l2tp: Fix inet_opt conversion.

We don't actually hold the socket lock at this point, so the
rcu_dereference_protected() isn't' correct.  Thanks to Eric
Dumazet for pointing this out.

Thankfully, we're only interested in fetching the faddr value
if srr is enabled, so we can simply make this an RCU sequence
and use plain rcu_dereference().

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

net: ibmveth: force reconfiguring checksum settings on startup

Commit b9367bf3ee6d ("net: ibmveth: convert to hw_features") accidentally
removed call to ibmveth_set_csum_offload() in ibmveth_probe(). Put the
call back where it was, but with additional error checking provided
by ibmveth_set_features().

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
[sfr: dev -> netdev]
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix memory leak in qlcnic_blink_led.

o Memory allocated in ETHTOOL_ACTIVE mode, is not getting freed. So,
  in ETHTOOL_ID_INACTIVE mode, return after freeing allocated memory.
o Using set bit instead of blink_down field, as it is also required
  in internal Loopback test and etc.

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

tg3: Convert u32 flag,flg2,flg3 uses to bitmap

Using a bitmap instead of separate u32 flags allows a consistent, simpler
and more extensible mechanism to determine capabilities.

Convert bitmasks to enum.
Add tg3_flag, tg3_flag_clear and tg3_flag_set.
Convert the flag & bitmask tests.

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

net: allow user to change NETIF_F_HIGHDMA

NETIF_F_HIGHDMA is like any other TX offloads, so allow user to toggle it.
This is needed later for bridge and bonding convertsion to hw_features.

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

bridge: convert br_features_recompute() to ndo_fix_features

Note: netdev_update_features() needs only rtnl_lock as br->port_list
is only changed while holding it.

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

net: fix netdev_increment_features()

Simplify and fix netdev_increment_features() to conform to what is
stated in netdevice.h comments about NETIF_F_ONE_FOR_ALL.
Include FCoE segmentation and VLAN-challedged flags in computation.

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

net:use help function of skb_checksum_start_offset to calculate offset

Although these are equivalent, but the skb_checksum_start_offset() is more readable.

Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

inet: add RCU protection to inet->opt

We lack proper synchronization to manipulate inet->opt ip_options

Problem is ip_make_skb() calls ip_setup_cork() and
ip_setup_cork() possibly makes a copy of ipc->opt (struct ip_options),
without any protection against another thread manipulating inet->opt.

Another thread can change inet->opt pointer and free old one under us.

Use RCU to protect inet->opt (changed to inet->inet_opt).

Instead of handling atomic refcounts, just copy ip_options when
necessary, to avoid cache line dirtying.

We cant insert an rcu_head in struct ip_options since its included in
skb->cb[], so this patch is large because I had to introduce a new
ip_options_rcu structure.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: filter: Just In Time compiler for x86-64

In order to speedup packet filtering, here is an implementation of a
JIT compiler for x86_64

It is disabled by default, and must be enabled by the admin.

echo 1 >/proc/sys/net/core/bpf_jit_enable

It uses module_alloc() and module_free() to get memory in the 2GB text
kernel range since we call helpers functions from the generated code.

EAX : BPF A accumulator
EBX : BPF X accumulator
RDI : pointer to skb   (first argument given to JIT function)
RBP : frame pointer (even if CONFIG_FRAME_POINTER=n)
r9d : skb->len - skb->data_len (headlen)
r8  : skb->data

To get a trace of generated code, use :

echo 2 >/proc/sys/net/core/bpf_jit_enable

Example of generated code :

# tcpdump -p -n -s 0 -i eth1 host 192.168.20.0/24

flen=18 proglen=147 pass=3 image=ffffffffa00b5000
JIT code: ffffffffa00b5000: 55 48 89 e5 48 83 ec 60 48 89 5d f8 44 8b 4f 60
JIT code: ffffffffa00b5010: 44 2b 4f 64 4c 8b 87 b8 00 00 00 be 0c 00 00 00
JIT code: ffffffffa00b5020: e8 24 7b f7 e0 3d 00 08 00 00 75 28 be 1a 00 00
JIT code: ffffffffa00b5030: 00 e8 fe 7a f7 e0 24 00 3d 00 14 a8 c0 74 49 be
JIT code: ffffffffa00b5040: 1e 00 00 00 e8 eb 7a f7 e0 24 00 3d 00 14 a8 c0
JIT code: ffffffffa00b5050: 74 36 eb 3b 3d 06 08 00 00 74 07 3d 35 80 00 00
JIT code: ffffffffa00b5060: 75 2d be 1c 00 00 00 e8 c8 7a f7 e0 24 00 3d 00
JIT code: ffffffffa00b5070: 14 a8 c0 74 13 be 26 00 00 00 e8 b5 7a f7 e0 24
JIT code: ffffffffa00b5080: 00 3d 00 14 a8 c0 75 07 b8 ff ff 00 00 eb 02 31
JIT code: ffffffffa00b5090: c0 c9 c3

BPF program is 144 bytes long, so native program is almost same size ;)

(000) ldh      [12]
(001) jeq      #0x800           jt 2    jf 8
(002) ld       [26]
(003) and      #0xffffff00
(004) jeq      #0xc0a81400      jt 16   jf 5
(005) ld       [30]
(006) and      #0xffffff00
(007) jeq      #0xc0a81400      jt 16   jf 17
(008) jeq      #0x806           jt 10   jf 9
(009) jeq      #0x8035          jt 10   jf 17
(010) ld       [28]
(011) and      #0xffffff00
(012) jeq      #0xc0a81400      jt 16   jf 13
(013) ld       [38]
(014) and      #0xffffff00
(015) jeq      #0xc0a81400      jt 16   jf 17
(016) ret      #65535
(017) ret      #0

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Cc: Hagen Paul Pfeifer <hagen@jauu.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: Kill RTO_CONN.

It's not used by anything in the kernel, and defined in net/route.h so
never exported to userspace.

Therefore we can safely remove it.

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

ipv4: Remove erroneous check in igmpv3_newpack() and igmp_send_report().

Output route resolution never returns a route with rt_src set to zero
(which is INADDR_ANY).

Even if the flow key for the output route lookup specifies INADDR_ANY
for the source address, the output route resolution chooses a real
source address to use in the final route.

This test has existed forever in igmp_send_report() and David Stevens
simply copied over the erroneous test when implementing support for
IGMPv3.

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

ipv4: Sanitize and simplify ip_route_{connect,newports}()

These functions are used together as a unit for route resolution
during connect().  They address the chicken-and-egg problem that
exists when ports need to be allocated during connect() processing,
yet such port allocations require addressing information from the
routing code.

It's currently more heavy handed than it needs to be, and in
particular we allocate and initialize a flow object twice.

Let the callers provide the on-stack flow object.  That way we only
need to initialize it once in the ip_route_connect() call.

Later, if ip_route_newports() needs to do anything, it re-uses that
flow object as-is except for the ports which it updates before the
route re-lookup.

Also, describe why this set of facilities are needed and how it works
in a big comment.

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

r8169: fix merge conflict fix.

- use adequate MAC_VER id
  (see 01dc7fec4025f6bb72b6b98ec88b375346b6dbbb)
- remove duplicate rtl_firmware_info record
- remove duplicate functions

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: clean up route lookup calls

Change the call to take the transport parameter and set the
cached 'dst' appropriately inside the get_dst() function calls.

This will allow us in the future  to clean up source address
storage as well.

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

sctp: remove useless arguments from get_saddr() call

There is no point in passing a destination address to
a get_saddr() call.

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

sctp: make sctp over IPv6 work with IPsec

SCTP never called xfrm_output after it's v6 route lookups so
that never really worked with ipsec.  Additioanlly, we never
passed port nubmers and protocol in the flowi, so any port
based policies were never applied as well.  Now that we can
fixed ipv6 routing lookup code, using ip6_dst_lookup_flow()
and pass port numbers.

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

sctp: cache the ipv6 source after route lookup

The ipv6 routing lookup does give us a source address,
but instead of filling it into the dst, it's stored in
the flowi.  We can use that instead of going through the
entire source address selection again.
Also the useless ->dst_saddr member of sctp_pf is removed.
And sctp_v6_dst_saddr() is removed, instead by introduce
sctp_v6_to_addr(), which can be reused to cleanup some dup
code.

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

sctp: fix sctp to work with ipv6 source address routing

In the below test case, using the source address routing,
sctp can not work.
Node-A
1)ifconfig eth0 inet6 add 2001:1::1/64
2)ip -6 rule add from 2001:1::1 table 100 pref 100
3)ip -6 route add 2001:2::1 dev eth0 table 100
4)sctp_darn -H 2001:1::1 -P 250 -l &
Node-B
1)ifconfig eth0 inet6 add 2001:2::1/64
2)ip -6 rule add from 2001:2::1 table 100 pref 100
3)ip -6 route add 2001:1::1 dev eth0 table 100
4)sctp_darn -H 2001:2::1 -P 250 -h 2001:1::1 -p 250 -s

root cause:
Node-A and Node-B use the source address routing, and
at begining, source address will be NULL,sctp will
search the  routing table by the destination address,
because using the source address routing table, and
the result dst_entry will be NULL.

solution:
walk through the bind address list to get the source
address and then lookup the routing table again to get
the correct dst_entry.

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

ixgbe: convert to ethtool set_phys_id

Based on the original patch submitted by Stephen Hemminger.

This patch makes the following changes:
- Change ETHTOOL_ID_INACTIVE return value to 2 (blinks/sec)
- Fix restoring of IXGBE_LEDCTL

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

ixgbe: fix X540 ethtool loopback test.

On X540 we need to set the MACC.FLU bit to 1 in order to force the link
up before entering MAC loopback.  This is only used in the ethtool loopback
test, which was failing.  This patch corrects it.

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

ixgbe: Bump version

Bump the driver version number to better match up with the out of tree
driver that has similar functionality.

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

ixgbe: enable SCTP checksum offload for X540

X540 supports SCTP checksum offload so enable it.  It was overlooked when X540
support was initially added to the driver.

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

ixgbe: do not clear FCoE DDP error status for received ABTS

The ddp->err is initialized to be 1 to make sure outstanding DDP context is
guaranteed to be invalidated when HW is not auto-invalidating it. However,
in case of receiving ABTS response for a DDPed I/O, the ddp->err was cleared,
bypassing the invalidating of the DDP context from upper protocol stack when
ixgbe_fcoe_ddp_put() is called. This bug is fixed here by updating the error
only when FCP_RSP is received.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: fix static functions

Define functions as static added C=1 (sparse) to my make line
brought these to my attention.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: remove ntuple display support

This change removes the ntuple display support from ixgbe.  The reason for
this change is to resolve a number of issues in the way display filtering
is handled.

I plan to add support for displaying these filters via the network flow
classifier interface.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Evan Swanson <evan.swanson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

ixgbe: add LED blink code for x540

Implement blink_led_start and blink_led_stop functions for x540
using the MACC register.

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

ixgbe: add support for 64k EEPROM for 82599

82599 supports up to 32k EEPROM addressing via EERD register. If we
wish to address larger EEPROM this have to be done via serial interface.
This patch adds function ixgbe_read_eeprom_82599 which selects the best
method to read the EEPROM.

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

ixgbe: register defines cleanup

Remove duplicates.
Fix incorrect defines.
Fix/Update comments.
Fix whitespace.
Add new register defines.

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

ixgbe: explicitly disable 100H for x540

100H is not supported on this HW, but the bit is set on the PHY.
This can result in link at 100F when advertising only 1000F.

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

e1000e: implement ethtool set_phys_id

Based on a patch from Stephen Hemminger <shemminger@vyatta.com>.

The new ethtool set_phys_id takes over controlling the LED for
identifying boards.  This fixes the lockout during that period.
For this device lots of extra infrastructure can also be removed by
using set_phys_id.

v2: - return blink frequency for parts that do not support blink in h/w
    - add blink_led function pointers for devices that do support blink
      in h/w to cleanup the test for this functionality

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

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

Resolved logic conflicts causing a build failure due to
drivers/net/r8169.c changes using a patch from Stephen Rothwell.

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

tg3: Update version to 3.118

This patch updates the tg3 version to 3.118.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Whitespace cleanups

This patch gets rid of some harmless whitespace errors.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add EEH support

This patch adds EEH support to the tg3 driver.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add TSO loopback test

This patch adds code to exercise the TSO portion of the device through
a phy loopback test.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Organize loopback test failure flags

As more test modes are added to each loopback mode, the need to
organise the results increases.  This patch groups the results by
loopback mode, and then by test mode.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Fix int generation hw bug for 5719 / 5720

On the 5719 and 5720, there is a bug where the hardware will
misinterpret a status tag update and leave interrupts permanently
disabled.  This patch enables a hardware fix that works around the
issue.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bluetooth: Fix use-before-initiailized var.

net/bluetooth/l2cap_core.c: In function ‘l2cap_recv_frame’:
net/bluetooth/l2cap_core.c:3612:15: warning: ‘sk’ may be used uninitialized in this function
net/bluetooth/l2cap_core.c:3612:15: note: ‘sk’ was declared here

Actually the problem is in the inline function l2cap_data_channel(), we
branch to the label 'done' which tests 'sk' before we set it to anything.

Initialize it to NULL to fix this.

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

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

be2net: Fixed a bug in be_cmd_get_regs().

Same WRB entry was being reused over different iterations of a
loop while issuing non-embedded IOCTL requests.Fixed couple of minor bugs
in this path as well.
Re-factored code to alloc/free memory for DMA outside of loop

Signed-off-by: Somnath Kotur <somnath.kotur@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: move processing of recv handlers into handle_frame()

Since now when bonding uses rx_handler, all traffic going into bond
device goes thru bond_handle_frame. So there's no need to go back into
bonding code later via ptype handlers. This patch converts
original ptype handlers into "bonding receive probes". These functions
are called from bond_handle_frame and they are registered per-mode.

Note that vlan packets are also handled because they are always untagged
thanks to vlan_untag()

Note that this also allows arpmon for eth-bond-bridge-vlan topology.

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

net: make WARN_ON in dev_disable_lro() useful

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: provide cow_metrics() methods to blackhole dst_ops

Since commit 62fa8a846d7d (net: Implement read-only protection and COW'ing
of metrics.) the kernel throws an oops.

[  101.620985] BUG: unable to handle kernel NULL pointer dereference at
           (null)
[  101.621050] IP: [<          (null)>]           (null)
[  101.621084] PGD 6e53c067 PUD 3dd6a067 PMD 0
[  101.621122] Oops: 0010 [#1] SMP
[  101.621153] last sysfs file: /sys/devices/virtual/ppp/ppp/uevent
[  101.621192] CPU 2
[  101.621206] Modules linked in: l2tp_ppp pppox ppp_generic slhc
l2tp_netlink l2tp_core deflate zlib_deflate twofish_x86_64
twofish_common des_generic cbc ecb sha1_generic hmac af_key
iptable_filter snd_pcm_oss snd_mixer_oss snd_seq snd_seq_device loop
snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_intel snd_hda_codec
snd_pcm snd_timer snd i2c_i801 iTCO_wdt psmouse soundcore snd_page_alloc
evdev uhci_hcd ehci_hcd thermal
[  101.621552]
[  101.621567] Pid: 5129, comm: openl2tpd Not tainted 2.6.39-rc4-Quad #3
Gigabyte Technology Co., Ltd. G33-DS3R/G33-DS3R
[  101.621637] RIP: 0010:[<0000000000000000>]  [<          (null)>]   (null)
[  101.621684] RSP: 0018:ffff88003ddeba60  EFLAGS: 00010202
[  101.621716] RAX: ffff88003ddb5600 RBX: ffff88003ddb5600 RCX:
0000000000000020
[  101.621758] RDX: ffffffff81a69a00 RSI: ffffffff81b7ee61 RDI:
ffff88003ddb5600
[  101.621800] RBP: ffff8800537cd900 R08: 0000000000000000 R09:
ffff88003ddb5600
[  101.621840] R10: 0000000000000005 R11: 0000000000014b38 R12:
ffff88003ddb5600
[  101.621881] R13: ffffffff81b7e480 R14: ffffffff81b7e8b8 R15:
ffff88003ddebad8
[  101.621924] FS:  00007f06e4182700(0000) GS:ffff88007fd00000(0000)
knlGS:0000000000000000
[  101.621971] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  101.622005] CR2: 0000000000000000 CR3: 0000000045274000 CR4:
00000000000006e0
[  101.622046] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[  101.622087] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7:
0000000000000400
[  101.622129] Process openl2tpd (pid: 5129, threadinfo
ffff88003ddea000, task ffff88003de9a280)
[  101.622177] Stack:
[  101.622191]  ffffffff81447efa ffff88007d3ded80 ffff88003de9a280
ffff88007d3ded80
[  101.622245]  0000000000000001 ffff88003ddebbb8 ffffffff8148d5a7
0000000000000212
[  101.622299]  ffff88003dcea000 ffff88003dcea188 ffffffff00000001
ffffffff81b7e480
[  101.622353] Call Trace:
[  101.622374]  [<ffffffff81447efa>] ? ipv4_blackhole_route+0x1ba/0x210
[  101.622415]  [<ffffffff8148d5a7>] ? xfrm_lookup+0x417/0x510
[  101.622450]  [<ffffffff8127672a>] ? extract_buf+0x9a/0x140
[  101.622485]  [<ffffffff8144c6a0>] ? __ip_flush_pending_frames+0x70/0x70
[  101.622526]  [<ffffffff8146fbbf>] ? udp_sendmsg+0x62f/0x810
[  101.622562]  [<ffffffff813f98a6>] ? sock_sendmsg+0x116/0x130
[  101.622599]  [<ffffffff8109df58>] ? find_get_page+0x18/0x90
[  101.622633]  [<ffffffff8109fd6a>] ? filemap_fault+0x12a/0x4b0
[  101.622668]  [<ffffffff813fb5c4>] ? move_addr_to_kernel+0x64/0x90
[  101.622706]  [<ffffffff81405d5a>] ? verify_iovec+0x7a/0xf0
[  101.622739]  [<ffffffff813fc772>] ? sys_sendmsg+0x292/0x420
[  101.622774]  [<ffffffff810b994a>] ? handle_pte_fault+0x8a/0x7c0
[  101.622810]  [<ffffffff810b76fe>] ? __pte_alloc+0xae/0x130
[  101.622844]  [<ffffffff810ba2f8>] ? handle_mm_fault+0x138/0x380
[  101.622880]  [<ffffffff81024af9>] ? do_page_fault+0x189/0x410
[  101.622915]  [<ffffffff813fbe03>] ? sys_getsockname+0xf3/0x110
[  101.622952]  [<ffffffff81450c4d>] ? ip_setsockopt+0x4d/0xa0
[  101.622986]  [<ffffffff813f9932>] ? sockfd_lookup_light+0x22/0x90
[  101.623024]  [<ffffffff814b61fb>] ? system_call_fastpath+0x16/0x1b
[  101.623060] Code:  Bad RIP value.
[  101.623090] RIP  [<          (null)>]           (null)
[  101.623125]  RSP <ffff88003ddeba60>
[  101.623146] CR2: 0000000000000000
[  101.650871] ---[ end trace ca3856a7d8e8dad4 ]---
[  101.651011] __sk_free: optmem leakage (160 bytes) detected.

The oops happens in dst_metrics_write_ptr()
include/net/dst.h:124: return dst->ops->cow_metrics(dst, p);

dst->ops->cow_metrics is NULL and causes the oops.

Provide cow_metrics() methods, like we did in commit 214f45c91bb
(net: provide default_advmss() methods to blackhole dst_ops)

Signed-off-by: Held Bernhard <berny156@gmx.de>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Conflicts:
drivers/net/wireless/iwlwifi/iwl-core.c
drivers/net/wireless/rt2x00/rt2x00queue.c
drivers/net/wireless/rt2x00/rt2x00queue.h

cdc_ncm: fix short packet issue on some devices

The default maximum transmit length for NCM USB frames should be so
that a short packet happens at the end if the device supports a length
greater than the defined maximum. This is achieved by adding 4 bytes
to the maximum length so that the existing logic can fit a short
packet there.

Signed-off-by: Hans Petter Selasky <hselasky@c2i.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'davem.r8169' of git://git./linux/kernel/git/romieu/netdev-2.6

net: Remove __KERNEL__ cpp checks from include/net

These header files are never installed to user consumption, so any
__KERNEL__ cpp checks are superfluous.

Projects should also not copy these files into their userland utility
sources and try to use them there.  If they insist on doing so, the
onus is on them to sanitize the headers as needed.

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

r8169: don't request firmware when there's no userspace.

The firmware is cached during the first successfull call to open() and
released once the network device is unregistered. The driver uses the
cached firmware between open() and unregister_netdev().

So far the firmware is optional : a failure to load the firmware does
not prevent open() to success. It is thus necessary to 1) unregister
all 816x / 810[23] devices and 2) force a driver probe to issue a new
firmware load.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Fixed-by: Ciprian Docan <docan@eden.rutgers.edu>
Cc: Realtek linux nic maintainers <nic_swsd@realtek.com>

bnx2x: fix UDP csum offload

Fixed packets parameters for FW in UDP checksum offload flow.

Do not dereference TCP headers on non TCP frames.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netconsole: fix deadlock when removing net driver that netconsole is using (v2)

A deadlock was reported to me recently that occured when netconsole was being
used in a virtual guest.  If the virtio_net driver was removed while netconsole
was setup to use an interface that was driven by that driver, the guest
deadlocked.  No backtrace was provided because netconsole was the only console
configured, but it became clear pretty quickly what the problem was.  In
netconsole_netdev_event, if we get an unregister event, we call
__netpoll_cleanup with the target_list_lock held and irqs disabled.
__netpoll_cleanup can, if pending netpoll packets are waiting call
cancel_delayed_work_sync, which is a sleeping path.  the might_sleep call in
that path gets triggered, causing a console warning to be issued.  The
netconsole write handler of course tries to take the target_list_lock again,
which we already hold, causing deadlock.

The fix is pretty striaghtforward.  Simply drop the target_list_lock and
re-enable irqs prior to calling __netpoll_cleanup, the re-acquire the lock, and
restart the loop.  Confirmed by myself to fix the problem reported.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: "David S. Miller" <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

inet: constify ip headers and in6_addr

Add const qualifiers to structs iphdr, ipv6hdr and in6_addr pointers
where possible, to make code intention more obvious.

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

Revert "bridge: Forward reserved group addresses if !STP"

This reverts commit 1e253c3b8a1aeed51eef6fc366812f219b97de65.

It breaks 802.3ad bonding inside of a bridge.

The commit was meant to support transport bridging, and specifically
virtual machines bridged to an ethernet interface connected to a
switch port wiht 802.1x enabled.

But this isn't the way to do it, it breaks too many other things.

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

atl1c: Fix work event interrupt/task races

The mechanism used to initiate work events from the interrupt
handler has a classic read/modify/write race between the interrupt
handler that sets the condition, and the worker task that reads and
clears the condition. Close these races by using atomic
bit fields.

Cc: stable@kernel.org
Cc: Jie Yang <jie.yang@atheros.com>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: increment work_counter in be_worker

The commit 609ff3b ("be2net: add code to display temperature of ASIC")
adds support to display temperature of ASIC but there is missing
increment of work_counter in be_worker. Because of this 1) the
function be_cmd_get_die_temperature is called every 1 second instead
of every 32 seconds 2) be_cmd_get_die_temperature is called, although
it is not supported.  This patch fixes this bug.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: Remove hoplimit initialization to -1

The changes introduced with git-commit a02e4b7d ("ipv6: Demark default
hoplimit as zero.") missed to remove the hoplimit initialization. As a
result, ipv6_get_mtu interprets the return value of dst_metric_raw
(-1) as 255 and answers ping6 with this hoplimit.  This patche removes
the line such that ping6 is answered with the hoplimit value
configured via sysctl.

Signed-off-by: Thomas Egerer <thomas.egerer@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix hw_features ethtool_ops->set_flags compatibility

__ethtool_set_flags() was not taking into account features set but not
user-toggleable.

Since GFLAGS returns masked dev->features, EINVAL is returned when
passed flags differ to it, and not to wanted_features.

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

tg3: Add additional EEE messaging

This patch adds link messages and an item to the sign-on banner to make
EEE status more visible.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add macro for SMDSP toggling

A common AUX CTRL operation in the driver is to enable and disable the
SMDSP.  This patch consolidates the code so that the details of the
operation are in one place.  This patch also adds code to make sure the
SMDSP is enabled before executing code that relies on it.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add write accessor for AUX CTRL phy reg

This patch adds a write accessor for the aux ctrl phy register.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Add read accessor for AUX CTRL phy reg

This patch adds a read accessor for the aux ctrl register.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Move phy accessor functions higher

Phy accessor functions should live closer to where the base phy read /
write routines are.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Only allow phy ioctls while netif_running

When tg3 was new, phy accesses through ioctl were allowable at any time.
Then, the driver started shutting down the phy when the device was
closed.  Phy accesses would be allowed when the driver first attached to
the device, but then would be forbidden after the device had been up'd
and down'd.  After that, management firmware made it illegal to access
the phy unless the driver "owned" the device.  Now that most firmware
is being moved over to the APE, it is less clear when phy accesses are
safe.

While it is possible to attempt to identify these conditions and code
the driver to navigate through the pitfalls, it could be perplexing to
the admin why phy accesses work in some cases and not others.  This
patch brings some uniformity to the problem by only allowing phy
accesses while the driver has control of the device.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Nullify RSS for loopback test

The loopback test assumes all traffic goes to the first rx queue.  There
is a 1 in 4 chance this won't be true if RSS is enabled though.  This
patch reprograms the RSS indirection table to route all rx packets to
the first queue.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Adjust rx prod ring bd replenish thresholds

The oldest tg3 devices had large rx producer ring BD caches.  Back then,
it made sense to make the BD cache replenish threshold only a function
of the number of rx buffers posted by the driver.  Since then, the BD
cache sizes have shrunk to 25% of their original size and, in some
cases, the ring sizes have quadrupled in size.  Under such conditions,
static BD cache replenish thresholds no longer match the hardware
constraints.

This patch attempts to factor in the BD cache size into the bd cache
replenish strategy, taking the existing hardware bugs into account.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Workaround rx_discards stat bug

The 5717, 5718, 5719 A0, and 5720 A0 has a bug where the rx_discards
statistic counter will increment when dropping unwanted multicast
frames.  This patch works around the problem by attempting to
recreate the data using other means.  The resulting value will not be
accurate, but it can still serve as a problem indicator.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

powerpc: Fix multicast problem in fs_enet driver

mac-fec.c was setting individual UDP address registers instead of multicast
group address registers when joining a multicast group.
This prevented from correctly receiving UDP multicast packets.
According to datasheet, replaced hash_table_high and hash_table_low
with grp_hash_table_high and grp_hash_table_low respectively.
Also renamed hash_table_* with grp_hash_table_* in struct fec declaration
for 8xx: these registers are used only for multicast there.

Tested on a MPC5121 based board.
Build tested also against mpc866_ads_defconfig.

Signed-off-by: Andrea Galbusera <gizero@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mv643xx_eth: Fix build regression.

From Stephen Rothwell:

--------------------
After merging the final tree, today's linux-next build (powerpc chrp32_defconfig)
failed like this:

drivers/net/mv643xx_eth.c: In function 'port_start':
drivers/net/mv643xx_eth.c:2250: error: 'dev' undeclared (first use in this function)

Caused by commit aad59c431b77 ("net: mv643xx: convert to hw_features").
--------------------

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

ipv6: udp: fix the wrong headroom check

At this point, skb->data points to skb_transport_header.
So, headroom check is wrong.

For some case:bridge(UFO is on) + eth device(UFO is off),
there is no enough headroom for IPv6 frag head.
But headroom check is always false.

This will bring about data be moved to there prior to skb->head,
when adding IPv6 frag header to skb.

Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: implement event notification SCTP_SENDER_DRY_EVENT

This patch implement event notification SCTP_SENDER_DRY_EVENT.
SCTP Socket API Extensions:

  6.1.9. SCTP_SENDER_DRY_EVENT

  When the SCTP stack has no more user data to send or retransmit, this
  notification is given to the user. Also, at the time when a user app
  subscribes to this event, if there is no data to be sent or
  retransmit, the stack will immediately send up this notification.

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

sctp: change auth event type name to SCTP_AUTHENTICATION_EVENT

This patch change the auth event type name to SCTP_AUTHENTICATION_EVENT,
which is based on API extension compliance.

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

sctp: implement socket option SCTP_GET_ASSOC_ID_LIST

This patch Implement socket option SCTP_GET_ASSOC_ID_LIST.
SCTP Socket API Extension:

  8.2.6. Get the Current Identifiers of Associations
         (SCTP_GET_ASSOC_ID_LIST)

  This option gets the current list of SCTP association identifiers of
  the SCTP associations handled by a one-to-many style socket.

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

iwl4965: fix skb usage after free

Since

commit a120e912eb51e347f36c71b60a1d13af74d30e83
Author: Stanislaw Gruszka <sgruszka@redhat.com>
Date:   Fri Feb 19 15:47:33 2010 -0800

    iwlwifi: sanity check before counting number of tfds can be free

we use skb->data after calling ieee80211_tx_status_irqsafe(), which
could free skb instantly.

On current kernels I do not observe practical problems related with
bug, but on 2.6.35.y it cause random system hangs when stressing
wireless link, making bisection of other problems impossible.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

iwlwifi: fix skb usage after free

Since

commit a120e912eb51e347f36c71b60a1d13af74d30e83
Author: Stanislaw Gruszka <sgruszka@redhat.com>
Date:   Fri Feb 19 15:47:33 2010 -0800

    iwlwifi: sanity check before counting number of tfds can be free

we use skb->data after calling ieee80211_tx_status_irqsafe(), which
could free skb instantly.

On current kernels I do not observe practical problems related with
bug, but on 2.6.35.y it cause random system hangs when stressing
wireless link.

Cc: stable@kernel.org # 2.6.32+
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix SMPS debugfs locking

The locking with SMPS requests means that the
debugs file should lock the mgd mutex, not the
iflist mutex. Calls to __ieee80211_request_smps()
need to hold that mutex, so add an assertion.

This has always been wrong, but for some reason
never been noticed, probably because the locking
error only happens while unassociated.

Cc: stable@kernel.org [2.6.34+]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

Merge branch 'master' of git://git./linux/kernel/git/padovan/bluetooth-2.6

can: add missing socket check in can/raw release

v2: added space after 'if' according code style.

We can get here with a NULL socket argument passed from userspace,
so we need to handle it accordingly.

Thanks to Dave Jones pointing at this issue in net/can/bcm.c

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

sctp: move chunk from retransmit queue to abandoned list

If there is still data waiting to retransmit and remain in
retransmit queue, while doing the next retransmit, if the
chunk is abandoned, we should move it to abandoned list.

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

sctp: make heartbeat information in sctp_make_heartbeat()

Make heartbeat information in sctp_make_heartbeat() instead
of make it in sctp_sf_heartbeat() directly for common using.

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

sctp: fix to check the source address of COOKIE-ECHO chunk

SCTP does not check whether the source address of COOKIE-ECHO
chunk is the original address of INIT chunk or part of the any
address parameters saved in COOKIE in CLOSED state. So even if
the COOKIE-ECHO chunk is from any address but with correct COOKIE,
the COOKIE-ECHO chunk still be accepted. If the COOKIE is not from
a valid address, the assoc should not be established.

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

sctp: handle ootb packet in chunk order as defined

Changed the order of processing SHUTDOWN ACK and COOKIE ACK
refer to section 8.4:Handle "Out of the Blue" Packets.

SHUTDOWN ACK chunk should be processed before processing
"Stale Cookie" ERROR or a COOKIE ACK.

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

sctp: bail from sctp_endpoint_lookup_assoc() if not bound

The sctp_endpoint_lookup_assoc() function uses a port hash
to lookup the association and then checks to see if any of
them are on the current endpoint.  However, if the current
endpoint is not bound, there can't be any associations on
it, thus we can bail early.

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

sctp: remove completely unsed EMPTY state

SCTP does not SCTP_STATE_EMPTY and we can never be in
that state.  Remove useless code.

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

sctp: remove redundant check when walking through a list of TLV parameters

When  pos.v <= (void *)chunk + end - ntohs(pos.p->length) and
ntohs(pos.p->length) >= sizeof(sctp_paramhdr_t) these two expressions are all true,
pos.v <= (void *)chunk + end - sizeof(sctp_paramhdr_t) *must* be true.

This patch removes this kind of redundant check.
It's same to  _sctp_walk_errors macro.

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

sctp: check invalid value of length parameter in error cause

RFC4960, section 3.3.7 said:
 If an endpoint receives an ABORT with a format error or no TCB is
 found, it MUST silently discard it.

When an endpoint receives ABORT that parameter value is invalid,
drop it.

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

sctp: check parameter value of length in ERROR chunk

When an endpoint receives ERROR that parameter value is invalid,
send an ABORT to peer with a Protocol Violation error code.

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

bonding: 802.3ad - fix agg_device_up

The slave member of struct aggregator does not necessarily point
to a slave which is part of the aggregator. It points to the
slave structure containing the aggregator structure, while
completely different slaves (or no slaves at all) may be part of
the aggregator.

The agg_device_up() function wrongly uses agg->slave to find the state
of the aggregator.  Use agg->lag_ports->slave instead. The bug has
been introduced by commit 4cd6fe1c6483cde93e2ec91f58b7af9c9eea51ad
("bonding: fix link down handling in 802.3ad mode").

Signed-off-by: Jiri Bohac <jbohac@suse.cz>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ehea: Fix a DLPAR bug on ehea_rereg_mrs().

We are currently continuing if ehea_restart_qps() fails, when we
do a memory DLPAR (remove or add more memory to the system).

This patch just let the NAPI disabled if the ehea_restart_qps()
fails.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: tun: convert to hw_features

This changes offload setting behaviour to what I think is correct:
 - offloads set via ethtool mean what admin wants to use (by default
   he wants 'em all)
 - offloads set via ioctl() mean what userspace is expecting to get
   (this limits which admin wishes are granted)
 - TUN_NOCHECKSUM is ignored, as it might cause broken packets when
   forwarded (ip_summed == CHECKSUM_UNNECESSARY means that checksum
   was verified, not that it can be ignored)

If TUN_NOCHECKSUM is implemented, it should set skb->csum_* and
skb->ip_summed (= CHECKSUM_PARTIAL) for known protocols and let others
be verified by kernel when necessary.

TUN_NOCHECKSUM handling was introduced by commit
f43798c27684ab925adde7d8acc34c78c6e50df8:

    tun: Allow GSO using virtio_net_hdr

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

net: xen-netback: convert to hw_features

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: qlcnic: convert to hw_features

Bit more than minimal conversion. There might be some issues because
of qlcnic_set_netdev_features() if it's called after netdev init.

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

net: ibmveth: convert to hw_features

A minimal conversion.

ibmveth_set_csum_offload() can be folded into ibmveth_set_features()
and adapter->rx_csum removed - left for later cleanup.

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

net: pch_gbe: convert to hw_features

This also fixes bug in xmit path, where TX checksum offload state was used
instead of skb->ip_summed to decide if the offload was needed.

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

net: infiniband/ulp/ipoib: convert to hw_features

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

net: s390: convert to hw_features

options.large_send was easy to get rid of. options.checksum_type has deeper
roots so is left for later cleanup.

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

net: infiniband/hw/nes: convert to hw_features

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

Staging: convert hv network driver to hw_features

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

net: batman-adv: remove rx_csum ethtool_ops

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

net: dsa: remove ethtool_ops->set_sg

Remove set_sg from DSA slave ethtool_ops. Features inheritance looks
broken/not fully implemented anyway.

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