openvswitch: Move dev pointer into vport itself")

The commit above added @dev and moved @rcu to the bottom of struct
vport, but the change was not reflected in the kernel doc. So let's
update the kernel doc as well.

Signed-off-by: Jean Sacren <sakiwit@gmail.com>
Cc: Thomas Graf <tgraf@suug.ch>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoopenvswitch: fix struct geneve_port member name
Jean Sacren [Sat, 9 Jan 2016 23:07:10 +0000 (16:07 -0700)]
openvswitch: fix struct geneve_port member name

commit 6b001e682e90 ("openvswitch: Use Geneve device.")

The commit above introduced 'port_no' as the name for the member of
struct geneve_port. The correct name should be 'dst_port' as described
in the kernel doc. Let's fix that member name and all the pertinent
instances so that both doc and code would be consistent.

Signed-off-by: Jean Sacren <sakiwit@gmail.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoopenvswitch: clean up unused function
Jean Sacren [Sat, 9 Jan 2016 23:07:09 +0000 (16:07 -0700)]
openvswitch: clean up unused function

commit 6b001e682e90 ("openvswitch: Use Geneve device.")

The commit above deleted the only call site of ovs_tunnel_route_lookup()
and now that function is not used any more. So let's delete the function
definition as well.

Signed-off-by: Jean Sacren <sakiwit@gmail.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ti: cpmac: Fix build error due to missed API change
Guenter Roeck [Sat, 9 Jan 2016 21:19:39 +0000 (13:19 -0800)]
net: ti: cpmac: Fix build error due to missed API change

Commit 7f854420fbfe ("phy: Add API for {un}registering an mdio device to
a bus") introduces an API to access mii_bus structures, but missed to
update the TI cpamc driver. This results in the following error message.

drivers/net/ethernet/ti/cpmac.c: In function 'cpmac_probe':
drivers/net/ethernet/ti/cpmac.c:1119:18: error:
'struct mii_bus' has no member named 'phy_map'

Fixes: 7f854420fbfe ("phy: Add API for {un}registering an mdio device to a bus")
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: tc35815: Drop unused variable
Guenter Roeck [Sun, 10 Jan 2016 15:10:45 +0000 (07:10 -0800)]
net: tc35815: Drop unused variable

Commit e7f4dc3536a4 ("mdio: Move allocation of interrupts into core")
removes some code from tc_mii_init(), but does not remove a now unused
variable. This results in the following build warning.

drivers/net/ethernet/toshiba/tc35815.c: In function 'tc_mii_init':
drivers/net/ethernet/toshiba/tc35815.c:670:6: warning: unused variable 'i'

Fixes: e7f4dc3536a4 ("mdio: Move allocation of interrupts into core")
Cc: Andrew Lunn <andrew@lunn.ch>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: tc35815: Fix build error due to missed API change
Guenter Roeck [Sun, 10 Jan 2016 15:10:44 +0000 (07:10 -0800)]
net: tc35815: Fix build error due to missed API change

Commit 7f854420fbfe ("phy: Add API for {un}registering an mdio device to
a bus") introduces an API to access mii_bus structures, but missed to
update the tc35815 driver. This results in the following error message.

drivers/net/ethernet/toshiba/tc35815.c: In function 'tc_mii_probe':
drivers/net/ethernet/toshiba/tc35815.c:617:18: error:
'struct mii_bus' has no member named 'phy_map'
drivers/net/ethernet/toshiba/tc35815.c:623:24: error:
'struct mii_bus' has no member named 'phy_map'

Instead of looping over the list of phy addresses to find a phy chip,
use phy_find_first(). While the intent of the original code was to return
an error if more than one phy was specified, this code path was never
executed because the loop aborted after finding the first phy. The
original code is therefore semantically identical to phy_find_first(),
thus it is simpler and more straightforward to use phy_find_first()
directly.

Fixes: 7f854420fbfe ("phy: Add API for {un}registering an mdio device to a bus")
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: phy: Add support for SMSC LAN8740 PHY
Joshua Henderson [Sat, 9 Jan 2016 11:54:21 +0000 (04:54 -0700)]
net: phy: Add support for SMSC LAN8740 PHY

LAN8740 has a different phy_id than LAN8710/LAN8720.

Signed-off-by: Joshua Henderson <joshua.henderson@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'wireless-drivers-next-for-davem-2016-01-09' of git://git.kernel.org/pub...
David S. Miller [Mon, 11 Jan 2016 04:10:10 +0000 (23:10 -0500)]
Merge tag 'wireless-drivers-next-for-davem-2016-01-09' of git://git./linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
brcmfmac

* query features through firmware command
* ARP offload through inet notifier
* force probe to succeed for debugging purposes
* random mac support for scheduled scan
* support wowl upon net detect

iwlwifi

* bug fixes and improvements for firmware debug system
* advertise support for Rx A-MSDU in A-MPDU
* support -20.ucode
* fix WoWLAN for iwldvm
* preparations towards multiple Rx queues
* platform power improvements for GO mode when no clients are associated
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: tcp: add rcu locking in tcp_v6_send_synack()
Eric Dumazet [Fri, 8 Jan 2016 17:35:51 +0000 (09:35 -0800)]
ipv6: tcp: add rcu locking in tcp_v6_send_synack()

When first SYNACK is sent, we already hold rcu_read_lock(), but this
is not true if a SYNACK is retransmitted, as a timer (soft) interrupt
does not hold rcu_read_lock()

Fixes: 45f6fad84cc30 ("ipv6: add complete rcu protection around np->opt")
Reported-by: Dave Jones <davej@codemonkey.org.uk>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: add scheduling point in recvmmsg/sendmmsg
Eric Dumazet [Fri, 8 Jan 2016 16:37:20 +0000 (08:37 -0800)]
net: add scheduling point in recvmmsg/sendmmsg

Applications often have to reduce number of datagrams
they receive or send per system call to avoid starvation problems.

Really the kernel should take care of this by using cond_resched(),
so that applications can experiment bigger batch sizes.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: always add flag an address that failed DAD with DADFAILED
Lubomir Rintel [Fri, 8 Jan 2016 12:47:23 +0000 (13:47 +0100)]
ipv6: always add flag an address that failed DAD with DADFAILED

The userspace needs to know why is the address being removed so that it can
perhaps obtain a new address.

Without the DADFAILED flag it's impossible to distinguish removal of a
temporary and tentative address due to DAD failure from other reasons (device
removed, manual address removal).

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: lpc_eth: Remove unused variables
Fabio Estevam [Fri, 8 Jan 2016 12:19:16 +0000 (10:19 -0200)]
net: lpc_eth: Remove unused variables

Commit e7f4dc3536a400 ("mdio: Move allocation of interrupts into core")
introduced the following build warnings:

drivers/net/ethernet/nxp/lpc_eth.c: In function 'lpc_mii_init':
drivers/net/ethernet/nxp/lpc_eth.c:865:1: warning: label 'err_out_1' defined but not used [-Wunused-label]
drivers/net/ethernet/nxp/lpc_eth.c:826:20: warning: unused variable 'i' [-Wunused-variable]

Remove the unused variables to fix them.

Reported-by: Olof's autobuilder <build@lixom.net>
Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobfin_mac: fix error path
Sudip Mukherjee [Fri, 8 Jan 2016 11:28:15 +0000 (16:58 +0530)]
bfin_mac: fix error path

While building blackfin defconfig we were getting a build warning:
warning: label 'out_err_irq_alloc' defined but not used.

Commit e7f4dc3536a4 ("mdio: Move allocation of interrupts into core")
removed the label out_err_mdiobus_register but then mistakenly jumped to
out_err_alloc. But it was actually supposed to jump to out_err_irq_alloc.

Fixes: e7f4dc3536a4 ("mdio: Move allocation of interrupts into core")
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agophy: fix blackfin build failure
Sudip Mukherjee [Fri, 8 Jan 2016 11:28:14 +0000 (16:58 +0530)]
phy: fix blackfin build failure

The build of blackfin defconfig is failing with the error:
error: 'struct mii_bus' has no member named 'phy_map'

A new API mdiobus_get_phy() was introduced and phy_map was removed but
it was not changed here.

Fixes: 7f854420fbfe ("phy: Add API for {un}registering an mdio device to a bus.")
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocxgb4: Fixes static checker warning in mps_tcam_show()
Hariprasad Shenai [Fri, 8 Jan 2016 06:51:25 +0000 (12:21 +0530)]
cxgb4: Fixes static checker warning in mps_tcam_show()

The commit 115b56af88b5 ("cxgb4: Update mps_tcam output to include T6
fields") from Dec 23, 2015, leads to the following static checker
warning:

        drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c:1735
mps_tcam_show()
        warn: we tested 'lookup_type' before and it was 'true'

Fixing it.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'emac-RK3036'
David S. Miller [Mon, 11 Jan 2016 03:21:32 +0000 (22:21 -0500)]
Merge branch 'emac-RK3036'

Xing Zheng says:

====================
Add support emac for the RK3036 SoC platform

  We have supported the emac for RK3066/RK3188, but the RK3036 have
some configuration different with them. We should let the driver of
emac_rockchip compatible with other Rockchip SoCs.

Changes in v2:
- Separate DTS from patch series.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: arc: Add support emac for RK3036
Xing Zheng [Fri, 8 Jan 2016 01:35:02 +0000 (09:35 +0800)]
net: ethernet: arc: Add support emac for RK3036

The RK3036's GRFs offset are different with RK3066/RK3188, and need to set
mac TX/RX clock before probe emac.

Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: arc: Keep emac compatibility for more Rockchip SoCs
Xing Zheng [Fri, 8 Jan 2016 01:35:01 +0000 (09:35 +0800)]
net: ethernet: arc: Keep emac compatibility for more Rockchip SoCs

On the RK3066/RK3188, there was fixed GRF offset configuration to set emac
and fixed DIV2 mac TX/RX clock. So, we need to easily set and fit to other
SoCs (RK3036) which maybe have different GRF offset, and need adjust mac
TX/RX clock.

Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: arc: Probe emac after set RMII clock
Xing Zheng [Fri, 8 Jan 2016 01:35:00 +0000 (09:35 +0800)]
net: ethernet: arc: Probe emac after set RMII clock

After enter arc_emac_probe, emac will get_phy_id, phy_poll_reset and
other connecting PHY via mdiobus_read, so we need to set correct
ref clock rate for emac before probe emac.

Signed-off-by: Xing Zheng <zhengxing@rock-chips.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bnxt_en-zeropad-fw-and-reset'
David S. Miller [Mon, 11 Jan 2016 03:19:19 +0000 (22:19 -0500)]
Merge branch 'bnxt_en-zeropad-fw-and-reset'

Michael Chan says:

====================
bnxt_en: Zero pad fw messages and add fw reset.

2 patches related to firmware for net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnxt_en: Reset embedded processor after applying firmware upgrade
Rob Swindell [Fri, 8 Jan 2016 00:56:58 +0000 (19:56 -0500)]
bnxt_en: Reset embedded processor after applying firmware upgrade

Use HWRM_FW_RESET command to request a self-reset of the embedded
processor(s) after successfully applying a firmware update. For boot
processor, the self-reset is currently deferred until the next PCIe reset.

Signed-off-by: Rob Swindell <swindell@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobnxt_en: Zero pad firmware messages to 128 bytes.
Michael Chan [Fri, 8 Jan 2016 00:56:57 +0000 (19:56 -0500)]
bnxt_en: Zero pad firmware messages to 128 bytes.

For future compatibility, zero pad all messages that the driver sends
to the firmware to 128 bytes.  If these messages are extended in the
future with new byte enables, zero padding these messages now will
guarantee future compatibility.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet, sched: add clsact qdisc
Daniel Borkmann [Thu, 7 Jan 2016 21:29:47 +0000 (22:29 +0100)]
net, sched: add clsact qdisc

This work adds a generalization of the ingress qdisc as a qdisc holding
only classifiers. The clsact qdisc works on ingress, but also on egress.
In both cases, it's execution happens without taking the qdisc lock, and
the main difference for the egress part compared to prior version of [1]
is that this can be applied with _any_ underlying real egress qdisc (also
classless ones).

Besides solving the use-case of [1], that is, allowing for more programmability
on assigning skb->priority for the mqprio case that is supported by most
popular 10G+ NICs, it also opens up a lot more flexibility for other tc
applications. The main work on classification can already be done at clsact
egress time if the use-case allows and state stored for later retrieval
f.e. again in skb->priority with major/minors (which is checked by most
classful qdiscs before consulting tc_classify()) and/or in other skb fields
like skb->tc_index for some light-weight post-processing to get to the
eventual classid in case of a classful qdisc. Another use case is that
the clsact egress part allows to have a central egress counterpart to
the ingress classifiers, so that classifiers can easily share state (e.g.
in cls_bpf via eBPF maps) for ingress and egress.

Currently, default setups like mq + pfifo_fast would require for this to
use, for example, prio qdisc instead (to get a tc_classify() run) and to
duplicate the egress classifier for each queue. With clsact, it allows
for leaving the setup as is, it can additionally assign skb->priority to
put the skb in one of pfifo_fast's bands and it can share state with maps.
Moreover, we can access the skb's dst entry (f.e. to retrieve tclassid)
w/o the need to perform a skb_dst_force() to hold on to it any longer. In
lwt case, we can also use this facility to setup dst metadata via cls_bpf
(bpf_skb_set_tunnel_key()) without needing a real egress qdisc just for
that (case of IFF_NO_QUEUE devices, for example).

The realization can be done without any changes to the scheduler core
framework. All it takes is that we have two a-priori defined minors/child
classes, where we can mux between ingress and egress classifier list
(dev->ingress_cl_list and dev->egress_cl_list, latter stored close to
dev->_tx to avoid extra cacheline miss for moderate loads). The egress
part is a bit similar modelled to handle_ing() and patched to a noop in
case the functionality is not used. Both handlers are now called
sch_handle_ingress() and sch_handle_egress(), code sharing among the two
doesn't seem practical as there are various minor differences in both
paths, so that making them conditional in a single handler would rather
slow things down.

Full compatibility to ingress qdisc is provided as well. Since both
piggyback on TC_H_CLSACT, only one of them (ingress/clsact) can exist
per netdevice, and thus ingress qdisc specific behaviour can be retained
for user space. This means, either a user does 'tc qdisc add dev foo ingress'
and configures ingress qdisc as usual, or the 'tc qdisc add dev foo clsact'
alternative, where both, ingress and egress classifier can be configured
as in the below example. ingress qdisc supports attaching classifier to any
minor number whereas clsact has two fixed minors for muxing between the
lists, therefore to not break user space setups, they are better done as
two separate qdiscs.

I decided to extend the sch_ingress module with clsact functionality so
that commonly used code can be reused, the module is being aliased with
sch_clsact so that it can be auto-loaded properly. Alternative would have been
to add a flag when initializing ingress to alter its behaviour plus aliasing
to a different name (as it's more than just ingress). However, the first would
end up, based on the flag, choosing the new/old behaviour by calling different
function implementations to handle each anyway, the latter would require to
register ingress qdisc once again under different alias. So, this really begs
to provide a minimal, cleaner approach to have Qdisc_ops and Qdisc_class_ops
by its own that share callbacks used by both.

Example, adding qdisc:

   # tc qdisc add dev foo clsact
   # tc qdisc show dev foo
   qdisc mq 0: root
   qdisc pfifo_fast 0: parent :1 bands 3 priomap  1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
   qdisc pfifo_fast 0: parent :2 bands 3 priomap  1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
   qdisc pfifo_fast 0: parent :3 bands 3 priomap  1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
   qdisc pfifo_fast 0: parent :4 bands 3 priomap  1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1
   qdisc clsact ffff: parent ffff:fff1

Adding filters (deleting, etc works analogous by specifying ingress/egress):

   # tc filter add dev foo ingress bpf da obj bar.o sec ingress
   # tc filter add dev foo egress  bpf da obj bar.o sec egress
   # tc filter show dev foo ingress
   filter protocol all pref 49152 bpf
   filter protocol all pref 49152 bpf handle 0x1 bar.o:[ingress] direct-action
   # tc filter show dev foo egress
   filter protocol all pref 49152 bpf
   filter protocol all pref 49152 bpf handle 0x1 bar.o:[egress] direct-action

A 'tc filter show dev foo' or 'tc filter show dev foo parent ffff:' will
show an empty list for clsact. Either using the parent names (ingress/egress)
or specifying the full major/minor will then show the related filter lists.

Prior work on a mqprio prequeue() facility [1] was done mainly by John Fastabend.

  [1] http://patchwork.ozlabs.org/patch/512949/

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoethernet: amd: au1000: Remove pointless warning
Andrew Lunn [Mon, 11 Jan 2016 00:13:21 +0000 (01:13 +0100)]
ethernet: amd: au1000: Remove pointless warning

The warning about being able to read any MDIO device, not just the
attached ethernet devices PHY applies to all MDIO drivers. So remove
it. This also removes a reference to a member in phy_device which has
moved.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostaging: netlogic: Fix build error due to missed API change
Andrew Lunn [Sun, 10 Jan 2016 23:57:43 +0000 (00:57 +0100)]
staging: netlogic: Fix build error due to missed API change

Fix a number of build errors due to moving the phy_map and centralizing
interrupt allocation.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: faraday: Use phy_find_first() instead of open coding it
Guenter Roeck [Sun, 10 Jan 2016 20:04:32 +0000 (12:04 -0800)]
net: ethernet: faraday: Use phy_find_first() instead of open coding it

Use phy_find_first() to find the first phy device instead of
open coding it.

Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet: broadcom: Fix build errors
Guenter Roeck [Sun, 10 Jan 2016 20:03:16 +0000 (12:03 -0800)]
net: ethernet: broadcom: Fix build errors

Commit 7f854420fbfe ("phy: Add API for {un}registering an mdio device to
a bus") introduces an API to access mii_bus structures, but missed to
update the sb1250 driver. This results in the following build error.

drivers/net/ethernet/broadcom/sb1250-mac.c: In function 'sbmac_mii_probe':
drivers/net/ethernet/broadcom/sb1250-mac.c:2360:24: error:
'struct mii_bus' has no member named 'phy_map'

Use phy_find_first() instead of open coding it.

Commit 2220943a21e2 ("phy: Centralise print about attached phy") introduces
the following build error.

drivers/net/ethernet/broadcom/sb1250-mac.c: In function 'sbmac_mii_probe':
drivers/net/ethernet/broadcom/sb1250-mac.c:2383:20: error: 'phydev' undeclared

Fixes: 7f854420fbfe ("phy: Add API for {un}registering an mdio device to a bus")
Fixes: 2220943a21e2 ("phy: Centralise print about attached phy")
Cc: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mdio-device-fixes'
David S. Miller [Sun, 10 Jan 2016 23:03:47 +0000 (18:03 -0500)]
Merge branch 'mdio-device-fixes'

Andrew Lunn says:

====================
Fix breakage from mdio device

These two patches fix MIPS platforms which got broken by
the recent mdio device patchset.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: ethernet-rgmii.c: Fix breakage from moving phdev bus
Andrew Lunn [Thu, 7 Jan 2016 20:55:52 +0000 (21:55 +0100)]
net: ethernet-rgmii.c: Fix breakage from moving phdev bus

The mdio device patches moved the bus member in phy_device into a
substructure. This driver got missed. Fix it.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: lantiq_etop.c: Use helper to find first phy
Andrew Lunn [Thu, 7 Jan 2016 20:55:51 +0000 (21:55 +0100)]
net: lantiq_etop.c: Use helper to find first phy

Make use of the helper to find the first phy device.
This also fixes the compile breakage.

Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agostmmac: Don't exit mdio registration when mdio subnode is not found in the DTS
Romain Perier [Thu, 7 Jan 2016 20:13:28 +0000 (21:13 +0100)]
stmmac: Don't exit mdio registration when mdio subnode is not found in the DTS

Originally, most of the platforms using this driver did not define an mdio subnode
in the devicetree. Commit e34d65 ("stmmac: create of compatible mdio bus for stmmac driver")
introduced a backward compatibily issue by using of_mdiobus_register explicitly
with an mdio subnode. This patch fixes the issue by calling the function
mdiobus_register, when mdio subnode is not found. The driver is now compatible
with both modes.

Fixes: e34d65696d2e ("stmmac: create of compatible mdio bus for stmmac driver")
Signed-off-by: Romain Perier <romain.perier@gmail.com>
Tested-by: Phil Reid <preid@electromag.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: sctp: prevent writes to cookie_hmac_alg from accessing invalid memory
Sasha Levin [Thu, 7 Jan 2016 19:52:43 +0000 (14:52 -0500)]
net: sctp: prevent writes to cookie_hmac_alg from accessing invalid memory

proc_dostring() needs an initialized destination string, while the one
provided in proc_sctp_do_hmac_alg() contains stack garbage.

Thus, writing to cookie_hmac_alg would strlen() that garbage and end up
accessing invalid memory.

Fixes: 3c68198e7 ("sctp: Make hmac algorithm selection for cookie generation dynamic")
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: qmi_wwan: Add SIMCom 7230E
Kristian Evensen [Thu, 7 Jan 2016 15:41:33 +0000 (16:41 +0100)]
net: qmi_wwan: Add SIMCom 7230E

SIMCom 7230E is a QMI LTE module with support for most "normal" bands.
Manual testing has showed that only interface five works.

Cc: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Kristian Evensen <kristian.evensen@gmail.com>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bpf-next'
David S. Miller [Sun, 10 Jan 2016 22:54:28 +0000 (17:54 -0500)]
Merge branch 'bpf-next'

Daniel Borkmann says:

====================
BPF update

Fixes a csum issue on ingress. As mentioned previously, net-next
seems just fine imho. Later on, will follow up with couple of
replacements like ovs_skb_postpush_rcsum() etc.

Thanks!

v1 -> v2:
  - Added patch 1 with helper
  - Implemented Hannes' idea to just use csum_partial, thanks!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: add skb_postpush_rcsum and fix dev_forward_skb occasions
Daniel Borkmann [Thu, 7 Jan 2016 14:50:23 +0000 (15:50 +0100)]
bpf: add skb_postpush_rcsum and fix dev_forward_skb occasions

Add a small helper skb_postpush_rcsum() and fix up redirect locations
that need CHECKSUM_COMPLETE fixups on ingress. dev_forward_skb() expects
a proper csum that covers also Ethernet header, f.e. since 2c26d34bbcc0
("net/core: Handle csum for CHECKSUM_COMPLETE VXLAN forwarding"), we
also do skb_postpull_rcsum() after pulling Ethernet header off via
eth_type_trans().

When using eBPF in a netns setup f.e. with vxlan in collect metadata mode,
I can trigger the following csum issue with an IPv6 setup:

  [  505.144065] dummy1: hw csum failure
  [...]
  [  505.144108] Call Trace:
  [  505.144112]  <IRQ>  [<ffffffff81372f08>] dump_stack+0x44/0x5c
  [  505.144134]  [<ffffffff81607cea>] netdev_rx_csum_fault+0x3a/0x40
  [  505.144142]  [<ffffffff815fee3f>] __skb_checksum_complete+0xcf/0xe0
  [  505.144149]  [<ffffffff816f0902>] nf_ip6_checksum+0xb2/0x120
  [  505.144161]  [<ffffffffa08c0e0e>] icmpv6_error+0x17e/0x328 [nf_conntrack_ipv6]
  [  505.144170]  [<ffffffffa0898eca>] ? ip6t_do_table+0x2fa/0x645 [ip6_tables]
  [  505.144177]  [<ffffffffa08c0725>] ? ipv6_get_l4proto+0x65/0xd0 [nf_conntrack_ipv6]
  [  505.144189]  [<ffffffffa06c9a12>] nf_conntrack_in+0xc2/0x5a0 [nf_conntrack]
  [  505.144196]  [<ffffffffa08c039c>] ipv6_conntrack_in+0x1c/0x20 [nf_conntrack_ipv6]
  [  505.144204]  [<ffffffff8164385d>] nf_iterate+0x5d/0x70
  [  505.144210]  [<ffffffff816438d6>] nf_hook_slow+0x66/0xc0
  [  505.144218]  [<ffffffff816bd302>] ipv6_rcv+0x3f2/0x4f0
  [  505.144225]  [<ffffffff816bca40>] ? ip6_make_skb+0x1b0/0x1b0
  [  505.144232]  [<ffffffff8160b77b>] __netif_receive_skb_core+0x36b/0x9a0
  [  505.144239]  [<ffffffff8160bdc8>] ? __netif_receive_skb+0x18/0x60
  [  505.144245]  [<ffffffff8160bdc8>] __netif_receive_skb+0x18/0x60
  [  505.144252]  [<ffffffff8160ccff>] process_backlog+0x9f/0x140
  [  505.144259]  [<ffffffff8160c4a5>] net_rx_action+0x145/0x320
  [...]

What happens is that on ingress, we push Ethernet header back in, either
from cls_bpf or right before skb_do_redirect(), but without updating csum.
The "hw csum failure" can be fixed by using the new skb_postpush_rcsum()
helper for the dev_forward_skb() case to correct the csum diff again.

Thanks to Hannes Frederic Sowa for the csum_partial() idea!

Fixes: 3896d655f4d4 ("bpf: introduce bpf_clone_redirect() helper")
Fixes: 27b29f63058d ("bpf: add bpf_redirect() helper")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet, sched: add skb_at_tc_ingress helper
Daniel Borkmann [Thu, 7 Jan 2016 14:50:22 +0000 (15:50 +0100)]
net, sched: add skb_at_tc_ingress helper

Add a skb_at_tc_ingress() as this will be needed elsewhere as well and
can hide the ugly ifdef.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'tcp-keepalive-namespaceify'
David S. Miller [Sun, 10 Jan 2016 22:32:09 +0000 (17:32 -0500)]
Merge branch 'tcp-keepalive-namespaceify'

Nikolay Borisov says:

====================
Namespaceify tcp keepalive machinery

The following patch series enables the tcp keepalive mechanism
to be configured per net namespace. This is especially useful
if you have multiple containers hosted on one node and one of
them is under DoS-  in such situations one thing which could
be done is to configure the tcp keepalive settings such that
connections for that particular container are being reset
faster.

Another scenario where not being able to control those knob
comes per container is problematic is occurs the value of
net.netfilter.nf_conntrack_tcp_timeout_established is set
below the keepalive interval, in such situations the server won't
send an RST packet resulting in applications not trying to
reconnect and stale connection waiting. Changing the global
keepalive value is a possible solution but it might interfere
with other containers.

The three patches gradually convert each of the affected knobs
to be per netns. I thought it would be easier for review than
put everything in one patch. If people deem it more appropriate
to squash everything in one patch (maybe after review) I'd
be more than happy to do it.

The patches have been compile-tested on 4.4 and functionally
tested on 3.12 and they work as expected.

These are based off 4.4-rc8
====================

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespecify the tcp_keepalive_intvl sysctl knob
Nikolay Borisov [Thu, 7 Jan 2016 14:38:45 +0000 (16:38 +0200)]
ipv4: Namespecify the tcp_keepalive_intvl sysctl knob

This is the final part required to namespaceify the tcp
keep alive mechanism.

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespecify tcp_keepalive_probes sysctl knob
Nikolay Borisov [Thu, 7 Jan 2016 14:38:44 +0000 (16:38 +0200)]
ipv4: Namespecify tcp_keepalive_probes sysctl knob

This is required to have full tcp keepalive mechanism namespace
support.

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: Namespaceify tcp_keepalive_time sysctl knob
Nikolay Borisov [Thu, 7 Jan 2016 14:38:43 +0000 (16:38 +0200)]
ipv4: Namespaceify tcp_keepalive_time sysctl knob

Different net namespaces might have different requirements as to
the keepalive time of tcp sockets. This might be required in cases
where different firewall rules are in place which require tcp
timeout sockets to be increased/decreased independently of the host.

Signed-off-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoudp: restrict offloads to one namespace
Hannes Frederic Sowa [Thu, 7 Jan 2016 13:28:39 +0000 (14:28 +0100)]
udp: restrict offloads to one namespace

udp tunnel offloads tend to aggregate datagrams based on inner
headers. gro engine gets notified by tunnel implementations about
possible offloads. The match is solely based on the port number.

Imagine a tunnel bound to port 53, the offloading will look into all
DNS packets and tries to aggregate them based on the inner data found
within. This could lead to data corruption and malformed DNS packets.

While this patch minimizes the problem and helps an administrator to find
the issue by querying ip tunnel/fou, a better way would be to match on
the specific destination ip address so if a user space socket is bound
to the same address it will conflict.

Cc: Tom Herbert <tom@herbertland.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlxsw-layer2-multicast'
David S. Miller [Sun, 10 Jan 2016 21:50:21 +0000 (16:50 -0500)]
Merge branch 'mlxsw-layer2-multicast'

Jiri Pirko says:

====================
mlxsw: Adding layer 2 multicast

Elad says:

This patchset add Linux hardware reflection for L2 multicast offload and add
MC support in mlxsw. For every bridge MDB entry insertion, either by IGMP
snooping or by static insertion/removal, a switchdev ops is been called.
In mlxsw, a new multicast group (MID) is been created and ports are assigned.
When all ports are removed, the multicast group is been deleted.

---
v1->v2:
- GFP_ATOMIC->GFP_KERNEL change in patch 7/8
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: Adding IGMP snooping documentation
Elad Raz [Sun, 10 Jan 2016 20:06:29 +0000 (21:06 +0100)]
switchdev: Adding IGMP snooping documentation

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: Adding layer 2 multicast support
Elad Raz [Sun, 10 Jan 2016 20:06:28 +0000 (21:06 +0100)]
mlxsw: Adding layer 2 multicast support

Add SWITCHDEV_OBJ_ID_PORT_MDB switchdev ops support. On first MDB insertion
creates a new multicast group (MID) and add members port to the MID. Also
add new MDB entry for the flooding-domain (fid-vid) and link the MDB entry
to the newly constructed MC group.

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: Adding VID to FID translatation
Elad Raz [Sun, 10 Jan 2016 20:06:27 +0000 (21:06 +0100)]
mlxsw: Adding VID to FID translatation

Adding a generic function that translate VID to FID.

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: Changing the maximum number of multicast group to a define
Elad Raz [Sun, 10 Jan 2016 20:06:26 +0000 (21:06 +0100)]
mlxsw: Changing the maximum number of multicast group to a define

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: reg: Adding SMID register
Elad Raz [Sun, 10 Jan 2016 20:06:25 +0000 (21:06 +0100)]
mlxsw: reg: Adding SMID register

Adding back SMID register definition and packing. For each MC group a new
SMID entry will be generated.

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: reg: Add definition of multicast record for SFD register
Elad Raz [Sun, 10 Jan 2016 20:06:24 +0000 (21:06 +0100)]
mlxsw: reg: Add definition of multicast record for SFD register

Multicast-related records have specific format in SFD register.

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobridge: Reflect MDB entries to hardware
Elad Raz [Sun, 10 Jan 2016 20:06:23 +0000 (21:06 +0100)]
bridge: Reflect MDB entries to hardware

Offload MDB changes per port to hardware

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: Adding MDB entry offload
Elad Raz [Sun, 10 Jan 2016 20:06:22 +0000 (21:06 +0100)]
switchdev: Adding MDB entry offload

Define HW multicast entry: MAC and VID.
Using a MAC address simplifies support for both IPV4 and IPv6.

Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge
David S. Miller [Sun, 10 Jan 2016 02:48:36 +0000 (21:48 -0500)]
Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge

Antonio Quartulli says:

====================
Included changes:
- increase internal module version
- increase BLA wait periods to 6
- purge BLA backbone table when it is disabled
- make sure post function is invoked only if sysfs value is changed
- simplify code by removing useless NULL checks
- various corrections to existing kerneldoc
- minor cleanups
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: remove FDB entry in case we get unknown object notification
Jiri Pirko [Thu, 7 Jan 2016 10:50:30 +0000 (11:50 +0100)]
mlxsw: spectrum: remove FDB entry in case we get unknown object notification

It may happen that we get notification for FDB entry for object (port,
lag, vport), which does not exist. Currently we ignore that, which only
causes this being re-sent in next notification. The entry will never
disappear. So get rid of it by simply removing it using SFD register.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: pass local_port to mlxsw_sp_port_fdb_uc_op
Jiri Pirko [Thu, 7 Jan 2016 10:50:29 +0000 (11:50 +0100)]
mlxsw: spectrum: pass local_port to mlxsw_sp_port_fdb_uc_op

Do not pass struct mlxsw_sp_port to mlxsw_sp_port_fdb_uc_op and rather
just pass local_port. This is needed in case this is called from SFN
process function and mlxsw_sp_port is not existent for particular
local_port.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agovxlan: fix test which detect duplicate vxlan iface
Nicolas Dichtel [Thu, 7 Jan 2016 10:26:53 +0000 (11:26 +0100)]
vxlan: fix test which detect duplicate vxlan iface

When a vxlan interface is created, the driver checks that there is not
another vxlan interface with the same properties. To do this, it checks
the existing vxlan udp socket. Since commit 1c51a9159dde, the creation of
the vxlan socket is done only when the interface is set up, thus it breaks
that test.

Example:
$ ip l a vxlan10 type vxlan id 10 group 239.0.0.10 dev eth0 dstport 0
$ ip l a vxlan11 type vxlan id 10 group 239.0.0.10 dev eth0 dstport 0
$ ip -br l | grep vxlan
vxlan10          DOWN           f2:55:1c:6a:fb:00 <BROADCAST,MULTICAST>
vxlan11          DOWN           7a:cb:b9:38:59:0d <BROADCAST,MULTICAST>

Instead of checking sockets, let's loop over the vxlan iface list.

Fixes: 1c51a9159dde ("vxlan: fix race caused by dropping rtnl_unlock")
Reported-by: Thomas Faivre <thomas.faivre@6wind.com>
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: plip: use new parport device model
Sudip Mukherjee [Thu, 7 Jan 2016 10:12:37 +0000 (15:42 +0530)]
net: plip: use new parport device model

Modify plip driver to use the new parallel port device model.

Signed-off-by: Sudip Mukherjee <sudip@vectorindia.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agocdc-acm: fix NULL pointer reference
Oliver Neukum [Thu, 7 Jan 2016 10:01:00 +0000 (11:01 +0100)]
cdc-acm: fix NULL pointer reference

The union descriptor must be checked. Its usage was conditional
before the parser was introduced. This is important, because
many RNDIS device, which also use the common parser, have
bogus extra descriptors.

Signed-off-by: Oliver Neukum <oneukum@suse.com>
Tested-by: Vasily Galkin <galkin-vv@yandex.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobatman-adv: Add kerneldoc for batadv_neigh_node::refcount
Sven Eckelmann [Sun, 6 Sep 2015 19:38:48 +0000 (21:38 +0200)]
batman-adv: Add kerneldoc for batadv_neigh_node::refcount

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <a@unstable.cc>
8 years agobatman-adv: Remove kerneldoc for missing struct members
Sven Eckelmann [Sun, 6 Sep 2015 19:38:47 +0000 (21:38 +0200)]
batman-adv: Remove kerneldoc for missing struct members

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <a@unstable.cc>
8 years agobatman-adv: Fix kerneldoc member names in for main structs
Sven Eckelmann [Sun, 6 Sep 2015 19:38:46 +0000 (21:38 +0200)]
batman-adv: Fix kerneldoc member names in for main structs

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <a@unstable.cc>