net: No dst refcounting in ip_queue_xmit()

TCP outgoing packets can avoid two atomic ops, and dirtying
of previously higly contended cache line using new refdst
infrastructure.

Note 1: loopback device excluded because of !IFF_XMIT_DST_RELEASE
Note 2: UDP packets dsts are built before ip_queue_xmit().

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

net: Use ip_route_input_noref() in input path

Use ip_route_input_noref() in ip fast path, to avoid two atomic ops per
incoming packet.

Note: loopback is excluded from this optimization in ip_rcv_finish()

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

net: implements ip_route_input_noref()

ip_route_input() is the version returning a refcounted dst, while
ip_route_input_noref() returns a non refcounted one.

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

net: add a noref bit on skb dst

Use low order bit of skb->_skb_dst to tell dst is not refcounted.

Change _skb_dst to _skb_refdst to make sure all uses are catched.

skb_dst() returns the dst, regardless of noref bit set or not, but
with a lockdep check to make sure a noref dst is not given if current
user is not rcu protected.

New skb_dst_set_noref() helper to set an notrefcounted dst on a skb.
(with lockdep check)

skb_dst_drop() drops a reference only if skb dst was refcounted.

skb_dst_force() helper is used to force a refcount on dst, when skb
is queued and not anymore RCU protected.

Use skb_dst_force() in __sk_add_backlog(), __dev_xmit_skb() if
!IFF_XMIT_DST_RELEASE or skb enqueued on qdisc queue, in
sock_queue_rcv_skb(), in __nf_queue().

Use skb_dst_force() in dev_requeue_skb().

Note: dst_use_noref() still dirties dst, we might transform it
later to do one dirtying per jiffies.

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

rps: avoid one atomic in enqueue_to_backlog

If CONFIG_SMP=y, then we own a queue spinlock, we can avoid the atomic
test_and_set_bit() from napi_schedule_prep().

We now have same number of atomic ops per netif_rx() calls than with
pre-RPS kernel.

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

drivers/net/usb/asix.c: Fix unaligned accesses

Using this driver can cause unaligned accesses in the IP layer
This has been fixed by aligning the skb data correctly using the
spare room left over by the 4 byte header inserted between packets
by the device.

Signed-off-by: Neil Jones <NeilJay@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ibmveth: Add suspend/resume support

Adds support for resuming from suspend for IBM virtual ethernet devices.
We may have lost an interrupt over the suspend, so we just kick the
interrupt handler to process anything that is outstanding.

Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6 addrlabel: permit deletion of labels assigned to removed dev

as addrlabels with an interface index are left alone when the
interface gets removed this results in addrlabels that can no
longer be removed.

Restrict validation of index to adding new addrlabels.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
include/linux/if_link.h

rtnetlink: make SR-IOV VF interface symmetric

Now we have a set of nested attributes:

  IFLA_VFINFO_LIST (NESTED)
    IFLA_VF_INFO (NESTED)
      IFLA_VF_MAC
      IFLA_VF_VLAN
      IFLA_VF_TX_RATE

This allows a single set to operate on multiple attributes if desired.
Among other things, it means a dump can be replayed to set state.

The current interface has yet to be released, so this seems like
something to consider for 2.6.34.

Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth: synchronize configuration interface

Synchronize access to the drivers configuration interface.
Also do not allow configuration changes during online/offline
transition.

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth: new message if OLM limit is reached

z/OS may activate Optimized Latency Mode (OLM) for a connection
through an OSA Express3 adapter, which reduces the number of
allowed concurrent connections, if adapter is used in shared mode.
Create a meaningful message, if activation of an OSA-connection fails
due to an active OLM-connection on the shared OSA-adapter.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth: exploit HW TX checksumming

OSA supports HW TX checksumming in layer 3 mode. Enable this
feature and remove software fallback used for TSO. Cleanup
checksum bits to indicate OSA can do checksumming only for
IPv4 TCP and UDP.

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: delete active ICMP proto unreachable timer when free transport

transport may be free before ICMP proto unreachable timer expire, so
we should delete active ICMP proto unreachable timer when transport
is going away.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: congestion notifications are not dropped packets

vlan/macvlan start_xmit() can inform caller of congestion with
NET_XMIT_CN return value. This doesnt mean packet was dropped.
Increment normal stat counters instead of tx_dropped.

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

net: Introduce sk_route_nocaps

TCP-MD5 sessions have intermittent failures, when route cache is
invalidated. ip_queue_xmit() has to find a new route, calls
sk_setup_caps(sk, &rt->u.dst), destroying the

sk->sk_route_caps &= ~NETIF_F_GSO_MASK

that MD5 desperately try to make all over its way (from
tcp_transmit_skb() for example)

So we send few bad packets, and everything is fine when
tcp_transmit_skb() is called again for this socket.

Since ip_queue_xmit() is at a lower level than TCP-MD5, I chose to use a
socket field, sk_route_nocaps, containing bits to mask on sk_route_caps.

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

tcp: fix MD5 (RFC2385) support

TCP MD5 support uses percpu data for temporary storage. It currently
disables preemption so that same storage cannot be reclaimed by another
thread on same cpu.

We also have to make sure a softirq handler wont try to use also same
context. Various bug reports demonstrated corruptions.

Fix is to disable preemption and BH.

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

net: Consistent skb timestamping

With RPS inclusion, skb timestamping is not consistent in RX path.

If netif_receive_skb() is used, its deferred after RPS dispatch.

If netif_rx() is used, its done before RPS dispatch.

This can give strange tcpdump timestamps results.

I think timestamping should be done as soon as possible in the receive
path, to get meaningful values (ie timestamps taken at the time packet
was delivered by NIC driver to our stack), even if NAPI already can
defer timestamping a bit (RPS can help to reduce the gap)

Tom Herbert prefer to sample timestamps after RPS dispatch. In case
sampling is expensive (HPET/acpi_pm on x86), this makes sense.

Let admins switch from one mode to another, using a new
sysctl, /proc/sys/net/core/netdev_tstamp_prequeue

Its default value (1), means timestamps are taken as soon as possible,
before backlog queueing, giving accurate timestamps.

Setting a 0 value permits to sample timestamps when processing backlog,
after RPS dispatch, to lower the load of the pre-RPS cpu.

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

xfrm: fix policy unreferencing on larval drop

I mistakenly had the error path to use num_pols to decide how
many policies we need to drop (cruft from earlier patch set
version which did not handle socket policies right).

This is wrong since normally we do not keep explicit references
(instead we hold reference to the cache entry which holds references
to policies). drop_pols is set to num_pols if we are holding the
references, so use that. Otherwise we eventually BUG_ON inside
xfrm_policy_destroy due to premature policy deletion.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: adjust handle_macvlan to pass port struct to hook

Now there's null check here and also again in the hook. Looking at bridge bits
which are simmilar, port structure is rcu_dereferenced right away in
handle_bridge and passed to hook. Looks nicer.

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

skge: use the DMA state API instead of the pci equivalents

This replace the PCI DMA state API (include/linux/pci-dma.h) with the
DMA equivalents since the PCI DMA state API will be obsolete.

No functional change.

For further information about the background:

http://marc.info/?l=linux-netdev&m=127037540020276&w=2

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: reserve ports for applications using fixed port numbers

(Dropped the infiniband part, because Tetsuo modified the related code,
I will send a separate patch for it once this is accepted.)

This patch introduces /proc/sys/net/ipv4/ip_local_reserved_ports which
allows users to reserve ports for third-party applications.

The reserved ports will not be used by automatic port assignments
(e.g. when calling connect() or bind() with port number 0). Explicit
port allocation behavior is unchanged.

Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sysctl: add proc_do_large_bitmap

The new function can be used to read/write large bitmaps via /proc. A
comma separated range format is used for compact output and input
(e.g. 1,3-4,10-10).

Writing into the file will first reset the bitmap then update it
based on the given input.

Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sysctl: refactor integer handling proc code

(Based on Octavian's work, and I modified a lot.)

As we are about to add another integer handling proc function a little
bit of cleanup is in order: add a few helper functions to improve code
readability and decrease code duplication.

In the process a bug is also fixed: if the user specifies a number
with more then 20 digits it will be interpreted as two integers
(e.g. 10000...13 will be interpreted as 100.... and 13).

Behavior for EFAULT handling was changed as well. Previous to this
patch, when an EFAULT error occurred in the middle of a write
operation, although some of the elements were set, that was not
acknowledged to the user (by shorting the write and returning the
number of bytes accepted). EFAULT is now treated just like any other
errors by acknowledging the amount of bytes accepted.

Signed-off-by: Octavian Purdila <opurdila@ixiacom.com>
Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/inaky/wimax

bridge: update sysfs link names if port device names have changed

Links for each port are created in sysfs using the device
name, but this could be changed after being added to the
bridge.

As well as being unable to remove interfaces after this
occurs (because userspace tools don't recognise the new
name, and the kernel won't recognise the old name), adding
another interface with the old name to the bridge will
cause an error trying to create the sysfs link.

This fixes the problem by listening for NETDEV_CHANGENAME
notifications and renaming the link.

https://bugzilla.kernel.org/show_bug.cgi?id=12743

Signed-off-by: Simon Arlott <simon@fire.lp0.eu>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: change console message interface

Use one set of macro's for all bridge messages.

Note: can't use netdev_XXX macro's because bridge is purely
virtual and has no device parent.

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

bridge: netpoll cleanup

Move code around so that the ifdef for NETPOLL_CONTROLLER don't have to
show up in main code path. The control functions should be in helpers
that are only compiled if needed.

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

rndis_host: Poll status channel before control channel

Some RNDIS devices don't respond on the control channel until polled
on the status channel.  In particular, this was reported to be the
case for the 2Wire HomePortal 1000SW.

This is roughly based on a patch by John Carr <john.carr@unrouted.co.uk>
which is reported to be needed for use with some Windows Mobile devices
and which is currently applied by Mandriva.

Reported-by: Mark Glassberg <vzeeaxwl@myfairpoint.net>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Tested-by: Mark Glassberg <vzeeaxwl@myfairpoint.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: xt_TEE depends on NF_CONNTRACK

Fix xt_TEE build for the case of NF_CONNTRACK=m and
NETFILTER_XT_TARGET_TEE=y:

xt_TEE.c:(.text+0x6df5c): undefined reference to `nf_conntrack_untracked'
4x

Built with all 4 m/y combinations.

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

Merge branch 'net-2.6' of git://git./linux/kernel/git/mst/vhost

qlcnic: add idc debug registers

When ever driver changes the device state, it should write
pci-func number and timestamp in debug registers.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: support quisce mode

Device can go to quiescent state, during which drivers
should refrain from using the device.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: check device class

pci-func class can be other than ethernet in Qlogic CNA device.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: check IDC version

Warn user if IDC version mismatch with different class of drivers.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: cleanup unused code

LRO ring, cut-thru mode and specific fw version are not
valid to Qlogic CNA device.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: cleanup dma mask setting

Device support 64 bit dma mask.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix caching window register

o Window register is not per pci-func, so caching can
result in expected result.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: remove obsolete register

MSI_MODE, CAPABILITIES_FW and SCRATCHPAD registers are obsolete.
Driver should not use them.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix context cleanup

Before going for recovery, every pci-func should check fw state,
irrespective of device state. This to avoid unnecssary sending
of command for ctx destroy.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: version 1.28

Version 1.28

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

sky2: Avoid allocating memory in sky2_resume

Allocating memory can fail, and since we have the memory we need
in sky2_resume when sky2_suspend is called, just stop the hardware
without freeing the memory it's using.

This avoids the possibility of failing because we can't allocate
memory in sky2_resume(), and allows sharing code with sky2_restart().

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: Refactor down/up code out of sky2_restart()

Code to bring down all sky2 interfaces and bring it up
again can be reused in sky2_suspend and sky2_resume.

Factor the code to bring the interfaces down into
sky2_all_down and the up code into sky2_all_up.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: Shut off interrupts before NAPI

Interrupts should be masked, then synchronized, and
finally NAPI should be disabled.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: Avoid race in sky2_change_mtu

netif_stop_queue does not ensure all in-progress transmits are complete,
 so use netif_tx_disable() instead.

Secondly, make sure NAPI polls are disabled before stopping the tx queue,
 otherwise sky2_status_intr might trigger a TX queue wakeup between when
 we stop the queue and NAPI is disabled.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: Restore multicast after restart

Multicast settings will be lost on reset, so restore them.

Signed-off-by: Mike McCormack <mikem@ring3k.org>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgb and e1000: Use new function for copybreak tests

There appears to be an off-by-1 defect in the maximum packet size
copied when copybreak is speified in these modules.

The copybreak module params are specified as:
"Maximum size of packet that is copied to a new buffer on receive"

The tests are changed from "< copybreak" to "<= copybreak"
and moved into new static functions for readability.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: cleanup unused parameters

During the cleanup pass after the removal of e1000e hardware from e1000 some
parameters were missed.  Remove them because it is just dead code.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: fix WARN_ON with mac-vlan

When adding more than 14 mac-vlan adapters on e1000 the driver
would fire a WARN_ON when adding the 15th.  The WARN_ON in this
case is completely un-necessary, as the code below the WARN_ON is
directly handling the value the WARN_ON triggered on.

CC: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net: Remove unnecessary returns from void function()s

This patch removes from drivers/net/ all the unnecessary
return; statements that precede the last closing brace of
void functions.

It does not remove the returns that are immediately
preceded by a label as gcc doesn't like that.

It also does not remove null void functions with return.

Done via:
$ grep -rP --include=*.[ch] -l "return;\n}" net/ | \
  xargs perl -i -e 'local $/ ; while (<>) { s/\n[ \t\n]+return;\n}/\n}/g; print; }'

with some cleanups by hand.

Compile tested x86 allmodconfig only.

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

ixgbe: Refactor common code between 82598 & 82599 to accommodate new hardware

Some of the following MAC functions are moved from 82598 & 82599 specific
hardware files to common.[ch] to accommodate new silicon changes. Also
fixed some white space issues
 * get_san_mac_addr, check_link, set_vmdq, clear_vmdq, clear_vfta,
 * set_vfta, fc_enable, init_uta_tables

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Use bool flag to see if the packet unmapping is delayed in HWRSC

We can't use zero magic "bad" value to check if IXGBE_RSC_CB(skb)->dma
is valid. It is only valid in x86/arm/m68k/alpha architectures and in
spark, powerPC and other architectures it should be ~0. As per
Benjamin Herrenschmidt feedback use a bool flag to decide if
the packet unmapping is delayed in hardware RSC till EOP is reached

Signed-off-by: Mallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: fix setting of promisc mode when using mac-vlans

Introduce uc_set_promisc flag to fix enabling of promisc mode
when exceeding the number of supported RAR entries.

Issue discovered by Ben Greear when using mac-vlans.

Reported-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-next-2.6

netfilter: remove unnecessary returns from void function()s

This patch removes from net/ netfilter files
all the unnecessary return; statements that precede the
last closing brace of void functions.

It does not remove the returns that are immediately
preceded by a label as gcc doesn't like that.

Done via:
$ grep -rP --include=*.[ch] -l "return;\n}" net/ | \
  xargs perl -i -e 'local $/ ; while (<>) { s/\n[ \t\n]+return;\n}/\n}/g; print; }'

Signed-off-by: Joe Perches <joe@perches.com>
[Patrick: changed to keep return statements in otherwise empty function bodies]
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: cleanup printk messages

Make sure all printk messages have a severity level.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: change NF_ASSERT to WARN_ON

Change netfilter asserts to standard WARN_ON. This has the
benefit of backtrace info and also causes netfilter errors
to show up on kerneloops.org.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: bridge-netfilter: fix crash in br_nf_forward_finish()

[ 4593.956206] BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
[ 4593.956219] IP: [<ffffffffa03357a4>] br_nf_forward_finish+0x154/0x170 [bridge]
[ 4593.956232] PGD 195ece067 PUD 1ba005067 PMD 0
[ 4593.956241] Oops: 0000 [#1] SMP
[ 4593.956248] last sysfs file:
/sys/devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/device:08/ATK0110:00/hwmon/hwmon0/temp2_label
[ 4593.956253] CPU 3
...
[ 4593.956380] Pid: 29512, comm: kvm Not tainted 2.6.34-rc7-net #195 P6T DELUXE/System Product Name
[ 4593.956384] RIP: 0010:[<ffffffffa03357a4>]  [<ffffffffa03357a4>] br_nf_forward_finish+0x154/0x170 [bridge]
[ 4593.956395] RSP: 0018:ffff880001e63b78  EFLAGS: 00010246
[ 4593.956399] RAX: 0000000000000608 RBX: ffff880057181700 RCX: ffff8801b813d000
[ 4593.956402] RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff880057181700
[ 4593.956406] RBP: ffff880001e63ba8 R08: ffff8801b9d97000 R09: ffffffffa0335650
[ 4593.956410] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8801b813d000
[ 4593.956413] R13: ffffffff81ab3940 R14: ffff880057181700 R15: 0000000000000002
[ 4593.956418] FS:  00007fc40d380710(0000) GS:ffff880001e60000(0000) knlGS:0000000000000000
[ 4593.956422] CS:  0010 DS: 002b ES: 002b CR0: 000000008005003b
[ 4593.956426] CR2: 0000000000000018 CR3: 00000001ba1d7000 CR4: 00000000000026e0
[ 4593.956429] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 4593.956433] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 4593.956437] Process kvm (pid: 29512, threadinfo ffff8801ba566000, task ffff8801b8003870)
[ 4593.956441] Stack:
[ 4593.956443]  0000000100000020 ffff880001e63ba0 ffff880001e63ba0 ffff880057181700
[ 4593.956451] <0> ffffffffa0335650 ffffffff81ab3940 ffff880001e63bd8 ffffffffa03350e6
[ 4593.956462] <0> ffff880001e63c40 000000000000024d ffff880057181700 0000000080000000
[ 4593.956474] Call Trace:
[ 4593.956478]  <IRQ>
[ 4593.956488]  [<ffffffffa0335650>] ? br_nf_forward_finish+0x0/0x170 [bridge]
[ 4593.956496]  [<ffffffffa03350e6>] NF_HOOK_THRESH+0x56/0x60 [bridge]
[ 4593.956504]  [<ffffffffa0335282>] br_nf_forward_arp+0x112/0x120 [bridge]
[ 4593.956511]  [<ffffffff813f7184>] nf_iterate+0x64/0xa0
[ 4593.956519]  [<ffffffffa032f920>] ? br_forward_finish+0x0/0x60 [bridge]
[ 4593.956524]  [<ffffffff813f722c>] nf_hook_slow+0x6c/0x100
[ 4593.956531]  [<ffffffffa032f920>] ? br_forward_finish+0x0/0x60 [bridge]
[ 4593.956538]  [<ffffffffa032f800>] ? __br_forward+0x0/0xc0 [bridge]
[ 4593.956545]  [<ffffffffa032f86d>] __br_forward+0x6d/0xc0 [bridge]
[ 4593.956550]  [<ffffffff813c5d8e>] ? skb_clone+0x3e/0x70
[ 4593.956557]  [<ffffffffa032f462>] deliver_clone+0x32/0x60 [bridge]
[ 4593.956564]  [<ffffffffa032f6b6>] br_flood+0xa6/0xe0 [bridge]
[ 4593.956571]  [<ffffffffa032f800>] ? __br_forward+0x0/0xc0 [bridge]

Don't call nf_bridge_update_protocol() for ARP traffic as skb->nf_bridge isn't
used in the ARP case.

Reported-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Bart De Schuymer <bdschuym@pandora.be>
Signed-off-by: Patrick McHardy <kaber@trash.net>

cxgb4: report GRO stats with ethtool -S

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: report the PCIe link speed

Report the PCIe link speed (2.5 or 5 Gbps).

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: configure HW VLAN extraction through FW

HW VLAN extraction needs to be configured through FW to work correctly in
virtualization environments.  Remove the direct register manipulation and
rely on FW.

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: add PCI device id to enable support for 82567V-4

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: Fix/cleanup PHY reset code for ICHx/PCHx

i) Fixes a bug where e1000_sw_lcd_config_ich8lan() was calling
e1000_lan_init_done_ich8lan() to poll the STATUS.LAN_INIT_DONE bit to
make sure the MAC had completed the PHY configuration.  However,
e1000_lan_init_done_ich8lan() had already been called in one of the two
places where PHY reset occurs for ICHx/PCHx parts, which caused the second
call to busy-wait for 150 msec because the LAN_INIT_DONE bit had already
been checked and cleared.

ii) Cleanup the two separate PHY reset code paths, i.e. the full-chip reset
in e1000_reset_hw_ich8lan() and the PHY-only reset in
e1000_phy_hw_reset_ich8lan().  There was duplicate code in both paths to be
performed post-reset that are now combined into one new function -
e1000_post_phy_reset_ich8lan().  This cleanup also included moving the
clearing of the PHY Reset Asserted bit in the STATUS register (now done for
all ICH/PCH parts) and the check for the indication from h/w that basic
configuration has completed back to where it previously was in
e1000_get_cfg_done_ich8lan().

iii) Corrected a few comments

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: move settting of flow control refresh timer to link setup code

The flow control refresh timer value needs to be saved off so that it can
be programmed into the approrpiate register when applicable but without a
reset, e.g. when changing flow control parameters via ethtool.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: fix checks for manageability enabled and management pass-through

The mac->arc_subsystem was being incorrectly used to flag whether or not
manageability was enabled when it should only be used to state whether the
ARC (Host interface) subsystem is available on a particular MAC _and_ only
valid when any manageability is enabled. The ARC subsystem is currently
only available on 80003es2lan and 82573 parts supported by the driver.

A new flag, has_fwsm, is introduced to be used when checking if
manageability is enabled but only on parts that acutally have an FWSM
register. While the above parts have an FWSM register, there are other
parts that have FWSM but do not have support for the ARC subsystem,
namely 82571/2 and ICHx/PCH.

And then there are parts that have manageability, but do not have either
FWSM register or support for the ARC subsystem - these are 82574 and 82583.

For 80003es2lan, 82571/2/3 and ICH/PCH parts, this patch makes no
functional changes, it only corrects the usage of the manageability flags.
For 82574 and 82583, it fixes the incorrect accesses of the non-existent
FWSM register and ARC subsystem as well as corrects the check for
management pass-through.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: Incorrect function pointer set for force_speed_duplex on 82577

The force_speed_duplex function pointer was incorrectly set.  Instead of
calling the 82577-specific version it was calling the m88 version which,
among other incorrect things, reset the PHY causing autonegotiation to be
re-enabled in the PHY resulting in the link defaulting to half-duplex.
The 82577-specific force_speed_duplex function also had an issue where
it disabled Auto-MDI-X which caused the link to not come up.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: Cleanup e1000_sw_lcd_config_ich8lan()

After every reset all ICH/PCH parts call this function which acquires the
swflag, performs a workaround on applicable parts and releases the swflag.
There is no reason for parts for which this workaround is not applicable
to acquire and release the swflag so the function should just return
without doing anything for these parts.  This also provides for the
indentation of most of the function contents to be shifted left cleaning up
the code.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: Remove EN_MAC_ADDR_FILTER check from enable_mng_pass_thru check

Patch addresses issues when manageability passthrough is enabled, but the
MAC_ADDR_FILTER bit is not set in the MANC register.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: cleanup multiple common exit points

...in e1000_update_nvm_checksum_ich8lan().

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: s/w initiated LSC MSI-X interrupts not generated; no transmit

In MSI-X mode when an IMPI SoL session was active (i.e. the PHY reset was
blocked), the LSC interrupt generated by s/w to start the watchdog which
started the transmitter was not getting fired by the hardware because bit
24 (the 'other' cause bit) also needed to be set.  Without an active SoL
session, the PHY was reset which caused the h/w to fire the LSC interrupt.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: initialize manageability (IPMI) pass-through in 82574/82583

82574/82583 uses different registers/bits to setup manageability filters
than all other parts supported by e1000e; set them accordingly for IPMI
pass-through.  Rename the function to better reflect what it does.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: bad state after running ethtool diagnostics with AMT enabled

When running ethtool online diagnostics with no open interface, there is a
short period of time where the driver relinquishes control of the adapter
during which time AMT (manageability firmware) can put the adapter into an
unknown state resulting in such things as link test failure, hardware hang,
reporting an incorrect link speed, etc.  Resetting the adapter during an
open() resolves this by putting the adapter into a quiescent state.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000e: use static params to save stack space (part 2)

A couple stack cleanups missed in an earlier patch from Jesse.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/kaber/ipmr-2.6

netxen: handle queue manager access

Check the access by tools for hardware queue engine and handle it
separately than other block registers, otherwise incorrect data
is returned.

Support for only NX3031 based cards.

Acked-by: Dhananjay Phadke <dhananjay.phadke@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netxen: to fix onchip memory access.

Remove unnecessary remap of the region in bar 0 to access onhip memory
for NX3031.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netxen: remove unnecessary size checks

NX3031 have 64bit on card memory. Fix the limit check to
64MB and remove unnecessary 128bit read/write check.

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netxen: fix register usage

o For NX3031, MSI_MODE, CAPABILITIES_FW and SCRATCHPAD registers
  are obsolete. These register addresses can be used for different
  purpose.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining tipc_msg_* routines

Convert tipc_msg_* inline routines that are more than one line into
standard functions, thereby eliminating some repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining buf_acquire routine

Convert buf_acquire inline routine that is more than one line into
a standard function, thereby eliminating some repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining bearer congestion routine

Convert bearer congestion inline routine that is more than one line into
a standard function, thereby eliminating some repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining port list routines

Converts port list inline routines that are more than one line into
standard functions, thereby eliminating a significant amount of
repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining nmap routines

Converts nmap inline routines that are more than one line into standard
functions, thereby eliminating a significant amount of repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Reduce footprint by un-inlining address routines

Convert address-related inline routines that are more than one
line into standard functions, thereby eliminating a significant
amount of repeated code.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: add tipc_ prefix to fcns targeted for un-inlining

These functions have enough code in them such that they
seem like sensible targets for un-inlining.  Prior to doing
that, this adds the tipc_ prefix to the functions, so that
in the event of a panic dump or similar, the subsystem from
which the functions come from is immediately clear.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Relocate trivial link status functions to header file

Rather than live in link.c where they can only be used in that file alone,
these helper routines are better served by being in link.h

Relocated are the following:

link_working_working
link_working_unknown
link_reset_unknown
link_reset_reset
link_blocked
link_congested

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: remove abstraction for link_max_pkt

This is just a straight return of a field; there is no
value in the abstraction of hiding it behind a function.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Update commenting in TIPC API

Eliminate comments in TIPC's main API files that are either obsolete,
incorrect, misleading, or unhelpful.  It also adds in one new comment.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Add support for "-s" configuration option

Provide initial support for displaying overall TIPC status/statistics
information at runtime.  Currently, only version info for the TIPC
kernel module is displayed.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Rename "multicast-link" to "broadcast-link"

Make a cosmetic change to the name displayed for the broadcast link,
to better reflect its true nature. Since TIPC utilizes this link to
distribute name table information, in addition to multicast messages
sent by user applications, the prior name "multicast-link" is
no longer appropriate.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Eliminate unnecessary initialization in native API send routines

Eliminate a couple of instances where TIPC's native API send routines
were doing pointless initialization of local variables.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Prune unused data structures from configuration service

Eliminate some unused data structures in the TIPC
configuration service that relate to the handling of link
subscriptions, which were not supported when TIPC 1.5 was
introduced.  If and when support for link subscriptions is
offered in TIPC, these elements may need to be re-introduced.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Eliminate unused argument in print statement

Eliminate an argument in a print statement that has no corresponding
format specification.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: Eliminate obsolete port's "congested_link" field

Eliminate a field of the TIPC port structure that is populated,
but never referenced.

Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netfilter: xtables: add missing depends for xt_TEE

Aviod these link-time errors when IPV6=m, XT_TEE=y:

net/built-in.o: In function `tee_tg_route6':
xt_TEE.c:(.text+0x45ca5): undefined reference to `ip6_route_output'
net/built-in.o: In function `tee_tg6':
xt_TEE.c:(.text+0x45d79): undefined reference to `ip6_local_out'

Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vhost: fix barrier pairing

According to memory-barriers.txt, an smp memory barrier in guest
should always be paired with an smp memory barrier in host,
and I quote "a lack of appropriate pairing is almost certainly an
error". In case of vhost, failure to flush out used index
update before looking at the interrupt disable flag
could result in missed interrupts, resulting in
networking hang under stress.

This might happen when flags read bypasses used index write.
So we see interrupts disabled and do not interrupt, at the
same time guest writes flags value to enable interrupt,
reads an old used index value, thinks that
used ring is empty and waits for interrupt.

Note: the barrier we pair with here is in
drivers/virtio/virtio_ring.c, function
vring_enable_cb.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Juan Quintela <quintela@redhat.com>

Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
Documentation/feature-removal-schedule.txt
drivers/net/wireless/ath/ar9170/usb.c
drivers/scsi/iscsi_tcp.c
net/ipv4/ipmr.c

Merge branch 'hwmon-for-linus' of git://git./linux/kernel/git/jdelvare/staging

* 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
  hwmon: (applesmc) Correct sysfs fan error handling
  hwmon: (asc7621) Bug fixes

Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  kprobes/x86: Fix removed int3 checking order
  perf: Fix static strings treated like dynamic ones

drivers/gpu/drm/i915/i915_irq.c:i915_error_object_create(): use correct kmap-atomic slot

i915_error_object_create() is called from the timer interrupt and hence
can corrupt the KM_USER0 slot.  Use KM_IRQ0 instead.

Reported-by: Jaswinder Singh Rajput <jaswinderlinux@gmail.com>
Tested-by: Jaswinder Singh Rajput <jaswinderlinux@gmail.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>