Commit cacc0621 ("sctp: use GFP_USER for user-controlled kmalloc")
missed two other spots.

For connectx, as it's more likely to be used by kernel users of the API,
it detects if GFP_USER should be used or not.

Fixes: cacc0621 ("sctp: use GFP_USER for user-controlled kmalloc")
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In the original code, if we succeeded on the last iteration through the
loop then we still returned failure.

Fixes: 389e4e04 ('qlcnic: fix a timeout loop')
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

NCM buffer sizes are negotiated with the device independently of
the network device MTU.  The RX buffers are allocated by the
usbnet framework based on the rx_urb_size value set by cdc_ncm. A
single RX buffer can hold a number of MTU sized packets.

The default usbnet change_mtu ndo only modifies rx_urb_size if it
is equal to hard_mtu.  And the cdc_ncm driver will set rx_urb_size
and hard_mtu independently of each other, based on dwNtbInMaxSize
and dwNtbOutMaxSize respectively. It was therefore assumed that
usbnet_change_mtu() would never touch rx_urb_size.  This failed to
consider the case where dwNtbInMaxSize and dwNtbOutMaxSize happens
to be equal.

Fix by implementing an NCM specific change_mtu ndo, modifying the
netdev MTU without touching the buffer size settings.
Signed-off-by: NBjørn Mork <bjorn@mork.no>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently the needed_headroom field for the geneve device is left
to the default value.

This patch set it to space required for basic geneve encapsulation,
so that we can avoid the skb head re-allocation on xmit.

This give a 6% speedup for unsegment traffic on geneve tunnel.

v1 -> v2:
  - add ETH_HLEN for the lower device to the needed headroom
Signed-off-by: NPaolo Abeni <pabeni@redhat.com>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Marc Haber reported we don't honor interface indexes when we receive link
local router addresses in router advertisements. Luckily the non-strict
version of ipv6_chk_addr already does the correct job here, so we can
simply use it to lighten the checks and use those addresses by default
without any configuration change.

Link: <http://permalink.gmane.org/gmane.linux.network/391348>
Reported-by: NMarc Haber <mh+netdev@zugschlus.de>
Cc: Marc Haber <mh+netdev@zugschlus.de>
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When sysctl performs restrict writes, it allows to write from
a middle position of a sysctl file, which requires us to initialize
the table data before calling proc_dostring() for the write case.

Fixes: 3d1bec99 ("ipv6: introduce secret_stable to ipv6_devconf")
Reported-by: NSasha Levin <sasha.levin@oracle.com>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Tested-by: NSasha Levin <sasha.levin@oracle.com>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Steffen Klassert says:

====================
pull request (net): ipsec 2015-12-22

Just one patch to fix dst_entries_init with multiple namespaces.
From Dan Streetman.

Please pull or let me know if there are problems.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ip6addrlbl_get() has never worked. If ip6addrlbl_hold() succeeded,
ip6addrlbl_get() will exit with '-ESRCH'. If ip6addrlbl_hold() failed,
ip6addrlbl_get() will use about to be free ip6addrlbl_entry pointer.

Fix this by inverting ip6addrlbl_hold() check.

Fixes: 2a8cc6c8 ("[IPV6] ADDRCONF: Support RFC3484 configurable address selection policy table.")
Signed-off-by: NAndrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: NCong Wang <cwang@twopensource.com>
Acked-by: NYOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The bridge's ageing time is offloaded to hardware when:
	1) A port joins a bridge
	2) The ageing time of the bridge is changed

In the first case the ageing time is offloaded as jiffies, but in the
second case it's offloaded as clock_t, which is what existing switchdev
drivers expect to receive.

Fixes: 6ac311ae ("Adding switchdev ageing notification on port bridged")
Signed-off-by: NIdo Schimmel <idosch@mellanox.com>
Signed-off-by: NJiri Pirko <jiri@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 1299653a ("sh_eth: fix descriptor access endianness") only
addressed the 32-bit buffer address field byte-swapping  but the driver
still accesses 16-bit frame/buffer length descriptor fields without the
necessary byte-swapping -- which should affect the big-endian kernels.
In order to be able to use {cpu|edmac}_to_{edmac|cpu}(), we need to declare
the RX/TX descriptor word 1 as a 32-bit field and use shifts/masking to
access the 16-bit subfields (which gets rid of the ugly #ifdef'ery too)...
Signed-off-by: NSergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Packets that arrive from real hardware devices have ip_summed ==
CHECKSUM_UNNECESSARY if the hardware verified the checksums, or
CHECKSUM_NONE if the packet is bad or it was unable to verify it. The
current version of veth will replace CHECKSUM_NONE with
CHECKSUM_UNNECESSARY, which causes corrupt packets routed from hardware to
a veth device to be delivered to the application. This caused applications
at Twitter to receive corrupt data when network hardware was corrupting
packets.

We believe this was added as an optimization to skip computing and
verifying checksums for communication between containers. However, locally
generated packets have ip_summed == CHECKSUM_PARTIAL, so the code as
written does nothing for them. As far as we can tell, after removing this
code, these packets are transmitted from one stack to another unmodified
(tcpdump shows invalid checksums on both sides, as expected), and they are
delivered correctly to applications. We didn’t test every possible network
configuration, but we tried a few common ones such as bridging containers,
using NAT between the host and a container, and routing from hardware
devices to containers. We have effectively deployed this in production at
Twitter (by disabling RX checksum offloading on veth devices).

This code dates back to the first version of the driver, commit
<e314dbdc> ("[NET]: Virtual ethernet device driver"), so I
suspect this bug occurred mostly because the driver API has evolved
significantly since then. Commit <0b796750> ("net/veth: Fix
packet checksumming") (in December 2010) fixed this for packets that get
created locally and sent to hardware devices, by not changing
CHECKSUM_PARTIAL. However, the same issue still occurs for packets coming
in from hardware devices.
Co-authored-by: NEvan Jones <ej@evanjones.ca>
Signed-off-by: NEvan Jones <ej@evanjones.ca>
Cc: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Cc: Phil Sutter <phil@nwl.cc>
Cc: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Cc: netdev@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: NVijay Pandurangan <vijayp@vijayp.ca>
Acked-by: NCong Wang <cwang@twopensource.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains two netfilter fixes:

1) Oneliner from Florian to dump missing NFT_CT_L3PROTOCOL netlink
   attribute, from Florian Westphal.

2) Another oneliner for nf_tables to use skb->protocol from the new
   netdev family, we can't assume ethernet there.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Daniele Palmas says:

====================
net: usb: cdc_ncm: Adding support for two new Dell devices

This patch series add support in the cdc_ncm driver for two devices
based on the same platform, that are different only for carrier
customization.

V2: Added comment for highlighting FLAG_NOARP usage for those devices
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Unlike DW5550, Dell DW5813 is a mobile broadband card with no ARP
capabilities: the patch makes this device to use wwan_noarp_info struct
Signed-off-by: NDaniele Palmas <dnlplm@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Unlike DW5550, Dell DW5812 is a mobile broadband card with no ARP
capabilities: the patch makes this device to use wwan_noarp_info struct
Signed-off-by: NDaniele Palmas <dnlplm@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

refill_rx() and start_tx() do not check if mapping dma memory succeed.
The patch adds the checks and failure handling.

Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: NAlexey Khoroshilov <khoroshilov@ispras.ru>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The commit c6ff5268 ("rhashtable:
Fix walker list corruption") causes a suspicious RCU usage warning
because we no longer hold ht->mutex when we dereference ht->tbl.

However, this is a false positive because we now hold ht->lock
which also guarantees that ht->tbl won't disppear from under us.

This patch kills the warning by using rcu_dereference_protected.
Reported-by: Nkernel test robot <ying.huang@linux.intel.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In a set action tunnel attributes should be encoded in a
nested action.

I noticed this because ovs-dpctl was reporting an error
when dumping flows due to the incorrect encoding of tunnel attributes
in a set action.

Fixes: fc4099f1 ("openvswitch: Fix egress tunnel info.")
Signed-off-by: NSimon Horman <simon.horman@netronome.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

FW has a rare corner case in which a fragmented packet using lots
of frags would not be linearized, causing the FW to assert while trying
to transmit the packet.

To prevent this, we need to make sure the window of fragements containing
MSS worth of data contains 1 BD less than for regular packets due to
the additional parsing BD.
Signed-off-by: NYuval Mintz <Yuval.Mintz@qlogic.com>
Signed-off-by: NAriel Elior <Ariel.Elior@qlogic.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

IP-TTL case is already handled in ip_tunnel_ioctl() API.
Signed-off-by: NPravin B Shelar <pshelar@nicira.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Need to do the unregister_device() after all references to the driver
private have been done.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Need to do the unregister_device() after all references to the driver
private have been done.

Also we need to use del_timer_sync() for the timers so that we don't
have any asynchronous references after the unregister.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the first request_irq fails in be_msix_register, i value
would be zero. The current code decrements the i value and
accesses the eq object without validating the decremented
"i" value. This can cause an "invalid memory address access"
violation.

This patch fixes the problem by accessing the eq object after
validating the "i" value.
Signed-off-by: NVenkat Duvvuru <venkatkumar.duvvuru@avagotech.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Or Gerlitz says:

====================
Mellanox mlx4 driver fixes

Two small fixes from Jenny for code flows that deal with time-stamping.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

mlx4_en_init_timestamp was called before creation of netdev and port
init, thus used uninitialized values.  Specifically - NIC frequency was
incorrect causing wrong calculations and later wrong HW timestamps.

Fixes: 1ec4864b ('net/mlx4_en: Fixed crash when port type is changed')
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NMarina Varshaver <marinav@mellanox.com>
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Service task is responsible for other tasks in addition to timestamping
overflow check. Launch it even if timestamping is not supported by device.

Fixes: 07841f9d ('net/mlx4_en: Schedule napi when RX buffers allocation fails')
Signed-off-by: NEugenia Emantayev <eugenia@mellanox.com>
Signed-off-by: NEran Ben Elisha <eranbe@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

David Rivshin says:

====================
drivers: net: cpsw: Fix bugs in fixed-link PHY DT parsing

Commit 1f71e8c9 ("drivers: net: cpsw:
Add support for fixed-link PHY") added initial fixed-link PHY support
for CPSW, but missed a few considerations.

This series is based on the tip of the net tree. The first two patches
fix user-visible errors in different hardware configurations. The third
patch is for an internal reference counting issue. They are logically
independent changes, but in the same function, so must be applied in
order to apply cleanly.

The first patch was originally submitted by Pascal Speck on December 4,
but was not picked up by patchwork. I suspect that is because the patch
was mangled by the mailer. I fixed the mangling and am including it in
this series, as I believe it is the correct change.

I have tested on the following hardware configurations:
 - (EVMSK) dual emac with two real MDIO-connected phys using RGMII-TXID
 - single emac with fixed-link using RGMII
Testing of other CPSW emac configurations that folks may have would
be appreciated.

Changes from v1 [1]:
 - Split into 3 smaller patches.
 - Maintain 1f71e8c9's preference for fixed-link over phy_id if
   they are both (incorrectly) specified in the slave node.
 - Update binding documentation to no longer say that phy_mode is also
   mutually exclusive with fixed-link.
 - Dropped unnecessary include of phy_fixed.h.

[1] https://patchwork.ozlabs.org/patch/554989/
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a fixed-link sub-node exists in a slave node, the slave node
is also the PHY node. Since this is a separate use of the slave node,
of_node_get() should be used to increment the reference count.

Fixes: 1f71e8c9 ("drivers: net: cpsw: Add support for fixed-link PHY")
Signed-off-by: NDavid Rivshin <drivshin@allworx.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 1f71e8c9 ("drivers: net: cpsw: Add
support for fixed-link PHY") did not parse the "phy-mode" property in
the case of a fixed-link PHY, leaving slave_data->phy_if with its default
of PHY_INTERFACE_MODE_NA(0). This later gets passed to phy_connect() in
cpsw_slave_open(), and eventually to cpsw_phy_sel() where it hits a default
case that configures the MAC for MII mode.

The user visible symptom is that while kernel log messages seem to indicate
that the interface is set up, there is no network communication. Eventually
a watchdog error occurs:
    NETDEV WATCHDOG: eth0 (cpsw): transmit queue 0 timed out

Fixes: 1f71e8c9 ("drivers: net: cpsw: Add support for fixed-link PHY")
Signed-off-by: NDavid Rivshin <drivshin@allworx.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When using more than one slave with ti cpsw and fixed phy the pd->phy_id
will be always zero, but slave_data->phy_id must be unique. pd->phy_id
means a "phy hardware id" whereas slave_data->phy_id means an "unique id",
so we should use pd->addr which has the same unique meaning.

Fixes: 1f71e8c9 ("drivers: net: cpsw: Add support for fixed-link PHY")
Signed-off-by: NPascal Speck <kernel@iktek.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

one nft userspace test case fails with

'ct l3proto original ipv4' mismatches 'ct l3proto ipv4'

... because NFTA_CT_DIRECTION attr is missing.
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Otherwise we may end up with incorrect network and transport header for
other protocols.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>

Pull networking fixes from David Miller:

 1) Fix uninitialized variable warnings in nfnetlink_queue, a lot of
    people reported this...  From Arnd Bergmann.

 2) Don't init mutex twice in i40e driver, from Jesse Brandeburg.

 3) Fix spurious EBUSY in rhashtable, from Herbert Xu.

 4) Missing DMA unmaps in mvpp2 driver, from Marcin Wojtas.

 5) Fix race with work structure access in pppoe driver causing
    corruptions, from Guillaume Nault.

 6) Fix OOPS due to sh_eth_rx() not checking whether netdev_alloc_skb()
    actually succeeded or not, from Sergei Shtylyov.

 7) Don't lose flags when settifn IFA_F_OPTIMISTIC in ipv6 code, from
    Bjørn Mork.

 8) VXLAN_HD_RCO defined incorrectly, fix from Jiri Benc.

 9) Fix clock source used for cookies in SCTP, from Marcelo Ricardo
    Leitner.

10) aurora driver needs HAS_DMA dependency, from Geert Uytterhoeven.

11) ndo_fill_metadata_dst op of vxlan has to handle ipv6 tunneling
    properly as well, from Jiri Benc.

12) Handle request sockets properly in xfrm layer, from Eric Dumazet.

13) Double stats update in ipv6 geneve transmit path, fix from Pravin B
    Shelar.

14) sk->sk_policy[] needs RCU protection, and as a result
    xfrm_policy_destroy() needs to free policies using an RCU grace
    period, from Eric Dumazet.

15) SCTP needs to clone ipv6 tx options in order to avoid use after
    free, from Eric Dumazet.

16) Missing kbuild export if ila.h, from Stephen Hemminger.

17) Missing mdiobus_alloc() return value checking in mdio-mux.c, from
    Tobias Klauser.

18) Validate protocol value range in ->create() methods, from Hannes
    Frederic Sowa.

19) Fix early socket demux races that result in illegal dst reuse, from
    Eric Dumazet.

20) Validate socket address length in pptp code, from WANG Cong.

21) skb_reorder_vlan_header() uses incorrect offset and can corrupt
    packets, from Vlad Yasevich.

22) Fix memory leaks in nl80211 registry code, from Ola Olsson.

23) Timeout loop count handing fixes in mISDN, xgbe, qlge, sfc, and
    qlcnic.  From Dan Carpenter.

24) msg.msg_iocb needs to be cleared in recvfrom() otherwise, for
    example, AF_ALG will interpret it as an async call.  From Tadeusz
    Struk.

25) inetpeer_set_addr_v4 forgets to initialize the 'vif' field, from
    Eric Dumazet.

26) rhashtable enforces the minimum table size not early enough,
    breaking how we calculate the per-cpu lock allocations.  From
    Herbert Xu.

27) Fix FCC port lockup in 82xx driver, from Martin Roth.

28) FOU sockets need to be freed using RCU, from Hannes Frederic Sowa.

29) Fix out-of-bounds access in __skb_complete_tx_timestamp() and
    sock_setsockopt() wrt.  timestamp handling.  From WANG Cong.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (117 commits)
  net: check both type and procotol for tcp sockets
  drivers: net: xgene: fix Tx flow control
  tcp: restore fastopen with no data in SYN packet
  af_unix: Revert 'lock_interruptible' in stream receive code
  fou: clean up socket with kfree_rcu
  82xx: FCC: Fixing a bug causing to FCC port lock-up
  gianfar: Don't enable RX Filer if not supported
  net: fix warnings in 'make htmldocs' by moving macro definition out of field declaration
  rhashtable: Fix walker list corruption
  rhashtable: Enforce minimum size on initial hash table
  inet: tcp: fix inetpeer_set_addr_v4()
  ipv6: automatically enable stable privacy mode if stable_secret set
  net: fix uninitialized variable issue
  bluetooth: Validate socket address length in sco_sock_bind().
  net_sched: make qdisc_tree_decrease_qlen() work for non mq
  ser_gigaset: remove unnecessary kfree() calls from release method
  ser_gigaset: fix deallocation of platform device structure
  ser_gigaset: turn nonsense checks into WARN_ON
  ser_gigaset: fix up NULL checks
  qlcnic: fix a timeout loop
  ...

Dmitry reported the following out-of-bound access:

Call Trace:
 [<ffffffff816cec2e>] __asan_report_load4_noabort+0x3e/0x40
mm/kasan/report.c:294
 [<ffffffff84affb14>] sock_setsockopt+0x1284/0x13d0 net/core/sock.c:880
 [<     inline     >] SYSC_setsockopt net/socket.c:1746
 [<ffffffff84aed7ee>] SyS_setsockopt+0x1fe/0x240 net/socket.c:1729
 [<ffffffff85c18c76>] entry_SYSCALL_64_fastpath+0x16/0x7a
arch/x86/entry/entry_64.S:185

This is because we mistake a raw socket as a tcp socket.
We should check both sk->sk_type and sk->sk_protocol to ensure
it is a tcp socket.

Willem points out __skb_complete_tx_timestamp() needs to fix as well.
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Acked-by: NWillem de Bruijn <willemb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently the Tx flow control is based on reading the hardware state,
which is not accurate since it may not reflect the descriptors that
are not yet reached the memory.

To accurately control the Tx flow, changing it to be software based.
Signed-off-by: NIyappan Subramanian <isubramanian@apm.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Yuchung tracked a regression caused by commit 57be5bda ("ip: convert
tcp_sendmsg() to iov_iter primitives") for TCP Fast Open.

Some Fast Open users do not actually add any data in the SYN packet.

Fixes: 57be5bda ("ip: convert tcp_sendmsg() to iov_iter primitives")
Reported-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

With b3ca9b02, the AF_UNIX SOCK_STREAM
receive code was changed from using mutex_lock(&u->readlock) to
mutex_lock_interruptible(&u->readlock) to prevent signals from being
delayed for an indefinite time if a thread sleeping on the mutex
happened to be selected for handling the signal. But this was never a
problem with the stream receive code (as opposed to its datagram
counterpart) as that never went to sleep waiting for new messages with the
mutex held and thus, wouldn't cause secondary readers to block on the
mutex waiting for the sleeping primary reader. As the interruptible
locking makes the code more complicated in exchange for no benefit,
change it back to using mutex_lock.
Signed-off-by: NRainer Weikusat <rweikusat@mobileactivedefense.com>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull drm fixes from Dave Airlie:
 "Some i915 fixes, one omap fix, one core regression fix.

  Not even enough fixes for a twelve days of xmas song, which seemms
  good"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm: Don't overwrite UNVERFIED mode status to OK
  drm/omap: fix fbdev pix format to support all platforms
  drm/i915: Do a better job at disabling primary plane in the noatomic case.
  drm/i915/skl: Double RC6 WRL always on
  drm/i915/skl: Disable coarse power gating up until F0
  drm/i915: Remove incorrect warning in context cleanup

The Cavium guys reported a soft lockup on their arm64 machine, caused by
commit c55a6ffa ("locking/osq: Relax atomic semantics"):

    mutex_optimistic_spin+0x9c/0x1d0
    __mutex_lock_slowpath+0x44/0x158
    mutex_lock+0x54/0x58
    kernfs_iop_permission+0x38/0x70
    __inode_permission+0x88/0xd8
    inode_permission+0x30/0x6c
    link_path_walk+0x68/0x4d4
    path_openat+0xb4/0x2bc
    do_filp_open+0x74/0xd0
    do_sys_open+0x14c/0x228
    SyS_openat+0x3c/0x48
    el0_svc_naked+0x24/0x28

This is because in osq_lock we initialise the node for the current CPU:

    node->locked = 0;
    node->next = NULL;
    node->cpu = curr;

and then publish the current CPU in the lock tail:

    old = atomic_xchg_acquire(&lock->tail, curr);

Once the update to lock->tail is visible to another CPU, the node is
then live and can be both read and updated by concurrent lockers.

Unfortunately, the ACQUIRE semantics of the xchg operation mean that
there is no guarantee the contents of the node will be visible before
lock tail is updated.  This can lead to lock corruption when, for
example, a concurrent locker races to set the next field.

Fixes: c55a6ffa ("locking/osq: Relax atomic semantics"):
Reported-by: NDavid Daney <ddaney@caviumnetworks.com>
Reported-by: NAndrew Pinski <andrew.pinski@caviumnetworks.com>
Tested-by: NAndrew Pinski <andrew.pinski@caviumnetworks.com>
Acked-by: NDavidlohr Bueso <dave@stgolabs.net>
Signed-off-by: NWill Deacon <will.deacon@arm.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1449856001-21177-1-git-send-email-will.deacon@arm.comSigned-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Pull libnvdimm fixes from Dan Williams:

 - Two bug fixes for misuse of PAGE_MASK in scatterlist and dma-debug.
   These are tagged for -stable.  The scatterlist impact is potentially
  corrupted dma addresses on HIGHMEM enabled platforms.

 - A minor locking fix for the NFIT hot-add implementation that is new
   in 4.4-rc.  This would only trigger in the case a hot-add raced
   driver removal.

* 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  dma-debug: Fix dma_debug_entry offset calculation
  Revert "scatterlist: use sg_phys()"
  nfit: acpi_nfit_notify(): Do not leave device locked