Dave Miller <davem@davemloft.net> provided a detailed description of
why the way IPoIB is using neighbours for its own ipoib_neigh struct
is buggy:

    Any time an ipoib_neigh is changed, a sequence like the following is made:

    			spin_lock_irqsave(&priv->lock, flags);
    			/*
    			 * It's safe to call ipoib_put_ah() inside
    			 * priv->lock here, because we know that
    			 * path->ah will always hold one more reference,
    			 * so ipoib_put_ah() will never do more than
    			 * decrement the ref count.
    			 */
    			if (neigh->ah)
    				ipoib_put_ah(neigh->ah);
    			list_del(&neigh->list);
    			ipoib_neigh_free(dev, neigh);
    			spin_unlock_irqrestore(&priv->lock, flags);
    			ipoib_path_lookup(skb, n, dev);

    This doesn't work, because you're leaving a stale pointer to the freed up
    ipoib_neigh in the special neigh->ha pointer cookie.  Yes, it even fails
    with all the locking done to protect _changes_ to *ipoib_neigh(n), and
    with the code in ipoib_neigh_free() that NULLs out the pointer.

    The core issue is that read side calls to *to_ipoib_neigh(n) are not
    being synchronized at all, they are performed without any locking.  So
    whether we hold the lock or not when making changes to *ipoib_neigh(n)
    you still can have threads see references to freed up ipoib_neigh
    objects.

    	cpu 1			cpu 2
    	n = *ipoib_neigh()
    				*ipoib_neigh() = NULL
    				kfree(n)
    	n->foo == OOPS

    [..]

    Perhaps the ipoib code can have a private path database it manages
    entirely itself, which holds all the necessary information and is
    looked up by some generic key which is available easily at transmit
    time and does not involve generic neighbour entries.

See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and
<http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b>
for the full discussion.

This patch aims to solve the race conditions found in the IPoIB driver.

The patch removes the connection between the core networking neighbour
structure and the ipoib_neigh structure.  In addition to avoiding the
race described above, it allows us to handle SKBs carrying IP packets
that don't have any associated neighbour.

We add an ipoib_neigh hash table with N buckets where the key is the
destination hardware address.  The ipoib_neigh is fetched from the
hash table and instead of the stashed location in the neighbour
structure. The hash table uses both RCU and reference counting to
guarantee that no ipoib_neigh instance is ever deleted while in use.

Fetching the ipoib_neigh structure instance from the hash also makes
the special code in ipoib_start_xmit that handles remote and local
bonding failover redundant.

Aged ipoib_neigh instances are deleted by a garbage collection task
that runs every M seconds and deletes every ipoib_neigh instance that
was idle for at least 2*M seconds. The deletion is safe since the
ipoib_neigh instances are protected using RCU and reference count
mechanisms.

The number of buckets (N) and frequency of running the GC thread (M),
are taken from the exported arb_tbl.

Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

Commit 36a8f01cd24b ("IB/qib: Add congestion control agent
implementation") tries to store the value 1984 in a u8, which leads to
truncation.  Fix this by making the member big enough.

This bug was detected by a smatch warning.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Ramkrishna Vepa <ramkrishna.vepa@intel.com>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

Suggested by scripts/coccinelle/api/memdup_user.cocci.

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

It looks like one check was accidentally duplicated, and the other 3
checks were left out.  This was detected by scripts/coccinelle/tests/doubletest.cocci:

    drivers/infiniband/hw/ocrdma/ocrdma_verbs.c:895:6-54: duplicated argument to && or ||

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>

Suggested by scripts/coccinelle/api/ptr_ret.cocci.

Signed-off-by: Roland Dreier <roland@purestorage.com>

Pull InfiniBand/RDMA changes from Roland Dreier:
 - Updates to the qib low-level driver
 - First chunk of changes for SR-IOV support for mlx4 IB
 - RDMA CM support for IPv6-only binding
 - Other misc cleanups and fixes

Fix up some add-add conflicts in include/linux/mlx4/device.h and
drivers/net/ethernet/mellanox/mlx4/main.c

* tag 'rdma-for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (30 commits)
  IB/qib: checkpatch fixes
  IB/qib: Add congestion control agent implementation
  IB/qib: Reduce sdma_lock contention
  IB/qib: Fix an incorrect log message
  IB/qib: Fix QP RCU sparse warnings
  mlx4: Put physical GID and P_Key table sizes in mlx4_phys_caps struct and paravirtualize them
  mlx4_core: Allow guests to have IB ports
  mlx4_core: Implement mechanism for reserved Q_Keys
  net/mlx4_core: Free ICM table in case of error
  IB/cm: Destroy idr as part of the module init error flow
  mlx4_core: Remove double function declarations
  IB/mlx4: Fill the masked_atomic_cap attribute in query device
  IB/mthca: Fill in sq_sig_type in query QP
  IB/mthca: Warning about event for non-existent QPs should show event type
  IB/qib: Fix sparse RCU warnings in qib_keys.c
  net/mlx4_core: Initialize IB port capabilities for all slaves
  mlx4: Use port management change event instead of smp_snoop
  IB/qib: RCU locking for MR validation
  IB/qib: Avoid returning EBUSY from MR deregister
  IB/qib: Fix UC MR refs for immediate operations
  ...

Pull networking changes from David S Miller:

 1) Remove the ipv4 routing cache.  Now lookups go directly into the FIB
    trie and use prebuilt routes cached there.

    No more garbage collection, no more rDOS attacks on the routing
    cache.  Instead we now get predictable and consistent performance,
    no matter what the pattern of traffic we service.

    This has been almost 2 years in the making.  Special thanks to
    Julian Anastasov, Eric Dumazet, Steffen Klassert, and others who
    have helped along the way.

    I'm sure that with a change of this magnitude there will be some
    kind of fallout, but such things ought the be simple to fix at this
    point.  Luckily I'm not European so I'll be around all of August to
    fix things :-)

    The major stages of this work here are each fronted by a forced
    merge commit whose commit message contains a top-level description
    of the motivations and implementation issues.

 2) Pre-demux of established ipv4 TCP sockets, saves a route demux on
    input.

 3) TCP SYN/ACK performance tweaks from Eric Dumazet.

 4) Add namespace support for netfilter L4 conntrack helpers, from Gao
    Feng.

 5) Add config mechanism for Energy Efficient Ethernet to ethtool, from
    Yuval Mintz.

 6) Remove quadratic behavior from /proc/net/unix, from Eric Dumazet.

 7) Support for connection tracker helpers in userspace, from Pablo
    Neira Ayuso.

 8) Allow userspace driven TX load balancing functions in TEAM driver,
    from Jiri Pirko.

 9) Kill off NLMSG_PUT and RTA_PUT macros, more gross stuff with
    embedded gotos.

10) TCP Small Queues, essentially minimize the amount of TCP data queued
    up in the packet scheduler layer.  Whereas the existing BQL (Byte
    Queue Limits) limits the pkt_sched --> netdevice queuing levels,
    this controls the TCP --> pkt_sched queueing levels.

    From Eric Dumazet.

11) Reduce the number of get_page/put_page ops done on SKB fragments,
    from Alexander Duyck.

12) Implement protection against blind resets in TCP (RFC 5961), from
    Eric Dumazet.

13) Support the client side of TCP Fast Open, basically the ability to
    send data in the SYN exchange, from Yuchung Cheng.

    Basically, the sender queues up data with a sendmsg() call using
    MSG_FASTOPEN, then they do the connect() which emits the queued up
    fastopen data.

14) Avoid all the problems we get into in TCP when timers or PMTU events
    hit a locked socket.  The TCP Small Queues changes added a
    tcp_release_cb() that allows us to queue work up to the
    release_sock() caller, and that's what we use here too.  From Eric
    Dumazet.

15) Zero copy on TX support for TUN driver, from Michael S. Tsirkin.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1870 commits)
  genetlink: define lockdep_genl_is_held() when CONFIG_LOCKDEP
  r8169: revert "add byte queue limit support".
  ipv4: Change rt->rt_iif encoding.
  net: Make skb->skb_iif always track skb->dev
  ipv4: Prepare for change of rt->rt_iif encoding.
  ipv4: Remove all RTCF_DIRECTSRC handliing.
  ipv4: Really ignore ICMP address requests/replies.
  decnet: Don't set RTCF_DIRECTSRC.
  net/ipv4/ip_vti.c: Fix __rcu warnings detected by sparse.
  ipv4: Remove redundant assignment
  rds: set correct msg_namelen
  openvswitch: potential NULL deref in sample()
  tcp: dont drop MTU reduction indications
  bnx2x: Add new 57840 device IDs
  tcp: avoid oops in tcp_metrics and reset tcpm_stamp
  niu: Change niu_rbr_fill() to use unlikely() to check niu_rbr_add_page() return value
  niu: Fix to check for dma mapping errors.
  net: Fix references to out-of-scope variables in put_cmsg_compat()
  net: ethernet: davinci_emac: add pm_runtime support
  net: ethernet: davinci_emac: Remove unnecessary #include
  ...

Pull target updates from Nicholas Bellinger:
 "There have been lots of work in a number of areas this past round.
  The highlights include:

   - Break out target_core_cdb.c emulation into SPC/SBC ops (hch)
   - Add a parse_cdb method to target backend drivers (hch)
   - Move sync_cache + write_same + unmap into spc_ops (hch)
   - Use target_execute_cmd for WRITEs in iscsi_target + srpt (hch)
   - Offload WRITE I/O backend submission in tcm_qla2xxx + tcm_fc (hch +
     nab)
   - Refactor core_update_device_list_for_node() into enable/disable
     funcs (agrover)
   - Replace the TCM processing thread with a TMR work queue (hch)
   - Fix regression in transport_add_device_to_core_hba from TMR
     conversion (DanC)
   - Remove racy, now-redundant check of sess_tearing_down with qla2xxx
     (roland)
   - Add range checking, fix reading of data len + possible underflow in
     UNMAP (roland)
   - Allow for target_submit_cmd() returning errors + convert fabrics
     (roland + nab)
   - Drop bogus struct file usage for iSCSI/SCTP (viro)"

* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/nab/target-pending: (54 commits)
  iscsi-target: Drop bogus struct file usage for iSCSI/SCTP
  target: NULL dereference on error path
  target: Allow for target_submit_cmd() returning errors
  target: Check number of unmap descriptors against our limit
  target: Fix possible integer underflow in UNMAP emulation
  target: Fix reading of data length fields for UNMAP commands
  target: Add range checking to UNMAP emulation
  target: Add generation of LOGICAL BLOCK ADDRESS OUT OF RANGE
  target: Make unnecessarily global se_dev_align_max_sectors() static
  target: Remove se_session.sess_wait_list
  qla2xxx: Remove racy, now-redundant check of sess_tearing_down
  target: Check sess_tearing_down in target_get_sess_cmd()
  sbp-target: Consolidate duplicated error path code in sbp_handle_command()
  target: Un-export target_get_sess_cmd()
  qla2xxx: Get rid of redundant qla_tgt_sess.tearing_down
  target: Make core_disable_device_list_for_node use pre-refactoring lock ordering
  target: refactor core_update_device_list_for_node()
  target: Eliminate else using boolean logic
  target: Misc retval cleanups
  target: Remove hba param from core_dev_add_lun
  ...