Pablo says:

====================
This is the second batch of Netfilter updates for net-next. It contains the
kernel changes for the new user-space connection tracking helper
infrastructure.

More details on this infrastructure are provides here:
http://lwn.net/Articles/500196/

Still, I plan to provide some official documentation through the
conntrack-tools user manual on how to setup user-space utilities for this.
So far, it provides two helper in user-space, one for NFSv3 and another for
Oracle/SQLnet/TNS. Yet in my TODO list.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix code style - place the asterisk where it belongs.
Signed-off-by: NEldad Zack <eldad@fogrefinery.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There are good reasons to supports helpers in user-space instead:

* Rapid connection tracking helper development, as developing code
  in user-space is usually faster.

* Reliability: A buggy helper does not crash the kernel. Moreover,
  we can monitor the helper process and restart it in case of problems.

* Security: Avoid complex string matching and mangling in kernel-space
  running in privileged mode. Going further, we can even think about
  running user-space helpers as a non-root process.

* Extensibility: It allows the development of very specific helpers (most
  likely non-standard proprietary protocols) that are very likely not to be
  accepted for mainline inclusion in the form of kernel-space connection
  tracking helpers.

This patch adds the infrastructure to allow the implementation of
user-space conntrack helpers by means of the new nfnetlink subsystem
`nfnetlink_cthelper' and the existing queueing infrastructure
(nfnetlink_queue).

I had to add the new hook NF_IP6_PRI_CONNTRACK_HELPER to register
ipv[4|6]_helper which results from splitting ipv[4|6]_confirm into
two pieces. This change is required not to break NAT sequence
adjustment and conntrack confirmation for traffic that is enqueued
to our user-space conntrack helpers.

Basic operation, in a few steps:

1) Register user-space helper by means of `nfct':

 nfct helper add ftp inet tcp

 [ It must be a valid existing helper supported by conntrack-tools ]

2) Add rules to enable the FTP user-space helper which is
   used to track traffic going to TCP port 21.

For locally generated packets:

 iptables -I OUTPUT -t raw -p tcp --dport 21 -j CT --helper ftp

For non-locally generated packets:

 iptables -I PREROUTING -t raw -p tcp --dport 21 -j CT --helper ftp

3) Run the test conntrackd in helper mode (see example files under
   doc/helper/conntrackd.conf

 conntrackd

4) Generate FTP traffic going, if everything is OK, then conntrackd
   should create expectations (you can check that with `conntrack':

 conntrack -E expect

    [NEW] 301 proto=6 src=192.168.1.136 dst=130.89.148.12 sport=0 dport=54037 mask-src=255.255.255.255 mask-dst=255.255.255.255 sport=0 dport=65535 master-src=192.168.1.136 master-dst=130.89.148.12 sport=57127 dport=21 class=0 helper=ftp
[DESTROY] 301 proto=6 src=192.168.1.136 dst=130.89.148.12 sport=0 dport=54037 mask-src=255.255.255.255 mask-dst=255.255.255.255 sport=0 dport=65535 master-src=192.168.1.136 master-dst=130.89.148.12 sport=57127 dport=21 class=0 helper=ftp

This confirms that our test helper is receiving packets including the
conntrack information, and adding expectations in kernel-space.

The user-space helper can also store its private tracking information
in the conntrack structure in the kernel via the CTA_HELP_INFO. The
kernel will consider this a binary blob whose layout is unknown. This
information will be included in the information that is transfered
to user-space via glue code that integrates nfnetlink_queue and
ctnetlink.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

This attribute can be used to modify and to dump the internal
protocol information.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

User-space programs that receive traffic via NFQUEUE may mangle packets.
If NAT is enabled, this usually puzzles sequence tracking, leading to
traffic disruptions.

With this patch, nfnl_queue will make the corresponding NAT TCP sequence
adjustment if:

1) The packet has been mangled,
2) the NFQA_CFG_F_CONNTRACK flag has been set, and
3) NAT is detected.

There are some records on the Internet complaning about this issue:
http://stackoverflow.com/questions/260757/packet-mangling-utilities-besides-iptables

By now, we only support TCP since we have no helpers for DCCP or SCTP.
Better to add this if we ever have some helper over those layer 4 protocols.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

This patch allows you to include the conntrack information together
with the packet that is sent to user-space via NFQUEUE.

Previously, there was no integration between ctnetlink and
nfnetlink_queue. If you wanted to access conntrack information
from your libnetfilter_queue program, you required to query
ctnetlink from user-space to obtain it. Thus, delaying the packet
processing even more.

Including the conntrack information is optional, you can set it
via NFQA_CFG_F_CONNTRACK flag with the new NFQA_CFG_FLAGS attribute.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

This patch uses the new variable length conntrack extensions.

Instead of using union nf_conntrack_help that contain all the
helper private data information, we allocate variable length
area to store the private helper data.

This patch includes the modification of all existing helpers.
It also includes a couple of include header to avoid compilation
warnings.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

We can now define conntrack extensions of variable size. This
patch is useful to get rid of these unions:

union nf_conntrack_help
union nf_conntrack_proto
union nf_conntrack_nat_help
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

This patch modifies the struct nf_conntrack_helper to allocate
the room for the helper name. The maximum length is 16 bytes
(this was already introduced in 2.6.24).

For the maximum length for expectation policy names, I have
also selected 16 bytes.

This patch is required by the follow-up patch to support
user-space connection tracking helpers.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Conflicts:
	net/ipv6/route.c

Pull in 'net' again to get the revert of Thomas's change
which introduced regressions.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts commit 2a0c451a.

It causes crashes, because now ip6_null_entry is used before
it is initialized.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The mtu should be a __be32, not the mark.
Reported-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Conflicts:
	net/ipv6/route.c

This deals with a merge conflict between the net-next addition of the
inetpeer network namespace ops, and Thomas Graf's bug fix in
2a0c451a which makes sure we don't
register /proc/net/ipv6_route before it is actually safe to do so.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

/proc/net/ipv6_route reflects the contents of fib_table_hash. The proc
handler is installed in ip6_route_net_init() whereas fib_table_hash is
allocated in fib6_net_init() _after_ the proc handler has been installed.

This opens up a short time frame to access fib_table_hash with its pants
down.

fib6_init() as a whole can't be moved to an earlier position as it also
registers the rtnetlink message handlers which should be registered at
the end. Therefore split it into fib6_init() which is run early and
fib6_init_late() to register the rtnetlink message handlers.
Signed-off-by: NThomas Graf <tgraf@suug.ch>
Reviewed-by: NNeil Horman <nhorman@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The adapter->npars[] array has QLCNIC_MAX_PCI_FUNC elements.  We
allocate it that way a few lines earlier in the function.  So this test
is off by one.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NAnirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Orphaning skb in dev_hard_start_xmit() makes bonding behavior
unfriendly for applications sending big UDP bursts : Once packets
pass the bonding device and come to real device, they might hit a full
qdisc and be dropped. Without orphaning, the sender is automatically
throttled because sk->sk_wmemalloc reaches sk->sk_sndbuf (assuming
sk_sndbuf is not too big)

We could try to defer the orphaning adding another test in
dev_hard_start_xmit(), but all this seems of little gain,
now that BQL tends to make packets more likely to be parked
in Qdisc queues instead of NIC TX ring, in cases where performance
matters.

Reverts commits :
fc6055a5 net: Introduce skb_orphan_try()
87fd308c net: skb_tx_hash() fix relative to skb_orphan_try()
and removes SKBTX_DRV_NEEDS_SK_REF flag
Reported-and-bisected-by: NJean-Michel Hautbois <jhautbois@gmail.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Tested-by: NOliver Hartkopp <socketcan@hartkopp.net>
Acked-by: NOliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is a off by one error in the minimal number of BD in
bnx2x_start_xmit() and bnx2x_tx_int() before stopping/resuming tx queue.

A full size GSO packet, with data included in skb->head really needs
(MAX_SKB_FRAGS + 4) BDs, because of bnx2x_tx_split()

This error triggers if BQL is disabled and heavy TCP transmit traffic
occurs.

bnx2x_tx_split() definitely can be called, remove a wrong comment.
Reported-by: NTomas Hruby <thruby@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Cc: Yaniv Rosner <yanivr@broadcom.com>
Cc: Merav Sicron <meravs@broadcom.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Robert Evans <evansr@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

(CAN_CTRLMODE_LISTENONLY & CAN_CTRLMODE_LOOPBACK) is (0x02 & 0x01) which
is zero so the condition is never true.  The intent here was to test
that both flags were set.

Cc: <stable@kernel.org> # 2.6.39+
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Acked-by: NOliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: NMarc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

One tricky issue on the ipv6 side vs. ipv4 is that the ICMP callouts
to handle the error pass the 32-bit info cookie in network byte order
whereas ipv4 passes it around in host byte order.

Like the ipv4 side, we have two helper functions.  One for when we
have a socket context and one for when we do not.

ip6ip6 tunnels are not handled here, because they handle PMTU events
by essentially relaying another ICMP packet-too-big message back to
the original sender.

This patch allows us to get rid of rt6_do_pmtu_disc().  It handles all
kinds of situations that simply cannot happen when we do the PMTU
update directly using a fully resolved route.

In fact, the "plen == 128" check in ip6_rt_update_pmtu() can very
likely be removed or changed into a BUG_ON() check.  We should never
have a prefixed ipv6 route when we get there.

Another piece of strange history here is that TCP and DCCP, unlike in
ipv4, never invoke the update_pmtu() method from their ICMP error
handlers.  This is incredibly astonishing since this is the context
where we have the most accurate context in which to make a PMTU
update, namely we have a fully connected socket and associated cached
socket route.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With ip_rt_frag_needed() removed, we have to explicitly update PMTU
information in every ICMP error handler.

Create two helper functions to facilitate this.

1) ipv4_sk_update_pmtu()

   This updates the PMTU when we have a socket context to
   work with.

2) ipv4_update_pmtu()

   Raw version, used when no socket context is available.  For this
   interface, we essentially just pass in explicit arguments for
   the flow identity information we would have extracted from the
   socket.

   And you'll notice that ipv4_sk_update_pmtu() is simply implemented
   in terms of ipv4_update_pmtu()

Note that __ip_route_output_key() is used, rather than something like
ip_route_output_flow() or ip_route_output_key().  This is because we
absolutely do not want to end up with a route that does IPSEC
encapsulation and the like.  Instead, we only want the route that
would get us to the node described by the outermost IP header.
Reported-by: NSteffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull kvm fix from Marcelo Tosatti:
 "Fix a spurious warning on CPU offline path"

* git://git.kernel.org/pub/scm/virt/kvm/kvm:
  x86: kvmclock: remove check_and_clear_guest_paused warning

Merge tag 'pinctrl-fixes-for-v3.5' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl

Pull pinctrl fixes from Linus Walleij:
 - section markup fixes
 - clk_prepare() fix to conform to the clk API
 - memory leaks
 - incorrect debug messages
 - bad errorpaths
 - typos

* tag 'pinctrl-fixes-for-v3.5' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: pinctrl-mxs: set platform driver data to NULL at errpath and at unregister
  pinctrl: pinctrl-mxs: Take care of frees if the kzalloc fails
  pinctrl: pinctrl-imx: fix incorrect debug message of maps
  pinctrl: pinctrl-imx: free if of_get_parent fails to get the parent node
  pinctrl: pinctrl-imx: free allocated pinctrl_map structure only once and use kernel facilities for IMX_PMX_DUMP
  pinctrl: nomadik: fix up typo
  pinctrl: nomadik: add clk_prepare() call
  pinctrl: fix a minor harmless typo
  pinctrl: sirf: mark of_device_id match table as __devinitconst

Pull sound fixes from Takashi Iwai:

 - Fix a regression of USB-audio PCM assignment since 3.4
 - A few VGA-switcheroo-related fixes for proper HDMI audio enablement
 - Fixed the missing initializations of HD-audio verbs, which may have
   resulted in various breakage
 - Some driver-specific ASoC updates
 - A few fixes for the dynamic PCM code
 - The addition of pinctrl support for the i.MX audmux which didn't make
   it into -rc1 due to cross tree dependency issues
 - A few minor fixes in compress API codes

* tag 'sound-3.5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Don't forget to call init verbs added by fixup list
  ALSA: HDA: Pin fixup for Zotac Z68 motherboard
  ALSA: compress_core: cleanup pointers on stop
  ALSA: compress_core: don't wake up on pause
  ALSA: hda - Fix detection of Creative SoundCore3D controllers
  vga_switcheroo: Enable/disable audio clients at the right time
  ALSA: hda - HDMI Audio init all connectors when VGA-switcheroo is off
  vga_switcheroo: Fix error without CONFIG_VGA_SWITCHEROO
  ALSA: hda - Fix uninitialized HDMI controllers with VGA-switcheroo
  vga_switcheroo: Add a helper function to get the client state
  ALSA: usb-audio: Fix substream assignments
  ASoC: tegra: add MODULE_DEVICE_TABLE to tegra30_ahub
  ASoC: wm2000: Always use a 4s timeout for the firmware
  ASoC: dapm: Fix input list to use source widgets
  ASoC: dpcm: Fix dpcm_get_be() to check that DAI is BE
  ASoC: wm8994: Apply volume updates with clocks enabled
  ASoC: wm8994: Ensure all AIFnCLK events are run from the _late variants
  ASoC: imx-audmux: add pinctrl support
  ASoC: dapm: Fix connected widget capture path query.

Pull networking fixes from David S. Miller:

This has the fix for the wireless issues I ran into the other week as
well as:

 1) Fix CAN c_can driver transmit handling resulting in BUG check
    triggers, from AnilKumar Ch.

 2) Fix packet drop monitor sleeping in atomic context, from Eric
    Dumazet.

 3) Fix mv643xx_eth driver build regression, from Andrew Lunn.

 4) Inetpeer freeing needs an RCU grace period in order to avoid races
    during tree invalidation.  From Eric Dumazet.

 5) Fix endianness bugs in xt_HMARK netfilter module, from Hans
    Schillstrom.

 6) Add proper module refcounting to l2tp_eth to avoid crash on module
    unload, from Eric Dumazet.

 7) Fix truncation of neighbour entry dumps due to logic errors in
    neigh_dump_info() and friends, from Eric Dumazet.

 8) The conversion of fib6_age() to dst_neigh_lookup() accidently
    reversed the logic of a flags test, fix from Thomas Graf.

 9) Fix checksum configuration in newer sky2 chips, from Stephen
    Hemminger.

10) Revert BQL support in NIU driver, doesn't work.

11) l2tp_ip_sendmsg() illegally uses a route without a proper reference.
    From Eric Dumazet.

12) be2net driver references an SKB after it's potentially been freed,
    also from Eric Dumazet.

13) Fix RCU stalls in dummy net driver init.  Also from Eric Dumazet.

14) lpc_eth has several bugs in it's transmit engine leading to packet
    leaks and improper queue wakes, from Eric Dumazet.

15) Apply short DMA workaround to more tg3 chips, from Matt Carlson.

16) Add tilegx network driver.

17) Bonding queue mapping for a packet can get corrupted, fix from Eric
    Dumazet.

18) Fix bug in netpoll_send_udp() SKB management that can leave garbage
    in the payload in certain situations.  From Eric Dumazet.

19) bnx2x driver interprets chip RX checksum offload incorrectly in
    encapsulation situations.  Fix from Eric Dumazet.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (75 commits)
  bnx2x: fix checksum validation
  netpoll: fix netpoll_send_udp() bugs
  bonding: Fix corrupted queue_mapping
  bonding:record primary when modify it via sysfs
  tilegx network driver: initial support
  tg3: Apply short DMA frag workaround to 5906
  net: stmmac: Fix clock en-/disable calls
  lpc_eth: fix tx completion
  lpc_eth: add missing ndo_change_mtu()
  dummy: fix rcu_sched self-detected stalls
  net: Reorder initialization in ip_route_output to fix gcc warning
  virtio-net: fix a race on 32bit arches
  r8169: avoid NAPI scheduling delay.
  net: Make linux/tcp.h C++ friendly (trivial)
  netdev: fix drivers/net/phy/ kernel-doc warnings
  net/core: fix kernel-doc warnings
  be2net: fix a race in be_xmit()
  l2tp: fix a race in l2tp_ip_sendmsg()
  mac80211: add back channel change flag
  NFC: Fix possible NULL ptr deref when getting the name of a socket
  ...

This patch fixes a potential Rx timestamp deadlock that causes the Rx
timestamping to stall indefinitely. The issue could occur when a PTP packet is
timestamped by hardware but never reaches the Rx queue. In order to prevent a
permanent loss of timestamping, the RXSTMP(L/H) registers have to be read to
unlock them. (This used to only occur when a packet that was timestamped
reached the software.) However the registers can't be read early otherwise
there is no way to correlate them to the packet.

This patch introduces a filter function which can be used to determine if a
packet should have been timestamped. Supplied with the filter setup by the
hwtstamp ioctl, check to make sure the PTP protocol and message type match the
expected values. If so, then read the timestamp registers (to free them.) At
this point check the descriptor bit, if the bit is set then we know this
packet correlates to the timestamp stored in the RXTSTAMP registers.
Otherwise, assume that packet was dropped by the hardware, and ignore this
timestamp value. However, we have at least unlocked the rxtstamp registers for
future timestamping.

Due to the way the driver handles skb data, it cannot be directly accessed. In
order to work around this, a copy of the skb data into a linear buffer is
made. From this buffer it becomes possible to read the data correctly
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Reviewed-by: NRichard Cochran <richardcochran@gmail.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

When enabling the hwtstamp mode for Rx timestamping the V2 ptp event type
specific modes (Delay Request and Sync) have been rolled into the V2 all event
packet modes, in order to more accurately represent what hardware is doing.
Hardware always timestamps the Path delay packets when a V2 mode is selected,
regardless of what type was selected (in order to always support Path delay
mode). However this means the user selected modes of timestamping only Sync or
Delay Request is not truly supported. This patch correctly sets the mode for
the hwtstamp config and returns to the user that all V2 event packets will be
timestamped.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

This patch fixes two minor nits from Richard Cochran. The first is a case of
ambitious line wrapping that wasn't necessary. The second is to re-order the
flag checks for PPS support. Previously, the hardware test was done first, and
the interrupt flag test was done second. Now, test the interrupt flag and use
the unlikely macro.
Signed-off-by: NJacob Keller <jacob.e.keller@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe_sysfs.c is only needed when CONFIG_IXGBE_HWMON is configured in the
kernel.
Signed-off-by: NEmil Tantilov <emil.s.tantilov@intel.com>
Acked-by: NDon Skidmore <Donald.c.skidmore@intel.com>
Tested-by: NPhil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

The flow control DV macros are used to calculate the flow control
high and low thresholds. This patch annotates these macros slightly
better and fixes the issues below.

The macro variables are renamed LINK to _max_frame_link and TC to
_max_frame_tc. This was to avoid confusion and make them more
readable. It was found that people auditing the code read TC to be
'traffic class' in the 802.1Q definition instead of the max frame
size of the tc. Hopefully it is clear now.

This audit also found the following real deviations from the
theoretical values. Fixed in this patch.

  * I multiplied the DV calculations by (36/25) which always
    evaluates to 1. This does not match the intended theoretical
    value of 1.44.

  * IXGBE_BT2KB added 1023 to account for rounding however this
    really should be 8 * 1023 - 1 to account for division by 8k.

  * x2 multiplication of max frame in DV calculations to account
    for updated hardware recommendations.

With this patch the DV values are inline with the recommendations
in the 82599 and 82598 data sheets. Its worth noting I did not
see any dropped frames with flow control on in my experiments without
this patch. However aligning with the hardware specs and
recommendations seems like a good idea here to account for worst
case scenarios.
Signed-off-by: NJohn Fastabend <john.r.fastabend@intel.com>
Tested-by: NRoss Brattain <ross.b.brattain@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Based on a report from Ethan Zhao, before calling register_netdev() the
driver should be using logging macros that do not display the potentially
confusing "(unregistered net_device)" yet still display the useful driver
name and PCI bus/device/function.
Reported-by: NEthan Zhao <ethan.kernel@gmail.com>
Signed-off-by: NBruce Allan <bruce.w.allan@intel.com>
Tested-by: NAaron Brown <aaron.f.brown@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>

Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When packets are dropped in TX path, its better to use kfree_skb()
instead of dev_kfree_skb() to give proper drop_monitor events.

Also move the kfree_skb() call after read_unlock() in bond_alb_xmit()
and bond_xmit_activebackup()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

bnx2x driver incorrectly sets ip_summed to CHECKSUM_UNNECESSARY on
encapsulated segments. TCP stack happily accepts frames with bad
checksums, if they are inside a GRE or IPIP encapsulation.

Our understanding is that if no IP or L4 csum validation was done by the
hardware, we should leave ip_summed as is (CHECKSUM_NONE), since
hardware doesn't provide CHECKSUM_COMPLETE support in its cqe.

Then, if IP/L4 checksumming was done by the hardware, set
CHECKSUM_UNNECESSARY if no error was flagged.

Patch based on findings and analysis from Robert Evans
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Eilon Greenstein <eilong@broadcom.com>
Cc: Yaniv Rosner <yanivr@broadcom.com>
Cc: Merav Sicron <meravs@broadcom.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Robert Evans <evansr@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Acked-by: NEilon Greenstein <eilong@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Bogdan Hamciuc diagnosed and fixed following bug in netpoll_send_udp() :

"skb->len += len;" instead of "skb_put(skb, len);"

Meaning that _if_ a network driver needs to call skb_realloc_headroom(),
only packet headers would be copied, leaving garbage in the payload.

However the skb_realloc_headroom() must be avoided as much as possible
since it requires memory and netpoll tries hard to work even if memory
is exhausted (using a pool of preallocated skbs)

It appears netpoll_send_udp() reserved 16 bytes for the ethernet header,
which happens to work for typicall drivers but not all.

Right thing is to use LL_RESERVED_SPACE(dev)
(And also add dev->needed_tailroom of tailroom)

This patch combines both fixes.

Many thanks to Bogdan for raising this issue.
Reported-by: NBogdan Hamciuc <bogdan.hamciuc@freescale.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Tested-by: NBogdan Hamciuc <bogdan.hamciuc@freescale.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Neil Horman <nhorman@tuxdriver.com>
Reviewed-by: NNeil Horman <nhorman@tuxdriver.com>
Reviewed-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

EMSGSIZE - ran out of space while constructing message
EOPNOTSUPP - driver/hardware does not support operation
ENODEV - network device not found
EINVAL - invalid message
Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>