GitHub/moto-9609/android_kernel_motorola_exynos9610.git
10 years agonetlabel: directly return netlbl_unlabel_genl_init()
Fabian Frederick [Wed, 8 Oct 2014 18:37:01 +0000 (20:37 +0200)]
netlabel: directly return netlbl_unlabel_genl_init()

No need to store netlbl_unlabel_genl_init result and test it before returning.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: add netdev_txq_bql_{enqueue, complete}_prefetchw() helpers
Eric Dumazet [Wed, 8 Oct 2014 15:19:27 +0000 (08:19 -0700)]
net: add netdev_txq_bql_{enqueue, complete}_prefetchw() helpers

Add two helpers so that drivers do not have to care of BQL being
available or not.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Jim Davis <jim.epost@gmail.com>
Fixes: 29d40c903247 ("net/mlx4_en: Use prefetch in tx path")
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: description of dma_cookie cause make xmldocs warning
Masanari Iida [Wed, 8 Oct 2014 14:53:39 +0000 (23:53 +0900)]
net: description of dma_cookie cause make xmldocs warning

In commit 7bced397510ab569d31de4c70b39e13355046387,
dma_cookie was removed from struct skbuff.
But the description of dma_cookie still exist.
So the "make xmldocs" output following warning.

Warning(.//include/linux/skbuff.h:609): Excess struct/union
/enum/typedef member 'dma_cookie' description in 'sk_buff'

Remove description of dma_cookie fix the symptom.

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb4: clean up a type issue
Dan Carpenter [Wed, 8 Oct 2014 13:44:34 +0000 (16:44 +0300)]
cxgb4: clean up a type issue

The tx_desc struct holds 8 __be64 values.  The original code in
ring_tx_db() took a tx_desc pointer then casted it to an int pointer and
then casted it to a u64 pointer.  It was confusing and triggered some
static checker warnings.

I have changed the cxgb_pio_copy() function to only take tx_desc
pointers.  This isn't really a loss of flexibility because anything else
was buggy to begin with.

I also removed the casting on the destination pointer since that was
unnecessary and a bit messy.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agocxgb4: potential shift wrapping bug
Dan Carpenter [Wed, 8 Oct 2014 13:43:17 +0000 (16:43 +0300)]
cxgb4: potential shift wrapping bug

"cntxt_id" is an unsigned int but "udb" is a u64 so there is a potential
shift wrapping bug here.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoi40e: skb->xmit_more support
Eric Dumazet [Tue, 7 Oct 2014 20:30:23 +0000 (13:30 -0700)]
i40e: skb->xmit_more support

Support skb->xmit_more in i40e is straightforward : we need to move
around i40e_maybe_stop_tx() call to correctly test netif_xmit_stopped()
before taking the decision to not kick the NIC.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'fs_enet_napi'
David S. Miller [Wed, 8 Oct 2014 20:01:46 +0000 (16:01 -0400)]
Merge branch 'fs_enet_napi'

Christophe Leroy says:

====================
net: fs_enet: Remove non NAPI RX and add NAPI for TX

When using a MPC8xx as a router, 'perf' shows a significant time spent in
fs_enet_interrupt() and fs_enet_start_xmit().
'perf annotate' shows that the time spent in fs_enet_start_xmit is indeed spent
between spin_unlock_irqrestore() and the following instruction, hence in
interrupt handling. This is due to the TX complete interrupt that fires after
each transmitted packet.
This patchset first remove all non NAPI handling as NAPI has become the only
mode for RX, then adds NAPI for handling TX complete.
This improves NAT TCP throughput by 21% on MPC885 with FEC.

Tested on MPC885 with FEC.

[PATCH 1/2] net: fs_enet: Remove non NAPI RX
[PATCH 2/2] net: fs_enet: Add NAPI TX

Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fs_enet: Add NAPI TX
LEROY Christophe [Tue, 7 Oct 2014 13:05:02 +0000 (15:05 +0200)]
net: fs_enet: Add NAPI TX

When using a MPC8xx as a router, 'perf' shows a significant time spent in
fs_enet_interrupt() and fs_enet_start_xmit().
'perf annotate' shows that the time spent in fs_enet_start_xmit is indeed spent
between spin_unlock_irqrestore() and the following instruction, hence in
interrupt handling. This is due to the TX complete interrupt that fires after
each transmitted packet.
This patch modifies the handling of TX complete to use NAPI.
With this patch, my NAT router offers a throughput improved by 21%

Original performance:

[root@localhost tmp]# scp toto pgs:/tmp
toto                                          100%  256MB   2.8MB/s   01:31

Performance with the patch:

[root@localhost tmp]# scp toto pgs:/tmp
toto                                          100%  256MB   3.4MB/s   01:16

Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fs_enet: Remove non NAPI RX
LEROY Christophe [Tue, 7 Oct 2014 13:04:57 +0000 (15:04 +0200)]
net: fs_enet: Remove non NAPI RX

In the probe function, use_napi is inconditionnaly set to 1. This patch removes
all the code which is conditional to !use_napi, and removes use_napi which has
then become useless.

Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8169:add support for RTL8168EP
Chun-Hao Lin [Tue, 7 Oct 2014 07:10:41 +0000 (15:10 +0800)]
r8169:add support for RTL8168EP

RTL8168EP is Realtek PCIe Gigabit Ethernet controller with DASH support.
It is a successor chip of RTL8168DP.

For RTL8168EP, the read/write ocp register is via eri channel type 2, so I
move ocp_xxx() related functions under rtl_eri_xxx. And use r8168dp_ocp_xxx()
for RTL8168DP ocp read/write, r8168ep_ocp_xxx() for RTL8168EP ocp read/write.

The way of checking dash enable is different with RTL8168DP. I use
r8168dp_check_dash()for RTL8168DP and r8168ep_check_dash() for RTL8168EP,
to check if dash is enabled.

The driver_start() and driver_stop() of RTL8168EP is also different with
RTL8168DP. I use rtl8168dp_driver_xxx() for RTL8168DP and
rtl8168ep_driver_xxx for RTL8168EP.

Right now, RTL8168EP phy mcu did not need firmware code patch, so I did not
add firmware code for it.
so I did not add firmware code for it.

Signed-off-by: Chun-Hao Lin <hau@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agowimax: convert printk to pr_foo()
Fabian Frederick [Tue, 7 Oct 2014 20:12:03 +0000 (22:12 +0200)]
wimax: convert printk to pr_foo()

Use current logging functions and add module name prefix.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoaf_unix: remove 0 assignment on static
Fabian Frederick [Tue, 7 Oct 2014 21:02:15 +0000 (23:02 +0200)]
af_unix: remove 0 assignment on static

static values are automatically initialized to 0

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: Do not warn for informational ICMP messages, regardless of type.
David S. Miller [Tue, 7 Oct 2014 20:33:53 +0000 (16:33 -0400)]
ipv6: Do not warn for informational ICMP messages, regardless of type.

There is no reason to emit a log message for these.

Based upon a suggestion from Hannes Frederic Sowa.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
10 years agotipc: fix bug in multicast congestion handling
Jon Maloy [Tue, 7 Oct 2014 18:12:34 +0000 (14:12 -0400)]
tipc: fix bug in multicast congestion handling

One aim of commit 50100a5e39461b2a61d6040e73c384766c29975d ("tipc:
use pseudo message to wake up sockets after link congestion") was
to handle link congestion abatement in a uniform way for both unicast
and multicast transmit. However, the latter doesn't work correctly,
and has been broken since the referenced commit was applied.

If a user now sends a burst of multicast messages that is big
enough to cause broadcast link congestion, it will be put to sleep,
and not be waked up when the congestion abates as it should be.

This has two reasons. First, the flag that is used, TIPC_WAKEUP_USERS,
is set correctly, but in the wrong field. Instead of setting it in the
'action_flags' field of the arrival node struct, it is by mistake set
in the dummy node struct that is owned by the broadcast link, where it
will never tested for. Second, we cannot use the same flag for waking
up unicast and multicast users, since the function tipc_node_unlock()
needs to pick the wakeup pseudo messages to deliver from different
queues. It must hence be able to distinguish between the two cases.

This commit solves this problem by adding a new flag
TIPC_WAKEUP_BCAST_USERS, and a new function tipc_bclink_wakeup_user().
The latter is to be called by tipc_node_unlock() when the named flag,
now set in the correct field, is encountered.

v2: using explicit 'unsigned int' declaration instead of 'uint', as
per comment from David Miller.

Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: better IFF_XMIT_DST_RELEASE support
Eric Dumazet [Mon, 6 Oct 2014 01:38:35 +0000 (18:38 -0700)]
net: better IFF_XMIT_DST_RELEASE support

Testing xmit_more support with netperf and connected UDP sockets,
I found strange dst refcount false sharing.

Current handling of IFF_XMIT_DST_RELEASE is not optimal.

Dropping dst in validate_xmit_skb() is certainly too late in case
packet was queued by cpu X but dequeued by cpu Y

The logical point to take care of drop/force is in __dev_queue_xmit()
before even taking qdisc lock.

As Julian Anastasov pointed out, need for skb_dst() might come from some
packet schedulers or classifiers.

This patch adds new helper to cleanly express needs of various drivers
or qdiscs/classifiers.

Drivers that need skb_dst() in their ndo_start_xmit() should call
following helper in their setup instead of the prior :

dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
->
netif_keep_dst(dev);

Instead of using a single bit, we use two bits, one being
eventually rebuilt in bonding/team drivers.

The other one, is permanent and blocks IFF_XMIT_DST_RELEASE being
rebuilt in bonding/team. Eventually, we could add something
smarter later.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: remove NETDEV_TX_BUSY
Eric Dumazet [Mon, 6 Oct 2014 16:30:35 +0000 (09:30 -0700)]
net/mlx4_en: remove NETDEV_TX_BUSY

Drivers should avoid NETDEV_TX_BUSY as much as possible.

They should stop the tx queue before qdisc even tries to push another
packet, to avoid requeues.

For a driver supporting skb->xmit_more, this is likely to be a prereq
anyway, otherwise we could have a tx deadlock : We need to force a
doorbell if TX ring is full.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: bcmgenet: fix Tx ring priority programming
Petri Gynther [Tue, 7 Oct 2014 16:30:01 +0000 (09:30 -0700)]
net: bcmgenet: fix Tx ring priority programming

GENET MAC has three Tx ring priority registers:
- GENET_x_TDMA_PRIORITY0 for queues 0-5
- GENET_x_TDMA_PRIORITY1 for queues 6-11
- GENET_x_TDMA_PRIORITY2 for queues 12-16

Fix bcmgenet_init_multiq() to program them correctly.

Signed-off-by: Petri Gynther <pgynther@google.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: phy: adjust fixed_phy_register() return value
Petri Gynther [Mon, 6 Oct 2014 18:38:30 +0000 (11:38 -0700)]
net: phy: adjust fixed_phy_register() return value

Adjust fixed_phy_register() to return struct phy_device *, so that
it becomes easy to use fixed PHYs without device tree support:

  phydev = fixed_phy_register(PHY_POLL, &fixed_phy_status, NULL);
  fixed_phy_set_link_update(phydev, fixed_phy_link_update);
  phy_connect_direct(netdev, phydev, handler_fn, phy_interface);

This change is a prerequisite for modifying bcmgenet driver to work
without a device tree on Broadcom's MIPS-based 7xxx platforms.

Signed-off-by: Petri Gynther <pgynther@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet_sched: fix unused variables in __gnet_stats_copy_basic_cpu()
WANG Cong [Tue, 7 Oct 2014 00:01:33 +0000 (17:01 -0700)]
net_sched: fix unused variables in __gnet_stats_copy_basic_cpu()

Probably not a big deal, but we'd better just use the
one we get in retry loop.

Fixes: commit 22e0f8b9322cb1a48b1357e8 ("net: sched: make bstats per cpu and estimator RCU safe")
Reported-by: Joe Perches <joe@perches.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: fix a compilation error when CONFIG_INET is not setW!
Andy Zhou [Mon, 6 Oct 2014 22:15:14 +0000 (15:15 -0700)]
openvswitch: fix a compilation error when CONFIG_INET is not setW!

Fix a openvswitch compilation error when CONFIG_INET is not set:

=====================================================
   In file included from include/net/geneve.h:4:0,
                       from net/openvswitch/flow_netlink.c:45:
          include/net/udp_tunnel.h: In function 'udp_tunnel_handle_offloads':
  >> include/net/udp_tunnel.h:100:2: error: implicit declaration of function 'iptunnel_handle_offloads' [-Werror=implicit-function-declaration]
  >>      return iptunnel_handle_offloads(skb, udp_csum, type);
  >>           ^
  >>           >> include/net/udp_tunnel.h:100:2: warning: return makes pointer from integer without a cast
  >>           >>    cc1: some warnings being treated as errors

=====================================================

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: fix a sparse warning
Andy Zhou [Mon, 6 Oct 2014 20:22:51 +0000 (13:22 -0700)]
openvswitch: fix a sparse warning

Fix a sparse warning introduced by commit:
f5796684069e0c71c65bce6a6d4766114aec1396 (openvswitch: Add support for
Geneve tunneling.) caught by kbuild test robot:

reproduce:
  # apt-get install sparse
  #   git checkout f5796684069e0c71c65bce6a6d4766114aec1396
  #     make ARCH=x86_64 allmodconfig
  #       make C=1 CF=-D__CHECK_ENDIAN__
  #
  #
  #       sparse warnings: (new ones prefixed by >>)
  #
  #       >> net/openvswitch/vport-geneve.c:109:15: sparse: incorrect type in assignment (different base types)
  #          net/openvswitch/vport-geneve.c:109:15:    expected restricted __be16 [usertype] sport
  #             net/openvswitch/vport-geneve.c:109:15:    got int
  #             >> net/openvswitch/vport-geneve.c:110:56: sparse: incorrect type in argument 3 (different base types)
  #                net/openvswitch/vport-geneve.c:110:56:    expected unsigned short [unsigned] [usertype] value
  #                   net/openvswitch/vport-geneve.c:110:56:    got restricted __be16 [usertype] sport

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fix a sparse warning
Andy Zhou [Mon, 6 Oct 2014 20:22:50 +0000 (13:22 -0700)]
net: fix a sparse warning

Fix a sparse warning introduced by Commit
0b5e8b8eeae40bae6ad7c7e91c97c3c0d0e57882 (net: Add Geneve tunneling
protocol driver) caught by kbuild test robot:

  # apt-get install sparse
  #   git checkout 0b5e8b8eeae40bae6ad7c7e91c97c3c0d0e57882
  #     make ARCH=x86_64 allmodconfig
  #       make C=1 CF=-D__CHECK_ENDIAN__
  #
  #
  #       sparse warnings: (new ones prefixed by >>)
  #
  #       >> net/ipv4/geneve.c:230:42: sparse: incorrect type in assignment (different base types)
  #          net/ipv4/geneve.c:230:42:    expected restricted __be32 [addressable] [assigned] [usertype] s_addr
  #             net/ipv4/geneve.c:230:42:    got unsigned long [unsigned] <noident>
  #

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'ipv6-next'
David S. Miller [Tue, 7 Oct 2014 04:02:49 +0000 (00:02 -0400)]
Merge branch 'ipv6-next'

Hannes Frederic Sowa says:

====================
ipv6: cleanup after rt6_genid removal

Leftover patches after rt6_genid removal after 705f1c869d577c ("ipv6:
remove rt6i_genid").

Major two changes are:
* keep fib6_sernum per namespace to reduce number of flushes in case
  system has high number of namespaces
* make fn_sernum updates cheaper

v2: Incorporated feedback from Cong Wang, thanks a lot!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: don't walk node's leaf during serial number update
Hannes Frederic Sowa [Mon, 6 Oct 2014 17:58:38 +0000 (19:58 +0200)]
ipv6: don't walk node's leaf during serial number update

Cc: YOSHIFUJI Hideaki <hideaki@yoshifuji.org>
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: make fib6 serial number per namespace
Hannes Frederic Sowa [Mon, 6 Oct 2014 17:58:37 +0000 (19:58 +0200)]
ipv6: make fib6 serial number per namespace

Try to reduce number of possible fn_sernum mutation by constraining them
to their namespace.

Also remove rt_genid which I forgot to remove in 705f1c869d577c ("ipv6:
remove rt6i_genid").

Cc: YOSHIFUJI Hideaki <hideaki@yoshifuji.org>
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: only generate one new serial number per fib mutation
Hannes Frederic Sowa [Mon, 6 Oct 2014 17:58:36 +0000 (19:58 +0200)]
ipv6: only generate one new serial number per fib mutation

Cc: YOSHIFUJI Hideaki <hideaki@yoshifuji.org>
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: make rt_sernum atomic and serial number fields ordinary ints
Hannes Frederic Sowa [Mon, 6 Oct 2014 17:58:35 +0000 (19:58 +0200)]
ipv6: make rt_sernum atomic and serial number fields ordinary ints

Cc: YOSHIFUJI Hideaki <hideaki@yoshifuji.org>
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: minor fib6 cleanups like type safety, bool conversion, inline removal
Hannes Frederic Sowa [Mon, 6 Oct 2014 17:58:34 +0000 (19:58 +0200)]
ipv6: minor fib6 cleanups like type safety, bool conversion, inline removal

Also renamed struct fib6_walker_t to fib6_walker and enum fib_walk_state_t
to fib6_walk_state as recommended by Cong Wang.

Cc: Cong Wang <cwang@twopensource.com>
Cc: YOSHIFUJI Hideaki <hideaki@yoshifuji.org>
Cc: Martin Lau <kafai@fb.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: validate_xmit_vlan() is static
Eric Dumazet [Mon, 6 Oct 2014 18:26:27 +0000 (11:26 -0700)]
net: validate_xmit_vlan() is static

Marking this as static allows compiler to inline it.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fix rcu access on phonet_routes
Fabian Frederick [Mon, 6 Oct 2014 18:15:20 +0000 (20:15 +0200)]
net: fix rcu access on phonet_routes

-Add __rcu annotation on table to fix sparse warnings:
net/phonet/pn_dev.c:279:25: warning: incorrect type in assignment (different address spaces)
net/phonet/pn_dev.c:279:25:    expected struct net_device *<noident>
net/phonet/pn_dev.c:279:25:    got void [noderef] <asn:4>*<noident>
net/phonet/pn_dev.c:376:17: warning: incorrect type in assignment (different address spaces)
net/phonet/pn_dev.c:376:17:    expected struct net_device *volatile <noident>
net/phonet/pn_dev.c:376:17:    got struct net_device [noderef] <asn:4>*<noident>
net/phonet/pn_dev.c:392:17: warning: incorrect type in assignment (different address spaces)
net/phonet/pn_dev.c:392:17:    expected struct net_device *<noident>
net/phonet/pn_dev.c:392:17:    got void [noderef] <asn:4>*<noident>

-Access table with rcu_access_pointer (fixes the following sparse errors):
net/phonet/pn_dev.c:278:25: error: incompatible types in comparison expression (different address spaces)
net/phonet/pn_dev.c:391:17: error: incompatible types in comparison expression (different address spaces)

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'net_sched-next'
David S. Miller [Mon, 6 Oct 2014 22:02:40 +0000 (18:02 -0400)]
Merge branch 'net_sched-next'

John Fastabend says:

====================
net sched rcu updates

This fixes the use of tcf_proto from RCU callbacks it requires
moving the unbind calls out of the callbacks and removing the
tcf_proto argument from the tcf_em_tree_destroy().

This is a rework of two previous series and addresses comments
from Cong. And should apply against latest net-next.

The previous series links below for reference:

(1/2) net: sched: do not use tcf_proto 'tp' argument from call_rcu
http://patchwork.ozlabs.org/patch/396149/

(2/2) net: sched: replace ematch calls to use struct net
http://patchwork.ozlabs.org/patch/396150/

net: sched: cls_cgroup tear down exts and ematch from rcu callback
http://patchwork.ozlabs.org/patch/396307/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sched: do not use tcf_proto 'tp' argument from call_rcu
John Fastabend [Mon, 6 Oct 2014 04:28:52 +0000 (21:28 -0700)]
net: sched: do not use tcf_proto 'tp' argument from call_rcu

Using the tcf_proto pointer 'tp' from inside the classifiers callback
is not valid because it may have been cleaned up by another call_rcu
occuring on another CPU.

'tp' is currently being used by tcf_unbind_filter() in this patch we
move instances of tcf_unbind_filter outside of the call_rcu() context.
This is safe to do because any running schedulers will either read the
valid class field or it will be zeroed.

And all schedulers today when the class is 0 do a lookup using the
same call used by the tcf_exts_bind(). So even if we have a running
classifier hit the null class pointer it will do a lookup and get
to the same result. This is particularly fragile at the moment because
the only way to verify this is to audit the schedulers call sites.

Reported-by: Cong Wang <xiyou.wangconf@gmail.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sched: cls_cgroup tear down exts and ematch from rcu callback
John Fastabend [Mon, 6 Oct 2014 04:28:20 +0000 (21:28 -0700)]
net: sched: cls_cgroup tear down exts and ematch from rcu callback

It is not RCU safe to destroy the action chain while there
is a possibility of readers accessing it. Move this code
into the rcu callback using the same rcu callback used in the
code patch to make a change to head.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sched: remove tcf_proto from ematch calls
John Fastabend [Mon, 6 Oct 2014 04:27:53 +0000 (21:27 -0700)]
net: sched: remove tcf_proto from ematch calls

This removes the tcf_proto argument from the ematch code paths that
only need it to reference the net namespace. This allows simplifying
qdisc code paths especially when we need to tear down the ematch
from an RCU callback. In this case we can not guarentee that the
tcf_proto structure is still valid.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: introduce netdevice gso_min_segs attribute
Eric Dumazet [Sun, 5 Oct 2014 17:11:27 +0000 (10:11 -0700)]
net: introduce netdevice gso_min_segs attribute

Some TSO engines might have a too heavy setup cost, that impacts
performance on hosts sending small bursts (2 MSS per packet).

This patch adds a device gso_min_segs, allowing drivers to set
a minimum segment size for TSO packets, according to the NIC
performance.

Tested on a mlx4 NIC, this allows to get a ~110% increase of
throughput when sending 2 MSS per packet.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agor8152: nway reset after setting eee
hayeswang [Mon, 6 Oct 2014 02:36:04 +0000 (10:36 +0800)]
r8152: nway reset after setting eee

Restart autonegotiation is necessary after setting EEE.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv4: igmp: fix v3 general query drop monitor false positive
Daniel Borkmann [Sun, 5 Oct 2014 15:27:50 +0000 (17:27 +0200)]
ipv4: igmp: fix v3 general query drop monitor false positive

In case we find a general query with non-zero number of sources, we
are dropping the skb as it's malformed.

RFC3376, section 4.1.8. Number of Sources (N):

  This number is zero in a General Query or a Group-Specific Query,
  and non-zero in a Group-and-Source-Specific Query.

Therefore, reflect that by using kfree_skb() instead of consume_skb().

Fixes: d679c5324d9a ("igmp: avoid drop_monitor false positives")
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: Simplify the xmit function for modes that use xmit_hash
Mahesh Bandewar [Sun, 5 Oct 2014 00:45:01 +0000 (17:45 -0700)]
bonding: Simplify the xmit function for modes that use xmit_hash

Earlier change to use usable slave array for TLB mode had an additional
performance advantage. So extending the same logic to all other modes
that use xmit-hash for slave selection (viz 802.3AD, and XOR modes).
Also consolidating this with the earlier TLB change.

The main idea is to build the usable slaves array in the control path
and use that array for slave selection during xmit operation.

Measured performance in a setup with a bond of 4x1G NICs with 200
instances of netperf for the modes involved (3ad, xor, tlb)
cmd: netperf -t TCP_RR -H <TargetHost> -l 60 -s 5

Mode        TPS-Before   TPS-After

802.3ad   : 468,694      493,101
TLB (lb=0): 392,583      392,965
XOR       : 475,696      484,517

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: display xmit_hash_policy for non-dynamic-tlb mode
Mahesh Bandewar [Sun, 5 Oct 2014 00:44:44 +0000 (17:44 -0700)]
bonding: display xmit_hash_policy for non-dynamic-tlb mode

It's a trivial fix to display xmit_hash_policy for this new TLB mode
since it uses transmit-hash-poilicy as part of bonding-master info
(/proc/net/bonding/<bonding-interface).

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'mlx4-next'
David S. Miller [Mon, 6 Oct 2014 05:04:21 +0000 (01:04 -0400)]
Merge branch 'mlx4-next'

Amir Vadai says:

====================
net/mlx4_en: Optimizations to TX flow

This patchset contains optimizations to TX flow in mlx4_en driver. It also introduce
setting/getting tx copybreak, to enable controlling inline threshold dynamically.

TX flow optimizations was authored and posted to the mailing list by Eric
Dumazet [1] as a single patch. I splitted this patch to smaller patches,
Reviewed it and tested.
Changed from original patch:
- s/iowrite32be/iowrite32/, since ring->doorbell_qpn is stored as be32

The tx copybreak patch was also suggested by Eric Dumazet, and was edited and
reviewed by me. User space patch will be sent after kernel code is ready.

I am sending this patchset now since the merge window is near and don't want to
miss it.

More work need to do:
- Disable BF when xmit_more is in use
- Make TSO use xmit_more too. Maybe by splitting small TSO packets in the
  driver itself, to avoid extra cpu/memory costs of GSO before the driver
- Fix mlx4_en_xmit buggy handling of queue full in the middle of a burst
  partially posted to send queue using xmit_more

Eric, I edited the patches to have you as the Author and the first
signed-off-by. I hope it is ok with you (I wasn't sure if it is ok to sign by
you), anyway all the credit to those changes should go to you.

Patchset was tested and applied over commit 1e203c1 "(net: sched:
suspicious RCU usage in qdisc_watchdog")

[1] - https://patchwork.ozlabs.org/patch/394256/
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Use the new tx_copybreak to set inline threshold
Eric Dumazet [Sun, 5 Oct 2014 09:35:22 +0000 (12:35 +0300)]
net/mlx4_en: Use the new tx_copybreak to set inline threshold

Instead of setting inline threshold using module parameter only on
driver load, use set_tunable() to set it dynamically.
No need to store the threshold per ring, using instead the netdev global
priv->prof->inline_thold
Initial value still is set using the module parameter, therefore
backward compatability is kept.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoethtool: Ethtool parameter to dynamically change tx_copybreak
Eric Dumazet [Sun, 5 Oct 2014 09:35:21 +0000 (12:35 +0300)]
ethtool: Ethtool parameter to dynamically change tx_copybreak

Use new ethtool [sg]et_tunable() to set tx_copybread (inline threshold)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Enable the compiler to make is_inline() inlined
Eric Dumazet [Sun, 5 Oct 2014 09:35:20 +0000 (12:35 +0300)]
net/mlx4_en: Enable the compiler to make is_inline() inlined

Reorganize code to call is_inline() once, so compiler can inline it

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: tx_info->ts_requested was not cleared
Eric Dumazet [Sun, 5 Oct 2014 09:35:19 +0000 (12:35 +0300)]
net/mlx4_en: tx_info->ts_requested was not cleared

Properly clear tx_info->ts_requested

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Use local var for skb_headlen(skb)
Eric Dumazet [Sun, 5 Oct 2014 09:35:18 +0000 (12:35 +0300)]
net/mlx4_en: Use local var for skb_headlen(skb)

Access skb_headlen() once in tx flow

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Use local var in tx flow for skb_shinfo(skb)
Eric Dumazet [Sun, 5 Oct 2014 09:35:17 +0000 (12:35 +0300)]
net/mlx4_en: Use local var in tx flow for skb_shinfo(skb)

Acces skb_shinfo(skb) once in tx flow.
Also, rename @i variable to @i_frag to avoid confusion, as the "goto
tx_drop_unmap;" relied on this @i variable.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: mlx4_en_xmit() reads ring->cons once, and ahead of time to avoid stalls
Eric Dumazet [Sun, 5 Oct 2014 09:35:16 +0000 (12:35 +0300)]
net/mlx4_en: mlx4_en_xmit() reads ring->cons once, and ahead of time to avoid stalls

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Avoid false sharing in mlx4_en_en_process_tx_cq()
Eric Dumazet [Sun, 5 Oct 2014 09:35:15 +0000 (12:35 +0300)]
net/mlx4_en: Avoid false sharing in mlx4_en_en_process_tx_cq()

mlx4_en_process_tx_cq() carefully fetches and writes ring->last_nr_txbb
and ring->cons only one time to avoid false sharing

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Use prefetch in tx path
Eric Dumazet [Sun, 5 Oct 2014 09:35:14 +0000 (12:35 +0300)]
net/mlx4_en: Use prefetch in tx path

mlx4_en_free_tx_desc() uses a prefetchw(&skb->users) to speed up
consume_skb()
prefetchw(&ring->tx_queue->dql) to speed up BQL update

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Avoid a cache line miss in TX completion for single frag skb's
Eric Dumazet [Sun, 5 Oct 2014 09:35:13 +0000 (12:35 +0300)]
net/mlx4_en: Avoid a cache line miss in TX completion for single frag skb's

Add frag0_dma/frag0_byte_count into mlx4_en_tx_info to avoid a cache
line miss in TX completion for frames having one dma element.  (We avoid
reading back the tx descriptor)

Note this could be extended to 2/3 dma elements later, as we have free
room in mlx4_en_tx_info

Also, mlx4_en_free_tx_desc() no longer accesses skb_shinfo(). We use a
new nr_maps fields in mlx4_en_tx_info to avoid 2 or 3 cache misses.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: tx_info allocated with kmalloc() instead of vmalloc()
Eric Dumazet [Sun, 5 Oct 2014 09:35:12 +0000 (12:35 +0300)]
net/mlx4_en: tx_info allocated with kmalloc() instead of vmalloc()

Try to allocate using kmalloc_node() first, only on failure use
vmalloc()

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Avoid calling bswap in tx fast path
Eric Dumazet [Sun, 5 Oct 2014 09:35:11 +0000 (12:35 +0300)]
net/mlx4_en: Avoid calling bswap in tx fast path

- doorbell_qpn is stored in the cpu_to_be32() way to avoid bswap() in fast
  path.
- mdev->mr.key stored in ring->mr_key to also avoid bswap() and access to
  cold cache line.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Align tx path structures to cache lines
Eric Dumazet [Sun, 5 Oct 2014 09:35:10 +0000 (12:35 +0300)]
net/mlx4_en: Align tx path structures to cache lines

Reorganize struct mlx4_en_tx_ring to have:
- One cache line containing last_nr_txbb & cons & wake_queue, used by tx
  completion.
- One cache line containing fields dirtied by mlx4_en_xmit()
- Following part is read mostly and shared by cpus.

Align struct mlx4_en_tx_info to a cache line

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Code cleanups in tx path
Eric Dumazet [Sun, 5 Oct 2014 09:35:09 +0000 (12:35 +0300)]
net/mlx4_en: Code cleanups in tx path

- Remove unused variable ring->poll_cnt
- No need to set some fields if using blueflame
- Add missing const's
- Use unlikely
- Remove unneeded new line
- Make some comments more precise
- struct mlx4_bf @offset field reduced to unsigned int to save space

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sched: avoid costly atomic operation in fq_dequeue()
Eric Dumazet [Sat, 4 Oct 2014 17:11:31 +0000 (10:11 -0700)]
net: sched: avoid costly atomic operation in fq_dequeue()

Standard qdisc API to setup a timer implies an atomic operation on every
packet dequeue : qdisc_unthrottled()

It turns out this is not really needed for FQ, as FQ has no concept of
global qdisc throttling, being a qdisc handling many different flows,
some of them can be throttled, while others are not.

Fix is straightforward : add a 'bool throttle' to
qdisc_watchdog_schedule_ns(), and remove calls to qdisc_unthrottled()
in sch_fq.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofec: Fix fec_enet_alloc_buffers() error path
Fabio Estevam [Sat, 4 Oct 2014 16:40:01 +0000 (13:40 -0300)]
fec: Fix fec_enet_alloc_buffers() error path

When fec_enet_alloc_buffers() fails we should better undo the previous actions,
which consists of: disabling the FEC clocks and putting the FEC pins into
inactive state.

The error path for fec_enet_mii_probe() is kept unchanged.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/phy/Kconfig: Let MDIO_BCM_UNIMAC depend on HAS_IOMEM
Chen Gang [Sat, 4 Oct 2014 09:54:33 +0000 (17:54 +0800)]
drivers/net/phy/Kconfig: Let MDIO_BCM_UNIMAC depend on HAS_IOMEM

MDIO_BCM_UNIMAC needs HAS_IOMEM, so depend on it, the related error (
with allmodconfig under um):

    MODPOST 1205 modules
  ERROR: "devm_ioremap" [drivers/net/phy/mdio-bcm-unimac.ko] undefined!

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: skb_segment() provides list head and tail
Eric Dumazet [Sat, 4 Oct 2014 03:59:19 +0000 (20:59 -0700)]
net: skb_segment() provides list head and tail

Its unfortunate we have to walk again skb list to find the tail
after segmentation, even if data is probably hot in cpu caches.

skb_segment() can store the tail of the list into segs->prev,
and validate_xmit_skb_list() can immediately get the tail.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'geneve'
David S. Miller [Mon, 6 Oct 2014 04:32:25 +0000 (00:32 -0400)]
Merge branch 'geneve'

Andy Zhou says:

====================
Add Geneve tunnel protocol support

This patch series adds kernel support for Geneve (Generic Network
Virtualization Encapsulation) based on Geneve IETF draft:
http://www.ietf.org/id/draft-gross-geneve-01.txt

Patch 1 implements Geneve tunneling protocol driver

Patch 2-6 adds openvswitch support for creating and using
Geneve tunnels by OVS user space.

v1->v2:   Style fixes: use tab instead space for Kconfig
  Patch 2-6 are reviewed by Pravin Shetty, add him to acked-by
  Patch 6 was reviewed by Thomas Graf when commiting
    to openvswitch.org, add him to acked-by.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Add support for Geneve tunneling.
Jesse Gross [Fri, 3 Oct 2014 22:35:33 +0000 (15:35 -0700)]
openvswitch: Add support for Geneve tunneling.

The Openvswitch implementation is completely agnostic to the options
that are in use and can handle newly defined options without
further work. It does this by simply matching on a byte array
of options and allowing userspace to setup flows on this array.

Signed-off-by: Jesse Gross <jesse@nicira.com>
Singed-off-by: Ansis Atteka <aatteka@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Factor out allocation and verification of actions.
Jesse Gross [Fri, 3 Oct 2014 22:35:32 +0000 (15:35 -0700)]
openvswitch: Factor out allocation and verification of actions.

As the size of the flow key grows, it can put some pressure on the
stack. This is particularly true in ovs_flow_cmd_set(), which needs several
copies of the key on the stack. One of those uses is logically separate,
so this factors it out to reduce stack pressure and improve readibility.

Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Wrap struct ovs_key_ipv4_tunnel in a new structure.
Jesse Gross [Fri, 3 Oct 2014 22:35:31 +0000 (15:35 -0700)]
openvswitch: Wrap struct ovs_key_ipv4_tunnel in a new structure.

Currently, the flow information that is matched for tunnels and
the tunnel data passed around with packets is the same. However,
as additional information is added this is not necessarily desirable,
as in the case of pointers.

This adds a new structure for tunnel metadata which currently contains
only the existing struct. This change is purely internal to the kernel
since the current OVS_KEY_ATTR_IPV4_TUNNEL is simply a compressed version
of OVS_KEY_ATTR_TUNNEL that is translated at flow setup.

Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Add support for matching on OAM packets.
Jesse Gross [Fri, 3 Oct 2014 22:35:30 +0000 (15:35 -0700)]
openvswitch: Add support for matching on OAM packets.

Some tunnel formats have mechanisms for indicating that packets are
OAM frames that should be handled specially (either as high priority or
not forwarded beyond an endpoint). This provides support for allowing
those types of packets to be matched.

Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoopenvswitch: Eliminate memset() from flow_extract.
Jesse Gross [Fri, 3 Oct 2014 22:35:29 +0000 (15:35 -0700)]
openvswitch: Eliminate memset() from flow_extract.

As new protocols are added, the size of the flow key tends to
increase although few protocols care about all of the fields. In
order to optimize this for hashing and matching, OVS uses a variable
length portion of the key. However, when fields are extracted from
the packet we must still zero out the entire key.

This is no longer necessary now that OVS implements masking. Any
fields (or holes in the structure) which are not part of a given
protocol will be by definition not part of the mask and zeroed out
during lookup. Furthermore, since masking already uses variable
length keys this zeroing operation automatically benefits as well.

In principle, the only thing that needs to be done at this point
is remove the memset() at the beginning of flow. However, some
fields assume that they are initialized to zero, which now must be
done explicitly. In addition, in the event of an error we must also
zero out corresponding fields to signal that there is no valid data
present. These increase the total amount of code but very little of
it is executed in non-error situations.

Removing the memset() reduces the profile of ovs_flow_extract()
from 0.64% to 0.56% when tested with large packets on a 10G link.

Suggested-by: Pravin Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Add Geneve tunneling protocol driver
Andy Zhou [Fri, 3 Oct 2014 22:35:28 +0000 (15:35 -0700)]
net: Add Geneve tunneling protocol driver

This adds a device level support for Geneve -- Generic Network
Virtualization Encapsulation. The protocol is documented at
http://tools.ietf.org/html/draft-gross-geneve-01

Only protocol layer Geneve support is provided by this driver.
Openvswitch can be used for configuring, set up and tear down
functional Geneve tunnels.

Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fec: fix build error at m68k platform
Frank Li [Fri, 3 Oct 2014 21:29:14 +0000 (14:29 -0700)]
net: fec: fix build error at m68k platform

reproduce:
  wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
  chmod +x ~/bin/make.cross
  git checkout 1b7bde6d659d30f171259cc2dfba8e5dab34e735

  make.cross ARCH=m68k m5275evb_defconfig
  make.cross ARCH=m68k

All error/warnings:

   drivers/net/ethernet/freescale/fec_main.c: In function 'fec_enet_rx_queue':
>> drivers/net/ethernet/freescale/fec_main.c:1470:3: error: implicit declaration of function 'prefetch' [-Werror=implicit-function-declaration]
      prefetch(skb->data - NET_IP_ALIGN);
      ^
   cc1: some warnings being treated as errors

missed included prefetch.h

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Frank Li <Frank.Li@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: bcmgenet: improve bcmgenet_mii_setup()
Petri Gynther [Fri, 3 Oct 2014 19:25:01 +0000 (12:25 -0700)]
net: bcmgenet: improve bcmgenet_mii_setup()

bcmgenet_mii_setup() is called from the PHY state machine every 1-2 seconds
when the PHYs are in PHY_POLL mode.

Improve bcmgenet_mii_setup() so that it touches the MAC registers only when
the link is up and there was a change to link, speed, duplex, or pause status.

Signed-off-by: Petri Gynther <pgynther@google.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'altera_tse'
David S. Miller [Mon, 6 Oct 2014 01:39:58 +0000 (21:39 -0400)]
Merge branch 'altera_tse'

Walter Lozano says:

====================
Altera TSE with no PHY

In some scenarios there is no PHY chip present, for example in optical links.
This serie of patches moves PHY get addr and MDIO create to a new function and
avoids PHY and MDIO probing in these cases.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoAltera TSE: Add support for no PHY
Walter Lozano [Fri, 3 Oct 2014 18:09:01 +0000 (15:09 -0300)]
Altera TSE: Add support for no PHY

This patch avoids PHY and MDIO probing if no PHY chip is present.
This is the case mainly in optical links where there is no need for
PHY chip, and therefore no need of MDIO. In this scenario Ethernet
MAC is directly connected to an optical module through an external
SFP transceiver.

Signed-off-by: Walter Lozano <walter@vanguardiasur.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoAltera TSE: Move PHY get addr and MDIO create
Walter Lozano [Fri, 3 Oct 2014 18:09:00 +0000 (15:09 -0300)]
Altera TSE: Move PHY get addr and MDIO create

Move PHY get addr and MDIO create to a new function to improve readability
and make it easier to avoid its usage. This will be useful for example in
the case where there is no PHY chip.

Signed-off-by: Walter Lozano <walter@vanguardiasur.com.ar>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'master-2014-10-02' of git://git.kernel.org/pub/scm/linux/kernel/git/linvil...
David S. Miller [Mon, 6 Oct 2014 01:34:39 +0000 (21:34 -0400)]
Merge tag 'master-2014-10-02' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
pull request: wireless-next 2014-10-03

Please pull tihs batch of updates intended for the 3.18 stream!

For the iwlwifi bits, Emmanuel says:

"I have here a few things that depend on the latest mac80211's changes:
RRM, TPC, Quiet Period etc...  Eyal keeps improving our rate control
and we have a new device ID. This last patch should probably have
gone to wireless.git, but at that stage, I preferred to send it to
-next and CC stable."

For (most of) the Atheros bits, Kalle says:

"The only new feature is testmode support from me. Ben added a new method
to crash the firmware with an assert for debug purposes. As usual, we
have lots of smaller fixes from Michal. Matteo fixed a Kconfig
dependency with debugfs. I fixed some warnings recently added to
checkpatch."

For the NFC bits, Samuel says:

"We've had major updates for TI and ST Microelectronics drivers, and a
few NCI related changes.

For TI's trf7970a driver:

- Target mode support for trf7970a
- Suspend/resume support for trf7970a
- DT properties additions to handle different quirks
- A bunch of fixes for smartphone IOP related issues

For ST Microelectronics' ST21NFCA and ST21NFCB drivers:

- ISO15693 support for st21nfcb
- checkpatch and sparse related warning fixes
- Code cleanups and a few minor fixes

Finally, Marvell added ISO15693 support to the NCI stack, together with a
couple of NCI fixes."

For the Bluetooth bits, Johan says:

"This 3.18 pull request replaces the one I did on Monday ("bluetooth-next
2014-09-22", which hasn't been pulled yet). The additions since the last
request are:

 - SCO connection fix for devices not supporting eSCO
 - Cleanups regarding the SCO establishment logic
 - Remove unnecessary return value from logging functions
 - Header compression fix for 6lowpan
 - Cleanups to the ieee802154/mrf24j40 driver

Here's a copy from previous request that this one replaces:

'
Here are some more patches for 3.18. They include various fixes to the
btusb HCI driver, a fix for LE SMP, as well as adding Jukka to the
MAINTAINERS file for generic 6LoWPAN (as requested by Alexander Aring).

I've held on to this pull request a bit since we were waiting for a SCO
related fix to get sorted out first. However, since the merge window is
getting closer I decided not to wait for it. If we do get the fix sorted
out there'll probably be a second small pull request later this week.
'"

And,

"Unless 3.17 gets delayed this will probably be our last -next pull request for
3.18. We've got:

  - New Marvell hardware supportr
  - Multicast support for 6lowpan
  - Several of 6lowpan fixes & cleanups
  - Fix for a (false-positive) lockdep warning in L2CAP
  - Minor btusb cleanup"

On top of all that comes the usual sort of updates to ath5k, ath9k,
ath10k, brcmfmac, mwifiex, and wil6210.  This time around there are
also a number of rtlwifi updates to enable some new hardware and
to reconcile the in-kernel drivers with some newer releases of the
Realtek vendor drivers.  Also of note is some device tree work for
the bcma bus.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Mon, 6 Oct 2014 01:32:37 +0000 (21:32 -0400)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
Netfilter/IPVS updates for net-next

The following patchset contains another batch with Netfilter/IPVS updates
for net-next, they are:

1) Add abstracted ICMP codes to the nf_tables reject expression. We
   introduce four reasons to reject using ICMP that overlap in IPv4
   and IPv6 from the semantic point of view. This should simplify the
   maintainance of dual stack rule-sets through the inet table.

2) Move nf_send_reset() functions from header files to per-family
   nf_reject modules, suggested by Patrick McHardy.

3) We have to use IS_ENABLED(CONFIG_BRIDGE_NETFILTER) everywhere in the
   code now that br_netfilter can be modularized. Convert remaining spots
   in the network stack code.

4) Use rcu_barrier() in the nf_tables module removal path to ensure that
   we don't leave object that are still pending to be released via
   call_rcu (that may likely result in a crash).

5) Remove incomplete arch 32/64 compat from nft_compat. The original (bad)
   idea was to probe the word size based on the xtables match/target info
   size, but this assumption is wrong when you have to dump the information
   back to userspace.

6) Allow to filter from prerouting and postrouting in the nf_tables bridge.
   In order to emulate the ebtables NAT chains (which are actually simple
   filter chains with no special semantics), we have support filtering from
   this hooks too.

7) Add explicit module dependency between xt_physdev and br_netfilter.
   This provides a way to detect if the user needs br_netfilter from
   the configuration path. This should reduce the breakage of the
   br_netfilter modularization.

8) Cleanup coding style in ip_vs.h, from Simon Horman.

9) Fix crash in the recently added nf_tables masq expression. We have
   to register/unregister the notifiers to clean up the conntrack table
   entries from the module init/exit path, not from the rule addition /
   deletion path. From Arturo Borrero.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'bridge_default_pvid'
David S. Miller [Mon, 6 Oct 2014 01:21:44 +0000 (21:21 -0400)]
Merge branch 'bridge_default_pvid'

Vladislav Yasevich says:

====================
bridge: Add vlan filtering support for default pvid

This series adds default pvid support to vlan filtering in the bridge.
VLAN 1 (as recommended by 802.1q spec) is used as default pvid on ports.
Then the user can over-ride this configuration by configuring their
own vlan information.
The user can additionally change the default value through the
sysfs interface (netlink coming shortly).
The user can turn off default pvid functionality by setting default
pvid to 0.
This series changes the default behavior of the bridge when
vlan filtering is turned on.  Currently, ports without any vlan
filtering configured will not recevie any traffic at all.  This patch
changes the behavior of the above ports to receive only untagged traffic.

Since v3:
- allocated 'changed' bitmap on the heap and re-arrange code to clean it up.
- remove extra blank lines.
- Fix patch1 to build by itself.
- Fix error recover to not add vlan 0.
- Restructure nbp_vlan_init to remove uneeded variable.

Since v2:
- Fix handling of invalid values in sysfs interface.
- Add some additional log messages.
- Fix default_pvid handling when vlan filtering is compiled out.
- Fix sparse issues with new code.
- Fix how we located the old default pvid (added a helper function).

Since v1:
- Add ability to turn off default_pvid settings.
- Drop the automiatic filtering support based on configured vlan devices (will
  be its own series)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Add filtering support for default_pvid
Vlad Yasevich [Fri, 3 Oct 2014 15:29:18 +0000 (11:29 -0400)]
bridge: Add filtering support for default_pvid

Currently when vlan filtering is turned on on the bridge, the bridge
will drop all traffic untill the user configures the filter.  This
isn't very nice for ports that don't care about vlans and just
want untagged traffic.

A concept of a default_pvid was recently introduced.  This patch
adds filtering support for default_pvid.   Now, ports that don't
care about vlans and don't define there own filter will belong
to the VLAN of the default_pvid and continue to receive untagged
traffic.

This filtering can be disabled by setting default_pvid to 0.

Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Simplify pvid checks.
Vlad Yasevich [Fri, 3 Oct 2014 15:29:17 +0000 (11:29 -0400)]
bridge: Simplify pvid checks.

Currently, if the pvid is not set, we return an illegal vlan value
even though the pvid value is set to 0.  Since pvid of 0 is currently
invalid, just return 0 instead.  This makes the current and future
checks simpler.

Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobridge: Add a default_pvid sysfs attribute
Vlad Yasevich [Fri, 3 Oct 2014 15:29:16 +0000 (11:29 -0400)]
bridge: Add a default_pvid sysfs attribute

This patch allows the user to set and retrieve default_pvid
value.  A new value can only be stored when vlan filtering
is disabled.

Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: pxa168_eth: avoid using signed char for bitops
Antoine Ténart [Fri, 3 Oct 2014 15:08:19 +0000 (17:08 +0200)]
net: pxa168_eth: avoid using signed char for bitops

Signedness bugs may occur when using signed char for bitops,
depending on if the highest bit is ever used.

Signed-off-by: Antoine Tenart <antoine.tenart@free-electrons.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'isdn-next'
David S. Miller [Mon, 6 Oct 2014 01:17:56 +0000 (21:17 -0400)]
Merge branch 'isdn-next'

Tilman Schmidt says:

====================
ISDN patches for net-next

Here's a series of patches for the ISDN CAPI subsystem and the
Gigaset ISDN driver.  Please merge via net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/gigaset: use USB API function usb_endpoint_num()
Tilman Schmidt [Fri, 3 Oct 2014 15:03:32 +0000 (17:03 +0200)]
isdn/gigaset: use USB API function usb_endpoint_num()

Use function usb_endpoint_num() for the bulk endpoint and store
the endpoint number in the cardstate structure instead of the raw
endpoint address value.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/gigaset: drop unused cardstate structure member
Tilman Schmidt [Fri, 3 Oct 2014 15:03:32 +0000 (17:03 +0200)]
isdn/gigaset: drop unused cardstate structure member

Field int_in_endpointAddr was set but never used. Drop it.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/gigaset: improve error handling when leaving DLE mode
Tilman Schmidt [Fri, 3 Oct 2014 15:03:32 +0000 (17:03 +0200)]
isdn/gigaset: improve error handling when leaving DLE mode

Avoid cascading warnings when leaving DLE mode fails by clearing
the DLE flag before entering recovery.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn/capi: drop two dead if branches
Tilman Schmidt [Fri, 3 Oct 2014 15:03:32 +0000 (17:03 +0200)]
isdn/capi: drop two dead if branches

The last branch in command_2_index() cannot be reached since
c==0xff is already caught by the first "if".
The empty second branch makes no difference since no other branch
will be taken for c<0x0f.

Signed-off-by: Tilman Schmidt <tilman@imap.cc>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: sched: suspicious RCU usage in qdisc_watchdog
John Fastabend [Fri, 3 Oct 2014 05:43:09 +0000 (22:43 -0700)]
net: sched: suspicious RCU usage in qdisc_watchdog

Suspicious RCU usage in qdisc_watchdog call needs to be done inside
rcu_read_lock/rcu_read_unlock. And then Qdisc destroy operations
need to ensure timer is cancelled before removing qdisc structure.

[ 3992.191339] ===============================
[ 3992.191340] [ INFO: suspicious RCU usage. ]
[ 3992.191343] 3.17.0-rc6net-next+ #72 Not tainted
[ 3992.191345] -------------------------------
[ 3992.191347] include/net/sch_generic.h:272 suspicious rcu_dereference_check() usage!
[ 3992.191348]
[ 3992.191348] other info that might help us debug this:
[ 3992.191348]
[ 3992.191351]
[ 3992.191351] rcu_scheduler_active = 1, debug_locks = 1
[ 3992.191353] no locks held by swapper/1/0.
[ 3992.191355]
[ 3992.191355] stack backtrace:
[ 3992.191358] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 3.17.0-rc6net-next+ #72
[ 3992.191360] Hardware name:                  /DZ77RE-75K, BIOS GAZ7711H.86A.0060.2012.1115.1750 11/15/2012
[ 3992.191362]  0000000000000001 ffff880235803e48 ffffffff8178f92c 0000000000000000
[ 3992.191366]  ffff8802322224a0 ffff880235803e78 ffffffff810c9966 ffff8800a5fe3000
[ 3992.191370]  ffff880235803f30 ffff8802359cd768 ffff8802359cd6e0 ffff880235803e98
[ 3992.191374] Call Trace:
[ 3992.191376]  <IRQ>  [<ffffffff8178f92c>] dump_stack+0x4e/0x68
[ 3992.191387]  [<ffffffff810c9966>] lockdep_rcu_suspicious+0xe6/0x130
[ 3992.191392]  [<ffffffff8167213a>] qdisc_watchdog+0x8a/0xb0
[ 3992.191396]  [<ffffffff810f93f2>] __run_hrtimer+0x72/0x420
[ 3992.191399]  [<ffffffff810f9bcd>] ? hrtimer_interrupt+0x7d/0x240
[ 3992.191403]  [<ffffffff816720b0>] ? tc_classify+0xc0/0xc0
[ 3992.191406]  [<ffffffff810f9c4f>] hrtimer_interrupt+0xff/0x240
[ 3992.191410]  [<ffffffff8109e4a5>] ? __atomic_notifier_call_chain+0x5/0x140
[ 3992.191415]  [<ffffffff8103577b>] local_apic_timer_interrupt+0x3b/0x60
[ 3992.191419]  [<ffffffff8179c2b5>] smp_apic_timer_interrupt+0x45/0x60
[ 3992.191422]  [<ffffffff8179a6bf>] apic_timer_interrupt+0x6f/0x80
[ 3992.191424]  <EOI>  [<ffffffff815ed233>] ? cpuidle_enter_state+0x73/0x2e0
[ 3992.191432]  [<ffffffff815ed22e>] ? cpuidle_enter_state+0x6e/0x2e0
[ 3992.191437]  [<ffffffff815ed567>] cpuidle_enter+0x17/0x20
[ 3992.191441]  [<ffffffff810c0741>] cpu_startup_entry+0x3d1/0x4a0
[ 3992.191445]  [<ffffffff81106fc6>] ? clockevents_config_and_register+0x26/0x30
[ 3992.191448]  [<ffffffff81033c16>] start_secondary+0x1b6/0x260

Fixes: b26b0d1e8b1 ("net: qdisc: use rcu prefix and silence sparse warnings")
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: dsa: do not call phy_start_aneg
Florian Fainelli [Fri, 3 Oct 2014 01:56:03 +0000 (18:56 -0700)]
net: dsa: do not call phy_start_aneg

Commit f7f1de51edbd ("net: dsa: start and stop the PHY state machine")
add calls to phy_start() in dsa_slave_open() respectively phy_stop() in
dsa_slave_close().

We also call phy_start_aneg() in dsa_slave_create(), and this call is
messing up with the PHY state machine, since we basically start the
auto-negotiation, and later on restart it when calling phy_start().
phy_start() does not currently handle the PHY_FORCING or PHY_AN states
properly, but such a fix would be too invasive for this window.

Fixes: f7f1de51edbd ("net: dsa: start and stop the PHY state machine")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoRemoved unused inet6 address state
Sébastien Barré [Thu, 2 Oct 2014 19:15:22 +0000 (21:15 +0200)]
Removed unused inet6 address state

the inet6 state INET6_IFADDR_STATE_UP only appeared in its definition.

Cc: Christoph Paasch <christoph.paasch@uclouvain.be>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sébastien Barré <sebastien.barre@uclouvain.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Cleanup skb cloning by adding SKB_FCLONE_FREE
Vijay Subramanian [Thu, 2 Oct 2014 17:00:43 +0000 (10:00 -0700)]
net: Cleanup skb cloning by adding SKB_FCLONE_FREE

SKB_FCLONE_UNAVAILABLE has overloaded meaning depending on type of skb.
1: If skb is allocated from head_cache, it indicates fclone is not available.
2: If skb is a companion fclone skb (allocated from fclone_cache), it indicates
it is available to be used.

To avoid confusion for case 2 above, this patch  replaces
SKB_FCLONE_UNAVAILABLE with SKB_FCLONE_FREE where appropriate. For fclone
companion skbs, this indicates it is free for use.

SKB_FCLONE_UNAVAILABLE will now simply indicate skb is from head_cache and
cannot / will not have a companion fclone.

Signed-off-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agomlx4: add a new xmit_more counter
Eric Dumazet [Thu, 2 Oct 2014 15:24:21 +0000 (08:24 -0700)]
mlx4: add a new xmit_more counter

ethtool -S reports a new counter, tracking number of time doorbell
was not triggered, because skb->xmit_more was set.

$ ethtool -S eth0 | egrep "tx_packet|xmit_more"
     tx_packets: 2413288400
     xmit_more: 666121277

I merged the tso_packet false sharing avoidance in this patch as well.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'gudp'
David S. Miller [Fri, 3 Oct 2014 23:53:36 +0000 (16:53 -0700)]
Merge branch 'gudp'

Tom Herbert says:

====================
net: Generic UDP Encapsulation

Generic UDP Encapsulation (GUE) is UDP encapsulation protocol which
encapsulates packets of various IP protocols. The GUE protocol is
described in http://tools.ietf.org/html/draft-herbert-gue-01.

The receive path of GUE is implemented in the FOU over UDP module (FOU).
This includes a UDP encap receive function for GUE as well as GUE
specific GRO functions. Management and configuration of GUE ports shares
most of the same code with FOU.

For the transmit path, the previous FOU support for IPIP, sit, and GRE
was simply extended for GUE (when GUE is enabled insert the GUE
header on transmit in addition to UDP header inserted for FOU).

Semantically GUE is the same as FOU in that the encapsulation (UDP
and GUE headers) that are inserted on transmission and removed on
reception so that IP packet is processed with the inner header.

This patch set includes:
 - Some fixes to FOU, removal of IPv4,v6 specific GRO functions
 - Support to configure a GUE receive port
 - Implementation of GUE receive path (normal and GRO)
 - Additions to ip_tunnel netlink to configure GUE
 - GUE header inserion in ip_tunnel transmit path

v2:
 - Include net/gue.h in patch set

Testing:

I ran performance numbers using netperf TCP_RR with 200 streams,
comparing encapsulation without GUE, encapsulation with GUE, and
encapsulation with FOU.

 GRE
    TCP_STREAM
      IPv4, FOU, UDP checksum enabled
        14.04% TX CPU utilization
        13.17% RX CPU utilization
        9211 Mbps
      IPv4, GUE, UDP checksum enabled
        14.99% TX CPU utilization
        13.79% RX CPU utilization
        9185 Mbps
      IPv4, FOU, UDP checksum disabled
        13.14% TX CPU utilization
        23.18% RX CPU utilization
        9277 Mbps
      IPv4, GUE, UDP checksum disabled
        13.66% TX CPU utilization
        23.57% RX CPU utilization
        9184 Mbps
    TCP_RR
      IPv4, FOU, UDP checksum enabled
        94.2% CPU utilization
        155/249/460 90/95/99% latencies
        1.17018e+06 tps
      IPv4, GUE, UDP checksum enabled
        93.9% CPU utilization
        158/253/472 90/95/99% latencies
        1.15045e+06 tps

  IPIP
    TCP_STREAM
      FOU, UDP checksum enabled
        15.28% TX CPU utilization
        13.92% RX CPU utilization
        9342 Mbps
      GUE, UDP checksum enabled
        13.99% TX CPU utilization
        13.34% RX CPU utilization
        9210 Mbps
      FOU, UDP checksum disabled
        15.08% TX CPU utilization
        24.64% RX CPU utilization
        9226 Mbps
      GUE, UDP checksum disabled
        15.90% TX CPU utilization
        24.77% RX CPU utilization
        9197 Mbps
    TCP_RR
      FOU, UDP checksum enabled
        94.23% CPU utilization
        149/237/429 90/95/99% latencies
        1.19553e+06 tps
      GUE, UDP checksum enabled
        93.75% CPU utilization
        152/243/442 90/95/99% latencies
        1.17027e+06 tps

  SIT
    TCP_STREAM
      FOU, UDP checksum enabled
        14.47% TX CPU utilization
        14.58% RX CPU utilization
        9106 Mbps
      GUE, UDP checksum enabled
        15.09% TX CPU utilization
        14.84% RX CPU utilization
        9080 Mbps
      FOU, UDP checksum disabled
        15.70% TX CPU utilization
        27.93% RX CPU utilization
        9097 Mbps
      GUE, UDP checksum disabled
        15.04% TX CPU utilization
        27.54% RX CPU utilization
        9073 Mbps
    TCP_RR
      FOU, UDP checksum enabled
        96.9% CPU utilization
        170/281/581 90/95/99% latencies
        1.03372e+06 tps
      GUE, UDP checksum enabled
        97.16% CPU utilization
        172/286/576 90/95/99% latencies
        1.00469e+06 tps
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoip_tunnel: Add GUE support
Tom Herbert [Fri, 3 Oct 2014 22:48:10 +0000 (15:48 -0700)]
ip_tunnel: Add GUE support

This patch allows configuring IPIP, sit, and GRE tunnels to use GUE.
This is very similar to fou excpet that we need to insert the GUE header
in addition to the UDP header on transmit.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agogue: Receive side for Generic UDP Encapsulation
Tom Herbert [Fri, 3 Oct 2014 22:48:09 +0000 (15:48 -0700)]
gue: Receive side for Generic UDP Encapsulation

This patch adds support receiving for GUE packets in the fou module. The
fou module now supports direct foo-over-udp (no encapsulation header)
and GUE. To support this a type parameter is added to the fou netlink
parameters.

For a GUE socket we define gue_udp_recv, gue_gro_receive, and
gue_gro_complete to handle the specifics of the GUE protocol. Most
of the code to manage and configure sockets is common with the fou.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agofou: eliminate IPv4,v6 specific GRO functions
Tom Herbert [Fri, 3 Oct 2014 22:48:08 +0000 (15:48 -0700)]
fou: eliminate IPv4,v6 specific GRO functions

This patch removes fou[46]_gro_receive and fou[46]_gro_complete
functions. The v4 or v6 variants were chosen for the UDP offloads
based on the address family of the socket this is not necessary
or correct. Alternatively, this patch adds is_ipv6 to napi_gro_skb.
This is set in udp6_gro_receive and unset in udp4_gro_receive. In
fou_gro_receive the value is used to select the correct inet_offloads
for the protocol of the outer IP header.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoip_tunnel: Account for secondary encapsulation header in max_headroom
Tom Herbert [Fri, 3 Oct 2014 22:48:07 +0000 (15:48 -0700)]
ip_tunnel: Account for secondary encapsulation header in max_headroom

When adjusting max_header for the tunnel interface based on egress
device we need to account for any extra bytes in secondary encapsulation
(e.g. FOU).

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: do not export skb_gro_receive()
Eric Dumazet [Thu, 2 Oct 2014 14:38:46 +0000 (07:38 -0700)]
net: do not export skb_gro_receive()

skb_gro_receive() is only called from tcp_gro_receive() which is
not in a module.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/irda/Kconfig: Let SH_IRDA depend on HAS_IOMEM
Chen Gang [Thu, 2 Oct 2014 14:32:56 +0000 (22:32 +0800)]
drivers/net/irda/Kconfig: Let SH_IRDA depend on HAS_IOMEM

SH_IRDA needs HAS_IOMEM, so depend on it. The related error(with
allmodconfig under um):

    CC [M]  drivers/net/irda/sh_irda.o
  drivers/net/irda/sh_irda.c: In function ‘sh_irda_probe’:
  drivers/net/irda/sh_irda.c:776:2: error: implicit declaration of function ‘ioremap_nocache’ [-Werror=implicit-function-declaration]
    self->membase = ioremap_nocache(res->start, resource_size(res));
    ^
  drivers/net/irda/sh_irda.c:776:16: warning: assignment makes pointer from integer without a cast [enabled by default]
    self->membase = ioremap_nocache(res->start, resource_size(res));
                  ^
  drivers/net/irda/sh_irda.c:821:2: error: implicit declaration of function ‘iounmap’ [-Werror=implicit-function-declaration]
    iounmap(self->membase);
    ^

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/ethernet/marvell/Kconfig: Let PXA168_ETH depend on HAS_IOMEM
Chen Gang [Thu, 2 Oct 2014 14:23:33 +0000 (22:23 +0800)]
drivers/net/ethernet/marvell/Kconfig: Let PXA168_ETH depend on HAS_IOMEM

PXA168_ETH need HAS_IOMEM, so depend on it, the related error (with
allmodconfig under um):

    CC [M]  drivers/net/ethernet/marvell/pxa168_eth.o
  drivers/net/ethernet/marvell/pxa168_eth.c: In function ‘pxa168_eth_probe’:
  drivers/net/ethernet/marvell/pxa168_eth.c:1605:2: error: implicit declaration of function ‘iounmap’ [-Werror=implicit-function-declaration]
    iounmap(pep->base);
    ^

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/dsa/Kconfig: Let NET_DSA_BCM_SF2 depend on HAS_IOMEM
Chen Gang [Thu, 2 Oct 2014 14:14:04 +0000 (22:14 +0800)]
drivers/net/dsa/Kconfig: Let NET_DSA_BCM_SF2 depend on HAS_IOMEM

NET_DSA_BCM_SF2 need HAS_IOMEM, so depend on it, the related error (with
allmodconfig under um):

    CC [M]  drivers/net/dsa/bcm_sf2.o
  drivers/net/dsa/bcm_sf2.c: In function ‘bcm_sf2_sw_setup’:
  drivers/net/dsa/bcm_sf2.c:487:3: error: implicit declaration of function ‘iounmap’ [-Werror=implicit-function-declaration]
     iounmap(*base);
     ^

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers/net/can/Kconfig: Let CAN_AT91 depend on HAS_IOMEM
Chen Gang [Thu, 2 Oct 2014 14:01:42 +0000 (22:01 +0800)]
drivers/net/can/Kconfig: Let CAN_AT91 depend on HAS_IOMEM

CAN_AT91 needs HAS_IOMEM, so depends on it. The related error (with
allmodconfig under um):

    CC [M]  drivers/net/can/at91_can.o
  drivers/net/can/at91_can.c: In function ‘at91_can_probe’:
  drivers/net/can/at91_can.c:1329:2: error: implicit declaration of function ‘ioremap_nocache’ [-Werror=implicit-function-declaration]
  addr = ioremap_nocache(res->start, resource_size(res));
    ^
  drivers/net/can/at91_can.c:1329:7: warning: assignment makes pointer from integer without a cast [enabled by default]
    addr = ioremap_nocache(res->start, resource_size(res));
         ^
  drivers/net/can/at91_can.c:1384:2: error: implicit declaration of function ‘iounmap’ [-Werror=implicit-function-declaration]
    iounmap(addr);
    ^

Signed-off-by: Chen Gang <gang.chen.5i5j@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Fri, 3 Oct 2014 22:43:50 +0000 (15:43 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2014-10-02

This series contains updates to fm10k, igb, ixgbe and i40e.

Alex provides two updates to the fm10k driver.  First reduces the buffer
size to 2k for all page sizes, since most frames only have a 1500 MTU
so supporting a buffer size larger than this is somewhat wasteful.
Second fixes an issue where the number of transmit queues was not being
updated, so added the lines necessary to update the number of transmit
queues.

Rick Jones provides two patches to convert ixgbe, igb and i40e to use
dev_consume_skb_any().

Emil provides two patches for ixgbe, first cleans up a couple of wait
loops on auto-negotiation that were not needed.  Second fixes an issue
reported by Fujitsu/Red Hat, which consolidates the logic behind the
dynamically setting of TXDCTL.WTHRESH depending on interrupt throttle
rate (ITR) setting regardless of BQL.

Ethan Zhao provides a cleanup patch for ixgbe where he noticed a
duplicate define.

Bernhard Kaindl provides a patch for igb to remove a source of latency
spikes by not calling code that uses mdelay() for feeding a PHY stat
while being called with a spinlock held.

Todd bumps the igb version based on the recent changes.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'mlx5-next'
David S. Miller [Fri, 3 Oct 2014 22:42:37 +0000 (15:42 -0700)]
Merge branch 'mlx5-next'

Eli Cohen says:

====================
mlx5 update for 3.18

This series integrates a new mechanism for populating and extracting field values
used in the driver/firmware interaction around command mailboxes.

Changes from V1:
 - Remove unused definition of memcpy_cpu_to_be32()
 - Remove definitions of non_existent_*() and use BUILD_BUG_ON() instead.
 - Added a patch one line patch to add support for ConnectX-4 devices.

Changes from V0:
 - trimmed the auto-generated file to a minimum, as required by the reviewers.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx5_core: Add ConnectX-4 to list of supported devices
Eli Cohen [Thu, 2 Oct 2014 09:19:46 +0000 (12:19 +0300)]
net/mlx5_core: Add ConnectX-4 to list of supported devices

Add the upcoming ConnectX-4 device to the list of supported devices by then
mlx5 driver.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>