net-next: Fix an overflow bug in vmxnet3 Tx descriptor

Fix an overflow bug in vmxnet3 Tx descriptor

This patch fixes a bug where a 16K buffer on a Tx descriptor was overflowing
into the 'gen' bit in the descriptor thereby corrupting the descriptor and
stalling the transmit ring.

Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: Matthew Delco <delcoM@vmware.com>
Signed-off-by: Ronghua Zhang <ronghua@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3c59x: Add ethtool WOL support

This patch adds wrappers for ethtool to get or set wake-on-LAN
setting without re-inserting the kernel module.

Signed-off-by: Andrew O. Shadoura <andrew@beldisplaytech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix diag resource allocation

netif_device_attach missing from error path in qlcnic_diag_alloc_res

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

qlcnic: fix loopback test

o Loopback not supported for virtual function.

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

net: pskb_expand_head() optimization

Move frags[] at the end of struct skb_shared_info, and make
pskb_expand_head() copy only the used part of it instead of whole array.

This should avoid kmemcheck warnings and speedup pskb_expand_head() as
well, avoiding a lot of cache misses.

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

sysfs: add attribute to indicate hw address assignment type

Add addr_assign_type to struct net_device and expose it via sysfs.
This new attribute has the purpose of giving user-space the ability to
distinguish between different assignment types of MAC addresses.

For example user-space can treat NICs with randomly generated MAC
addresses differently than NICs that have permanent (locally assigned)
MAC addresses.
For the former udev could write a persistent net rule by matching the
device path instead of the MAC address.
There's also the case of devices that 'steal' MAC addresses from slave
devices. In which it is also be beneficial for user-space to be aware
of the fact.

This patch also introduces a helper function to assist adoption of
drivers that generate MAC addresses randomly.

Signed-off-by: Stefan Assmann <sassmann@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: fix ethtool stats

In latest changes about 64bit stats on 32bit arches,
[commit 28172739f0a276eb8 (net: fix 64 bit counters on 32 bit arches)],
I missed ixgbe uses a bit of magic in its ixgbe_gstrings_stats
definition.

IXGBE_NETDEV_STAT() must now assume offsets relative to
rtnl_link_stats64, not relative do dev->stats.

As a bonus, we also get 64bit stats on ethtool -S

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix bandwidth check

Fix maximum and minmum bandwith value.

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

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

Conflicts:
drivers/net/wireless/iwlwifi/iwl-commands.h

3c59x: Fix call to mdio_sync() with the wrong argument

commit a095cfc40ec7ebe63e9532383c5b5c2a27b14075
"3c59x: Specify window explicitly for access to windowed registers"
changed the first parameter to mdio_sync(), from a pointer to the
register mapping, to a pointer to the vortex_private structure,
and changed all but one of the call sites.  Fix that last one.

Reported-by: Luca Falavigna <dktrkranz@debian.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: s2io: fix buffer overflow

vpd_data[] is allocated as kmalloc(256, GFP_KERNEL), so if cnt = 255
then (cnt + 3) overflows 256. memset() is executed without checking.
vpd_data[cnt+2] must be less than 256-cnt-2 as the latter is number of
vpd_data[] elements to copy.

Do not fill with zero the beginning of nic->serial_num as it will
be filled with vpd_data[].

String in product_name[] should be terminated by '\0'.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mv643xx_eth: potential null dereference

We assume that "pd" can be null on the previous line, and throughout the
function so we should check it here as well.  This was introduced by
9b2c2ff7a1c0 "mv643xx_eth: use sw csum for big packets"

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: 3c59x: fix leak of iomaps

If vortex_probe1() fails we should unmap ioaddr mapped earlier.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: core: don't use own hex_to_bin() method

Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

wireless: use newly introduced hex_to_bin()

Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Corey Thomas <coreythomas@charter.net>
Cc: "John W. Linville" <linville@tuxdriver.com>
Cc: linux-wireless@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

usb: usbnet: use newly introduced hex_to_bin()

Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: linux-usb@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers: net: use newly introduced hex_to_bin()

Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers: atm: don't use private copy of hex_to_bin()

Signed-off-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Chas Williams <chas@cmf.nrl.navy.mil>
Cc: linux-atm-general@lists.sourceforge.net
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: bump the driver version number

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: variable name changes

This patch changes names of some variables.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: supress printing error when mac query fails for VF

When a virtual function driver in initialized, the network mac
query command can fail. Skip display of error message in that case.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: code to support tx rate configuration on virtual functions

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: add vlan support for sriov virtual functions

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: add support to get vf config

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: change to call pmac_del only if necessary

If a mac address has not been configured for a VF,
there is no need to call be_cmd_pmac_del.

Signed-off-by: Ajit Khaparde <ajitk@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth: avoid useless removal of multicast addresses

Function qeth_l2_remove_device invokes qeth_l2_del_all_mc at the end.
This is needless, because it is already called in the offline function.
And even more this is invalid, because multicast addresses cannot be
removed in DOWN state. Thus this patch deletes invocation of
qeth_l2_del_all_mc in function qeth_l2_remove_device.

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

qeth: Use memdup_user when user data is immediately copied into the allocated region.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression from,to,size,flag;
position p;
identifier l1,l2;
@@

-  to = \(kmalloc@p\|kzalloc@p\)(size,flag);
+  to = memdup_user(from,size);
   if (
-      to==NULL
+      IS_ERR(to)
                 || ...) {
   <+... when != goto l1;
-  -ENOMEM
+  PTR_ERR(to)
   ...+>
   }
-  if (copy_from_user(to, from, size) != 0) {
-    <+... when != goto l2;
-    -EFAULT
-    ...+>
-  }
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qeth: return zero from reply callback functions

Reply callback functions in qeth should return zero if command
response consists of one part only, otherwise qeth continues
waiting for further parts of the command response.

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

claw: A claw device is a group of just 2 ccw devices

When creating a claw device, just 2 subchannels have to be grouped.

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

qeth: avoid loop if ipa command response is missing

If qeth issues an ipa command, but for some reasons the response
never comes back, qeth reaches a timeout.
Reset the irq_pending flag of the write channel in timeout handling
code and trigger a recovery to avoid endless looping for the following
ipa command.

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

qeth: serialize sysfs-triggered device configurations

This patch serializes device removal and other sysfs-triggered
configurations by moving removal of sysfs-attributes to the beginning
of the remove functions. And it serializes online/offline setting
and discipline-switching (causing reestablishing of the net_device)
by making use of a new discipline mutex.

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

qeth: Clear mac_bits field when switching between Layer 2 and Layer 3

This patch fixes a problem that occurs when switching from layer 3 to layer 2
mode. Resetting this mac_bits makes sure that we retrieve our mac address from
the card, otherwise the interface simply would'nt work.

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

qeth: IP address takeover flag setting

The qeth IP address flag setting is possible when device is
offline. When setting device online afterwards the current set
IP addresses have to be correctly registered with the device
regarding the IP address takeover attribute.

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

stmmac: handle allocation errors in setup functions

If the allocations fail in either dwmac1000_setup() or dwmac100_setup()
then return NULL.  These are called from stmmac_mac_device_setup().  The
check for NULL returns in stmmac_mac_device_setup() needed to be moved
forward a couple lines.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

caif: precedence bug

Negate has precedence over comparison so the original assert only
checked that "rfml->fragment_size" was larger than 1 or 0.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: don't lock when copying/clearing VLAN list on slave

When copying VLAN information to or removing from a slave
during slave addition or removal, the bonding code currently holds
the bond->lock for write to prevent concurrent modification of the
vlan_list / vlgrp.

This is unnecessary, as all of these operations occur under
RTNL.  Holding the bond->lock also caused might_sleep issues for
some drivers' ndo_vlan_* functions.  This patch removes the extra
locking.

Problem reported by Michael Chan <mchan@broadcom.com>

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Cc: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: change test for presence of VLANs

After commit ad1afb00393915a51c21b1ae8704562bf036855f
("vlan_dev: VLAN 0 should be treated as "no vlan tag" (802.1p packet)")
it is now regular practice for a VLAN "add vid" for VLAN 0 to
arrive prior to any VLAN registration or creation of a vlan_group.

This patch updates the bonding code that tests for the presence
of VLANs configured above bonding.  The new logic tests for bond->vlgrp
to determine if a registration has occured, instead of testing that
bonding's internal vlan_list is empty.

The old code would panic when vlan_list was not empty, but
vlgrp was still NULL (because only an "add vid" for VLAN 0 had occured).

Bonding still adds VLAN 0 to its internal list so that 802.1p
frames are handled correctly on transmit when non-VLAN accelerated
slaves are members of the bond.  The test against bond->vlan_list
remains in bond_dev_queue_xmit for this reason.

Modification to the bond->vlgrp now occurs under lock (in
addition to RTNL), because not all inspections of it occur under RTNL.

Additionally, because 8021q will never issue a "kill vid" for
VLAN 0, there is now logic in bond_uninit to release any remaining
entries from vlan_list.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Cc: Pedro Garcia <pedro.netdev@dondevamos.com>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/fec: restore interrupt mask after software-reset in fec_stop()

After the change from mdio polling to irq, it became necessary to
restore the interrupt mask after resetting the chip in fec_stop().
Otherwise, with all irqs disabled, no communication with the PHY will be
possible after e.g. un-/replugging the cable and the device gets
stalled.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

3c59x: handle pci_iomap() errors

pci_iomap() can fail, handle this case and return -ENOMEM from probe
function.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: RTA_MARK addition

Add a new rt attribute, RTA_MARK, and use it in
rt_fill_info()/inet_rtm_getroute() to support following commands :

ip route get 192.168.20.110 mark NUMBER
ip route get 192.168.20.108 from 192.168.20.110 iif eth1 mark NUMBER
ip route list cache [192.168.20.110] mark NUMBER

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

CAN: Add Flexcan CAN controller driver

This core is found on some Freescale SoCs and also some Coldfire
SoCs. Support for Coldfire is missing though at the moment as
they have an older revision of the core which does not have RX FIFO
support.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

r6040: Fix args to phy_mii_ioctl().

Reported by Stephen Rothwell.

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

vhost net: Fix warning.

Reported by Stephen Rothwell.

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

net: remove last uses of __attribute__((packed))

Network code uses the __packed macro instead of __attribute__((packed)).

Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: David S. Miller <davem@davemloft.net>

irda: Use __packed annotation instead IRDA_PACKED macro

Remove IRDA_PACKED macro, which maps to __attribute__((packed)). IRDA is
one of the last users of __attribute__((packet)). Networking code uses
__packed now.

Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/irda: use for_each_pci_dev()

Use for_each_pci_dev() to simplify the code.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/qlge: Use pr_<level>, shrink text a bit

Add and use a few neatening macros
Remove PFX
Add #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
Convert printk(KERN_ERR to pr_err(

$ size drivers/net/qlge/built-in.o.*
  text    data     bss     dec     hex filename
 116456    2312   25712  144480   23460 drivers/net/qlge/built-in.o.old
 114909    2312   25728  142949   22e65 drivers/net/qlge/built-in.o.new

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

mac80211: proper IBSS locking

IBSS has never had locking, instead relying on some
memory barriers etc. That's hard to get right, and
I think we had it wrong too until the previous patch.
Since this is not performance sensitive, it doesn't
make sense to have the maintenance overhead of that,
so add proper locking.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix IBSS lockdep complaint

Bob reported a lockdep complaint originating in
the mac80211 IBSS code due to the common work
struct patch. The reason is that the IBSS and
station mode code have different locking orders
for the cfg80211 wdev lock and the work struct
(where "locking" implies running/canceling).

Fix this by simply not canceling the work in
the IBSS code, it is not necessary since when
the REQ_RUN bit is cleared, the work will run
without effect if it runs. When the interface
is set down, it is flushed anyway, so there's
no concern about it running after memory has
been invalidated either.

This fixes
https://bugzilla.kernel.org/show_bug.cgi?id=16419

Additionally, looking into this I noticed that
there's a small window while the IBSS is torn
down in which the work may be rescheduled and
the REQ_RUN bit be set again after leave() has
cleared it when a scan finishes at exactly the
same time. Avoid that by setting the ssid_len
to zero before clearing REQ_RUN which signals
to the scan finish code that this interface is
not active.

Reported-by: Bob Copeland <me@bobcopeland.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: refuse shared key auth when WEP is unavailable

When WEP is not available, we should reject shared
key authentication because it could never succeed.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cfg80211: fix race between sysfs and cfg80211

device_add() is called before adding the phy to the cfg80211 device
list.

So if a userspace program uses sysfs uevents to detect new phy
devices, and queries nl80211 to get phy info, it can get ENODEV even
though the phy exists in sysfs.

An easy workaround is to hold the cfg80211 mutex until the phy is
present in sysfs/cfg80211/debugfs.

Signed-off-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: silence phy_n sparse warnings

drivers/net/wireless/b43/phy_n.c:512:53: warning: cast truncates bits from constant value (ffff0fff becomes fff)
drivers/net/wireless/b43/phy_n.c:765:66: warning: cast truncates bits from constant value (ffff7fff becomes 7fff)
drivers/net/wireless/b43/phy_n.c:1012:38: warning: cast truncates bits from constant value (ffff00ff becomes ff)
drivers/net/wireless/b43/phy_n.c:1119:38: warning: cast truncates bits from constant value (ffff0fff becomes fff)
drivers/net/wireless/b43/phy_n.c:2458:56: warning: cast truncates bits from constant value (ffff7fff becomes 7fff)
drivers/net/wireless/b43/phy_n.c:2933:38: warning: cast truncates bits from constant value (ffff0fff becomes fff)
drivers/net/wireless/b43/phy_n.c:3294:57: warning: cast truncates bits from constant value (ffff3fff becomes 3fff)

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: silence most sparse warnings

  CHECK   drivers/net/wireless/b43/main.c
drivers/net/wireless/b43/main.c:111:5: warning: symbol 'b43_modparam_pio' was not declared. Should it be static?
  CHECK   drivers/net/wireless/b43/phy_g.c
drivers/net/wireless/b43/phy_g.c:975:56: warning: cast truncates bits from constant value (ffff7fff becomes 7fff)
  CHECK   drivers/net/wireless/b43/phy_lp.c
drivers/net/wireless/b43/phy_lp.c:2701:6: warning: symbol 'b43_lpphy_op_switch_analog' was not declared. Should it be static?
drivers/net/wireless/b43/phy_lp.c:1148:30: warning: cast truncates bits from constant value (ffff1fff becomes 1fff)
drivers/net/wireless/b43/phy_lp.c:1525:30: warning: cast truncates bits from constant value (ffff1fff becomes 1fff)
drivers/net/wireless/b43/phy_lp.c:1529:30: warning: cast truncates bits from constant value (ffff1fff becomes 1fff)
  CHECK   drivers/net/wireless/b43/wa.c
drivers/net/wireless/b43/wa.c:385:60: warning: cast truncates bits from constant value (ffff00ff becomes ff)
drivers/net/wireless/b43/wa.c:403:55: warning: cast truncates bits from constant value (ffff00ff becomes ff)
drivers/net/wireless/b43/wa.c:405:55: warning: cast truncates bits from constant value (ffff00ff becomes ff)
drivers/net/wireless/b43/wa.c:415:71: warning: cast truncates bits from constant value (ffff0fff becomes fff)

AFAICT, none of these amount to real bugs.  But this reduces warning
spam from sparse w/o significantly affecting readability of the code (IMHO).

Signed-off-by: John W. Linville <linville@tuxdriver.com>

arch/um/drivers: remove duplicate structure field initialization

There are two initializations of ndo_set_mac_address, one to a local
function that is not used otherwise and one to a function that is defined
elsewhere.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
identifier I, s, fld;
position p0,p;
expression E;
@@

struct I s =@p0 { ... .fld@p = E, ...};

@s@
identifier I, s, r.fld;
position r.p0,p;
expression E;
@@

struct I s =@p0 { ... .fld@p = E, ...};

@script:python@
p0 << r.p0;
fld << r.fld;
ps << s.p;
pr << r.p;
@@

if int(ps[0].line)<int(pr[0].line) or int(ps[0].column)<int(pr[0].column):
  cocci.print_main(fld,p0)
// </smpl>

akpm:

- Use the standard eth_mac_addr() in uml_net_set_mac()

- Remove unneeded and racy local set_ether_mac()

- Remove duplicated (and incorrect)
  uml_netdev_ops.ndo_set_mac_address initializer.

Fixes 8bb95b39a16ed55226810596f92216c53329d2fe ("uml: convert network
device to netdevice ops").

[akpm@linux-foundation.org: rework as above]
Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/cxgb3/t3_hw.c: use new hex_to_bin() method

Get rid of own implementation of hex_to_bin().

Signed-off-by: Andy Shevchenko <ext-andriy.shevchenko@nokia.com>
Acked-by: Divy Le Ray <divy@chelsio.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Conflicts:
drivers/vhost/net.c
net/bridge/br_device.c

Fix merge conflict in drivers/vhost/net.c with guidance from
Stephen Rothwell.

Revert the effects of net-2.6 commit 573201f36fd9c7c6d5218cdcd9948cee700b277d
since net-next-2.6 has fixes that make bridge netpoll work properly thus
we don't need it disabled.

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

rtl8180: improve signal reporting for rtl8185 hardware

The existing code seemed to be somewhat based on the datasheet, but
varied substantially from the vendor-provided driver.  This mirrors the
handling of the rtl8185 case from that driver, but still neglects the
specifics for the rtl8180 hardware.  Those details are a bit muddled...

Signed-off-by: John W. Linville <linville@tuxdriver.com>

mwl8k: correct/silence sparse warnings

drivers/net/wireless/mwl8k.c:1541:21: warning: incorrect type in assignment (different base types)
drivers/net/wireless/mwl8k.c:1541:21:    expected restricted __le16 [usertype] result
drivers/net/wireless/mwl8k.c:1541:21:    got int
drivers/net/wireless/mwl8k.c:1575:42:    expected unsigned short [unsigned] [usertype] cmd
drivers/net/wireless/mwl8k.c:1575:42:    got restricted __le16 [usertype] code
drivers/net/wireless/mwl8k.c:1587:50: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1587:50:    expected unsigned short [unsigned] [usertype] cmd
drivers/net/wireless/mwl8k.c:1587:50:    got restricted __le16 [usertype] code
drivers/net/wireless/mwl8k.c:1592:50: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1592:50:    expected unsigned short [unsigned] [usertype] cmd
drivers/net/wireless/mwl8k.c:1592:50:    got restricted __le16 [usertype] code
drivers/net/wireless/mwl8k.c:1845:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1845:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1845:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:1848:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1848:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1848:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:1851:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1851:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1851:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:1854:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1854:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1854:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:1857:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1857:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1857:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:1860:27: warning: incorrect type in argument 1 (different base types)
drivers/net/wireless/mwl8k.c:1860:27:    expected unsigned int [unsigned] [usertype] <noident>
drivers/net/wireless/mwl8k.c:1860:27:    got restricted __le32 [usertype] <noident>
drivers/net/wireless/mwl8k.c:3055:20: warning: incorrect type in assignment (different base types)
drivers/net/wireless/mwl8k.c:3055:20:    expected restricted __le16 [usertype] ht_caps
drivers/net/wireless/mwl8k.c:3055:20:    got unsigned short [unsigned] [usertype] cap

At least the last one looks like a real bug...

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Acked-by: Lennert Buytenhek <buytenh@wantstofly.org>

wireless: remove unnecessary reg_same_country_ie_hint

"Might as well remove  reg_same_country_ie_hint() completely since we
already dealt with suspend/resume through the regulatory hint
disconnect." -- Luis

Reported-by: Luis R. Rodriguez <mcgrof@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rt2x00: correct sparse warning in rt2x00debug.c

  CHECK   drivers/net/wireless/rt2x00/rt2x00debug.c
drivers/net/wireless/rt2x00/rt2x00debug.c:193:28: warning: incorrect type in assignment (different base types)
drivers/net/wireless/rt2x00/rt2x00debug.c:193:28:    expected restricted __le32 [usertype] chip_rev
drivers/net/wireless/rt2x00/rt2x00debug.c:193:28:    got restricted __le16 [usertype] <noident>

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>

libipw: correct sparse warnings and mark some variables static

  CHECK   drivers/net/wireless/ipw2x00/libipw_module.c
drivers/net/wireless/ipw2x00/libipw_module.c:65:21: warning: symbol 'libipw_config_ops' was not declared. Should it be static?
drivers/net/wireless/ipw2x00/libipw_module.c:66:6: warning: symbol 'libipw_wiphy_privid' was not declared. Should it be static?
  CHECK   drivers/net/wireless/ipw2x00/libipw_wx.c
drivers/net/wireless/ipw2x00/libipw_wx.c:415:17: warning: symbol 'ssid' shadows an earlier one
drivers/net/wireless/ipw2x00/libipw_wx.c:324:9: originally declared here

Signed-off-by: John W. Linville <linville@tuxdriver.com>

ipw2100: mark ipw2100_pm_qos_req static

  CHECK   drivers/net/wireless/ipw2x00/ipw2100.c
drivers/net/wireless/ipw2x00/ipw2100.c:177:28: warning: symbol 'ipw2100_pm_qos_req' was not declared. Should it be static?

Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: correct sparse identified endian bug in ath_paprd_calibrate

drivers/net/wireless/ath/ath9k/main.c:282:26: warning: incorrect type in assignment (different base types)
drivers/net/wireless/ath/ath9k/main.c:282:26:    expected restricted __le16 [usertype] duration_id
drivers/net/wireless/ath/ath9k/main.c:282:26:    got int

Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: mark cfg80211_is_all_idle as static

  CHECK   net/wireless/sme.c
net/wireless/sme.c:38:6: warning: symbol 'cfg80211_is_all_idle' was not declared. Should it be static?

It is not used elsewhere, so mark it static.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: correct sparse warning in generated regdb.c

  CHECK   net/wireless/regdb.c
net/wireless/regdb.c:8:34: warning: symbol 'reg_regdb' was not declared.  Should it be static?
net/wireless/regdb.c:11:5: warning: symbol 'reg_regdb_size' was not declared. Should it be static?

Simply include the also generated regdb.h.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: correct sparse warning in wext-compat.c

  CHECK   net/wireless/wext-compat.c
net/wireless/wext-compat.c:1434:5: warning: symbol 'cfg80211_wext_siwpmksa' was not declared. Should it be static?

Add declaration in cfg80211.h.  Also add an EXPORT_SYMBOL_GPL, since all
the peer functions have it.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: correct sparse warning in lib80211_crypt_tkip.c

  CHECK   net/wireless/lib80211_crypt_tkip.c
net/wireless/lib80211_crypt_tkip.c:581:27: warning: cast to restricted __le16

Caused by dereferencing a "u8 *" and passing it to le16_to_cpu...

Signed-off-by: John W. Linville <linville@tuxdriver.com>

wireless: only use alpha2 regulatory information from country IE

The meaning and/or usage of the country IE is somewhat poorly defined.
In practice, this means that regulatory rulesets in a country IE are
often incomplete and might be untrustworthy.  This removes the code
associated with interpreting those rulesets while preserving respect
for country "alpha2" codes also contained in the country IE.

Signed-off-by: John W. Linville <linville@tuxdriver.com>

drop_monitor: convert some kfree_skb call sites to consume_skb

Convert a few calls from kfree_skb to consume_skb

Noticed while I was working on dropwatch that I was detecting lots of internal
skb drops in several places.  While some are legitimate, several were not,
freeing skbs that were at the end of their life, rather than being discarded due
to an error.  This patch converts those calls sites from using kfree_skb to
consume_skb, which quiets the in-kernel drop_monitor code from detecting them as
drops.  Tested successfully by myself

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

drop_monitor: Add error code to detect duplicate state changes

Patch to add -EAGAIN error to dropwatch netlink message handling code.
-EAGAIN will be returned anytime userspace attempts to transition the state of
the drop monitor service to a state that its already in.  That allows user space
to detect this condition, so it doesn't wait for a success ACK that will never
arrive.  Tested successfully by me

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

__dst_free(): put EXPORT_SYMBOLS after the fct

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phy: add suspend/resume in the ic+

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mac80211: set carrier on for monitor interfaces on ieee80211_open

If a station interface is reused as monitor interface it is possible that
the carrier is still set to off. This breaks packet injection on that
monitor interface.
Force the carrier on in monitor interface initialisation like it is also done
for other interface types (e.g. adhoc, mesh point, ap).

Signed-off-by: David Gnedt <david.gnedt@davizone.at>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: move QoS-enable to BSS info

Ever since

commit e1b3ec1a2a336c328c336cfa5485a5f0484cc90d
Author: Stanislaw Gruszka <sgruszka@redhat.com>
Date:   Mon Mar 29 12:18:34 2010 +0200

    mac80211: explicitly disable/enable QoS

mac80211 is telling drivers, in particular
iwlwifi, whether QoS is enabled or not.

However, this is only relevant for station mode,
since only then will any device send nullfunc
frames and need to know whether they should be
QoS frames or not. In other modes, there are
(currently) no frames the device is supposed to
send.

When you now consider virtual interfaces, it
becomes apparent that the current mechanism is
inadequate since it enables/disables QoS on a
global scale, where for nullfunc frames it has
to be on a per-interface scale.

Due to the above considerations, we can change
the way mac80211 advertises the QoS state to
drivers to only ever advertise it as "off" in
station mode, and make it a per-BSS setting.

Tested-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix aggregation action frame handling with AP VLANs

When aggregation related action frames are enqueued for further work,
and they originate from a STA that is part of an AP VLAN, they are
currently enqueued for the AP interface. This breaks the sta_info_get()
lookup in the actual work function, and because of that, aggregation
sessions are not established for this STA.

Fix this by replacing the sta_info_get call with a call to
sta_info_get_bss.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

cxgb4vf: add maintainer entry for cxgb4vf

Adding myself as the official maintainer of the Chelsio T4 Virtual function
Driver (cxgb4vf).

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4vf: Fix bug where we were only allocating one queue in MSI mode

Fix bug in setup_sge_queues() where we were incorrectly only allocating a
single "Queue Set" for MSI mode.

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4vf: Fix off-by-one error checking for the end of the mailbox delay array

Fix off-by-one error in checking for the end of the mailbox response delay
array.  We ended up walking off the end and, if we were unlucky, we'd end up
pulling in a 0 and never terminate the mailbox response delay loop ...

Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: Partially disable netpoll support

The new netpoll code in bridging contains use-after-free bugs
that are non-trivial to fix.

This patch fixes this by removing the code that uses skbs after
they're freed.

As a consequence, this means that we can no longer call bridge
from the netpoll path, so this patch also removes the controller
function in order to disable netpoll.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Thanks,
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: Make IP6CB(skb)->nhoff 16-bit.

Even with jumbograms I cannot see any way in which we would need
to records a larger than 65535 valued next-header offset.

The maximum extension header length is (256 << 3) == 2048.
There are only a handful of extension headers specified which
we'd even accept (say 5 or 6), therefore the largest next-header
offset we'd ever have to contend with is something less than
say 16k.

Therefore make it a u16 instead of a u32.

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

bnx2: Update version to 2.0.17.

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

bnx2: Remove some unnecessary smp_mb() in tx fast path.

smp_mb() inside bnx2_tx_avail() is used twice in the normal
bnx2_start_xmit() path (see illustration below).  The full memory
barrier is only necessary during race conditions with tx completion.
We can speed up the tx path by replacing smp_mb() in bnx2_tx_avail()
with a compiler barrier.  The compiler barrier is to force the
compiler to fetch the tx_prod and tx_cons from memory.

In the race condition between bnx2_start_xmit() and bnx2_tx_int(),
we have the following situation:

bnx2_start_xmit()                       bnx2_tx_int()
    if (!bnx2_tx_avail())
            BUG();

    ...

    if (!bnx2_tx_avail())
            netif_tx_stop_queue();          update_tx_index();
            smp_mb();                       smp_mb();
            if (bnx2_tx_avail())            if (netif_tx_queue_stopped() &&
                    netif_tx_wake_queue();      bnx2_tx_avail())

With smp_mb() removed from bnx2_tx_avail(), we need to add smp_mb() to
bnx2_start_xmit() as shown above to properly order netif_tx_stop_queue()
and bnx2_tx_avail() to check the ring index.  If it is not strictly
ordered, the tx queue can be stopped forever.

This improves performance by about 5% with 2 ports running bi-directional
64-byte packets.

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

bnx2: Call pci_enable_msix() with actual number of vectors.

Based on original patch by Breno Leitão <leitao@linux.vnet.ibm.com>.

Allocate the actual number of vectors and make use of fewer vectors
if pci_enable_msix() returns > 0.  We must allocate one additional
vector for the cnic driver.

Cc: Breno Leitão <leitao@linux.vnet.ibm.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Reviewed-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2: Use proper counter for net_device_stats->multicast.

We were using the wrong tx multicast counter instead of the rx multicast
counter.

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

ixgbe: fix version string for ixgbe

Bump the version string to better reflect what is in the driver.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: use GFP_ATOMIC when allocating FCoE DDP context from the dma pool

The FCoE protocol stack may hold a lock when this gets called.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: properly toggling netdev feature flags when disabling FCoE

When FCoE is disabled, there is a race condition that FCoE offload is
turned off but the FCoE protocol driver is still queuing I/O thinking
offload support still exists. This patch toggles off corresponding FCoE
netdev feature flags and notify the FCoE stack first, allowing FCoE
protocol stack driver to update its flags upon NETDEV_FEAT_CHANGE so no
I/O will be using offload.

Also, indicate FCoE offload flags in vlan_features in ixgbe_probe once
and do not toggle them in ixgbe_fcoe_enable/disable so when FCoE is
created on the VLAN interface, vlan_transfer_features() would properly
update the VLAN netdev features flag and notify the FCoE protocol driver
for NETDEV_FEAT_CHANGE.

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

ixgbe: drop support for UDP in RSS hash generation

This change removes UDP from the supported protocols for RSS hashing.  The
reason for removing this protocol is because IP fragmentation was causing a
network flow to be broken into two streams, one for fragmented, and one for
non-fragmented and this in turn was causing out-of-order issues.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ixgbe: dcb, set DPF bit when PFC is enabled

Set the DPF bit when PFC is enabled.  This will discard
PFC frames so they do not get passed up the stack.

The DPF bit is set for flow control, but not priority
flow control this brings pfc inline with fc.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: allow option to limit number of descriptors down to 48 per ring

This change makes it possible to limit the number of descriptors down to 48
per ring.  The reason for this change is to address a variation on hardware
errata 10 for 82546GB in which descriptors will be lost if more than 32
descriptors are fetched and the PCI-X MRBC is 512.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Emil Tantilov <emil.s.tantilov@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/82596.c: fix warning

drivers/net/82596.c: In function 'i596_open':
drivers/net/82596.c:1044: warning: label 'err_irq_dev' defined but not used

Caused by "82596: free resources on error"

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ks8842: Fix ks8842_tx_frame() for 16bit case.

As reported by Andrew:

drivers/net/ks8842.c: In function 'ks8842_handle_rx':
drivers/net/ks8842.c:428: warning: 'status' may be used uninitialized in this function

Just use the 32-bit status for all reads, and delete the useless
cast to 'int' when reading a u16 into 'len'.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: this_cpu_xxx conversions

Use modern this_cpu_xxx() api, saving few bytes on x86

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

smsc911x: Add spinlocks around registers access

On SMP systems, the SMSC911x registers may be accessed by multiple CPUs
and this seems to put the chip in an inconsistent state. The patch adds
spinlocks to the smsc911x_reg_read, smsc911x_reg_write,
smsc911x_rx_readfifo and smsc911x_tx_writefifo functions.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bonding: avoid a warning

drivers/net/bonding/bond_main.c:179:12: warning: ‘disable_netpoll’
defined but not used

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

net: Removing dead ARCH_PNX010X

ARCH_PNX010X doesn't exist in Kconfig, therefore removing all
references for it from the source code/Kconfig.

Signed-off-by: Christoph Egger <siccegge@cs.fau.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

net-next: vmxnet3 fixes [5/5] Respect the interrupt type in VM configuration

Respect the interrupt type set in VM configuration.

When interrupt type is not auto, do not ignore the interrupt type set from
VM configuration.

Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net-next: vmxnet3 fixes [4/5] Do not reset when the device is not opened

Hold rtnl_lock to get the right link state.

While asynchronously resetting the device, hold rtnl_lock to get the
right value from netif_running. If a reset is scheduled, and the device
goes thru close and open, it may happen that reset and open may run in
parallel. Holding rtnl_lock will avoid this.

Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tcp: fix crash in tcp_xmit_retransmit_queue

It can happen that there are no packets in queue while calling
tcp_xmit_retransmit_queue(). tcp_write_queue_head() then returns
NULL and that gets deref'ed to get sacked into a local var.

There is no work to do if no packets are outstanding so we just
exit early.

This oops was introduced by 08ebd1721ab8fd (tcp: remove tp->lost_out
guard to make joining diff nicer).

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Reported-by: Lennart Schulte <lennart.schulte@nets.rwth-aachen.de>
Tested-by: Lennart Schulte <lennart.schulte@nets.rwth-aachen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

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