spider_net: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in
spider_net_release_tx_chain which can be called in hard irq and other
contexts.

dev_consume_skb_any was choosen as it preserves the current
dev_kfree_skb semantics (dev_kfree_skb is consume_skb) and
is because it is correct most of the time as most packets
will have been successfully transmitted not dropeed.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

tilepro: Call dev_consume_skb_any instead of kfree_skb.

Replace kfree_skb with dev_consume_skb_any in tile_net_tx and
tile_net_tx_tso which can be called in hard irq and other contexts.

At the point where the skbs are freed a packet has been successfully
transmitted so dev_consume_skb_any is the appropriate variant to use.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sungem: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in gem_tx which can be
called in hard irq and other contexts.  gem_tx handles successfully
transmitted packets.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

stmmac: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in stmmac_tx_clean that can
be called in hard irq and other contexts.  stmmac_tx_clean handles
freeing successfully transmitted packets.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

smsc911x: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in smsc911x_hard_xmit
which can be called in hard irq and other contexts. smsc911x_hard_xmit
always transmits and consumes the specified skb.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

smc91x: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in
smc_hardware_send_pkt that can be called in hard irq and other
contexts, and handles successfully transmitted packets.

Replace dev_kfree_skb with dev_kfree_skb_any in smc_hard_start_xmit which
can be called in hard irq and other contexts, and only frees skbs
when dropping them.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

smc911x: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in functions that can
be called in hard irq and other contexts.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sis900: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in functions that can
be called in hard irq and other contexts.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sc92031: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in sc92031_start_xmit
that can be called in hard irq and other contexts.

Using dev_consume_skb_any preserves the current semantics (as
dev_kfree_skb is just consume_skb) and since packet drops
are rare is usually accurate.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

forcedeth: Call dev_kfree_skb_any instead of kfree_skb.

Replace kfree_skb with dev_kfree_skb_any in functions that can
be called in hard irq and other contexts.

Every location changes is a drop making dev_kfree_skby_any appropriate.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

vxge: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in vxge_xmit that can
be called in hard irq and other contexts.

vxge_xmit only calls dev_kfree_skb_any when errors result in dropping
skbs.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

s2io: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in s2io_xmit that can
be called in hard irq and other contexts.

All instances that are changed are packet drops.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

ksz884x: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in copy_old_skb that can
be called in hard irq and other contexts.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sky2: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in sky2_xmit_frame that
can be called in hard irq and other contexts.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

skge: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any skge_xmit_free that can
be called in hard irq and other contexts, on the path that
handles dropped packets.

Replace dev_kfree_skb with dev_consume_skb_any in skge_tx_done that can
be called in hard irq and other contexts, on the path that handles
successfully transmitted skbs.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

mv643xx_eth: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in mv643xx_eth_xmit and
txq_submit_skb that can be called in hard irq and other contexts,
on paths where the skbs are dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

jme: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in jme_expand_header that
can be called in hard irq and other contexts, on the failure
path where the skb is dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

ibmveth: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in ibmveth_start_xmit
that can be called in hard irq and other contexts.

In this code path the packet can have either been transmitted
or dropped, dev_consume_skb_any was choosen because that preserves
the existing semantics of the code, and a transmitted packet is
more likely.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

ehea: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in functions that can
be called in hard irq and other contexts.

None of the locations was a packet drop so dev_kfree_skb_any is
inappropriate.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

i825xx: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in i596_start_xmit that
can be called in hard irq and other contexts, when the skb is dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

ucc_geth: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in ucc_geth_tx that can
be called in hard irq and other contexts, when processing the
tx completion event.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

fec: Call dev_kfree_skb_any instead of kfree_skb.

Replace kfree_skb with dev_kfree_skb_any in fec_enet_start_xmit that
can be called in hard irq and other contexts, when the packet is
dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sundance: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in start_tx that can
be called in hard irq and other contexts, when the skb is dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

uli526x: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in uli562x_start_xmit
that can be called in hard irq and other contexts, when the packet is
dropped.

Replace dev_kfree_skb with dev_consume_skb_any in uli562x_start_xmit
that can be called in hard irq and other contexts, when the packet is
transmitted.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

dmfe: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in dmfe_start_xmit that
can be called in hard irq and other contexts, when the packet is
dropped.

Replace dev_kfree_skb with dev_consume_skb_any in dmfe_start_xmit that
can be called in hard irq and other contexts, when the packet is
transmitted.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

dm9000: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in dm9000_start_xmit
that can be called in hard irq and other contexts, on the path
that successfully transmits the packet.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

enic: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in enic_hard_start_xmit
that can be called in hard irq and other contexts.

enic_hard_start_xmit only frees the skb when dropping it.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

cs89x0: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in net_send_packet that
can be called in hard irq and other contexts.

net_send_packet consumes (not drops) the skb of interest.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

cxfb4vf: Call dev_kfree/consume_skb_any instead of [dev_]kfree_skb.

Replace kfree_skb with dev_consume_skb_any in free_tx_desc that can be
called in hard irq and other contexts. dev_consume_skb_any is used
as this function consumes successfully transmitted skbs.

Replace dev_kfree_skb with dev_kfree_skb_any in t4vf_eth_xmit that can
be called in hard irq and other contexts, on paths that drop the skb.

Replace dev_kfree_skb with dev_consume_skb_any in t4vf_eth_xmit that can
be called in hard irq and other contexts, on paths that successfully
transmit the skb.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

cxgb4: Call dev_kfree/consume_skb_any instead of [dev_]kfree_skb.

Replace kfree_skb with dev_consume_skb_any in free_tx_desc that can be
called in hard irq and other contexts. dev_consume_skb_any is used
as this function consumes successfully transmitted skbs.

Replace dev_kfree_skb with dev_kfree_skb_any in t4_eth_xmit that can
be called in hard irq and other contexts, on paths that drop the skb.

Replace dev_kfree_skb with dev_consume_skb_any in t4_eth_xmit that can
be called in hard irq and other contexts, on paths that successfully
transmit the skb.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

cxgb3: Call dev_kfree/consume_skb_any instead of [dev_]kfree_skb.

Replace kfree_skb with dev_consume_skb_any in free_tx_desc, and
write_tx_pkt_wr that can be called in hard irq and other contexts.

Replace dev_kfree_skb with dev_kfree_skb_any in t3_eth_xmit that can
be called in hard irq and other contexts.

dev_kfree_skb is replaced with dev_kfree_skb_any in t3_eth_xmit as
that location is a packet drop, while kfree_skb in free_tx_desc,
and in write_tx_pkt_wr are places where packets are consumed
in a healthy manner.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

xgmac: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in xgmac_tx_complete
that can be called in hard irq and other contexts.

Replace dev_kfree_skb with dev_kfree_skb_any in xgmac_xmit that can
be called in hard irq and other contexts.

dev_consume_skb_any is used in xgamc_tx_complete as skbs that reach
there have been successfully transmitted, dev_kfree_skby_any is used
in xgmac_xmit as skbs that are freed there are being dropped.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

macb: Call dev_kfree_skb_any instead of kfree_skb.

Replace kfree_skb with dev_kfree_skb_any in macb_start_xmit that can
be called in hard irq and other contexts.

macb_start_xmit only frees skbs when dropping them so
dev_kfree_skb_any is used.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

bnad: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in bnad_start_xmit that
can be called in hard irq and other contexts.

dev_kfree_skb_any is used as bnad_start_xmit only frees skbs when to
drop them, normally transmitted packets are handled elsewhere.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

atl1c: Call dev_kfree/consume_skb_any instead of dev_kfree_skb.

The call path: atl1c_xmit_frame, atlc_tx_rollback, atl1c_clean_buffer
can not be tell at compile time if it will be invoked from hard irq
or other context, as atl1c_xmit_frame does not know.  So remove
the logic that  passes the compile time knowledge into al1c_clean_buffer
and figure out it out at runtime with dev_consume_skb_any.

Replace dev_kfree_skb with dev_kfree_skb_any in atl1c_xmit_frame that
can be called in hard irq and other contexts.

Replace dev_kfree_skb and dev_kfree_skb_irq with dev_consume_skb_any
in atl1c_clean_buffer that can be called in hard irq and other
contexts.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

alx: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in alx_start_xmit that
can be called in hard irq and other contexts.

dev_kfree_skb_any is used as alx_start_xmit only frees skbs
when dropping them.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

pcnet32: Call dev_kfree_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_kfree_skb_any in pcnet32_start_xmit
that can be called in hard irq and other contexts.

dev_kfree_skb_any is used as pcnet32_start_xmit only frees an
skb when it drops a packet during transmit.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

lance: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in lance_start_xmit
that can be called in hard irq and other contexts.

dev_consume_skb_any is used as lance_start_xmit always immediately
consumes the skb.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

am79c961a: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in am79c961 that can
be called in hard irq and other contexts.

dev_consume_skb_any is used as am79c961_sendpacket always
immediately consumes the skb.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

sun4i-emac: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in emacs_start_xmit
which can be called in hard irq and other contexts.

emac_start_xmit always transmits the packet making dev_consume_skb
the appropriate function to call.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

bfin_mac: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in _tx_reclaim_skb that
can be called in hard irq and other contexts.

dev_consume_skb is used as _tx_reclaim_skb is called after a packet
has been successfully transmitted.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

8390: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in __ei_start_xmit that
can be called in hard irq and other contexts.

dev_consume_skb is used as in this simple driver the skb is always
immediately consumed, there are no drops.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

3c59x: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in vortex_start_xmit
as it can be called in hard irq and other contexts.

dev_consume_skb_any is used when vortext_start_xmit directly consumes
the packet instead of dmaing it to the device.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

3c509: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in el3_start_xmit
as it can be called in hard irq and other contexts.

dev_consume_skb_any is used as on this simple hardware the
skb is consumed directly by the start_xmit function.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

uml/net_kern: Call dev_consume_skb_any instead of dev_kfree_skb.

Replace dev_kfree_skb with dev_consume_skb_any in uml_net_start_xmit
as it can be called in hard irq and other contexts.

dev_consume_skb_any is used as uml_net_start_xmit typically
consumes (not drops) packets.

Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>

Merge git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

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

This series contains updates to igb, ixgbe, ixgbevf, i40e and i40evf.

Anjali provides a i40e/i40evf patch to add Energy Efficient Ethernet
Low Power Idle stats and a fix for i40e to change the string
"Side Band" to "Sideband" for consistency.

Mitch provides 2 patches for i40evf to fix if the driver encounters
an error while communicating with the PF driver, do not shut down the
admin queue unconditionally.  Add an error message when the admin
queue message never completes and fix formatting on another message
that was unnecessarily wrapped.

Mark provides a ixgbe patch and five ixgbevf patches.  Fix a possible
infinite recursion when an adapter is removed and registers all read
as all one's in ixgbe_clear_vmdq_generic() and ixgbe_clear_rar_generic().
Converts macros to static inline functions to align kernel coding standard
and prepare for adding Live Error Recovery (LER) to ixgbevf.  Change the
ethtool register test to use the normal register accessor functions and
eliminate macors used for calling register test functions to make error
exits more clear.  Checks all register reads for adapter removal by checking
the status register after any register read that returns all F's since the
status register will never return 0xFFFFFFFF unless the adapter is removed.

Jacob implements SIOCGHWTSTAMP ioctl for igb which enables user processes
to read the current hardware stamp config settings non-destructively.

Todd adds the initial register read and write for surprise removal (LER)
for igb.

Christian Engelmayer fixes an igb memory leak in the igb_get_module_eeprom()
error handling path.

Ken Ichikawa provides a fix for igb, specifically for 82575 hardware to
specify -1 to the phc_index for ethtool's get_ts_info, otherwise a wrong
value will be set to the phc_index.

Christopher Paasch fixes a null pointer dereference in igb and makes sure
to unset the HAS_MSIX flag when the driver falls back to MSI only.
====================

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

ipv4: remove ip_rt_dump from route.c

ip_rt_dump do nothing after IPv4 route caches removal, so we can remove it.

Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

atheros/atlx: use SET_ETHTOOL_OPS directly

As commit a6e28b34205b("staging/et131x: use SET_ETHTOOL_OPS
directly"), using a wrapper around SET_ETHTOOL_OPS macro is
not actually required, remove and use SET_ETHTOOL_OPS directly.

Signed-off-by: Wang Weidong <wangweidong1@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Altera TSE: Correct two typos in original submission

This patch addresses two typos in the original driver submission. One derived
from a cut & paste error, and another is a misspelling.

Signed-off-by: Vince Bridgers <vbridgers2013@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Altera TSE: Correct typecast issue detected by kbuild test robot

This patch addresses a portable pointer arithmetic issue in the
original submission found by the kbuild test robot.

config: make ARCH=i386 allyesconfig

   altera_sgdma.c: In function 'sgdma_txphysaddr':
>> altera_sgdma.c:393:33: warning: cast from
>> pointer to integer of different size [-Wpointer-to-int-cast]
     dma_addr_t offs = (dma_addr_t)((dma_addr_t)desc -
                                    ^
>> altera_sgdma.c:394:5: warning: cast from
>> pointer to integer of different size [-Wpointer-to-int-cast]
        (dma_addr_t)priv->tx_dma_desc);
        ^
   altera_sgdma.c: In function 'sgdma_rxphysaddr':
>> altera_sgdma.c:403:33: warning: cast from
>> pointer to integer of different size [-Wpointer-to-int-cast]
     dma_addr_t offs = (dma_addr_t)((dma_addr_t)desc -
                                    ^
>> altera_sgdma.c:404:5: warning: cast from
>> pointer to integer of different size [-Wpointer-to-int-cast]
        (dma_addr_t)priv->rx_dma_desc);
        ^

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Vince Bridgers <vbridgers2013@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Altera TSE: Set version number by driver's get regs

Set the version number returned by the driver's get regs routine
invoked by ethtool so formatting can be dependent on the version
number returned, and any interesting formatted output can check
the version number for specific types of register data returned.

Signed-off-by: Vince Bridgers <vbridgers2013@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'qlcnic-next'

Shahed Shaikh says:

====================
This patch series containes following changes -

* TSO over IPv4 and IPv6, Tx checksum offload for VXLAN
* Rx checksum offload for VXLAN and support for .ndo_{add|del}_vxlan_port
  netdev ops.
====================

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

qlcnic: Update version to 5.3.57

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

qlcnic: Add VXLAN Rx offload support

This patch adds Rx checksum offload support for VXLAN.
Implements .ndo_{add|del}_vxlan_port netdev ops.
Adapter supports only one VXLAN port, so program adapter with
very first UDP port which VXLAN driver is listening to.

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

qlcnic: Add VXLAN Tx offload support

This patch adds LSO, LSO6 and Tx checksum offload support for VXLAN
encapsulated packets on 83xx/84xx series adapters.

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

gianfar: Fix P1010 config regression (SQ polling)

The P1010 device tree restricts the number of
supported interrupt groups to 1, although the eth
controller can support 2 interrupt groups and the
driver assumes the Multi-Group mode ("fsl,etsec2" model).

So, in this case the assumption that the Multi-Group
mode (MQ_MG_MODE) devices always support 2 interrupt
groups is false.  To fix this, a check for the actual
number of interrupt groups enabled in the board's
device tree has been added in gfar_probe for the
"fsl,etsec2" devices.

Without this fix, P1010 based boards claim support for
2 Tx queues to the net stack but only one is actually
allocated, leading to NULL access in xmit.  This issue
was introduced by enabling Single-Queue polling for
the P1010 devices.
(71ff9e3 gianfar: Use Single-Queue polling for
"fsl,etsec2")

Fixes: 71ff9e3df7e1c5d3293af6b595309124e8c97412
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: remove ipv4_ifdown_dst from route.c

ipv4_ifdown_dst does nothing after IPv4 route caches removal,
so we can remove it.

Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: optimize csum_replace2()

When changing one 16bit value by another in IP header, we can adjust
the IP checksum by doing a simple operation described in RFC 1624, as
reminded by David.

csum_partial() is a complex function on x86_64, not really suited for
small number of checksummed bytes.

I spotted csum_partial() being in the top 20 most consuming functions
(more than 1 %) in a GRO workload, which was rather unexpected.

The caller was inet_gro_complete() doing a csum_replace2() when
building the new IP header for the GRO packet.

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

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

Antonio Quartulli says:

====================
Included changes:
- use ether_addr_copy instead of memcpy when possible
- implement new multicast packet optimisation
- improve several kerneldoc sections
- minor code cleanups

here you have our patchset for net-next/linux-3.15.  They are 16
patches but most of them are just small cleanups and kerneldoc
improvements.

The only big change is the one from patch 8 to 13 by Linus Lüssing
that introduces a new multicast packets optimisation. This new
component aims to reduce the air overhead by sending multicast packets
as bat-unicast when only one destination exists or by dropping them
directly at the source if the multicast group is totally empty.

In patch 11 Linus introduces an atomic_t variable, that like others
that we already have is only object of write and read, thus making the
atomic characteristic totally useless.  Unfortunately this is part of
our sysfs framework, that helps the developer to introduce new knobs
by using few macros only. For this reason we decided to keep Linus'
new knob for now, but I'd like to let you know that we are in the
process of re-working such framework in order to convert all the
current (useless) atomic_t to boolean in one go.
====================

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

batman-adv: Start new development cycle

Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: improve DAT documentation

Add missing documentation for BATADV_DAT_ADDR_MAX and
convert an existing documentation to kerneldoc

Signed-off-by: Antonio Quartulli <antonio@open-mesh.com>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>

batman-adv: improve the TT flags documentation

Convert the current documentation for the TT flags in proper
kerneldoc and improve it by adding an explanation for each
of the flags.

Signed-off-by: Antonio Quartulli <antonio@open-mesh.com>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>

batman-adv: Send multicast packets to nodes with a WANT_ALL flag

With this patch a node sends IPv4 multicast packets to nodes which
have a BATADV_MCAST_WANT_ALL_IPV4 flag set and IPv6 multicast packets
to nodes which have a BATADV_MCAST_WANT_ALL_IPV6 flag set, too.

Why is this needed? There are scenarios involving bridges where
multicast report snooping and multicast TT announcements are not
sufficient, which would lead to packet loss for some nodes otherwise:

MLDv1 and IGMPv1/IGMPv2 have a suppression mechanism
for multicast listener reports. When we have an MLDv1/IGMPv1/IGMPv2
querier behind a bridge then our snooping bridge is potentially not
going to see any reports even though listeners exist because according
to RFC4541 such reports are only forwarded to multicast routers:

-----------------------------------------------------------
            ---------------
{Querier}---|Snoop. Switch|----{Listener}
            ---------------
                       \           ^
                      -------
                      | br0 |  <  ???
                      -------
                          \
                     _-~---~_
                 _-~/        ~-_
                ~   batman-adv  \-----{Sender}
                \~_   cloud    ~/
                   -~~__-__-~_/

I)  MLDv1 Query:  {Querier}  -> flooded
II) MLDv1 Report: {Listener} -> {Querier}

-> br0 cannot detect the {Listener}
=> Packets from {Sender} need to be forwarded to all
   detected listeners and MLDv1/IGMPv1/IGMPv2 queriers.

-----------------------------------------------------------

Note that we do not need to explicitly forward to MLDv2/IGMPv3 queriers,
because these protocols have no report suppression: A bridge has no
trouble detecting MLDv2/IGMPv3 listeners.

Even though we do not support bridges yet we need to provide the
according infrastructure already to not break compatibility later.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: Add IPv4 link-local/IPv6-ll-all-nodes multicast support

With this patch a node may additionally perform the dropping or
unicasting behaviour for a link-local IPv4 and link-local-all-nodes
IPv6 multicast packet, too.

The extra counter and BATADV_MCAST_WANT_ALL_UNSNOOPABLES flag is needed
because with a future bridge snooping support integration a node with a
bridge on top of its soft interface is not able to reliably detect its
multicast listeners for IPv4 link-local and the IPv6
link-local-all-nodes addresses anymore (see RFC4541, section 2.1.2.2
and section 3).

Even though this new flag does make "no difference" now, it'll ensure
a seamless integration of multicast bridge support without needing to
break compatibility later.

Also note, that even with multicast bridge support it won't be possible
to optimize 224.0.0.x and ff02::1 towards nodes with bridges, they will
always receive these ranges.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: Modified forwarding behaviour for multicast packets

With this patch a multicast packet is not always simply flooded anymore,
the behaviour for the following cases is changed to reduce
unnecessary overhead:

If all nodes within the horizon of a certain node have signalized
multicast listener announcement capability then an IPv6 multicast packet
with a destination of IPv6 link-local scope (excluding ff02::1) coming
from the upstream of this node...

* ...is dropped if there is no according multicast listener in the
  translation table,
* ...is forwarded via unicast if there is a single node with interested
  multicast listeners
* ...and otherwise still gets flooded.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: Announce new capability via multicast TVLV

If the soft interface of a node is not part of a bridge then a node
announces a new multicast TVLV: The existence of this TVLV
signalizes that this node is announcing all of its multicast listeners
via the translation table infrastructure.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: introduce capability initialization bitfield

The new bitfield allows us to keep track whether capability subsets of
an originator have gone through their initialization phase yet.

The translation table is the only user right now, but a new one will be
added soon.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: Multicast Listener Announcements via Translation Table

With this patch a node which has no bridge interface on top of its soft
interface announces its local multicast listeners via the translation
table.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: add kerneldoc for dst_hint argument

Some helper functions used along the TX path have now a new
"dst_hint" argument but the kerneldoc was missing.

Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>

batman-adv: call unregister_netdev() to have it handle the locking for us

Reported-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: fix a few kerneldoc inconsistencies

Reported-by: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: prefer ether_addr_copy to memcpy

On some architectures ether_addr_copy() is slightly faster
than memcpy() therefore use the former when possible.

Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>

batman-adv: remove obsolete skb_reset_mac_header() in batadv_bla_tx()

Our .ndo_start_xmit handler (batadv_interface_tx()) can rely on having
the skb mac header pointer set correctly since the following commit
present in kernels >= 3.9:

"net: reset mac header in dev_start_xmit()" (6d1ccff627)

Therefore this commit removes the according, now redundant,
skb_reset_mac_header() call in batadv_bla_tx().

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: use vlan_/eth_hdr() instead of skb->data in interface_tx path

Our .ndo_start_xmit handler (batadv_interface_tx()) can rely on having
the skb mac header pointer set correctly since the following commit
present in kernels >= 3.9:

"net: reset mac header in dev_start_xmit()" (6d1ccff627)

Therefore we can safely use eth_hdr() and vlan_eth_hdr() instead of
skb->data now, which spares us some ugly type casts.

At the same time set the mac_header in batadv_dat_snoop_incoming_arp_request()
before sending the skb along the TX path.

Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

batman-adv: fix coccinelle warnings

net/batman-adv/network-coding.c:1535:1-7: Replace memcpy with struct assignment

Generated by: coccinelle/misc/memcpy-assign.cocci
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Martin Hundebøll <martin@hundeboll.net>
Signed-off-by: Marek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: Antonio Quartulli <antonio@meshcoding.com>

Merge branch 'ptp-next'

Richard Cochran says:

====================
ptp: dynamic pin control

This patch series introduces a way of changing the auxiliary PTP
Hardware Clock functions (periodic output signals and time stamping
external signals) at run time. In the past on the netdev list, we have
discussed other ways to handle this, such as module parameters and
ethtool. This series implements a new PHC ioctl because that is the
most natural way. Users already activate the auxiliary functions via
the ioctls. The sysfs interface has also been expanded so that the pin
configuration can be programmed using shell scripts.

The first patch adds the new ioctls. The PHC subsystem does most of
the work of maintaining the function-to-pin mapping. Drivers will only
need to allocate and initialize a pin configuration table and also
provide a new method that validates a particular assignment.

Patches 5 and 6 just clean up a couple of issues in the phyter driver,
and the remaining patches actually hook the phyter's pins into the new
system.

* ChangeLog
** V3
   - simplify locking in the set pin logic
** V2
   - fix bug in sysfs code on init error path
   - rename ptp_setpin() to ptp_set_pinfunc()
   - rename .setpin() to .verify() in the driver interface
   - simplify ptp_find_pin() logic
   - use correct test when checking whether the pin with the
     calibration function is being reprogrammed
====================

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

dp83640: let the periodic pin from the module parameter be a default.

This patch changes the driver use the new pin configuration method when
programming the periodic output signal.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dp83640: let external input pins from the module parameters be defaults.

This patch changes the driver to use the new pin configuration method when
programming the external time stamp input signals.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dp83640: implement programmable pin functions.

This patch adapts the dp83640 driver to allow reconfiguration of which
auxiliary function goes on which pin. The functions may be reassigned
freely with the one exception of the calibration function.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dp83640: correct the periodic output frequency

The phyter driver incorrectly feeds the value of the period into what
is in fact a pulse width register, resulting in the actual period
being twice the dialed value. This patch fixes the issue and renames a
variable to make the code at bit more clear.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

dp83640: trivial fixes

This patch cleans up the input checking code on the external time stamp
function by using an unsigned rather than a signed channel index.

Also, this patch corrects the author's email address. When this macro
was last changed, the top level domain part of the email address was
left behind.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: drivers: set the number of programmable pins.

This patch updates the many PTP Hardware Clock drivers with the
newly introduced field that advertises the number of programmable
pins. Some of these devices do have programmable pins, but the
implementation will have to wait for follow on patches.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: expose the programmable pins via sysfs

This patch adds the sysfs hooks needed in order to get and set the
programmable pin settings.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: add the pin GET/SETFUNC ioctls to the testptp program.

This patch adds a option to the test program that lists the
programmable pins on a PTP Hardware Clock device, assuming there
are any such pins. A second option lets the user reprogram the
auxiliary function of a single pin.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ptp: introduce programmable pins.

This patch adds a pair of new ioctls to the PTP Hardware Clock device
interface. Using the ioctls, user space programs can query each pin to
find out its current function and also reprogram a different function
if desired.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

igb: Unset IGB_FLAG_HAS_MSIX-flag when falling back to msi-only

Prior to cd14ef54d25 (igb: Change to use statically allocated array for
MSIx entries), having msix_entries different from NULL was an indicator
that MSIX is enabled.
In igb_set_interrupt_capabiliy we may fall back to MSI-only. Prior to
the above patch msix_entries was set to NULL by
igb_reset_interrupt_capability.

However, now we are checking the flag for IGB_FLAG_HAS_MSIX and so the
stack gets completly confused:

[   42.659791] ------------[ cut here ]------------
[   42.715032] WARNING: CPU: 7 PID: 0 at net/sched/sch_generic.c:264 dev_watchdog+0x15c/0x1fb()
[   42.848263] NETDEV WATCHDOG: eth0 (igb): transmit queue 0 timed out
[   42.923253] Modules linked in:
[   42.959875] CPU: 7 PID: 0 Comm: swapper/7 Not tainted 3.14.0-rc2-mptcp #437
[   43.043184] Hardware name: HP ProLiant DL165 G7, BIOS O37 01/26/2011
[   43.119215]  0000000000000108 ffff88023fdc3da8 ffffffff81487847 0000000000000108
[   43.208165]  ffff88023fdc3df8 ffff88023fdc3de8 ffffffff81034e7d ffff88023fdc3dd8
[   43.297120]  ffffffff813fff10 ffff880236018000 ffff880236b178c0 0000000000000008
[   43.386071] Call Trace:
[   43.415303]  <IRQ>  [<ffffffff81487847>] dump_stack+0x49/0x62
[   43.484174]  [<ffffffff81034e7d>] warn_slowpath_common+0x77/0x91
[   43.556049]  [<ffffffff813fff10>] ? dev_watchdog+0x15c/0x1fb
[   43.623759]  [<ffffffff81034f2b>] warn_slowpath_fmt+0x41/0x43
[   43.692511]  [<ffffffff813fff10>] dev_watchdog+0x15c/0x1fb
[   43.758141]  [<ffffffff813ffdb4>] ? __netdev_watchdog_up+0x64/0x64
[   43.832091]  [<ffffffff8103cd04>] call_timer_fn+0x17/0x6f
[   43.896682]  [<ffffffff8103cebe>] run_timer_softirq+0x162/0x1a2
[   43.967511]  [<ffffffff81038520>] __do_softirq+0xcd/0x1cc
[   44.032104]  [<ffffffff81038689>] irq_exit+0x3a/0x48
[   44.091492]  [<ffffffff81026d43>] smp_apic_timer_interrupt+0x43/0x50
[   44.167525]  [<ffffffff8148c24a>] apic_timer_interrupt+0x6a/0x70
[   44.239392]  <EOI>  [<ffffffff8100992c>] ? default_idle+0x6/0x8
[   44.310343]  [<ffffffff81009b31>] arch_cpu_idle+0x13/0x18
[   44.374934]  [<ffffffff81066126>] cpu_startup_entry+0xa7/0x101
[   44.444724]  [<ffffffff81025660>] start_secondary+0x1b2/0x1b7
[   44.513472] ---[ end trace a5a075fd4e7f854f ]---
[   44.568753] igb 0000:04:00.0 eth0: Reset adapter
[   46.206945] random: nonblocking pool is initialized
[   46.465670] irq 44: nobody cared (try booting with the "irqpoll" option)
[   46.545862] CPU: 7 PID: 0 Comm: swapper/7 Tainted: G        W    3.14.0-rc2-mptcp #437
[   46.640610] Hardware name: HP ProLiant DL165 G7, BIOS O37 01/26/2011
[   46.716641]  ffff8802363f8c84 ffff88023fdc3e38 ffffffff81487847 00000000a03cdb6d
[   46.805598]  ffff8802363f8c00 ffff88023fdc3e68 ffffffff81068489 0000007f81825400
[   46.894539]  ffff8802363f8c00 0000000000000000 0000000000000000 ffff88023fdc3ea8
[   46.983484] Call Trace:
[   47.012714]  <IRQ>  [<ffffffff81487847>] dump_stack+0x49/0x62
[   47.081585]  [<ffffffff81068489>] __report_bad_irq+0x35/0xc1
[   47.149295]  [<ffffffff81068683>] note_interrupt+0x16e/0x1ea
[   47.217006]  [<ffffffff8106679e>] handle_irq_event_percpu+0x116/0x12e
[   47.294075]  [<ffffffff810667e9>] handle_irq_event+0x33/0x4f
[   47.361787]  [<ffffffff81068c95>] handle_fasteoi_irq+0x83/0xd1
[   47.431577]  [<ffffffff81003d5b>] handle_irq+0x1f/0x28
[   47.493047]  [<ffffffff81003567>] do_IRQ+0x4e/0xd4
[   47.550358]  [<ffffffff8148b06a>] common_interrupt+0x6a/0x6a
[   47.618066]  <EOI>  [<ffffffff8100992c>] ? default_idle+0x6/0x8
[   47.689016]  [<ffffffff81009b31>] arch_cpu_idle+0x13/0x18
[   47.753605]  [<ffffffff81066126>] cpu_startup_entry+0xa7/0x101
[   47.823397]  [<ffffffff81025660>] start_secondary+0x1b2/0x1b7
[   47.892146] handlers:
[   47.919301] [<ffffffff812fbd7d>] igb_intr

So, this patch unsets the flag to indicate that we are not using MSIX.
This patch does exactly this: Unsetting the flag when falling back to MSI.

Fixes: cd14ef54d25b (igb: Change to use statically allocated array for MSIx entries)
Cc: Carolyn Wyborny <carolyn.wyborny@intel.com>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: Fix Null-pointer dereference in igb_reset_q_vector

When igb_set_interrupt_capability() calls
igb_reset_interrupt_capability() (e.g., because CONFIG_PCI_MSI is unset),
num_q_vectors has been set but no vector has yet been allocated.

igb_reset_interrupt_capability() will then call igb_reset_q_vector,
which assumes that the vector is allocated. As this is not the case, we
are accessing a NULL-pointer.

This patch fixes it by checking that q_vector is indeed different from
NULL.

Fixes: 02ef6e1d0b0023 (igb: Fix queue allocation method to accommodate changing during runtime)
Cc: Carolyn Wyborny <carolyn.wyborny@intel.com>
Signed-off-by: Christoph Paasch <christoph.paasch@uclouvain.be>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: specify phc_index of 82575 for get_ts_info

82575 has only software timestamping capability and it has
no PTP Hardware Clocks. Therefore, -1 has to be specified
to the phc_index for ethtool's get_ts_info, otherwise a wrong
value will be set to the phc_index.

v2: move the if (adapter->ptp_clock) section specifying phc_index
    to above the switch statement as suggested by Matthew Vick.
    adapter->ptpclock will always be NULL for 82575.

Signed-off-by: Ken ICHIKAWA <ichikawa.ken@jp.fujitsu.com>
Acked-by: Matthew Vick <matthew.vick@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: Fix memory leak in igb_get_module_eeprom()

Fix a memory leak in the igb_get_module_eeprom() error handling path.
Detected by Coverity: CID 1016508.

Signed-off-by: Christian Engelmayer <cengelma@gmx.at>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: add register rd/wr for surprise removal

Add initial register rd/wr for surprise removal (LER).

Signed-off-by: Todd Fujinaka <todd.fujinaka@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

igb: implement SIOCGHWTSTAMP ioctl

This patch adds support for the SIOCGHWTSTAMP ioctl which enables user
processes to read the current hwtstamp_config settings
non-destructively. Previously a process had to be privileged and could
only set values, it couldn't return what is currently set without
possibly overwriting the value.

This patch adds support for this new operation into igb by keeping a
shadow copy of the config in the adapter structure, which is returned
upon request.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Matthew Vick <matthew.vick@intel.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbevf: Additional adapter removal checks

Additional checks are needed for a detected removal not to cause
problems. Some involve simply avoiding a lot of stuff that can't
do anything good, and also cases where the phony return value can
cause problems. In addition, down the adapter when the removal is
sensed.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbevf: Check for adapter removal on register writes

Prevent writes to an adapter that has been detected as removed
by a previous failing read.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbevf: Check register reads for adapter removal

Check all register reads for adapter removal by checking the status
register after any register read that returns 0xFFFFFFFF. Since the
status register will never return 0xFFFFFFFF unless the adapter is
removed, such a value from a status register read confirms the
removal. Since this patch adds so much to ixgbe_read_reg, stop
inlining it, to reduce driver bloat.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbevf: Make the 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.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbevf: 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.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: Break recursion in case of removal

When an adapter is removed and registers all read as all one's,
an infinite recursion can happen between ixgbe_clear_vmdq_generic
and ixgbe_clear_rar_generic. Adding a check for removal breaks
this recursion.

Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf: clean up init error messages

Add an error message when the admin queue message never completes, and
fix formatting on another one that was unnecessarily wrapped.

Change-ID: I8b8a4eb7629d741f09357250144023cd4a72231f
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40evf: don't shut down admin queue on error

If the driver encounters an error while communicating with the PF
driver,  don't just shut down the admin queue unconditionally. The PF
may be delayed, and shutting down the admin queue causes it to fail
completely. If this happens, the VF will never complete initialization.

Change-ID: I6192e9d8caeefb738428c3597fa2f54fa400ce7f
Signed-off-by: Mitch Williams <mitch.a.williams@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

i40e: Fix a message string

Change string from "Side Band" to "Sideband" for consistency.

Change-ID: I45f05466bb5c63b0f999d743312bcb61b5bd6518
Signed-off-by: Anjali Singhai Jain <anjali.singhai@intel.com>
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Tested-by: Kavindya Deegala <kavindya.s.deegala@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>