ixgbe: Make ethtool register test use accessors

Make the ethtool register test use the normal register accessor
functions. Also eliminate macros used for calling register test
functions to make error exits clearer. Use boolean values for
boolean returns instead of 0 and 1.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Use static inlines instead of macros

Kernel coding standard prefers static inline functions instead
of macros, so use them for register accessors. This is to prepare
for adding LER, Live Error Recovery, checks to those accessors.

Temporarily provide macros for calling the new static inline
accessors until all references are changed.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixbge: Protect ixgbe_down with __IXGBE_DOWN bit

The ixgbe_down function can now prevent multiple executions by
doing test_and_set_bit on __IXGBE_DOWN. This did not work before
introduction of the __IXGBE_REMOVING bit, because of overloading
of __IXGBE_DOWN. Also add smp_mb__before_clear_bit call before
clearing the __IXGBE_DOWN bit.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: Indicate removal state explicitly

Add a bit, __IXGBE_REMOVING, to indicate that the module is being
removed. The __IXGBE_DOWN bit had been overloaded for this purpose,
but that leads to trouble. A few places now check both __IXGBE_DOWN
and __IXGBE_REMOVE. Notably, setting either bit will prevent service
task execution.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'i40e'

Aaron Brown says:

====================
Intel Wired LAN Driver Updates

This series contains updates to i40e that are primarily minor fixes or
general cleanup.

Shannon fixes a bug where the VMDq queue is not associated with the
right setup within the hardware.

Mitch provides a patch adjusting where the VF is reset and another
one adding meaningful context to a message.

Jesse cleans up white space comments and parenthesis.

Catherine bumps the version.
====================

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

i40e: Bump version number

Update the driver version to 0.3.30-k.

Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: trivial cleanup

Remove some un-necessary parenthesis.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: whitespace fixes

Fix some whitespace and comment issues.

Change-ID: I1587599e50ce66fd389965720e86f9e331d86643
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: make message meaningful

Make this message mean something, rather than just spitting out a VSI id
without any context whatsoever.

Change-ID: Iafb906c6db46d4b5dcbe84adc9ed44730d08bd42
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: associate VMDq queue with VM type

Fix a bug where the queue was not associated with the right set-up
within the hardware.  The fix is to use the right QTX_CTL VSI type
when associating it to the VSI.

Change-ID: I65ef6c5a8205601c640a6593e4b7e78d6ba45545
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

i40e: remove extra register write

This write done at the end of VF reset and should not be performed here.

Change-ID: I4d89813b68c6173184293868a6f26cf559bc2405
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: do not export tcp_gso_segment() and tcp_gro_receive()

tcp_gso_segment() and tcp_gro_receive() no longer need to be
exported. IPv4 and IPv6 offloads are statically linked.

Note that tcp_gro_complete() is still used by bnx2x, unfortunately.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'dev_get_by_index'

Ying Xue says:

====================
use appropriate APIs to get interfaces

Under rtnl_lock protection, we should use __dev_get_name/index()
rather than dev_get_name()/index() to find interface handlers
because the former interfaces can help us avoid to change interface
reference counter.

v2 changes:
 - Change return value of nl80211_set_wiphy() to 0 in patch #10
   by johannes's suggestion.
 - Add 'Acked-by' into several patches which were acknowledged by
   corresponding maintainers.
====================

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

net: nl80211: __dev_get_by_index instead of dev_get_by_index to find interface

As __cfg80211_rdev_from_attrs(), nl80211_dump_wiphy_parse() and
nl80211_set_wiphy() are all under rtnl_lock protection,
__dev_get_by_index() instead of dev_get_by_index() should be used
to find interface handler in them allowing us to avoid to change
interface reference counter.

Cc: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: use __dev_get_by_index instead of dev_get_by_index to find interface

As cgw_create_job() is always under rtnl_lock protection,
__dev_get_by_index() instead of dev_get_by_index() should be used to
find interface handler in it having us avoid to change interface
reference counter.

Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

caif: __dev_get_by_index instead of dev_get_by_index to find interface

The following call chains indicate that chnl_net_open() is under
rtnl_lock protection as __dev_open() is protected by rtnl_lock.
So if __dev_get_by_index() instead of dev_get_by_index() is used
to find interface handler in it, this would help us avoid to change
interface reference counter.

__dev_open()
  chnl_net_open()

Cc: Dmitry Tarnyagin <dmitry.tarnyagin@lockless.no>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

batman-adv: use __dev_get_by_index instead of dev_get_by_index to find interface

The following call chains indicate that batadv_is_on_batman_iface()
is always under rtnl_lock protection as call_netdevice_notifier()
is protected by rtnl_lock. So if __dev_get_by_index() rather than
dev_get_by_index() is used to find interface handler in it, this
would help us avoid to change interface reference counter.

call_netdevice_notifier()
  batadv_hard_if_event()
    batadv_hardif_add_interface()
      batadv_is_valid_iface()
        batadv_is_on_batman_iface()

Cc: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxlan: use __dev_get_by_index instead of dev_get_by_index to find interface

The following call chains indicate that vxlan_fdb_parse() is
under rtnl_lock protection. So if we use __dev_get_by_index()
instead of dev_get_by_index() to find interface handler in it,
this would help us avoid to change interface reference counter.

rtnetlink_rcv()
  rtnl_lock()
  netlink_rcv_skb()
    rtnl_fdb_add()
      vxlan_fdb_add()
        vxlan_fdb_parse()
  rtnl_unlock()

rtnetlink_rcv()
  rtnl_lock()
  netlink_rcv_skb()
    rtnl_fdb_del()
      vxlan_fdb_del()
        vxlan_fdb_parse()
  rtnl_unlock()

Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

decnet: use __dev_get_by_index instead of dev_get_by_index to find interface

The following call chain we can identify that dn_cache_getroute() is
protected under rtnl_lock. So if we use __dev_get_by_index() instead
of dev_get_by_index() to find interface handlers in it, this would help
us avoid to change interface reference counter.

rtnetlink_rcv()
  rtnl_lock()
    netlink_rcv_skb()
      dn_cache_getroute()
  rtnl_unlock()

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dcb: use __dev_get_by_name instead of dev_get_by_name to find interface

The following call chain indicates that dcb_doit() is protected
under rtnl_lock. So if we use __dev_get_by_name() instead of
dev_get_by_name() to find interface handlers in it, this would
help us avoid to change interface reference counter.

rtnetlink_rcv()
  rtnl_lock()
  netlink_rcv_skb()
    dcb_doit()
  rtnl_unlock()

Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

eql: use __dev_get_by_name instead of dev_get_by_name to find interface

The following call chain indicates that eql_ioctl(), eql_enslave(),
eql_emancipate(), eql_g_slave_cfg() and eql_s_slave_cfg() are
protected under rtnl_lock. So if we use __dev_get_by_name() instead
of dev_get_by_name() to find interface handlers in them, this would
help us avoid to change interface reference counters.

dev_ioctl()
  rtnl_lock()
    dev_ifsioc()
      eql_ioctl()
        eql_enslave()
eql_emancipate()
eql_g_slave_cfg()
eql_s_slave_cfg()
  rtnl_unlock()

Additionally we also change their return values from -EINVAL to
-ENODEV in case that interfaces are no found.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: use __dev_get_by_name instead of dev_get_by_name to find interface

The following call chain indicates that bond_do_ioctl() is protected
under rtnl_lock. If we use __dev_get_by_name() instead of
dev_get_by_name() to find interface handler in it, this would
help us avoid to change reference counter of interface once.

dev_ioctl()
  rtnl_lock()
  dev_ifsioc()
    bond_do_ioctl()
  rtnl_unlock()

Additionally we also change the coding style in bond_do_ioctl(),
letting it more readable for us.

Cc: Jay Vosburgh <fubar@us.ibm.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Drivers: Staging: cxt1e1: use __dev_get_name instead of dev_get_name to find interfaces

The following call chain denotes that both do_reset() and do_del_chan()
are protected under rtnl_lock. If we use __dev_get_by_name() instead of
dev_get_by_name() to find interface handlers in them, this would help
us avoid to change interface reference counter.

dev_ioctl()
  rtnl_lock()
  dev_ifsioc()
    c4_ioctl()
      do_reset()
      do_del_chan()
  rtnl_unlock()

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'r8152'

Hayes Wang says:

====================
r8152: remove limitation

Remove the limitation between ecm mode and vendor mode.

v2: replace the patch #3 with "ecm and vendor modes coexist".
====================

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

r8152: ecm and vendor modes coexist

Remove the limitation that the ecm and r8152 drivers couldn't coexist.
 - Remove the devices from the blacklist of relative drivers.
 - Remove usb_driver_set_configuration() from r8152 driver.
 - Modify the id_table of the r8152 driver for the vendor mode only.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

r8152: fix the warnings and a error from checkpatch.pl

Fix the following warnings and error:
 - WARNING: usb_free_urb(NULL) is safe this check is probably not required
 - WARNING: kfree(NULL) is safe this check is probably not required
 - ERROR: do not use C99 // comments

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

r8152: change the descriptor

The r8152 could support RTL8153. Update the relative descriptor.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bgmac: propagate error codes in bgmac_probe()

bgmac_mii_register() and register_netdev() both return appropriate error
codes for the failures they would encounter, propagate this error code
instead of overriding the value with -ENOTSUPP which is not the correct
error code to return.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPv6: move the anycast_src_echo_reply sysctl to netns_sysctl_ipv6

This change move anycast_src_echo_reply sysctl with other ipv6 sysctls.

Suggested-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Francois-Xavier Le Bail <fx.lebail@yahoo.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: remove a redundant NULL check

It confuses Smatch when we check "sinit" for NULL and then non-NULL and
that causes a false positive warning later.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: spelling fixes

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: addrconf spelling fixes

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'atl_stats'

Sabrina Dubroca says:

====================
atheros: modify statistics code

Following Ben Hutchings's advice on how to fill net_stats in alx [1],
this patch modifies the other atheros ethernet drivers
similarly. Minor whitespace/empty line changes in atl1c and atl1e to
make the code completely consistent between atl1c, atl1e, and alx.

I don't have this hardware, so these patches have only been
compile-tested.

v2 (changes only in atl1):
 - don't set soft_stats.rx_missed_errors (Ben)
 - add errors to soft_stats.{rx,tx}_packets (Ben)
 - add soft_stats.rx_dropped field and update soft_stats.rx_dropped
   instead of netdev->stats (overwritten) outside of the stats
   update function

Detail of the changes (v1):
* atl1/atl1c/atl1e
  - fix collisions computation
  - rx_dropped = rx_rrd_ov
  - rx_over_errors = 0
  - rx_missed_errors = 0
  - X_packets = X_ok + X_errors

* only atl1c/atl1e
  - add rx_rxf_ov to rx_errors

[1] http://www.spinics.net/lists/netdev/msg264930.html
====================

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

atl1: update statistics code

As Ben Hutchings pointed out for the stats in alx, some
hardware-specific stats aren't matched to the right net_device_stats
field. Also fix the collision field and include errors in the total
number of RX/TX packets. Add a rx_dropped field and use it where
netdev->stats was modified directly out of the stats update function.

Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

atl1e: update statistics code

As Ben Hutchings pointed out for the stats in alx, some
hardware-specific stats aren't matched to the right net_device_stats
field. Also fix the collision field and include errors in the total
number of RX/TX packets.

Minor whitespace fixes to match the style in alx.

Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atl1c: update statistics code

As Ben Hutchings pointed out for the stats in alx, some
hardware-specific stats aren't matched to the right net_device_stats
field. Also fix the collision field and include errors in the total
number of RX/TX packets.

Minor whitespace fixes to match the style in alx.

Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Spelling s/transmition/transmission/

Signed-off-by: Geert Uytterhoeven <geert+renesas@linux-m68k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: amd8111e: Spelling s/recive/receive/

Signed-off-by: Geert Uytterhoeven <geert+renesas@linux-m68k.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Correct default Tx switching behaviour

With this patch bnx2x will configure the PF to perform Tx switching on
out-going traffic as soon as SR-IOV is dynamically enabled and de-activate
it when it is disabled.
This will allow VFs to communicate with their parent PFs.

Signed-off-by: Yuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-davem' of git://gitorious.org/linux-can/linux-can-next

Marc Kleine-Budde says:

====================
this is a pull request of three patches for net-next/master.

Oleg Moroz added support for a new PCI card to the generic SJA1000 PCI
driver, Guenter Roeck's patch limits the flexcan driver to little
endian arm (and powerpc) and I fixed a sparse warning found by the
kbuild robot in the ti_hecc driver.
====================

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

net: replace macros net_random and net_srandom with direct calls to prandom

This patch removes the net_random and net_srandom macros and replaces
them with direct calls to the prandom ones. As new commits only seem to
use prandom_u32 there is no use to keep them around.
This change makes it easier to grep for users of prandom_u32.

Signed-off-by: Aruna-Hewapathirane <aruna.hewapathirane@gmail.com>
Suggested-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: copy traffic class from ping request to reply

Suggested-by: Simon Schneider <simon-schneider@gmx.net>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: register igmp_notifier even when !CONFIG_PROC_FS

We still need this notifier even when we don't config
PROC_FS.

It should be rare to have a kernel without PROC_FS,
so just for completeness.

Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'netdev_tracing'

Ben Hutchings says:

====================
Improve tracing at the driver/core boundary

These patches add static tracpeoints at the driver/core boundary which
record various skb fields likely to be useful for datapath debugging.
On the TX side the boundary is where the core calls ndo_start_xmit, and
on the RX side it is where any of the various exported receive functions
is called.

The set of skb fields is mostly based on what I thought would be
interesting for sfc.

These patches are basically the same as what I sent as an RFC in
November, but rebased.  They now depend on 'net: core: explicitly select
a txq before doing l2 forwarding', so please merge net into net-next
before trying to apply them.  The first patch fixes a code formatting
error left behind after that fix.
====================

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

net: Add trace events for all receive entry points, exposing more skb fields

The existing net/netif_rx and net/netif_receive_skb trace events
provide little information about the skb, nor do they indicate how it
entered the stack.

Add trace events at entry of each of the exported functions, including
most fields that are likely to be interesting for debugging driver
datapath behaviour.  Split netif_rx() and netif_receive_skb() so that
internal calls are not traced.

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

net: Add net_dev_start_xmit trace event, exposing more skb fields

The existing net/net_dev_xmit trace event provides little information
about the skb that has been passed to the driver, and it is not
simple to add more since the skb may already have been freed at
the point the event is emitted.

Add a separate trace event before the skb is passed to the driver,
including most fields that are likely to be interesting for debugging
driver datapath behaviour.

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

net: Fix indentation in dev_hard_start_xmit()

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

Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Merge branch 'skb_checksum_help'

Paul Durrant says:

====================
make skb_checksum_setup generally available

Both xen-netfront and xen-netback need to be able to set up the partial
checksum offset of an skb and may also need to recalculate the pseudo-
header checksum in the process. This functionality is currently private
and duplicated between the two drivers.

Patch #1 of this series moves the implementation into the core network code
as there is nothing xen-specific about it and it is potentially useful to
any network driver.
Patch #2 removes the private implementation from netback.
Patch #3 removes the private implementation from netfront.

v2:
- Put skb_checksum_setup in skbuff.c rather than dev.c
- remove inline
====================

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

xen-netfront: use new skb_checksum_setup function

Use skb_checksum_setup to set up partial checksum offsets rather
then a private implementation.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

xen-netback: use new skb_checksum_setup function

Use skb_checksum_setup to set up partial checksum offsets rather
then a private implementation.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add skb_checksum_setup

This patch adds a function to set up the partial checksum offset for IP
packets (and optionally re-calculate the pseudo-header checksum) into the
core network code.
The implementation was previously private and duplicated between xen-netback
and xen-netfront, however it is not xen-specific and is potentially useful
to any network driver.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Veaceslav Falico <vfalico@redhat.com>
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Cc: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: move br_net_exit() to br.c

And it can become static.

Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: usbnet: fix SG initialisation

Commit 60e453a940ac ("USBNET: fix handling padding packet")
added an extra SG entry in case padding is necessary, but
failed to update the initialisation of the list. This can
cause list traversal to fall off the end of the list,
resulting in an oops.

Fixes: 60e453a940ac ("USBNET: fix handling padding packet")
Reported-by: Thomas Kear <thomas@kear.co.nz>
Cc: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Tested-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: 3com: fix warning for incorrect type in argument

The commit c466a9b2b329f7d9982c14eedc83a923d3bc711c
(net: 3com: slight optimization of addr compare)
cause a warning: "passing argument 1 of 'ether_addr_equal'
from incompatible pointer type", so fix it.

I think julia will convert ether_addr_equal to ether_addr_equal_64bits later.

Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: qlcnic: fix warning for incorrect type in argument

The commit 6878f79a8b71e8c7b0587a1185584f54fd31f185
(net: qlcnic: slight optimization of addr compare)
cause a warning "sparse: incorrect type in argument 2
(different type sizes)", so fix it.

I think julia will convert ether_addr_equal to ether_addr_equal_64bits later.

Cc: Himanshu Madhani <himanshu.madhani@qlogic.com>
Cc: Rajesh Borundia <rajesh.borundia@qlogic.com>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sh_eth: fix garbled TX error message

sh_eth_error() in case of a TX error tries to print a message using 2 dev_err()
calls with the first string not finished by '\n', so that the resulting message
would inevitably come out garbled, with something like "3net eth0: " inserted
in the middle.  Avoid that by merging 2 calls into one.

While at it, insert an empty line after the nearby declaration.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next

Conflicts:
net/xfrm/xfrm_policy.c

Steffen Klassert says:

====================
This pull request has a merge conflict between commits be7928d20bab
("net: xfrm: xfrm_policy: fix inline not at beginning of declaration") and
da7c224b1baa ("net: xfrm: xfrm_policy: silence compiler warning") from
the net-next tree and commit 2f3ea9a95c58 ("xfrm: checkpatch erros with
inline keyword position") from the ipsec-next tree.

The version from net-next can be used, like it is done in linux-next.

1) Checkpatch cleanups, from Weilong Chen.

2) Fix lockdep complaints when pktgen is used with IPsec,
   from Fan Du.

3) Update pktgen to allow any combination of IPsec transport/tunnel mode
   and AH/ESP/IPcomp type, from Fan Du.

4) Make pktgen_dst_metrics static, Fengguang Wu.

5) Compile fix for pktgen when CONFIG_XFRM is not set,
   from Fan Du.
====================

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

inet_diag: fix inet_diag_dump_icsk() to use correct state for timewait sockets

Fix inet_diag_dump_icsk() to reflect the fact that both TCP_TIME_WAIT
and TCP_FIN_WAIT2 connections are represented by inet_timewait_sock
(not just TIME_WAIT), and for such sockets the tw_substate field holds
the real state, which can be either TCP_TIME_WAIT or TCP_FIN_WAIT2.

This brings the inet_diag state-matching code in line with the field
it uses to populate idiag_state. This is also analogous to the info
exported in /proc/net/tcp, where get_tcp4_sock() exports sk->sk_state
and get_timewait4_sock() exports tw->tw_substate.

Before fixing this, (a) neither "ss -nemoi" nor "ss -nemoi state
fin-wait-2" would return a socket in TCP_FIN_WAIT2; and (b) "ss -nemoi
state time-wait" would also return sockets in state TCP_FIN_WAIT2.

This is an old bug that predates 05dbc7b ("tcp/dccp: remove twchain").

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'bonding_rcu'

Veaceslav Falico says:

====================
bonding: fix bond_3ad RCU usage

While digging through bond_3ad.c I've found that the RCU usage there is
just wrong - it's used as a kind of mutex/spinlock instead of RCU.

v3->v4: remove useless goto and wrap __get_first_agg() in proper RCU.

v2->v3: make bond_3ad_set_carrier() use RCU read lock for the whole
function, so that all other functions will be protected by RCU as well.
This way we can use _rcu variants everywhere.

v1->v2: use generic primitives instead of _rcu ones cause we can hold RTNL
lock without RCU one, which is still safe.

This patchset is on top of bond_3ad.c cleanup:
http://www.spinics.net/lists/netdev/msg265447.html
====================

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

bonding: fix __get_active_agg() RCU logic

Currently, the implementation is meaningless - once again, we take the
slave structure and use it after we've exited RCU critical section.

Fix this by removing the rcu_read_lock() from __get_active_agg(), and
ensuring that all its callers are holding RCU.

Fixes: be79bd048 ("bonding: add RCU for bond_3ad_state_machine_handler()")
CC: dingtianhong@huawei.com
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: fix __get_first_agg RCU usage

Currently, the RCU read lock usage is just wrong - it gets the slave struct
under RCU and continues to use it when RCU lock is released.

However, it's still safe to do this cause we didn't need the
rcu_read_lock() initially - all of the __get_first_agg() callers are either
holding RCU read lock or the RTNL lock, so that we can't sync while in it.

Fixes: be79bd048 ("bonding: add RCU for bond_3ad_state_machine_handler()")
CC: dingtianhong@huawei.com
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: fix bond_3ad_set_carrier() RCU usage

Currently, its usage is just plainly wrong. It first gets a slave under
RCU, and, after releasing the RCU lock, continues to use it - whilst it can
be freed.

Fix this by ensuring that bond_3ad_set_carrier() holds RCU till it uses its
slave (or its agg).

Fixes: be79bd048ab ("bonding: add RCU for bond_3ad_state_machine_handler()")
CC: dingtianhong@huawei.com
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge tag 'batman-adv-for-davem' of git://git.open-mesh.org/linux-merge

Included changes:
- drop dependency against CRC16
- move to new release version
- add size check at compile time for packet structs
- update copyright years in every file
- implement new bonding/interface alternation feature

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

net: resort some Kbuild files to hopefully help avoid some conflicts

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

Merge branch 'qlcnic'

Shahed Shaikh says:

====================
This series includes following changes:
o SRIOV and VLAN filtering related enhancements which includes
   - Do MAC learning for PF
   - Restrict VF from configuring any VLAN mode
   - Enable flooding on PF
   - Turn on promiscuous mode for PF

o Bug fix in qlcnic_sriov_cleanup() introduced by commit
  154d0c81("qlcnic: VLAN enhancement for 84XX adapters")

o Beaconing support for 83xx and 84xx series adapters

o Allow 82xx adapter to perform IPv6 LRO even if destination IP address is not
  programmed.
====================

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

qlcnic: Update version to 5.3.54

Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Enable IPv6 LRO even if IP address is not programmed

o Enabling BIT_9 while configuring hardware LRO allows adapter to
  perform LRO even if destination IP address is not programmed in adapter.

Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Fix SR-IOV cleanup code path

o Add __QLCNIC_SRIOV_ENABLE bit check before doing SRIOV cleanup

Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Enable beaconing for 83xx/84xx Series adapter.

o Refactored code to handle beaconing test for all adapters.
o Use GET_LED_CONFIG mailbox command for 83xx/84xx series adapter
  to detect current beaconing state of the adapter.

Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Do MAC learning for SRIOV PF.

o MAC learning will be done for SRIOV PF to help program VLAN filters
  onto adapter. This will help VNIC traffic to flow through without
  flooding traffic.

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

qlcnic: Turn on promiscous mode for SRIOV PF.

o By default, SRIOV PF will have promiscous mode on.

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

qlcnic: Enable VF flood bit on PF.

o On enabling VF flood bit, PF driver will  be able to receive traffic
  from all its VFs.

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

qlcnic: Restrict VF from configuring any VLAN mode.

o Adapter should allow vlan traffic only for vlans configured on a VF.
  On configuring any vlan mode from VF, adapter will allow any vlan
  traffic to pass for that VF. Do not allow VF to configure this mode.

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

net: make dev_set_mtu() honor notification return code

Currently, after changing the MTU for a device, dev_set_mtu() calls
NETDEV_CHANGEMTU notification, however doesn't verify it's return code -
which can be NOTIFY_BAD - i.e. some of the net notifier blocks refused this
change, and continues nevertheless.

To fix this, verify the return code, and if it's an error - then revert the
MTU to the original one, notify again and pass the error code.

CC: Jiri Pirko <jiri@resnulli.us>
CC: "David S. Miller" <davem@davemloft.net>
CC: Eric Dumazet <edumazet@google.com>
CC: Alexander Duyck <alexander.h.duyck@intel.com>
CC: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Veaceslav Falico <vfalico@redhat.com>
Reviewed-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>

packet: doc: describe PACKET_MMAP with one packet socket for rx and tx

Document how to use one AF_PACKET mmap socket for RX and TX.

Signed-off-by: Norbert van Bolhuis <nvbolhuis@aimvalley.nl>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: make sctp_addto_chunk_fixed local

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Add of_phy_attach

10G PHYs don't currently support running the state machine, which
is implicitly setup via of_phy_connect(). Therefore, it is necessary
to implement an OF version of phy_attach(), which does everything
except start the state machine.

Signed-off-by: Andy Fleming <afleming@gmail.com>
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Support attaching to generic 10g driver

phy_attach_direct() may now attach to a generic 10G driver. It can
also be used exactly as phy_connect_direct(), which will be useful
when using of_mdio, as phy_connect (and therefore of_phy_connect)
start the PHY state machine, which is currently irrelevant for 10G
PHYs.

Signed-off-by: Andy Fleming <afleming@gmail.com>
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Add generic 10G driver

Very incomplete, but will allow for binding an ethernet controller
to it.

Signed-off-by: Andy Fleming <afleming@gmail.com>
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: turn genphy_driver to an array

Then other generic phy driver such as generic 10g phy driver can join it.

Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: introduce PHY_INTERFACE_MODE_XGMII for 10G PHY

Signed-off-by: Andy Fleming <afleming@gmail.com>
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Add Clause 45 read/write functions

Need an extra parameter to read or write Clause 45 PHYs, so
need a different API with the extra parameter.

Signed-off-by: Andy Fleming <afleming@gmail.com>
Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: make local functions static

Avoid needless export of local functions

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: namespace and dead code cleanups

Fix a bunch of whole lot of namespace issues with the Broadcom bnx2x driver
found by running 'make namespacecheck'

 * global variables must be prefixed with bnx2x_
    naming a variable int_mode, or num_queue is invitation to disaster

 * make local functions static

 * move some inline's used in one file out of header
   (this driver has a bad case of inline-itis)

 * remove resulting dead code fallout
   bnx2x_pfc_statistic,
 bnx2x_emac_get_pfc_stat
   bnx2x_init_vlan_mac_obj,
   Looks like vlan mac support in this driver was a botch from day one
   either never worked, or not implemented or missing support functions

Compile tested only.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers: net: silence compiler warning in smc91x.c

If used 64 bit compiler GCC warns that:

drivers/net/ethernet/smsc/smc91x.c:1897:7:
warning: cast from pointer to integer of different
size [-Wpointer-to-int-cast]

This patch fixes this by changing typecast from "unsigned int" to "unsigned long"

CC: "David S. Miller" <davem@davemloft.net>
CC: Jingoo Han <jg1.han@samsung.com>
CC: netdev@vger.kernel.org
Signed-off-by: Pankaj Dubey <pankaj.dubey@samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gre_offload: simplify GRE header length calculation in gre_gso_segment()

Simplify the GRE header length calculation in gre_gso_segment().
Switch to an approach that is simpler, faster, and more general. The
new approach will continue to be correct even if we add support for
the optional variable-length routing info that may be present in a GRE
header.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: H.K. Jerry Chu <hkchu@google.com>
Cc: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: act: remove struct tcf_act_hdr

It is not necessary at all.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: avoid casting void pointer

tp->root is a void* pointer, no need to cast it.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: optimize tcf_match_indev()

tcf_match_indev() is called in fast path, it is not wise to
search for a netdev by ifindex and then compare by its name,
just compare the ifindex.

Also, dev->name could be changed by user-space, therefore
the match would be always fail, but dev->ifindex could
be consistent.

BTW, this will also save some bytes from the core struct of u32.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: add struct net pointer to tcf_proto_ops->dump

It will be needed by the next patch.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: act: clean up notification functions

Refactor tcf_add_notify() and factor out tcf_del_notify().

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net_sched: act: move idx_gen into struct tcf_hashinfo

There is no need to store the index separatedly
since tcf_hashinfo is allocated statically too.

Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: gro: change GRO overflow strategy

GRO layer has a limit of 8 flows being held in GRO list,
for performance reason.

When a packet comes for a flow not yet in the list,
and list is full, we immediately give it to upper
stacks, lowering aggregation performance.

With TSO auto sizing and FQ packet scheduler, this situation
happens more often.

This patch changes strategy to simply evict the oldest flow of
the list. This works better because of the nature of packet
trains for which GRO is efficient. This also has the effect
of lowering the GRO latency if many flows are competing.

Tested :

Used a 40Gbps NIC, with 4 RX queues, and 200 concurrent TCP_STREAM
netperf.

Before patch, aggregate rate is 11Gbps (while a single flow can reach
30Gbps)

After patch, line rate is reached.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jerry Chu <hkchu@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/mlx4_en: call gro handler for encapsulated frames

In order to use the native GRO handling of encapsulated protocols on
mlx4, we need to call napi_gro_receive() instead of netif_receive_skb()
unless busy polling is in action.

While we are at it, rename mlx4_en_cq_ll_polling() to
mlx4_en_cq_busy_polling()

Tested with GRE tunnel : GRO aggregation is now performed on the
ethernet device instead of being done later on gre device.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Amir Vadai <amirv@mellanox.com>
Cc: Jerry Chu <hkchu@google.com>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Acked-By: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gre_offload: fix sparse non static symbol warning

Fixes the following sparse warning:

net/ipv4/gre_offload.c:253:5: warning:
 symbol 'gre_gro_complete' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ip_forward_pmtu'

Hannes Frederic Sowa says:

====================
path mtu hardening patches

After a lot of back and forth I want to propose these changes regarding
path mtu hardening and give an outline why I think this is the best way
how to proceed:

This set contains the following patches:
* ipv4: introduce ip_dst_mtu_maybe_forward and protect forwarding path against pmtu spoofing
* ipv6: introduce ip6_dst_mtu_forward and protect forwarding path with it
* ipv4: introduce hardened ip_no_pmtu_disc mode

The first one switches the forwarding path of IPv4 to use the interface
mtu by default and ignore a possible discovered path mtu. It provides
a sysctl to switch back to the original behavior (see discussion below).

The second patch does the same thing unconditionally for IPv6. I don't
provide a knob for IPv6 to switch to original behavior (please see
below).

The third patch introduces a hardened pmtu mode, where only pmtu
information are accepted where the protocol is able to do more stringent
checks on the icmp piggyback payload (please see the patch commit msg
for further details).

Why is this change necessary?

First of all, RFC 1191 4. Router specification says:
"When a router is unable to forward a datagram because it exceeds the
 MTU of the next-hop network and its Don't Fragment bit is set, the
 router is required to return an ICMP Destination Unreachable message
 to the source of the datagram, with the Code indicating
 "fragmentation needed and DF set". ..."

For some time now fragmentation has been considered problematic, e.g.:
* http://www.hpl.hp.com/techreports/Compaq-DEC/WRL-87-3.pdf
* http://tools.ietf.org/search/rfc4963

Most of them seem to agree that fragmentation should be avoided because
of efficiency, data corruption or security concerns.

Recently it was shown possible that correctly guessing IP ids could lead
to data injection on DNS packets:
<https://sites.google.com/site/hayashulman/files/fragmentation-poisoning.pdf>

While we can try to completly stop fragmentation on the end host
(this is e.g. implemented via IP_PMTUDISC_INTERFACE), we cannot stop
fragmentation completly on the forwarding path. On the end host the
application has to deal with MTUs and has to choose fallback methods
if fragmentation could be an attack vector. This is already the case for
most DNS software, where a maximum UDP packet size can be configured. But
until recently they had no control over local fragmentation and could
thus emit fragmented packets.

On the forwarding path we can just try to delay the fragmentation to
the last hop where this is really necessary. Current kernel already does
that but only because routers don't receive feedback of path mtus, these are
only send back to the end host system. But it is possible to maliciously
insert path mtu inforamtion via ICMP packets which have an icmp echo_reply
payload, because we cannot validate those notifications against local
sockets. DHCP clients which establish an any-bound RAW-socket could also
start processing unwanted fragmentation-needed packets.

Why does IPv4 has a knob to revert to old behavior while IPv6 doesn't?

IPv4 does fragmentation on the path while IPv6 does always respond with
packet-too-big errors. The interface MTU will always be greater than
the path MTU information. So we would discard packets we could actually
forward because of malicious information. After this change we would
let the hop, which really could not forward the packet, notify the host
of this problem.

IPv4 allowes fragmentation mid-path. In case someone does use a software
which tries to discover such paths and assumes that the kernel is handling
the discovered pmtu information automatically. This should be an extremly
rare case, but because I could not exclude the possibility this knob is
provided. Also this software could insert non-locked mtu information
into the kernel. We cannot distinguish that from path mtu information
currently. Premature fragmentation could solve some problems in wrongly
configured networks, thus this switch is provided.

One frag-needed packet could reduce the path mtu down to 522 bytes
(route/min_pmtu).

Misc:

IPv6 neighbor discovery could advertise mtu information for an
interface. These information update the ipv6-specific interface mtu and
thus get used by the forwarding path.

Tunnel and xfrm output path will still honour path mtu and also respond
with Packet-too-Big or fragmentation-needed errors if needed.

Changelog for all patches:
v2)
* enabled ip_forward_use_pmtu by default
* reworded
v3)
* disabled ip_forward_use_pmtu by default
* reworded
v4)
* renamed ip_dst_mtu_secure to ip_dst_mtu_maybe_forward
* updated changelog accordingly
* removed unneeded !!(... & ...) double negations

v2)
* by default we honour pmtu information
3)
* only honor interface mtu
* rewritten and simplified
* no knob to fall back to old mode any more

v2)
* reworded Documentation
====================

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

ipv4: introduce hardened ip_no_pmtu_disc mode

This new ip_no_pmtu_disc mode only allowes fragmentation-needed errors
to be honored by protocols which do more stringent validation on the
ICMP's packet payload. This knob is useful for people who e.g. want to
run an unmodified DNS server in a namespace where they need to use pmtu
for TCP connections (as they are used for zone transfers or fallback
for requests) but don't want to use possibly spoofed UDP pmtu information.

Currently the whitelisted protocols are TCP, SCTP and DCCP as they check
if the returned packet is in the window or if the association is valid.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: John Heffner <johnwheffner@gmail.com>
Suggested-by: Florian Weimer <fweimer@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: introduce ip6_dst_mtu_forward and protect forwarding path with it

In the IPv6 forwarding path we are only concerend about the outgoing
interface MTU, but also respect locked MTUs on routes. Tunnel provider
or IPSEC already have to recheck and if needed send PtB notifications
to the sending host in case the data does not fit into the packet with
added headers (we only know the final header sizes there, while also
using path MTU information).

The reason for this change is, that path MTU information can be injected
into the kernel via e.g. icmp_err protocol handler without verification
of local sockets. As such, this could cause the IPv6 forwarding path to
wrongfully emit Packet-too-Big errors and drop IPv6 packets.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: John Heffner <johnwheffner@gmail.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: introduce ip_dst_mtu_maybe_forward and protect forwarding path against pmtu spoofing

While forwarding we should not use the protocol path mtu to calculate
the mtu for a forwarded packet but instead use the interface mtu.

We mark forwarded skbs in ip_forward with IPSKB_FORWARDED, which was
introduced for multicast forwarding. But as it does not conflict with
our usage in unicast code path it is perfect for reuse.

I moved the functions ip_sk_accept_pmtu, ip_sk_use_pmtu and ip_skb_dst_mtu
along with the new ip_dst_mtu_maybe_forward to net/ip.h to fix circular
dependencies because of IPSKB_FORWARDED.

Because someone might have written a software which does probe
destinations manually and expects the kernel to honour those path mtus
I introduced a new per-namespace "ip_forward_use_pmtu" knob so someone
can disable this new behaviour. We also still use mtus which are locked on a
route for forwarding.

The reason for this change is, that path mtus information can be injected
into the kernel via e.g. icmp_err protocol handler without verification
of local sockets. As such, this could cause the IPv4 forwarding path to
wrongfully emit fragmentation needed notifications or start to fragment
packets along a path.

Tunnel and ipsec output paths clear IPCB again, thus IPSKB_FORWARDED
won't be set and further fragmentation logic will use the path mtu to
determine the fragmentation size. They also recheck packet size with
help of path mtu discovery and report appropriate errors.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: John Heffner <johnwheffner@gmail.com>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>