net: bonding: factor out rlock(bond->lock) in xmit path

Pull read_lock(&bond->lock) and BOND_IS_OK() to bond_start_xmit() from
mode-dependent xmit functions.

netif_running() is always true in hard_start_xmit.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

vlan: remove one synchronize_net() call

At VLAN dismantle phase, unregister_vlan_dev() makes one
synchronize_net() call after vlan_group_set_device(grp, vlan_id, NULL).

This call can be safely removed because we are calling
unregister_netdevice_queue() to queue device for deletion, and this
process needs at least one rcu grace period to complete.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ben Greear <greearb@candelatech.com>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Jesse Gross <jesse@nicira.com>
Cc: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Acked-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

garp: remove one synchronize_rcu() call

Speedup vlan dismantling in CONFIG_VLAN_8021Q_GVRP=y cases,
by using a call_rcu() to free the memory instead of waiting with
expensive synchronize_rcu() [ while RTNL is held ]

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ben Greear <greearb@candelatech.com>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: use batched device unregister in veth and macvlan

veth devices dont use the batched device unregisters yet.

Since veth are a pair of devices, it makes sense to use a batch of two
unregisters, this roughly divides dismantle time by two.

Fix this by changing dellink() callers to always provide a non NULL
head. (Idea from Michał Mirosław)

This patch also handles macvlan case : We now dismantle all macvlans on
top of a lower dev at once.

Reported-by: Alex Bligh <alex@alex.org.uk>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Michał Mirosław <mirqus@gmail.com>
Cc: Jesse Gross <jesse@nicira.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Ben Greear <greearb@candelatech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: Fix debug message args.

I messed things up when I converted over to the transport
flow, I passed the ipv4 address value instead of it's address.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Pass flow key down into ip_append_*().

This way rt->rt_dst accesses are unnecessary.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Pass flow keys down into datagram packet building engine.

This way ip_output.c no longer needs rt->rt_{src,dst}.

We already have these keys sitting, ready and waiting, on the stack or
in a socket structure.

Signed-off-by: David S. Miller <davem@davemloft.net>

udp: Use flow key information instead of rt->rt_{src,dst}

We have two cases.

Either the socket is in TCP_ESTABLISHED state and connect() filled
in the inet socket cork flow, or we looked up the route here and
used an on-stack flow.

Track which one it was, and use it to obtain src/dst addrs.

Signed-off-by: David S. Miller <davem@davemloft.net>

net: Allow ethtool to set interface in loopback mode.

This patch enables ethtool to set the loopback mode on a given interface.
By configuring the interface in loopback mode in conjunction with a policy
route / rule, a userland application can stress the egress / ingress path
exposing the flows of the change in progress and potentially help developer(s)
understand the impact of those changes without even sending a packet out
on the network.

Following set of commands illustrates one such example -
    a) ip -4 addr add 192.168.1.1/24 dev eth1
    b) ip -4 rule add from all iif eth1 lookup 250
    c) ip -4 route add local 0/0 dev lo proto kernel scope host table 250
    d) arp -Ds 192.168.1.100 eth1
    e) arp -Ds 192.168.1.200 eth1
    f) sysctl -w net.ipv4.ip_nonlocal_bind=1
    g) sysctl -w net.ipv4.conf.all.accept_local=1
    # Assuming that the machine has 8 cores
    h) taskset 000f netserver -L 192.168.1.200
    i) taskset 00f0 netperf -t TCP_CRR -L 192.168.1.100 -H 192.168.1.200 -l 30

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pktgen: use %pI6c for printing IPv6 addresses

I don't know why %pI6 doesn't compress, but the format specifier is
kernel-standard, so use it.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ethtool: Add 20G bit definitions

Add 20G supported and advertising bit definitions.
20G will be supported with the 57840 chips.

Signed-off-by: Yaniv Rosner <yanivr@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
------
 include/linux/ethtool.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'batman-adv/next' of git://git.open-mesh.org/ecsv/linux-merge

tcp: Use cork flow info instead of rt->rt_dst in tcp_v4_get_peer()

Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: Don't use rt->rt_{src,dst} in sctp_v4_xmit()

Now we can pick it out of the transport's flow key.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Don't use rt->rt_{src,dst} in ip_queue_xmit().

Now we can pick it out of the provided flow key.

Signed-off-by: David S. Miller <davem@davemloft.net>

inet: Pass flowi to ->queue_xmit().

This allows us to acquire the exact route keying information from the
protocol, however that might be managed.

It handles all of the possibilities, from the simplest case of storing
the key in inet->cork.fl to the more complex setup SCTP has where
individual transports determine the flow.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Use inet_csk_route_child_sock() in DCCP and TCP.

Operation order is now transposed, we first create the child
socket then we try to hook up the route.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Create inet_csk_route_child_sock().

This is just like inet_csk_route_req() except that it operates after
we've created the new child socket.

In this way we can use the new socket's cork flow for proper route
key storage.

This will be used by DCCP and TCP child socket creation handling.

Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: Store a flowi in transports to provide persistent keying.

Several future simplifications are possible now because of this.

For example, the sctp_addr unions can simply refer directly to
the flowi information.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Use cork flow in ip_queue_xmit()

All invokers of ip_queue_xmit() must make certain that the
socket is locked.  All of SCTP, TCP, DCCP, and L2TP now make
sure this is the case.

Therefore we can use the cork flow during output route lookup in
ip_queue_xmit() when the socket route check fails.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Use cork flow in inet_sk_{reselect_saddr,rebuild_header}()

These two functions must be invoked only when the socket is locked
(because socket identity modifications are made non-atomically).

Therefore we can use the cork flow for output route lookups.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Lock socket and use cork flow in ip4_datagram_connect().

This is to make sure that an l2tp socket's inet cork flow is
fully filled in, when it's encapsulated in UDP.

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Use cork flow in l2tp_ip_connect() and l2tp_ip_sendmsg()

Now that the socket is consistently locked in these two routines,
this transformation is legal.

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Fix locking in l2tp_core.c

l2tp_xmit_skb() must take the socket lock.  It makes use of ip_queue_xmit()
which expects to execute in a socket atomic context.

Since we execute this function in software interrupts, we cannot use the
usual lock_sock()/release_sock() sequence, instead we have to use
bh_lock_sock() and see if a user has the socket locked, and if so drop
the packet.

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Fix locking in l2tp_ip.c

Both l2tp_ip_connect() and l2tp_ip_sendmsg() must take the socket
lock.  They both modify socket state non-atomically, and in particular
l2tp_ip_sendmsg() increments socket private counters without using
atomic operations.
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: Use cork flow in tcp_v4_connect()

Since this is invoked from inet_stream_connect() the socket is locked
and therefore this usage is safe.

Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Use cork flow in dccp_v4_connect()

Since this is invoked from inet_stream_connect() the socket is locked
and therefore this usage is safe.

Signed-off-by: David S. Miller <davem@davemloft.net>

ethtool: remove phys_id from ethtool_ops

After that all the upstream kernel drivers now use phys_id,
and the old ethtool_ops interface (phys_id) can be removed.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

batman-adv: remove duplicate code from function is_bidirectional_neigh()

In function is_bidirectional_neigh the code that find out the one hop
neighbor is duplicated.

Signed-off-by: Daniele Furlan <daniele.furlan@gmail.com>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: Remove multiline comments from line ending

It is slightly irritating that comments after a long line span over
multiple lines without any code. It is easier to put them before the
actual code and reduce the number of lines which the eye has to read.

Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: rename everything from *hna* into *tt* (translation table)

To be coherent, all the functions/variables/constants have been renamed
to the TranslationTable style

Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: Fix refcount imbalance in find_router

Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: Avoid deadlock between rtnl_lock and s_active

The hard_if_event is called by the notifier with rtnl_lock and tries to
remove sysfs entries when a NETDEV_UNREGISTER event is received. This
will automatically take the s_active lock.

The s_active lock is also used when a new interface is added to a meshif
through sysfs. In that situation we cannot wait for the rntl_lock before
creating the actual batman-adv interface to prevent a deadlock. It is
still possible to try to get the rtnl_lock and immediately abort the
current operation when the trylock call failed.

Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: Remove unnecessary hardif_list_lock

hardif_list_lock is unneccessary because we already ensure that no
multiple admin operations can take place through rtnl_lock.
hardif_list_lock only adds additional overhead and complexity.

Critical functions now check whether they are called with rtnl_lock
using ASSERT_RTNL.

It indirectly fixes the problem that orig_hash_del_if() expects that
only one interface is deleted from hardif_list at a time, but
hardif_remove_interfaces() removes all at once and then calls
orig_hash_del_if().

Reported-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: multi vlan support for bridge loop detection

The bridge loop detection for batman-adv allows the bat0 interface
to be bridged into an ethernet segment which other batman-adv nodes
are connected to. In order to also allow multiple VLANs on top of
the bat0 interface to be bridged into the ethernet segment this
patch extends the aforementioned bridge loop detection.

Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

batman-adv: remove misplaced comment

Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>

Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/jkirsher/net-next-2.6

hamachi: Put back RX_CHECKSUM

I deleted it by mistake in the TX_CHECKSUM removal
commit.

Reported-by: Michał Mirosław <mirqus@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: add ethtool counters for OS2BMC

OS2BMC registers are available for X540.
This patch adds ethtool counters based on those registers.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: add rxhash support

feed RSS hash into skb->rxhash

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: convert to ethtool set_phys_id

Based on patch from Stephen Hemminger.
Convert igb driver to use new set_phys_id ethtool interface.

CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

e1000: convert to set_phys_id

Based on the original patch from Stephen Hemminger.
Convert to new LED control infrastucture and remove no longer
necessary bits.

CC: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

e100: implemenet set_phys_id

Based on the original patch from Stephen Hemminger.
Implement set_phys_id to control LED.

CC: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ipv4: Initialize cork->opt using NULL not 0.

Noticed by Joe Perches.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Initialize on-stack cork more efficiently.

ip_setup_cork() explicitly initializes every member of
inet_cork except flags, addr, and opt.  So we can simply
set those three members to zero instead of using a
memset() via an empty struct assignment.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>

inet: Decrease overhead of on-stack inet_cork.

When we fast path datagram sends to avoid locking by putting
the inet_cork on the stack we use up lots of space that isn't
necessary.

This is because inet_cork contains a "struct flowi" which isn't
used in these code paths.

Split inet_cork to two parts, "inet_cork" and "inet_cork_full".
Only the latter of which has the "struct flowi" and is what is
stored in inet_sock.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>

hamachi: Delete TX checksumming code commented out since 1999

TX checksumming support has been ifdef commented out of this driver
for more than 10 years, and it makes references to aspects of the IPv4
stack from back then as well.

If someone has one of these rare cards and wants to properly resurrect
TX checksumming support, they can still get at this code in the
version control history.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

Conflicts:
drivers/net/tg3.c

Merge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next-2.6

net: Add sendmmsg socket system call

This patch adds a multiple message send syscall and is the send
version of the existing recvmmsg syscall. This is heavily
based on the patch by Arnaldo that added recvmmsg.

I wrote a microbenchmark to test the performance gains of using
this new syscall:

http://ozlabs.org/~anton/junkcode/sendmmsg_test.c

The test was run on a ppc64 box with a 10 Gbit network card. The
benchmark can send both UDP and RAW ethernet packets.

64B UDP

batch   pkts/sec
1       804570
2       872800 (+ 8 %)
4       916556 (+14 %)
8       939712 (+17 %)
16      952688 (+18 %)
32      956448 (+19 %)
64      964800 (+20 %)

64B raw socket

batch   pkts/sec
1       1201449
2       1350028 (+12 %)
4       1461416 (+22 %)
8       1513080 (+26 %)
16      1541216 (+28 %)
32      1553440 (+29 %)
64      1557888 (+30 %)

We see a 20% improvement in throughput on UDP send and 30%
on raw socket send.

[ Add sparc syscall entries. -DaveM ]

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: fix SJA1000 dlc for RTR packets

RTR frames do have a valid data length code on CAN.
The driver for SJA1000 did not handle that situation properly.

Signed-off-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: call dev_alloc_name from register_netdevice

Force dev_alloc_name() to be called from register_netdevice() by
dev_get_valid_name(). That allows to remove multiple explicit
dev_alloc_name() calls.

The possibility to call dev_alloc_name in advance remains.

This also fixes veth creation regresion caused by
84c49d8c3e4abefb0a41a77b25aa37ebe8d6b743

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: function descriptions format fixed

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: update year to 2011 and version to 1.62.12-0

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: improve memory handling, low memory recovery flows

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Do storage mac address validation for SF mode.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: allow WoL on every function in MF modes

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: link report improvements

To avoid link notification duplication

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

stmmac: removed not used definitions

Reported-by: Karim Hamiti <karim.hamiti@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

usbnet: runtime pm: fix out of memory

This patch makes use of the EVENT_DEV_OPEN flag introduced recently to
fix one out of memory issue, which can be reproduced on omap3/4 based
pandaboard/beagle XM easily with steps below:

- enable runtime pm
echo auto > /sys/devices/platform/usbhs-omap.0/ehci-omap.0/usb1/1-1/1-1.1/power/control

- ifconfig eth0 up

- then out of memroy happened, see [1] for kernel message.

Follows my analysis:
- 'ifconfig eth0 up' brings eth0 out of suspend, and usbnet_resume
is called to schedule dev->bh, then rx urbs are submited to prepare for
recieving data;

- some usbnet devices will produce garbage rx packets flood if
info->reset is not called in usbnet_open.

- so there is no enough chances for usbnet_bh to handle and release
recieved skb buffers since many rx interrupts consumes cpu, so out of memory
for atomic allocation in rx_submit happened.

This patch fixes the issue by simply not allowing schedule of usbnet_bh until device
is opened.

[1], dmesg
[  234.712005] smsc95xx 1-1.1:1.0: rpm_resume flags 0x4
[  234.712066] usb 1-1.1: rpm_resume flags 0x0
[  234.712066] usb 1-1: rpm_resume flags 0x0
[  234.712097] usb usb1: rpm_resume flags 0x0
[  234.712127] usb usb1: usb auto-resume
[  234.712158] ehci-omap ehci-omap.0: resume root hub
[  234.754028] hub 1-0:1.0: hub_resume
[  234.754821] hub 1-0:1.0: port 1: status 0507 change 0000
[  234.756011] hub 1-0:1.0: state 7 ports 3 chg 0000 evt 0000
[  234.756042] hub 1-0:1.0: rpm_resume flags 0x4
[  234.756072] usb usb1: rpm_resume flags 0x0
[  234.756164] usb usb1: rpm_resume returns 1
[  234.756195] hub 1-0:1.0: rpm_resume returns 0
[  234.756195] hub 1-0:1.0: rpm_suspend flags 0x4
[  234.756225] hub 1-0:1.0: rpm_suspend returns 0
[  234.756256] usb usb1: rpm_resume returns 0
[  234.757141] usb 1-1: usb auto-resume
[  234.793151] ehci-omap ehci-omap.0: GetStatus port:1 status 001005 0  ACK POWER sig=se0 PE CONNECT
[  234.816558] usb 1-1: finish resume
[  234.817871] hub 1-1:1.0: hub_resume
[  234.818420] hub 1-1:1.0: port 1: status 0507 change 0000
[  234.820495] ehci-omap ehci-omap.0: reused qh eec50220 schedule
[  234.820495] usb 1-1: link qh256-0001/eec50220 start 1 [1/0 us]
[  234.820587] usb 1-1: rpm_resume returns 0
[  234.820800] hub 1-1:1.0: state 7 ports 5 chg 0000 evt 0000
[  234.820800] hub 1-1:1.0: rpm_resume flags 0x4
[  234.820831] hub 1-1:1.0: rpm_resume returns 0
[  234.820861] hub 1-1:1.0: rpm_suspend flags 0x4
[  234.820861] hub 1-1:1.0: rpm_suspend returns 0
[  234.821777] usb 1-1.1: usb auto-resume
[  234.868591] hub 1-1:1.0: state 7 ports 5 chg 0000 evt 0002
[  234.868591] hub 1-1:1.0: rpm_resume flags 0x4
[  234.868621] hub 1-1:1.0: rpm_resume returns 0
[  234.868652] hub 1-1:1.0: rpm_suspend flags 0x4
[  234.868652] hub 1-1:1.0: rpm_suspend returns 0
[  234.879486] usb 1-1.1: finish resume
[  234.880279] usb 1-1.1: rpm_resume returns 0
[  234.880310] smsc95xx 1-1.1:1.0: rpm_resume returns 0
[  238.880187] ksoftirqd/0: page allocation failure. order:0, mode:0x20
[  238.880218] Backtrace:
[  238.880249] [<c01b9800>] (dump_backtrace+0x0/0xf8) from [<c065e1dc>] (dump_stack+0x18/0x1c)
[  238.880249]  r6:00000000 r5:00000000 r4:00000020 r3:00000002
[  238.880310] [<c065e1c4>] (dump_stack+0x0/0x1c) from [<c026ece4>] (__alloc_pages_nodemask+0x620/0x724)
[  238.880340] [<c026e6c4>] (__alloc_pages_nodemask+0x0/0x724) from [<c02986d4>] (kmem_getpages.clone.34+0x34/0xc8)
[  238.880371] [<c02986a0>] (kmem_getpages.clone.34+0x0/0xc8) from [<c02988f8>] (cache_grow.clone.42+0x84/0x154)
[  238.880371]  r6:ef871aa4 r5:ef871a80 r4:ef81fd40 r3:00000020
[  238.880401] [<c0298874>] (cache_grow.clone.42+0x0/0x154) from [<c0298b64>] (cache_alloc_refill+0x19c/0x1f0)
[  238.880432] [<c02989c8>] (cache_alloc_refill+0x0/0x1f0) from [<c0299804>] (kmem_cache_alloc+0x90/0x190)
[  238.880462] [<c0299774>] (kmem_cache_alloc+0x0/0x190) from [<c052e260>] (__alloc_skb+0x34/0xe8)
[  238.880493] [<c052e22c>] (__alloc_skb+0x0/0xe8) from [<bf0509f4>] (rx_submit+0x2c/0x1d4 [usbnet])
[  238.880523] [<bf0509c8>] (rx_submit+0x0/0x1d4 [usbnet]) from [<bf050d38>] (rx_complete+0x19c/0x1b0 [usbnet])
[  238.880737] [<bf050b9c>] (rx_complete+0x0/0x1b0 [usbnet]) from [<bf006fd0>] (usb_hcd_giveback_urb+0xa8/0xf4 [usbcore])
[  238.880737]  r8:eeeced34 r7:eeecec00 r6:eeecec00 r5:00000000 r4:eec2dd20
[  238.880767] r3:bf050b9c
[  238.880859] [<bf006f28>] (usb_hcd_giveback_urb+0x0/0xf4 [usbcore]) from [<bf03c8f8>] (ehci_urb_done+0xb0/0xbc [ehci_hcd])
[  238.880859]  r6:00000000 r5:eec2dd20 r4:eeeced44 r3:eec2dd34
[  238.880920] [<bf03c848>] (ehci_urb_done+0x0/0xbc [ehci_hcd]) from [<bf040204>] (qh_completions+0x308/0x3bc [ehci_hcd])
[  238.880920]  r7:00000000 r6:eeda21a0 r5:ffdfe3c0 r4:eeda21ac
[  238.880981] [<bf03fefc>] (qh_completions+0x0/0x3bc [ehci_hcd]) from [<bf040ef8>] (scan_async+0xb0/0x16c [ehci_hcd])
[  238.881011] [<bf040e48>] (scan_async+0x0/0x16c [ehci_hcd]) from [<bf040fec>] (ehci_work+0x38/0x90 [ehci_hcd])
[  238.881042] [<bf040fb4>] (ehci_work+0x0/0x90 [ehci_hcd]) from [<bf042940>] (ehci_irq+0x300/0x34c [ehci_hcd])
[  238.881072]  r4:eeeced34 r3:00000001
[  238.881134] [<bf042640>] (ehci_irq+0x0/0x34c [ehci_hcd]) from [<bf006828>] (usb_hcd_irq+0x40/0xac [usbcore])
[  238.881195] [<bf0067e8>] (usb_hcd_irq+0x0/0xac [usbcore]) from [<c0239764>] (handle_irq_event_percpu+0xb8/0x240)
[  238.881225]  r6:eec504e0 r5:0000006d r4:eec504e0 r3:bf0067e8
[  238.881256] [<c02396ac>] (handle_irq_event_percpu+0x0/0x240) from [<c0239930>] (handle_irq_event+0x44/0x64)
[  238.881256] [<c02398ec>] (handle_irq_event+0x0/0x64) from [<c023bbd0>] (handle_level_irq+0xe0/0x114)
[  238.881286]  r6:0000006d r5:c080c14c r4:c080c100 r3:00020000
[  238.881317] [<c023baf0>] (handle_level_irq+0x0/0x114) from [<c01ab090>] (asm_do_IRQ+0x90/0xd0)
[  238.881317]  r5:00000000 r4:0000006d
[  238.881347] [<c01ab000>] (asm_do_IRQ+0x0/0xd0) from [<c06624d0>] (__irq_svc+0x50/0x134)
[  238.881378] Exception stack(0xef837e20 to 0xef837e68)
[  238.881378] 7e20: 00000001 00185610 016cc000 c00490c0 eb380000 ef800540 00000020 00004ae0
[  238.881408] 7e40: 00000020 bf0509f4 60000013 ef837e9c ef837e40 ef837e68 c0226f0c c0298ca0
[  238.881408] 7e60: 20000013 ffffffff
[  238.881408]  r5:fa240100 r4:ffffffff
[  238.881439] [<c0298bb8>] (__kmalloc_track_caller+0x0/0x1d0) from [<c052e284>] (__alloc_skb+0x58/0xe8)
[  238.881469] [<c052e22c>] (__alloc_skb+0x0/0xe8) from [<bf0509f4>] (rx_submit+0x2c/0x1d4 [usbnet])
[  238.881500] [<bf0509c8>] (rx_submit+0x0/0x1d4 [usbnet]) from [<bf0513d8>] (usbnet_bh+0x1b4/0x250 [usbnet])
[  238.881530] [<bf051224>] (usbnet_bh+0x0/0x250 [usbnet]) from [<c01f912c>] (tasklet_action+0xb0/0x1f8)
[  238.881530]  r6:00000000 r5:ef9757f0 r4:ef9757ec r3:bf051224
[  238.881561] [<c01f907c>] (tasklet_action+0x0/0x1f8) from [<c01f97ac>] (__do_softirq+0x140/0x290)
[  238.881561]  r8:00000006 r7:00000101 r6:00000000 r5:c0806098 r4:00000001
[  238.881591] r3:c01f907c
[  238.881622] [<c01f966c>] (__do_softirq+0x0/0x290) from [<c01f99cc>] (run_ksoftirqd+0xd0/0x1f4)
[  238.881622] [<c01f98fc>] (run_ksoftirqd+0x0/0x1f4) from [<c02113b0>] (kthread+0x90/0x98)
[  238.881652]  r7:00000013 r6:c01f98fc r5:00000000 r4:ef831efc
[  238.881683] [<c0211320>] (kthread+0x0/0x98) from [<c01f62f4>] (do_exit+0x0/0x374)
[  238.881713]  r6:c01f62f4 r5:c0211320 r4:ef831efc
[  238.881713] Mem-info:
[  238.881744] Normal per-cpu:
[  238.881744] CPU    0: hi:  186, btch:  31 usd:  38
[  238.881744] CPU    1: hi:  186, btch:  31 usd: 169
[  238.881774] HighMem per-cpu:
[  238.881774] CPU    0: hi:   90, btch:  15 usd:  66
[  238.881774] CPU    1: hi:   90, btch:  15 usd:  86
[  238.881805] active_anon:544 inactive_anon:71 isolated_anon:0
[  238.881805]  active_file:926 inactive_file:2538 isolated_file:0
[  238.881805]  unevictable:0 dirty:10 writeback:0 unstable:0
[  238.881805]  free:57782 slab_reclaimable:864 slab_unreclaimable:186898
[  238.881805]  mapped:632 shmem:144 pagetables:50 bounce:0
[  238.881835] Normal free:1328kB min:3532kB low:4412kB high:5296kB active_anon:0kB inactive_anon:0kB active_file:880kB inactive_file:848kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:780288kB mlocked:0kB dirty:36kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:3456kB slab_unreclaimable:747592kB kernel_stack:392kB pagetables:200kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
[  238.881866] lowmem_reserve[]: 0 1904 1904
[  238.881896] HighMem free:229800kB min:236kB low:508kB high:784kB active_anon:2176kB inactive_anon:284kB active_file:2824kB inactive_file:9304kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:243712kB mlocked:0kB dirty:4kB writeback:0kB mapped:2528kB shmem:576kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:0kB unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
[  238.881927] lowmem_reserve[]: 0 0 0
[  238.881958] Normal: 0*4kB 4*8kB 6*16kB 0*32kB 1*64kB 1*128kB 0*256kB 2*512kB 0*1024kB 0*2048kB 0*4096kB = 1344kB
[  238.882019] HighMem: 6*4kB 2*8kB 4*16kB 4*32kB 1*64kB 1*128kB 0*256kB 2*512kB 3*1024kB 0*2048kB 55*4096kB = 229800kB
[  238.882080] 3610 total pagecache pages
[  238.882080] 0 pages in swap cache
[  238.882080] Swap cache stats: add 0, delete 0, find 0/0
[  238.882110] Free swap  = 0kB
[  238.882110] Total swap = 0kB
[  238.933776] 262144 pages of RAM
[  238.933776] 58240 free pages
[  238.933776] 10503 reserved pages
[  238.933776] 187773 slab pages
[  238.933807] 2475 pages shared
[  238.933807] 0 pages swap cached

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Acked-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-next-2.6 into for-davem

Conflicts:
drivers/net/wireless/libertas/if_cs.c
drivers/net/wireless/rtlwifi/pci.c
net/bluetooth/l2cap_sock.c

mac80211: Fix a warning due to skipping tailroom reservation for IV

The devices that require IV generation in software need tailroom
reservation  for ICVs used in TKIP or WEP encryptions.
Currently, decision to skip the tailroom reservation in the tx
path was taken only on whether driver wants MMIC to be generated
in software or not. Following patch appends IV generation check for
such decisions and fixes the following warning.

WARNING: at net/mac80211/wep.c:101 ieee80211_wep_add_iv+0x56/0xf3()
Hardware name: 64756D6
Modules linked in: ath9k ath9k_common ath9k_hw
Pid: 0, comm: swapper Tainted: G        W   2.6.39-rc5-wl
 Call Trace:
[<c102fd29>] warn_slowpath_common+0x65/0x7a
[<c1465c4e>] ? ieee80211_wep_add_iv+0x56/0xf3
[<c102fd4d>] warn_slowpath_null+0xf/0x13
[<c1465c4e>] ieee80211_wep_add_iv+0x56/0xf3
[<c1466007>] ieee80211_crypto_wep_encrypt+0x63/0x88
[<c1478bf3>] ieee80211_tx_h_encrypt+0x2f/0x63
[<c1478cba>] invoke_tx_handlers+0x93/0xe1
[<c1478eda>] ieee80211_tx+0x4b/0x6d
[<c147907c>] ieee80211_xmit+0x180/0x188
[<c147779d>] ? ieee80211_skb_resize+0x95/0xd9
[<c1479edf>] ieee80211_subif_start_xmit+0x64f/0x668
[<c13956fc>] dev_hard_start_xmit+0x368/0x48c
[<c13a8bd6>] sch_direct_xmit+0x4d/0x101
[<c1395ae1>] dev_queue_xmit+0x2c1/0x43f
[<c13a74a2>] ? eth_header+0x1e/0x90
[<c13a7400>] ? eth_type_trans+0x91/0xc2
[<c13a7484>] ? eth_rebuild_header+0x53/0x53
[<c139f079>] neigh_resolve_output+0x223/0x27e
[<c13c6b23>] ip_finish_output2+0x1d4/0x1fe
[<c13c6bc6>] ip_finish_output+0x79/0x7d
[<c13c6cbe>] T.1075+0x43/0x48
[<c13c6e6e>] ip_output+0x75/0x7b
[<c13c4970>] dst_output+0xc/0xe
[<c13c62c9>] ip_local_out+0x17/0x1a
[<c13c67bb>] ip_queue_xmit+0x2aa/0x2f8
[<c138b742>] ? sk_setup_caps+0x21/0x92
[<c13d95ea>] ? __tcp_v4_send_check+0x7e/0xb7
[<c13d5d2e>] tcp_transmit_skb+0x6a1/0x6d7
[<c13d533b>] ? tcp_established_options+0x20/0x8b
[<c13d6f28>] tcp_retransmit_skb+0x43a/0x527
[<c13d8d6d>] tcp_retransmit_timer+0x32e/0x45d
[<c13d8f23>] tcp_write_timer+0x87/0x16c
[<c103a030>] run_timer_softirq+0x156/0x1f9
[<c13d8e9c>] ? tcp_retransmit_timer+0x45d/0x45d
[<c1034d65>] __do_softirq+0x97/0x14a
[<c1034cce>] ? irq_enter+0x4d/0x4d

Cc: Yogesh Powar <yogeshp@marvell.com>
Reported-by: Fabio Rossi <rossi.f@inwind.it>
Tested-by: Fabio Rossi <rossi.f@inwind.it>
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

can: rename can_try_module_get to can_get_proto

can: rename can_try_module_get to can_get_proto

can_try_module_get does return a struct can_proto.
The name explains what is done in so much detail that a caller
may not notice that a struct can_proto is locked/unlocked.

Signed-off-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: make struct can_proto const

commit 53914b67993c724cec585863755c9ebc8446e83b had the
same message. That commit did put everything in place but
did not make can_proto const itself.

Signed-off-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: ip_expire() must revalidate route

Commit 4a94445c9a5c (net: Use ip_route_input_noref() in input path)
added a bug in IP defragmentation handling, in case timeout is fired.

When a frame is defragmented, we use last skb dst field when building
final skb. Its dst is valid, since we are in rcu read section.

But if a timeout occurs, we take first queued fragment to build one ICMP
TIME EXCEEDED message. Problem is all queued skb have weak dst pointers,
since we escaped RCU critical section after their queueing. icmp_send()
might dereference a now freed (and possibly reused) part of memory.

Calling skb_dst_drop() and ip_route_input_noref() to revalidate route is
the only possible choice.

Reported-by: Denys Fedoryshchenko <denys@visp.net.lb>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/jkirsher/net-next-2.6

ipv6: Use flowi4->{daddr,saddr} in ipip6_tunnel_xmit().

Instead of rt->rt_{dst,src}

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Kill rt->rt_{src, dst} usage in IP GRE tunnels.

First, make callers pass on-stack flowi4 to ip_route_output_gre()
so they can get at the fully resolved flow key.

Next, use that in ipgre_tunnel_xmit() to avoid the need to use
rt->rt_{dst,src}.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Pass explicit saddr/daddr args to ipmr_get_route().

This eliminates the need to use rt->rt_{src,dst}.

Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Cleanup PCIe bus speed info

PCIe connections should be expressed as GT/s (GigaTransfers per second)
instead of the current Gb/s (Gigabits per second).  In addition, it is
incorrect because (due to PCIe gen 1 & 2 having a 20% overhead) the
actually data rate, when expressed in Gb/s, is only 80% of the rate of
GT/s.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: improve EEPROM read/write operations

Introduce buffered read/writes which greatly improves performance on
parts with large EEPROMs.

Previously reading/writing a word requires taking/releasing of synchronization
semaphores which adds 10ms to each operation. The optimization is to
read/write in buffers, but make sure the semaphore is not held for >500ms
according to the datasheet.

Since we can't read the EEPROM page size ixgbe_detect_eeprom_page_size() is
used to discover the EEPROM size when needed and keeps the result in
word_page_size for the rest of the run time.

Use buffered reads for ethtool -e.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: fix sparse warning

warning: symbol 'before' shadows an earlier one

Convert large macros to functions similar to e1000e.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Acked-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: fix typo error with software defined pins on 82599

Correcting a simple typo with enabling software defined pins.  I don't
believe this was causing any issues but this is how it was meant to be
implemented.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Use function pointer for ixgbe_acquire/release_swfw_sync()

Change remaining direct calls to function pointers.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igbvf: remove bogus phys_id

This device lies about supporting phys_id. Remove it and just
let the upper layer report not supported.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: Add check for invalid size to igb_get_invariants_82575()

Recent commits have changed how EEPROM size is checked and if the size
word is misconfigured, the driver will fail to load.  This patch adds a
check for invalid size word in the EEPROM and uses default size instead
for 82576 parts.

Reported-by: Stefan Assmann <sassmann@redhat.com>
Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgb: convert to set_phys_id

Based on the original patch sent by Stephen Hemminger.

This version incorporates the ethtool changes that Bruce Allan
submitted.

CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>

ipv4: In ip_build_and_send_pkt() use 'saddr' and 'daddr' args passed in.

Instead of rt->rt_{dst,src}

The only tricky part is source route option handling.

If the source route option is enabled we can't just use plain 'daddr',
we have to use opt->opt.faddr.

Signed-off-by: David S. Miller <davem@davemloft.net>

e100: fix build warning

In function 'e100_hw_init':
warning: 'err' may be used uninitialized in this function

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ipv4: Use flowi4->{daddr,saddr} in ipip_tunnel_xmit().

Instead of rt->rt_{dst,src}

Signed-off-by: David S. Miller <davem@davemloft.net>

x86/net: only select BPF_JIT when NET is enabled

Fix kconfig unmet dependency warning: HAVE_BPF_JIT depends on NET, so
make the "select" of it depend on NET also.

warning: (X86) selects HAVE_BPF_JIT which has unmet direct dependencies (NET)

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: Use flowi4's {saddr,daddr} in sctp_v4_dst_saddr() and sctp_v4_get_dst()

Instead of rt->rt_{src,dst}

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Use flowi4's {saddr,daddr} in igmpv3_newpack() and igmp_send_report()

Instead of rt->rt_{src,dst}

Signed-off-by: David S. Miller <davem@davemloft.net>

libcxgbi: Use flowi4's saddr in cxgbi_check_route().

Instead of rt->rt_src

Signed-off-by: David S. Miller <davem@davemloft.net>

pptp: Use flowi4's daddr/saddr in pptp_xmit().

Instead of rt->rt_{src,dst}

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Make caller provide on-stack flow key to ip_route_output_ports().

Signed-off-by: David S. Miller <davem@davemloft.net>

dccp: Use flowi4->saddr in dccp_v4_connect()

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Renamt struct rtable's rt_tos to rt_key_tos.

To more accurately reflect that it is purely a routing
cache lookup key and is used in no other context.

Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Rework ipmr_rt_fib_lookup() flow key initialization.

Use information from the skb as much as possible, currently
this means daddr, saddr, and TOS.

Signed-off-by: David S. Miller <davem@davemloft.net>

net/stmmac: Move "#include <linux/platform_device.h>" to linux/stmmac.h

stmmac.h uses struct platform_device and doesn't include
<linux/platform_device.h>. Whereas drivers/net/stmmac/stmmac.h includes it, but
doesn't directly use it. And so we get following compilation warning while using
this file:
warning: ‘struct platform_device’ declared inside parameter list

This patch includes <linux/platform_device.h> in linux/stmmac.h and removes it
from drivers/net/stmmac/stmmac.h

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sysctl: net: call unregister_net_sysctl_table where needed

ctl_table_headers registered with register_net_sysctl_table should
have been unregistered with the equivalent unregister_net_sysctl_table

Signed-off-by: Lucian Adrian Grijincu <lucian.grijincu@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Revert: veth: remove unneeded ifname code from veth_newlink()

84c49d8c3e4abefb0a41a77b25aa37ebe8d6b743 ("veth: remove unneeded
ifname code from veth_newlink()") caused regression on veth
creation. This patch reverts the original one.

Reported-by: Michał Mirosław <mirqus@gmail.com>
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

smsc95xx: fix reset check

The reset loop check should check the MII_BMCR register value for
BMCR_RESET rather than for MII_BMCR (the register address, which also
happens to be zero).

Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Fix failure to enable WoL by default when possible

tg3 is supposed to enable WoL by default on adapters which support
that, but it fails to do so unless the adapter's
/sys/devices/.../power/wakeup file contains 'enabled' during the
initialization of the adapter.  Fix that by making tg3 use
device_set_wakeup_enable() to enable wakeup automatically whenever
WoL should be enabled by default.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

networking: inappropriate ioctl operation should return ENOTTY

ioctl() calls against a socket with an inappropriate ioctl operation
are incorrectly returning EINVAL rather than ENOTTY:

  [ENOTTY]
      Inappropriate I/O control operation.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=33992
Signed-off-by: Lifeng Sun <lifongsun@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dont hold rtnl mutex during netlink dump callbacks

Four years ago, Patrick made a change to hold rtnl mutex during netlink
dump callbacks.

I believe it was a wrong move. This slows down concurrent dumps, making
good old /proc/net/ files faster than rtnetlink in some situations.

This occurred to me because one "ip link show dev ..." was _very_ slow
on a workload adding/removing network devices in background.

All dump callbacks are able to use RCU locking now, so this patch does
roughly a revert of commits :

1c2d670f366 : [RTNETLINK]: Hold rtnl_mutex during netlink dump callbacks
6313c1e0992 : [RTNETLINK]: Remove unnecessary locking in dump callbacks

This let writers fight for rtnl mutex and readers going full speed.

It also takes care of phonet : phonet_route_get() is now called from rcu
read section. I renamed it to phonet_route_get_rcu()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Remi Denis-Courmont <remi.denis-courmont@nokia.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'batman-adv/next' of git://git.open-mesh.org/ecsv/linux-merge

amd8111e: trivial typo spelling: Negotitate -> Negotiate

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Make sure flowi4->{saddr,daddr} are always set.

Slow path output route resolution always makes sure that
->{saddr,daddr} are set, and also if we trigger into IPSEC resolution
we initialize them as well, because xfrm_lookup() expects them to be
fully resolved.

But if we hit the fast path and flowi4->flowi4_proto is zero, we won't
do this initialization.

Therefore, move the IPSEC path initialization to the route cache
lookup fast path to make sure these are always set.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6