On ARM64, a BUG() is triggered in the eBPF JIT if a filter with a
constant shift that can't be encoded in the immediate field of the
UBFM/SBFM instructions is passed to the JIT.  Since these shifts
amounts, which are negative or >= regsize, are invalid, reject them in
the eBPF verifier and the classic BPF filter checker, for all
architectures.
Signed-off-by: NRabin Vincent <rabin@rab.in>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 1f718f0f ("bonding: populate neighbour's private on enslave")
undoes the fix provided by commit c2edacf8 ("bonding / ipv6: no addrconf
for slaves separately from master") by effectively setting the slave flag
after the slave has been opened.  If the slave comes up quickly enough, it
will go through the IPv6 addrconf before the slave flag has been set and
will get a link local IPv6 address.

In order to ensure that addrconf knows to ignore the slave devices on state
change, set IFF_SLAVE before dev_open() during bonding enslavement.

Fixes: 1f718f0f ("bonding: populate neighbour's private on enslave")
Signed-off-by: NKarl Heiss <kheiss@gmail.com>
Signed-off-by: NJay Vosburgh <jay.vosburgh@canonical.com>
Reviewed-by: NJarod Wilson <jarod@redhat.com>
Signed-off-by: NAndy Gospodarek <gospo@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit acf8dd0a ("udp: only allow UFO for packets from SOCK_DGRAM
sockets") disallows UFO for packets sent from raw sockets. We need to do
the same also for SOCK_DGRAM sockets with SO_NO_CHECK options, even if
for a bit different reason: while such socket would override the
CHECKSUM_PARTIAL set by ip_ufo_append_data(), gso_size is still set and
bad offloading flags warning is triggered in __skb_gso_segment().

In the IPv6 case, SO_NO_CHECK option is ignored but we need to disallow
UFO for packets sent by sockets with UDP_NO_CHECK6_TX option.
Signed-off-by: NMichal Kubecek <mkubecek@suse.cz>
Tested-by: NShannon Nelson <shannon.nelson@intel.com>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix possible null pointer dereference that may occur when calling
skb_reserve() on a null skb.

Fixes: 879c7220 ("net: pktgen: Observe needed_headroom of the device")
Signed-off-by: NJohn Fastabend <john.r.fastabend@intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Sergei Shtylyov says:

====================
Fix some dubious code in the Renesas Ethernet drivers

   Here's a set of 2 patches against DaveM's 'net.git' repo. While initializing
EMAC the code tries to respect the duplex mode both programmed into ECMR and
stored in its own private data -- this just can't be right.

[1/2] ravb: stop reading ECMR in ravb_emac_init()
[2/2] sh_eth: stop reading ECMR in sh_eth_dev_init()
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The code in sh_eth_dev_init()  twiddling the ECMR bits always looked a bit
strange to me:  if one intends to respect 'mdp->duplex', why save old value
of the ECMR.DM bit? As all the other bits are zeroed anyway, we don't really
need to read ECMR before writing to it.
Signed-off-by: NSergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The code in ravb_emac_init() twiddling the ECMR bits always looked a bit
strange to me: if one intends to respect 'priv->duplex', why save old value
of the ECMR.DM bit?   As all the other bits are zeroed anyway, we don't
really need to read ECMR before writing to it.
Signed-off-by: NSergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

only when user space passes the addresses should we consider their
presence
Signed-off-by: NJamal Hadi Salim <jhs@mojatatu.com>
Acked-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For tcp_yeah, use an ssthresh floor of 2, the same floor used by Reno
and CUBIC, per RFC 5681 (equation 4).

tcp_yeah_ssthresh() was sometimes returning a 0 or negative ssthresh
value if the intended reduction is as big or bigger than the current
cwnd. Congestion control modules should never return a zero or
negative ssthresh. A zero ssthresh generally results in a zero cwnd,
causing the connection to stall. A negative ssthresh value will be
interpreted as a u32 and will set a target cwnd for PRR near 4
billion.

Oleksandr Natalenko reported that a system using tcp_yeah with ECN
could see a warning about a prior_cwnd of 0 in
tcp_cwnd_reduction(). Testing verified that this was due to
tcp_yeah_ssthresh() misbehaving in this way.
Reported-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dmitry Vyukov reported a use-after-free in the code expanded by the
macro debug_post_sfx, which is caused by the use of the asoc pointer
after it was freed within sctp_side_effect() scope.

This patch fixes it by allowing sctp_side_effect to clear that asoc
pointer when the TCB is freed.

As Vlad explained, we also have to cover the SCTP_DISPOSITION_ABORT case
because it will trigger DELETE_TCB too on that same loop.

Also, there were places issuing SCTP_CMD_INIT_FAILED and ASSOC_FAILED
but returning SCTP_DISPOSITION_CONSUME, which would fool the scheme
above. Fix it by returning SCTP_DISPOSITION_ABORT instead.

The macro is already prepared to handle such NULL pointer.
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: NVlad Yasevich <vyasevich@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dumping the FDB (invoked with a process context) or handling FDB
notifications (polled periodicly in delayed work) might each entail
multiple EMAD transcations due to the number of entries.

While we only allow one EMAD transaction at a time, there is nothing
stopping the dump and notification processing sessions from
interleaving. However, this is forbidden by the hardware, so we need to
make sure only one of these sessions can run at a time.

Solve this by adding a mutex ('fdb_lock'), as both kernel threads can
sleep while waiting for the response EMAD.

Fixes: 56ade8fe ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
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>

It is possible for a process to allocate and accumulate far more FDs than
the process' limit by sending them over a unix socket then closing them
to keep the process' fd count low.

This change addresses this problem by keeping track of the number of FDs
in flight per user and preventing non-privileged processes from having
more FDs in flight than their configured FD limit.

Reported-by: socketpair@gmail.com
Reported-by: NTetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Mitigates: CVE-2013-4312 (Linux 2.0+)
Suggested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NWilly Tarreau <w@1wt.eu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When first SYNACK is sent, we already hold rcu_read_lock(), but this
is not true if a SYNACK is retransmitted, as a timer (soft) interrupt
does not hold rcu_read_lock()

Fixes: 45f6fad8 ("ipv6: add complete rcu protection around np->opt")
Reported-by: NDave Jones <davej@codemonkey.org.uk>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

proc_dostring() needs an initialized destination string, while the one
provided in proc_sctp_do_hmac_alg() contains stack garbage.

Thus, writing to cookie_hmac_alg would strlen() that garbage and end up
accessing invalid memory.

Fixes: 3c68198e ("sctp: Make hmac algorithm selection for cookie generation dynamic")
Signed-off-by: NSasha Levin <sasha.levin@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

SIMCom 7230E is a QMI LTE module with support for most "normal" bands.
Manual testing has showed that only interface five works.

Cc: Bjørn Mork <bjorn@mork.no>
Signed-off-by: NKristian Evensen <kristian.evensen@gmail.com>
Acked-by: NBjørn Mork <bjorn@mork.no>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

udp tunnel offloads tend to aggregate datagrams based on inner
headers. gro engine gets notified by tunnel implementations about
possible offloads. The match is solely based on the port number.

Imagine a tunnel bound to port 53, the offloading will look into all
DNS packets and tries to aggregate them based on the inner data found
within. This could lead to data corruption and malformed DNS packets.

While this patch minimizes the problem and helps an administrator to find
the issue by querying ip tunnel/fou, a better way would be to match on
the specific destination ip address so if a user space socket is bound
to the same address it will conflict.

Cc: Tom Herbert <tom@herbertland.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a vxlan interface is created, the driver checks that there is not
another vxlan interface with the same properties. To do this, it checks
the existing vxlan udp socket. Since commit 1c51a915, the creation of
the vxlan socket is done only when the interface is set up, thus it breaks
that test.

Example:
$ ip l a vxlan10 type vxlan id 10 group 239.0.0.10 dev eth0 dstport 0
$ ip l a vxlan11 type vxlan id 10 group 239.0.0.10 dev eth0 dstport 0
$ ip -br l | grep vxlan
vxlan10          DOWN           f2:55:1c:6a:fb:00 <BROADCAST,MULTICAST>
vxlan11          DOWN           7a:cb:b9:38:59:0d <BROADCAST,MULTICAST>

Instead of checking sockets, let's loop over the vxlan iface list.

Fixes: 1c51a915 ("vxlan: fix race caused by dropping rtnl_unlock")
Reported-by: NThomas Faivre <thomas.faivre@6wind.com>
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The union descriptor must be checked. Its usage was conditional
before the parser was introduced. This is important, because
many RNDIS device, which also use the common parser, have
bogus extra descriptors.
Signed-off-by: NOliver Neukum <oneukum@suse.com>
Tested-by: NVasily Galkin <galkin-vv@yandex.ru>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When the autosuspend is enabled and occurs before system suspend, we should
wake the device before running system syspend. Then, we could change the wake
event for system suspend. Otherwise, the device would resume the system when
receiving any packet.
Signed-off-by: NHayes Wang <hayeswang@realtek.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit 651df218 ("phy: micrel: Fix finding PHY properties in MAC
 node.") only fixes finding PHY properties in MAC node for KSZ9021. This
commit applies the same fix for KSZ9031.

Fixes: 8b63ec18 ("phylib: Make PHYs children of their MDIO bus, not the bus' parent.")
Acked-by: NAndrew Lunn <andrew@lunn.ch>
Signed-off-by: NHenri Roosen <henri.roosen@ginzinger.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Antonio Quartulli says:

====================
Included change:
- Fix invalid read while copying bat_iv.bcast_own by Sven Eckelmann
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

I noticed this code because it has a typo in the module_param(),
the "trs" at the end of "toim3232fliptrs" should be "rts".  But it's
dead code so instead of fixing it, I just deleted it.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

There is no "membase", it was "act_port" that was intended.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch fixes a bug writing to EEPROM in lan78xx_ethtool_set_eeprom()
when asked to write to OTP.
Signed-off-by: NWoojung Huh <woojung.huh@microchip.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

batadv_iv_ogm_orig_del_if removes a part of the bcast_own which previously
belonged to the now removed interface. This is done by copying all data
which comes before the removed interface and then appending all the data
which comes after the removed interface.

The address calculation for the position of the data which comes after the
removed interface assumed that the bat_iv.bcast_own is a pointer to a
single byte datatype. But it is a pointer to unsigned long and thus the
calculated position was wrong off factor sizeof(unsigned long).

Fixes: 83a8342678a0 ("more basic routing code added (forwarding packets /
bitarray added)")
Signed-off-by: NSven Eckelmann <sven@narfation.org>
Signed-off-by: NMarek Lindner <mareklindner@neomailbox.ch>
Signed-off-by: NAntonio Quartulli <a@unstable.cc>

Pull networking fixes from David Miller:
 "As usual, there are a couple straggler bug fixes:

   1) qlcnic_alloc_mbx_args() error returns are not checked in qlcnic
      driver.  Fix from Insu Yun.

   2) SKB refcounting bug in connector, from Florian Westphal.

   3) vrf_get_saddr() has to propagate fib_lookup() errors to it's
      callers, from David Ahern.

   4) Fix AF_UNIX splice/bind deadlock, from Rainer Weikusat.

   5) qdisc_rcu_free() fails to free the per-cpu qstats.  Fix from John
      Fastabend.

   6) vmxnet3 driver passes wrong page to dma_map_page(), fix from
     Shrikrishna Khare.

   7) Don't allow zero cwnd in tcp_cwnd_reduction(), from Yuchung Cheng"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  tcp: fix zero cwnd in tcp_cwnd_reduction
  Driver: Vmxnet3: Fix regression caused by 5738a09d
  net: qmi_wwan: Add WeTelecom-WPD600N
  mkiss: fix scribble on freed memory
  net: possible use after free in dst_release
  net: sched: fix missing free per cpu on qstats
  ARM: net: bpf: fix zero right shift
  6pack: fix free memory scribbles
  net: filter: make JITs zero A for SKF_AD_ALU_XOR_X
  bridge: Only call /sbin/bridge-stp for the initial network namespace
  af_unix: Fix splice-bind deadlock
  net: Propagate lookup failure in l3mdev_get_saddr to caller
  r8152: add reset_resume function
  connector: bump skb->users before callback invocation
  cxgb4: correctly handling failed allocation
  qlcnic: correctly handle qlcnic_alloc_mbx_args

Patch 3759824d ("tcp: PRR uses CRB mode by default and SS mode
conditionally") introduced a bug that cwnd may become 0 when both
inflight and sndcnt are 0 (cwnd = inflight + sndcnt). This may lead
to a div-by-zero if the connection starts another cwnd reduction
phase by setting tp->prior_cwnd to the current cwnd (0) in
tcp_init_cwnd_reduction().

To prevent this we skip PRR operation when nothing is acked or
sacked. Then cwnd must be positive in all cases as long as ssthresh
is positive:

1) The proportional reduction mode
   inflight > ssthresh > 0

2) The reduction bound mode
  a) inflight == ssthresh > 0

  b) inflight < ssthresh
     sndcnt > 0 since newly_acked_sacked > 0 and inflight < ssthresh

Therefore in all cases inflight and sndcnt can not both be 0.
We check invalid tp->prior_cwnd to avoid potential div0 bugs.

In reality this bug is triggered only with a sequence of less common
events.  For example, the connection is terminating an ECN-triggered
cwnd reduction with an inflight 0, then it receives reordered/old
ACKs or DSACKs from prior transmission (which acks nothing). Or the
connection is in fast recovery stage that marks everything lost,
but fails to retransmit due to local issues, then receives data
packets from other end which acks nothing.

Fixes: 3759824d ("tcp: PRR uses CRB mode by default and SS mode conditionally")
Reported-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NNeal Cardwell <ncardwell@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Reported-by: NBingkuo Liu <bingkuol@vmware.com>
Signed-off-by: NShrikrishna Khare <skhare@vmware.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The WeTelecom-WPD600N is an LTE module that, in addition to supporting most
"normal" bands, also supports LTE over 450MHz. Manual testing showed that
only interface number three replies to QMI messages.

Cc: Bjørn Mork <bjorn@mork.no>
Signed-off-by: NKristian Evensen <kristian.evensen@gmail.com>
Acked-by: NBjørn Mork <bjorn@mork.no>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit d79f16c0 fixed a user triggerable
scribble on free memory but added a new one which allows the user to
scribble even more and user controlled data into freed space.

As with 6pack we need to halt the queue before we free the buffers, because
the transmit logic is not protected by the semaphore.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

dst_release should not access dst->flags after decrementing
__refcnt to 0. The dst_entry may be in dst_busy_list and
dst_gc_task may dst_destroy it before dst_release gets a chance
to access dst->flags.

Fixes: d69bbf88 ("net: fix a race in dst_release()")
Fixes: 27b75c95 ("net: avoid RCU for NOCACHE dst")
Signed-off-by: NFrancesco Ruggeri <fruggeri@arista.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When a qdisc is using per cpu stats (currently just the ingress
qdisc) only the bstats are being freed. This also free's the qstats.

Fixes: b0ab6f92 ("net: sched: enable per cpu qstats")
Signed-off-by: NJohn Fastabend <john.r.fastabend@intel.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The LSR instruction cannot be used to perform a zero right shift since a
0 as the immediate value (imm5) in the LSR instruction encoding means
that a shift of 32 is perfomed.  See DecodeIMMShift() in the ARM ARM.

Make the JIT skip generation of the LSR if a zero-shift is requested.

This was found using american fuzzy lop.
Signed-off-by: NRabin Vincent <rabin@rab.in>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

commit acf673a3 fixed a user triggerable free
memory scribble but in doing so replaced it with a different one that allows
the user to control the data and scribble even more.

sixpack_close is called by the tty layer in tty context. The tty context is
protected by sp_get() and sp_put(). However network layer activity via
sp_xmit() is not protected this way. We must therefore stop the queue
otherwise the user gets to dump a buffer mostly of their choice into freed
kernel pages.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The SKF_AD_ALU_XOR_X ancillary is not like the other ancillary data
instructions since it XORs A with X while all the others replace A with
some loaded value.  All the BPF JITs fail to clear A if this is used as
the first instruction in a filter.  This was found using american fuzzy
lop.

Add a helper to determine if A needs to be cleared given the first
instruction in a filter, and use this in the JITs.  Except for ARM, the
rest have only been compile-tested.

Fixes: 34805931 ("net: filter: get rid of BPF_S_* enum")
Signed-off-by: NRabin Vincent <rabin@rab.in>
Acked-by: NDaniel Borkmann <daniel@iogearbox.net>
Acked-by: NAlexei Starovoitov <ast@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

[I stole this patch from Eric Biederman. He wrote:]

> There is no defined mechanism to pass network namespace information
> into /sbin/bridge-stp therefore don't even try to invoke it except
> for bridge devices in the initial network namespace.
>
> It is possible for unprivileged users to cause /sbin/bridge-stp to be
> invoked for any network device name which if /sbin/bridge-stp does not
> guard against unreasonable arguments or being invoked twice on the
> same network device could cause problems.

[Hannes: changed patch using netns_eq]

Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NHannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Pull tracing fixes from Steven Rostedt:
 "Two more fixes:

  1. The recordmcount change had an output that used sprintf()
     (incorrectly) when it should have been a fprintf() to stderr.

  2. The printk_formats file could crash if someone added a
     trace_printk() in the core kernel, and also added one in a module.
     This does not affect production kernels.  Only kernels where
     developers add trace_printk() for debugging can crash"

* tag 'trace-v4.4-rc4-3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Fix setting of start_index in find_next()
  ftrace/scripts: Fix incorrect use of sprintf in recordmcount

Pull tile bugfix from Chris Metcalf:
 "This fixes a bug that Sudip's buildbot found for tilepro allmodconfig.

  I've tagged it for stable only back to 3.19, which was when most of
  the other affected architectures added their support for working
  around this issue"

* 'stable' of git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  tile: provide CONFIG_PAGE_SIZE_64KB etc for tilepro

This allows the build system to know that it can't attempt to
configure the Lustre virtual block device, for example, when tilepro
is using 64KB pages (as it does by default).  The tilegx build
already provided those symbols.

Previously we required that the tilepro hypervisor be rebuilt with
a different hardcoded page size in its headers, and then Linux be
rebuilt using the updated hypervisor header.  Now we allow each of
the hypervisor and Linux to be built independently.  We still check
at boot time to ensure that the page size provided by the hypervisor
matches what Linux expects.
Signed-off-by: NChris Metcalf <cmetcalf@ezchip.com>
Cc: stable@vger.kernel.org [3.19+]

On 2015/11/06, Dmitry Vyukov reported a deadlock involving the splice
system call and AF_UNIX sockets,

http://lists.openwall.net/netdev/2015/11/06/24

The situation was analyzed as

(a while ago) A: socketpair()
B: splice() from a pipe to /mnt/regular_file
	does sb_start_write() on /mnt
C: try to freeze /mnt
	wait for B to finish with /mnt
A: bind() try to bind our socket to /mnt/new_socket_name
	lock our socket, see it not bound yet
	decide that it needs to create something in /mnt
	try to do sb_start_write() on /mnt, block (it's
	waiting for C).
D: splice() from the same pipe to our socket
	lock the pipe, see that socket is connected
	try to lock the socket, block waiting for A
B:	get around to actually feeding a chunk from
	pipe to file, try to lock the pipe.  Deadlock.

on 2015/11/10 by Al Viro,

http://lists.openwall.net/netdev/2015/11/10/4

The patch fixes this by removing the kern_path_create related code from
unix_mknod and executing it as part of unix_bind prior acquiring the
readlock of the socket in question. This means that A (as used above)
will sb_start_write on /mnt before it acquires the readlock, hence, it
won't indirectly block B which first did a sb_start_write and then
waited for a thread trying to acquire the readlock. Consequently, A
being blocked by C waiting for B won't cause a deadlock anymore
(effectively, both A and B acquire two locks in opposite order in the
situation described above).

Dmitry Vyukov(<dvyukov@google.com>) tested the original patch.
Signed-off-by: NRainer Weikusat <rweikusat@mobileactivedefense.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>