]> git.kernelconcepts.de Git - karo-tx-linux.git/log
karo-tx-linux.git
8 years agodp83640: Prune rx timestamp list before reading from it
Stefan Sørensen [Tue, 3 Nov 2015 08:34:06 +0000 (09:34 +0100)]
dp83640: Prune rx timestamp list before reading from it

The list of rx timestamps are currently only pruned of old entries when a
new entry is inserted. If no new entries are added, old timestamps may
survive beyond their lifetime, possible causing them to be attached to
packets with the same sequence number after a rollover.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodp83640: Delay scheduled work.
Stefan Sørensen [Tue, 3 Nov 2015 08:34:05 +0000 (09:34 +0100)]
dp83640: Delay scheduled work.

Currently rx_timestamp_work reschedules itself as a regular workqueue item,
effectively causing it run constantly as long as there are packets left in
the queue. Fix by using delayed workqueue items, limiting it to run only
every two jiffies.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodp83640: Include hash in timestamp/packet matching
Stefan Sørensen [Tue, 3 Nov 2015 08:34:04 +0000 (09:34 +0100)]
dp83640: Include hash in timestamp/packet matching

Only using the message type and sequence id for matching timestamps
with packets is error prone, as multiple clients may very well be
sending packets with the same messagetype and timestamp at the same
time. Fix by extending the check to include the hash of bytes 20-29
(source id in PTPv2) that is provided with the timestamps.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlx5-fixes'
David S. Miller [Tue, 3 Nov 2015 15:41:51 +0000 (10:41 -0500)]
Merge branch 'mlx5-fixes'

Or Gerlitz says:

====================
Mellanox mlx5e driver update, Nov 3 2015

This series contains bunch of small fixes to the mlx5e driver from Achiad.

Changes from V0:
  - removed the driver patch that dealt with IRQ affinity changes during
    NAPI poll, as this is a generic problem which needs generic solution.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Fix LSO vlan insertion
Achiad Shochat [Tue, 3 Nov 2015 06:07:24 +0000 (08:07 +0200)]
net/mlx5e: Fix LSO vlan insertion

Consider vlan insertion impact on headers copy size also for LSO
packets.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Re-eanble client vlan TX acceleration
Achiad Shochat [Tue, 3 Nov 2015 06:07:23 +0000 (08:07 +0200)]
net/mlx5e: Re-eanble client vlan TX acceleration

This reverts commit cd58c714acb9 "net/mlx5e: Disable client vlan TX acceleration".

Bring back client vlan insertion offload, the original
performance issue was found and fixed in the next patch.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Return error in case mlx5e_set_features() fails
Achiad Shochat [Tue, 3 Nov 2015 06:07:22 +0000 (08:07 +0200)]
net/mlx5e: Return error in case mlx5e_set_features() fails

In case mlx5e_set_features() fails, return the failure status rather
than 0.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Don't allow more than max supported channels
Achiad Shochat [Tue, 3 Nov 2015 06:07:21 +0000 (08:07 +0200)]
net/mlx5e: Don't allow more than max supported channels

Consider MLX5E_MAX_NUM_CHANNELS @ethtool set/get_channels

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5_core: Use the the real irqn in eq->irqn
Achiad Shochat [Tue, 3 Nov 2015 06:07:20 +0000 (08:07 +0200)]
net/mlx5_core: Use the the real irqn in eq->irqn

Instead of storing the msix array index in eq->irqn (vecidx),
store the real irq number.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Wait for RX buffers initialization in a more proper manner
Achiad Shochat [Tue, 3 Nov 2015 06:07:19 +0000 (08:07 +0200)]
net/mlx5e: Wait for RX buffers initialization in a more proper manner

Use jiffies rather than wait loop with msleep().

The wait loop didn't take into consideration time when the
process was not executing.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/mlx5e: Avoid NULL pointer access in case of configuration failure
Achiad Shochat [Tue, 3 Nov 2015 06:07:18 +0000 (08:07 +0200)]
net/mlx5e: Avoid NULL pointer access in case of configuration failure

In case a configuration operation that involves closing and re-opening
resources (e.g RX/TX queue size change) fails at the re-opening stage
these resources will remain closed.
So when executing (following) configuration operations (e.g ifconfig
down) we cannot assume that these resources are available.

Signed-off-by: Achiad Shochat <achiad@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet/core: generic support for disabling netdev features down stack
Jarod Wilson [Tue, 3 Nov 2015 02:55:59 +0000 (21:55 -0500)]
net/core: generic support for disabling netdev features down stack

There are some netdev features, which when disabled on an upper device,
such as a bonding master or a bridge, must be disabled and cannot be
re-enabled on underlying devices.

This is a rework of an earlier more heavy-handed appraoch, which simply
disables and prevents re-enabling of netdev features listed in a new
define in include/net/netdev_features.h, NETIF_F_UPPER_DISABLES. Any upper
device that disables a flag in that feature mask, the disabling will
propagate down the stack, and any lower device that has any upper device
with one of those flags disabled should not be able to enable said flag.

Initially, only LRO is included for proof of concept, and because this
code effectively does the same thing as dev_disable_lro(), though it will
also activate from the ethtool path, which was one of the goals here.

[root@dell-per730-01 ~]# ethtool -k bond0 |grep large
large-receive-offload: on
[root@dell-per730-01 ~]# ethtool -k p5p1 |grep large
large-receive-offload: on
[root@dell-per730-01 ~]# ethtool -K bond0 lro off
[root@dell-per730-01 ~]# ethtool -k bond0 |grep large
large-receive-offload: off
[root@dell-per730-01 ~]# ethtool -k p5p1 |grep large
large-receive-offload: off

dmesg dump:

[ 1033.277986] bond0: Disabling feature 0x0000000000008000 on lower dev p5p2.
[ 1034.067949] bnx2x 0000:06:00.1 p5p2: using MSI-X  IRQs: sp 74  fp[0] 76 ... fp[7] 83
[ 1034.753612] bond0: Disabling feature 0x0000000000008000 on lower dev p5p1.
[ 1035.591019] bnx2x 0000:06:00.0 p5p1: using MSI-X  IRQs: sp 62  fp[0] 64 ... fp[7] 71

This has been successfully tested with bnx2x, qlcnic and netxen network
cards as slaves in a bond interface. Turning LRO on or off on the master
also turns it on or off on each of the slaves, new slaves are added with
LRO in the same state as the master, and LRO can't be toggled on the
slaves.

Also, this should largely remove the need for dev_disable_lro(), and most,
if not all, of its call sites can be replaced by simply making sure
NETIF_F_LRO isn't included in the relevant device's feature flags.

Note that this patch is driven by bug reports from users saying it was
confusing that bonds and slaves had different settings for the same
features, and while it won't be 100% in sync if a lower device doesn't
support a feature like LRO, I think this is a good step in the right
direction.

CC: "David S. Miller" <davem@davemloft.net>
CC: Eric Dumazet <edumazet@google.com>
CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <gospo@cumulusnetworks.com>
CC: Jiri Pirko <jiri@resnulli.us>
CC: Nikolay Aleksandrov <razor@blackwall.org>
CC: Michal Kubecek <mkubecek@suse.cz>
CC: Alexander Duyck <alexander.duyck@gmail.com>
CC: netdev@vger.kernel.org
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agosh_eth: fix typo in RX descriptor bit name
Sergei Shtylyov [Mon, 2 Nov 2015 22:28:07 +0000 (01:28 +0300)]
sh_eth: fix typo in RX descriptor bit name

The correct name of the RX descriptor 0 bit 30 is RDLE (receive descriptor
list end),  not  RDEL.

Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bonding-actor-updates'
David S. Miller [Tue, 3 Nov 2015 03:52:24 +0000 (22:52 -0500)]
Merge branch 'bonding-actor-updates'

Mahesh Bandewar says:

====================
re-org actor admin/oper key updates

I was observing machines entering into weird LACP state when the
partner is in passive mode. This issue is not because of the partners
in passive state but probably because of some operational key update
which is pushing the state-machine is that weird state. This was
happening randomly on about 1% of the machine (when the sample size
is a large set of machines with a variety of NICs/ports bonded).

In this patch-series I'm attempting to unify the logic of actor-key
/ operational-key changes to one place to avoid possible errors in
update. Also this eliminates the need for the event-handler to decide
if the key needs update.

After this patch-set none of the machines (from same sample set) were
exhibiting LACP-weirdness that was observed earlier.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobonding: simplify / unify event handling code for 3ad mode.
Mahesh Bandewar [Sat, 31 Oct 2015 19:45:11 +0000 (12:45 -0700)]
bonding: simplify / unify event handling code for 3ad mode.

Old logic of updating state-machine is not required since
ad_update_actor_keys() does it implicitly. The only loss is
the notification differentiation between speed vs. duplex
change. Now only one unified notification is printed.

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobonding: unify all places where actor-oper key needs to be updated.
Mahesh Bandewar [Sat, 31 Oct 2015 19:45:06 +0000 (12:45 -0700)]
bonding: unify all places where actor-oper key needs to be updated.

actor_admin, and actor_oper key is changed at multiple locations in
the code. This patch brings all those updates into one location in
an attempt to avoid possible inconsistent updates causing LACP state
machine to go in weird state.

The unified place is ad_update_actor_key() with simple state-machine
logic -
  (a) If port is "duplex" then only it can participate in LACP
  (b) Speed change reinitializes the LACP state-machine.

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobonding: Simplify __get_duplex function.
Mahesh Bandewar [Sat, 31 Oct 2015 19:45:00 +0000 (12:45 -0700)]
bonding: Simplify __get_duplex function.

Eliminate 'else' clause by simply initializing variable

Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bpf-persistent'
David S. Miller [Tue, 3 Nov 2015 03:48:39 +0000 (22:48 -0500)]
Merge branch 'bpf-persistent'

Daniel Borkmann says:

====================
BPF updates

This set adds support for persistent maps/progs. Please see
individual patches for further details. A man-page update
to bpf(2) will be sent later on, also a iproute2 patch for
support in tc.

v1 -> v2:
  - Reworked most of patch 4 and 5
  - Rebased to latest net-next
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: add sample usages for persistent maps/progs
Daniel Borkmann [Thu, 29 Oct 2015 13:58:10 +0000 (14:58 +0100)]
bpf: add sample usages for persistent maps/progs

This patch adds a couple of stand-alone examples on how BPF_OBJ_PIN
and BPF_OBJ_GET commands can be used.

Example with maps:

  # ./fds_example -F /sys/fs/bpf/m -P -m -k 1 -v 42
  bpf: map fd:3 (Success)
  bpf: pin ret:(0,Success)
  bpf: fd:3 u->(1:42) ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/m -G -m -k 1
  bpf: get fd:3 (Success)
  bpf: fd:3 l->(1):42 ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/m -G -m -k 1 -v 24
  bpf: get fd:3 (Success)
  bpf: fd:3 u->(1:24) ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/m -G -m -k 1
  bpf: get fd:3 (Success)
  bpf: fd:3 l->(1):24 ret:(0,Success)

  # ./fds_example -F /sys/fs/bpf/m2 -P -m
  bpf: map fd:3 (Success)
  bpf: pin ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/m2 -G -m -k 1
  bpf: get fd:3 (Success)
  bpf: fd:3 l->(1):0 ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/m2 -G -m
  bpf: get fd:3 (Success)

Example with progs:

  # ./fds_example -F /sys/fs/bpf/p -P -p
  bpf: prog fd:3 (Success)
  bpf: pin ret:(0,Success)
  bpf sock:4 <- fd:3 attached ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/p -G -p
  bpf: get fd:3 (Success)
  bpf: sock:4 <- fd:3 attached ret:(0,Success)

  # ./fds_example -F /sys/fs/bpf/p2 -P -p -o ./sockex1_kern.o
  bpf: prog fd:5 (Success)
  bpf: pin ret:(0,Success)
  bpf: sock:3 <- fd:5 attached ret:(0,Success)
  # ./fds_example -F /sys/fs/bpf/p2 -G -p
  bpf: get fd:3 (Success)
  bpf: sock:4 <- fd:3 attached ret:(0,Success)

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: add support for persistent maps/progs
Daniel Borkmann [Thu, 29 Oct 2015 13:58:09 +0000 (14:58 +0100)]
bpf: add support for persistent maps/progs

This work adds support for "persistent" eBPF maps/programs. The term
"persistent" is to be understood that maps/programs have a facility
that lets them survive process termination. This is desired by various
eBPF subsystem users.

Just to name one example: tc classifier/action. Whenever tc parses
the ELF object, extracts and loads maps/progs into the kernel, these
file descriptors will be out of reach after the tc instance exits.
So a subsequent tc invocation won't be able to access/relocate on this
resource, and therefore maps cannot easily be shared, f.e. between the
ingress and egress networking data path.

The current workaround is that Unix domain sockets (UDS) need to be
instrumented in order to pass the created eBPF map/program file
descriptors to a third party management daemon through UDS' socket
passing facility. This makes it a bit complicated to deploy shared
eBPF maps or programs (programs f.e. for tail calls) among various
processes.

We've been brainstorming on how we could tackle this issue and various
approches have been tried out so far, which can be read up further in
the below reference.

The architecture we eventually ended up with is a minimal file system
that can hold map/prog objects. The file system is a per mount namespace
singleton, and the default mount point is /sys/fs/bpf/. Any subsequent
mounts within a given namespace will point to the same instance. The
file system allows for creating a user-defined directory structure.
The objects for maps/progs are created/fetched through bpf(2) with
two new commands (BPF_OBJ_PIN/BPF_OBJ_GET). I.e. a bpf file descriptor
along with a pathname is being passed to bpf(2) that in turn creates
(we call it eBPF object pinning) the file system nodes. Only the pathname
is being passed to bpf(2) for getting a new BPF file descriptor to an
existing node. The user can use that to access maps and progs later on,
through bpf(2). Removal of file system nodes is being managed through
normal VFS functions such as unlink(2), etc. The file system code is
kept to a very minimum and can be further extended later on.

The next step I'm working on is to add dump eBPF map/prog commands
to bpf(2), so that a specification from a given file descriptor can
be retrieved. This can be used by things like CRIU but also applications
can inspect the meta data after calling BPF_OBJ_GET.

Big thanks also to Alexei and Hannes who significantly contributed
in the design discussion that eventually let us end up with this
architecture here.

Reference: https://lkml.org/lkml/2015/10/15/925
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: consolidate bpf_prog_put{, _rcu} dismantle paths
Daniel Borkmann [Thu, 29 Oct 2015 13:58:08 +0000 (14:58 +0100)]
bpf: consolidate bpf_prog_put{, _rcu} dismantle paths

We currently have duplicated cleanup code in bpf_prog_put() and
bpf_prog_put_rcu() cleanup paths. Back then we decided that it was
not worth it to make it a common helper called by both, but with
the recent addition of resource charging, we could have avoided
the fix in commit ac00737f4e81 ("bpf: Need to call bpf_prog_uncharge_memlock
from bpf_prog_put") if we would have had only a single, common path.
We can simplify it further by assigning aux->prog only once during
allocation time.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: align and clean bpf_{map,prog}_get helpers
Daniel Borkmann [Thu, 29 Oct 2015 13:58:07 +0000 (14:58 +0100)]
bpf: align and clean bpf_{map,prog}_get helpers

Add a bpf_map_get() function that we're going to use later on and
align/clean the remaining helpers a bit so that we have them a bit
more consistent:

  - __bpf_map_get() and __bpf_prog_get() that both work on the fd
    struct, check whether the descriptor is eBPF and return the
    pointer to the map/prog stored in the private data.

    Also, we can return f.file->private_data directly, the function
    signature is enough of a documentation already.

  - bpf_map_get() and bpf_prog_get() that both work on u32 user fd,
    call their respective __bpf_map_get()/__bpf_prog_get() variants,
    and take a reference.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: abstract anon_inode_getfd invocations
Daniel Borkmann [Thu, 29 Oct 2015 13:58:06 +0000 (14:58 +0100)]
bpf: abstract anon_inode_getfd invocations

Since we're going to use anon_inode_getfd() invocations in more than just
the current places, make a helper function for both, so that we only need
to pass a map/prog pointer to the helper itself in order to get a fd. The
new helpers are called bpf_map_new_fd() and bpf_prog_new_fd().

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: fix percpu memory leaks
Eric Dumazet [Mon, 2 Nov 2015 17:03:11 +0000 (09:03 -0800)]
net: fix percpu memory leaks

This patch fixes following problems :

1) percpu_counter_init() can return an error, therefore
  init_frag_mem_limit() must propagate this error so that
  inet_frags_init_net() can do the same up to its callers.

2) If ip[46]_frags_ns_ctl_register() fail, we must unwind
   properly and free the percpu_counter.

Without this fix, we leave freed object in percpu_counters
global list (if CONFIG_HOTPLUG_CPU) leading to crashes.

This bug was detected by KASAN and syzkaller tool
(http://github.com/google/syzkaller)

Fixes: 6d7b857d541e ("net: use lib/percpu_counter API for fragmentation mem accounting")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoravb: use pdev rather than ndev for error messages
Simon Horman [Mon, 2 Nov 2015 01:40:17 +0000 (10:40 +0900)]
ravb: use pdev rather than ndev for error messages

This corrects what appear to be typos, making the code consistent with
itself, and allowing meaningful prefixes to be displayed with the errors in
question.

Before:
 (null): failed to initialize MDIO
 (null): Cannot allocate desc base address table (size 176 bytes)

After:
ravb e6800000.ethernet: failed to initialize MDIO
ravb e6800000.ethernet: Cannot allocate desc base address table (size 176 bytes)

Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: make skb_set_owner_w() more robust
Eric Dumazet [Sun, 1 Nov 2015 23:36:55 +0000 (15:36 -0800)]
net: make skb_set_owner_w() more robust

skb_set_owner_w() is called from various places that assume
skb->sk always point to a full blown socket (as it changes
sk->sk_wmem_alloc)

We'd like to attach skb to request sockets, and in the future
to timewait sockets as well. For these kind of pseudo sockets,
we need to take a traditional refcount and use sock_edemux()
as the destructor.

It is now time to un-inline skb_set_owner_w(), being too big.

Fixes: ca6fb0651883 ("tcp: attach SYNACK messages to request sockets instead of listener")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Bisected-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobridge: vlan: Use rcu_dereference instead of rtnl_dereference
Ido Schimmel [Sun, 1 Nov 2015 16:31:45 +0000 (18:31 +0200)]
bridge: vlan: Use rcu_dereference instead of rtnl_dereference

br_should_learn() is protected by RCU and not by RTNL, so use correct
flavor of nbp_vlan_group().

Fixes: 907b1e6e83ed ("bridge: vlan: use proper rcu for the vlgrp
member")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: lookup switch name
Vivien Didelot [Fri, 30 Oct 2015 23:39:48 +0000 (19:39 -0400)]
net: dsa: mv88e6xxx: lookup switch name

All the mv88e6xxx drivers use the exact same code in their probe
function to lookup the switch name given its ID. Thus introduce a
mv88e6xxx_switch_id structure and a mv88e6xxx_lookup_name function in
the common mv88e6xxx code.

In the meantime make __mv88e6xxx_reg_{read,write} static since we do not
need to expose these low-level r/w routines anymore.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: mv88e6xxx: assert SMI lock
Vivien Didelot [Fri, 30 Oct 2015 22:56:45 +0000 (18:56 -0400)]
net: dsa: mv88e6xxx: assert SMI lock

It's easy to forget to lock the smi_mutex before calling the low-level
_mv88e6xxx_reg_{read,write}, so add a assert_smi_lock function in them.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Acked-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobpf: convert hashtab lock to raw lock
Yang Shi [Fri, 30 Oct 2015 22:16:26 +0000 (15:16 -0700)]
bpf: convert hashtab lock to raw lock

When running bpf samples on rt kernel, it reports the below warning:

BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:917
in_atomic(): 1, irqs_disabled(): 128, pid: 477, name: ping
Preemption disabled at:[<ffff80000017db58>] kprobe_perf_func+0x30/0x228

CPU: 3 PID: 477 Comm: ping Not tainted 4.1.10-rt8 #4
Hardware name: Freescale Layerscape 2085a RDB Board (DT)
Call trace:
[<ffff80000008a5b0>] dump_backtrace+0x0/0x128
[<ffff80000008a6f8>] show_stack+0x20/0x30
[<ffff8000007da90c>] dump_stack+0x7c/0xa0
[<ffff8000000e4830>] ___might_sleep+0x188/0x1a0
[<ffff8000007e2200>] rt_spin_lock+0x28/0x40
[<ffff80000018bf9c>] htab_map_update_elem+0x124/0x320
[<ffff80000018c718>] bpf_map_update_elem+0x40/0x58
[<ffff800000187658>] __bpf_prog_run+0xd48/0x1640
[<ffff80000017ca6c>] trace_call_bpf+0x8c/0x100
[<ffff80000017db58>] kprobe_perf_func+0x30/0x228
[<ffff80000017dd84>] kprobe_dispatcher+0x34/0x58
[<ffff8000007e399c>] kprobe_handler+0x114/0x250
[<ffff8000007e3bf4>] kprobe_breakpoint_handler+0x1c/0x30
[<ffff800000085b80>] brk_handler+0x88/0x98
[<ffff8000000822f0>] do_debug_exception+0x50/0xb8
Exception stack(0xffff808349687460 to 0xffff808349687580)
7460: 4ca2b600 ffff8083 4a3a7000 ffff8083 49687620 ffff8083 0069c5f8 ffff8000
7480: 00000001 00000000 007e0628 ffff8000 496874b0 ffff8083 007e1de8 ffff8000
74a0: 496874d0 ffff8083 0008e04c ffff8000 00000001 00000000 4ca2b600 ffff8083
74c0: 00ba2e80 ffff8000 49687528 ffff8083 49687510 ffff8083 000e5c70 ffff8000
74e0: 00c22348 ffff8000 00000000 ffff8083 49687510 ffff8083 000e5c74 ffff8000
7500: 4ca2b600 ffff8083 49401800 ffff8083 00000001 00000000 00000000 00000000
7520: 496874d0 ffff8083 00000000 00000000 00000000 00000000 00000000 00000000
7540: 2f2e2d2c 33323130 00000000 00000000 4c944500 ffff8083 00000000 00000000
7560: 00000000 00000000 008751e0 ffff8000 00000001 00000000 124e2d1d 00107b77

Convert hashtab lock to raw lock to avoid such warning.

Signed-off-by: Yang Shi <yang.shi@linaro.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'bridge_vlan_fixes'
David S. Miller [Mon, 2 Nov 2015 20:40:11 +0000 (15:40 -0500)]
Merge branch 'bridge_vlan_fixes'

Nikolay Aleksandrov says:

====================
bridge: vlan: failure path and comment fixes

This is a set from Ido which takes care of one failure path error in
nbp_vlan_init (patch 1) and a few comment errors (patch 2).
I must admit I didn't expect the port init continues after a vlan init
failure but should've checked to make sure. Thanks to Ido for catching
these!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobridge: vlan: Use correct flag name in comment
Ido Schimmel [Fri, 30 Oct 2015 16:46:20 +0000 (17:46 +0100)]
bridge: vlan: Use correct flag name in comment

The flag used to indicate if a VLAN should be used for filtering - as
opposed to context only - on the bridge itself (e.g. br0) is called
'brentry' and not 'brvlan'.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobridge: vlan: Prevent possible use-after-free
Ido Schimmel [Fri, 30 Oct 2015 16:46:19 +0000 (17:46 +0100)]
bridge: vlan: Prevent possible use-after-free

When adding a port to a bridge we initialize VLAN filtering on it. We do
not bail out in case an error occurred in nbp_vlan_init, as it can be
used as a non VLAN filtering bridge.

However, if VLAN filtering is required and an error occurred in
nbp_vlan_init, we should set vlgrp to NULL, so that VLAN filtering
functions (e.g. br_vlan_find, br_get_pvid) will know the struct is
invalid and will not try to access it.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agotcp/dccp: fix ireq->pktopts race
Eric Dumazet [Fri, 30 Oct 2015 16:46:12 +0000 (09:46 -0700)]
tcp/dccp: fix ireq->pktopts race

IPv6 request sockets store a pointer to skb containing the SYN packet
to be able to transfer it to full blown socket when 3WHS is done
(ireq->pktopts -> np->pktoptions)

As explained in commit 5e0724d027f0 ("tcp/dccp: fix hashdance race for
passive sessions"), we must transfer the skb only if we won the
hashdance race, if multiple cpus receive the 'ack' packet completing
3WHS at the same time.

Fixes: e994b2f0fb92 ("tcp: do not lock listener to process SYN packets")
Fixes: 079096f103fa ("tcp/dccp: install syn_recv requests into ehash table")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRDS: convert bind hash table to re-sizable hashtable
santosh.shilimkar@oracle.com [Fri, 30 Oct 2015 15:49:10 +0000 (08:49 -0700)]
RDS: convert bind hash table to re-sizable hashtable

To further improve the RDS connection scalabilty on massive systems
where number of sockets grows into tens of thousands  of sockets, there
is a need of larger bind hashtable. Pre-allocated 8K or 16K table is
not very flexible in terms of memory utilisation. The rhashtable
infrastructure gives us the flexibility to grow the hashtbable based
on use and also comes up with inbuilt efficient bucket(chain) handling.

Reviewed-by: David Miller <davem@davemloft.net>
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: rds: changing the return type from int to void
Saurabh Sengar [Fri, 30 Oct 2015 14:16:44 +0000 (19:46 +0530)]
net: rds: changing the return type from int to void

as result of function rds_iw_flush_mr_pool is nowhere checked,
changing its return type from int to void.
also removing the unused variable rc as there is nothing to return

Signed-off-by: Saurabh Sengar <saurabh.truth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'encx24j600-fixes'
David S. Miller [Mon, 2 Nov 2015 20:28:57 +0000 (15:28 -0500)]
Merge branch 'encx24j600-fixes'

Javier Martinez Canillas says:

====================
net: encx24j600: Fix SPI driver module autoload

Recently I've been trying to fix module autoloading for all SPI drivers and
found that the encx24j600 driver does not fill module alias information due
missing a MODULE_DEVICE_TABLE() so module autload won't work and the driver
Kconfig symbol is tristate which means the driver can be built as a module.

But also the SPI id table is not correctly defined so this series fixes both
issues.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: encx24j600: Export missing SPI module alias information
Javier Martinez Canillas [Fri, 30 Oct 2015 12:49:18 +0000 (13:49 +0100)]
net: encx24j600: Export missing SPI module alias information

The driver Kconfig symbol is tristate which means that it can be built as
a module but the module alias information is not added to the module info
so module autoload won't work since user-space won't have the information.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: encx24j600: Fix SPI id table definition
Javier Martinez Canillas [Fri, 30 Oct 2015 12:49:17 +0000 (13:49 +0100)]
net: encx24j600: Fix SPI id table definition

A driver's SPI id table is expected to be an array of struct spi_device_id
that ends with a zero-initialized sentinel entry. But this driver defines
the table as a single struct spi_device_id and sets .id_table to a pointer
to this struct.

But spi_match_id() has a loop that iterates while the struct spi_device_id
.name[0] is not NULL, so not having a sentinel can cause a NULL pointer
deference error.

This patch defines the SPI id table correctly as all other SPI drivers do.

Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoenic: assign affinity hint to interrupts
Govindarajulu Varadarajan [Fri, 30 Oct 2015 11:22:51 +0000 (16:52 +0530)]
enic: assign affinity hint to interrupts

The affinity hint is used by the user space daemon, irqbalancer, to
indicate a preferred CPU mask for irqs. This patch sets the irq affinity
hint to local numa core first, when exausted we try non-local numa cores.

Also set tx xps cpus mask bassed on affinity hint.

v2: remove the global affinity policy.

Signed-off-by: Govindarajulu Varadarajan <_govind@gmx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: use l4 hash for locally generated multipath flows
Paolo Abeni [Thu, 29 Oct 2015 21:20:40 +0000 (22:20 +0100)]
ipv4: use l4 hash for locally generated multipath flows

This patch changes how the multipath hash is computed for locally
generated flows: now the hash comprises l4 information.

This allows better utilization of the available paths when the existing
flows have the same source IP and the same destination IP: with l3 hash,
even when multiple connections are in place simultaneously, a single path
will be used, while with l4 hash we can use all the available paths.

v2 changes:
- use get_hash_from_flowi4() instead of implementing just another l4 hash
  function

Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoUse 64-bit timekeeping
Tina Ruchandani [Fri, 30 Oct 2015 08:24:56 +0000 (01:24 -0700)]
Use 64-bit timekeeping

This patch changes the use of struct timespec in
dccp_probe to use struct timespec64 instead. timespec uses a 32-bit
seconds field which will overflow in the year 2038 and beyond. timespec64
uses a 64-bit seconds field. Note that the correctness of the code isn't
changed, since the original code only uses the timestamps to compute a
small elapsed interval. This patch is part of a larger attempt to remove
instances of 32-bit timekeeping structures (timespec, timeval, time_t)
from the kernel so it is easier to identify where the real 2038 issues
are.

Signed-off-by: Tina Ruchandani <ruchandani.tina@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: hisilicon: Remove .owner assignment from platform_driver
huangdaode [Fri, 30 Oct 2015 02:50:54 +0000 (10:50 +0800)]
net: hisilicon: Remove .owner assignment from platform_driver

platform_driver doesn't need to set .owner, because
platform_driver_register() will set it.

Signed-off-by: huangdaode <huangdaode@hisilicon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agonet: dsa: use switchdev obj for VLAN add/del ops
Vivien Didelot [Sun, 1 Nov 2015 17:33:55 +0000 (12:33 -0500)]
net: dsa: use switchdev obj for VLAN add/del ops

Simplify DSA by pushing the switchdev objects for VLAN add and delete
operations down to its drivers. Currently only mv88e6xxx is affected.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoVSOCK: define VSOCK_SS_LISTEN once only
Stefan Hajnoczi [Thu, 29 Oct 2015 11:57:42 +0000 (11:57 +0000)]
VSOCK: define VSOCK_SS_LISTEN once only

The SS_LISTEN socket state is defined by both af_vsock.c and
vmci_transport.c.  This is risky since the value could be changed in one
file and the other would be out of sync.

Rename from SS_LISTEN to VSOCK_SS_LISTEN since the constant is not part
of enum socket_state (SS_CONNECTED, ...).  This way it is clear that the
constant is vsock-specific.

The big text reflow in af_vsock.c was necessary to keep to the maximum
line length.  Text is unchanged except for s/SS_LISTEN/VSOCK_SS_LISTEN/.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'csum_partial_frags'
David S. Miller [Sun, 1 Nov 2015 17:01:37 +0000 (12:01 -0500)]
Merge branch 'csum_partial_frags'

Hannes Frederic Sowa says:

====================
net: clean up interactions of CHECKSUM_PARTIAL and fragmentation

This series fixes wrong checksums on the wire for IPv4 and IPv6. Large
send buffers and especially NFS lead to wrong checksums in both IPv4
and IPv6.

CHECKSUM_PARTIAL skbs should not receive the respective fragmentations
functions, so we add WARN_ON_ONCE to those functions to fix up those as
soon as they get reported.

Changelog:
v2: added v4 checks
v3: removed WARN_ON_ONCES (advice by Tom Herbert)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: add defensive check for CHECKSUM_PARTIAL skbs in ip_fragment
Hannes Frederic Sowa [Tue, 27 Oct 2015 21:40:42 +0000 (22:40 +0100)]
ipv6: add defensive check for CHECKSUM_PARTIAL skbs in ip_fragment

CHECKSUM_PARTIAL skbs should never arrive in ip_fragment. If we get one
of those warn about them once and handle them gracefully by recalculating
the checksum.

Fixes: commit 32dce968dd987 ("ipv6: Allow for partial checksums on non-ufo packets")
See-also: commit 72e843bb09d45 ("ipv6: ip6_fragment() should check CHECKSUM_PARTIAL")
Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: no CHECKSUM_PARTIAL on MSG_MORE corked sockets
Hannes Frederic Sowa [Tue, 27 Oct 2015 21:40:41 +0000 (22:40 +0100)]
ipv6: no CHECKSUM_PARTIAL on MSG_MORE corked sockets

We cannot reliable calculate packet size on MSG_MORE corked sockets
and thus cannot decide if they are going to be fragmented later on,
so better not use CHECKSUM_PARTIAL in the first place.

The IPv6 code also intended to protect and not use CHECKSUM_PARTIAL in
the existence of IPv6 extension headers, but the condition was wrong. Fix
it up, too. Also the condition to check whether the packet fits into
one fragment was wrong and has been corrected.

Fixes: commit 32dce968dd987 ("ipv6: Allow for partial checksums on non-ufo packets")
See-also: commit 72e843bb09d45 ("ipv6: ip6_fragment() should check CHECKSUM_PARTIAL")
Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: add defensive check for CHECKSUM_PARTIAL skbs in ip_fragment
Hannes Frederic Sowa [Tue, 27 Oct 2015 21:40:40 +0000 (22:40 +0100)]
ipv4: add defensive check for CHECKSUM_PARTIAL skbs in ip_fragment

CHECKSUM_PARTIAL skbs should never arrive in ip_fragment. If we get one
of those warn about them once and handle them gracefully by recalculating
the checksum.

Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv4: no CHECKSUM_PARTIAL on MSG_MORE corked sockets
Hannes Frederic Sowa [Tue, 27 Oct 2015 21:40:39 +0000 (22:40 +0100)]
ipv4: no CHECKSUM_PARTIAL on MSG_MORE corked sockets

We cannot reliable calculate packet size on MSG_MORE corked sockets
and thus cannot decide if they are going to be fragmented later on,
so better not use CHECKSUM_PARTIAL in the first place.

Cc: Eric Dumazet <edumazet@google.com>
Cc: Vlad Yasevich <vyasevich@gmail.com>
Cc: Benjamin Coddington <bcodding@redhat.com>
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Sun, 1 Nov 2015 04:15:30 +0000 (00:15 -0400)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Sat, 31 Oct 2015 22:19:36 +0000 (15:19 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client

Pull Ceph fix from Sage Weil:
 "This sets the stable pages flag on the RBD block device when we have
  CRCs enabled.  (This is necessary since the default assumption for
  block devices changed in 3.9)"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: require stable pages if message data CRCs are enabled

8 years agoMerge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszer...
Linus Torvalds [Sat, 31 Oct 2015 21:49:19 +0000 (14:49 -0700)]
Merge branch 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs

Pull overlayfs bug fixes from Miklos Szeredi:
 "This contains fixes for bugs that appeared in earlier kernels (all are
  marked for -stable)"

* 'overlayfs-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs:
  ovl: free lower_mnt array in ovl_put_super
  ovl: free stack of paths in ovl_fill_super
  ovl: fix open in stacked overlay
  ovl: fix dentry reference leak
  ovl: use O_LARGEFILE in ovl_copy_up()

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sat, 31 Oct 2015 18:52:20 +0000 (11:52 -0700)]
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix two regressions in ipv6 route lookups, particularly wrt output
    interface specifications in the lookup key.  From David Ahern.

 2) Fix checks in ipv6 IPSEC tunnel pre-encap fragmentation, from
    Herbert Xu.

 3) Fix mis-advertisement of 1000BASE-T on bcm63xx_enet, from Simon
    Arlott.

 4) Some smsc phys misbehave with energy detect mode enabled, so add a
    DT property and disable it on such switches.  From Heiko Schocher.

 5) Fix TSO corruption on TX in mv643xx_eth, from Philipp Kirchhofer.

 6) Fix regression added by removal of openvswitch vport stats, from
    James Morse.

 7) Vendor Kconfig options should be bool, not tristate, from Andreas
    Schwab.

 8) Use non-_BH() net stats bump in tcp_xmit_probe_skb(), otherwise we
    barf during TCP REPAIR operations.

 9) Fix various bugs in openvswitch conntrack support, from Joe
    Stringer.

10) Fix NETLINK_LIST_MEMBERSHIPS locking, from David Herrmann.

11) Don't have VSOCK do sock_put() in interrupt context, from Jorgen
    Hansen.

12) Fix skb_realloc_headroom() failures properly in ISDN, from Karsten
    Keil.

13) Add some device IDs to qmi_wwan, from Bjorn Mork.

14) Fix ovs egress tunnel information when using lwtunnel devices, from
    Pravin B Shelar.

15) Add missing NETIF_F_FRAGLIST to macvtab feature list, from Jason
    Wang.

16) Fix incorrect handling of throw routes when the result of the throw
    cannot find a match, from Xin Long.

17) Protect ipv6 MTU calculations from wrap-around, from Hannes Frederic
    Sowa.

18) Fix failed autonegotiation on KSZ9031 micrel PHYs, from Nathan
    Sullivan.

19) Add missing memory barries in descriptor accesses or xgbe driver,
    from Thomas Lendacky.

20) Fix release conditon test in pppoe_release(), from Guillaume Nault.

21) Fix gianfar bugs wrt filter configuration, from Claudiu Manoil.

22) Fix violations of RX buffer alignment in sh_eth driver, from Sergei
    Shtylyov.

23) Fixing missing of_node_put() calls in various places around the
    networking, from Julia Lawall.

24) Fix incorrect leaf now walking in ipv4 routing tree, from Alexander
    Duyck.

25) RDS doesn't check pskb_pull()/pskb_trim() return values, from
    Sowmini Varadhan.

26) Fix VLAN configuration in mlx4 driver, from Jack Morgenstein.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (79 commits)
  ipv6: protect mtu calculation of wrap-around and infinite loop by rounding issues
  Revert "Merge branch 'ipv6-overflow-arith'"
  net/mlx4: Copy/set only sizeof struct mlx4_eqe bytes
  net/mlx4_en: Explicitly set no vlan tags in WQE ctrl segment when no vlan is present
  vhost: fix performance on LE hosts
  bpf: sample: define aarch64 specific registers
  amd-xgbe: Fix race between access of desc and desc index
  RDS-TCP: Recover correctly from pskb_pull()/pksb_trim() failure in rds_tcp_data_recv
  forcedeth: fix unilateral interrupt disabling in netpoll path
  openvswitch: Fix skb leak using IPv6 defrag
  ipv6: Export nf_ct_frag6_consume_orig()
  openvswitch: Fix double-free on ip_defrag() errors
  fib_trie: leaf_walk_rcu should not compute key if key is less than pn->key
  net: mv643xx_eth: add missing of_node_put
  ath6kl: add missing of_node_put
  net: phy: mdio: add missing of_node_put
  netdev/phy: add missing of_node_put
  net: netcp: add missing of_node_put
  net: thunderx: add missing of_node_put
  ipv6: gre: support SIT encapsulation
  ...

8 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 31 Oct 2015 01:49:44 +0000 (18:49 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input

Pull input layer fixes from Dmitry Torokhov:

 - a change to the ALPS driver where we had limit the quirk for
   trackstick handling from being active on all Dells to just a few
   models

 - a fix for a build dependency issue in the sur40 driver

 - a small clock handling fixup in the LPC32xx touchscreen driver

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: alps - only the Dell Latitude D420/430/620/630 have separate stick button bits
  Input: sur40 - add dependency on VIDEO_V4L2
  Input: lpc32xx_ts - fix warnings caused by enabling unprepared clock

8 years agoMerge tag 'pci-v4.3-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Linus Torvalds [Sat, 31 Oct 2015 01:47:18 +0000 (18:47 -0700)]
Merge tag 'pci-v4.3-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci

Pull PCI fix from Bjorn Helgaas:
 "Sorry for this last-minute update; it's been in -next for quite a
  while, but I forgot about it until I started getting ready for the
  merge window.

  It's small and fixes a way a user could cause a panic via sysfs, so I
  think it's worth getting it in v4.3.

  NUMA:
    - Prevent out of bounds access in sysfs numa_node override (Sasha Levin)"

* tag 'pci-v4.3-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: Prevent out of bounds access in numa_node override

8 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 30 Oct 2015 23:57:55 +0000 (16:57 -0700)]
Merge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "Apologies for this being so late, but we've uncovered a few nasty
  issues on arm64 which didn't settle down until yesterday and the fixes
  all look suitable for 4.3.  Of the four patches, three of them are
  Cc'd to stable, with the remaining patch fixing an issue that only
  took effect during the merge window.

  Summary:

   - Fix corruption in SWP emulation when STXR fails due to contention
   - Fix MMU re-initialisation when resuming from a low-power state
   - Fix stack unwinding code to match what ftrace expects
   - Fix relocation code in the EFI stub when DRAM base is not 2MB aligned"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64/efi: do not assume DRAM base is aligned to 2 MB
  Revert "ARM64: unwind: Fix PC calculation"
  arm64: kernel: fix tcr_el1.t0sz restore on systems with extended idmap
  arm64: compat: fix stxr failure case in SWP emulation

8 years agoMerge tag 'please-pull-syscalls' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 30 Oct 2015 23:56:44 +0000 (16:56 -0700)]
Merge tag 'please-pull-syscalls' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux

Pull ia64 kcmp syscall from Tony Luck:
 "Missed adding the kcmp() syscall a long time ago.  Now it seems that
  it is essential to build systemd"

* tag 'please-pull-syscalls' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux:
  [IA64] Wire up kcmp syscall

8 years agorbd: require stable pages if message data CRCs are enabled
Ronny Hegewald [Thu, 15 Oct 2015 18:50:46 +0000 (18:50 +0000)]
rbd: require stable pages if message data CRCs are enabled

rbd requires stable pages, as it performs a crc of the page data before
they are send to the OSDs.

But since kernel 3.9 (patch 1d1d1a767206fbe5d4c69493b7e6d2a8d08cc0a0
"mm: only enforce stable page writes if the backing device requires
it") it is not assumed anymore that block devices require stable pages.

This patch sets the necessary flag to get stable pages back for rbd.

In a ceph installation that provides multiple ext4 formatted rbd
devices "bad crc" messages appeared regularly (ca 1 message every 1-2
minutes on every OSD that provided the data for the rbd) in the
OSD-logs before this patch. After this patch this messages are pretty
much gone (only ca 1-2 / month / OSD).

Cc: stable@vger.kernel.org # 3.9+, needs backporting
Signed-off-by: Ronny Hegewald <Ronny.Hegewald@online.de>
[idryomov@gmail.com: require stable pages only in crc case, changelog]
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
8 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
David S. Miller [Fri, 30 Oct 2015 11:51:56 +0000 (20:51 +0900)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next

Steffen Klassert says:

====================
pull request (net-next): ipsec-next 2015-10-30

1) The flow cache is limited by the flow cache limit which
   depends on the number of cpus and the xfrm garbage collector
   threshold which is independent of the number of cpus. This
   leads to the fact that on systems with more than 16 cpus
   we hit the xfrm garbage collector limit and refuse new
   allocations, so new flows are dropped. On systems with 16
   or less cpus, we hit the flowcache limit. In this case, we
   shrink the flow cache instead of refusing new flows.

   We increase the xfrm garbage collector threshold to INT_MAX
   to get the same behaviour, independent of the number of cpus.

2) Fix some unaligned accesses on sparc systems.
   From Sowmini Varadhan.

3) Fix some header checks in _decode_session4. We may call
   pskb_may_pull with a negative value converted to unsigened
   int from pskb_may_pull. This can lead to incorrect policy
   lookups. We fix this by a check of the data pointer position
   before we call pskb_may_pull.

4) Reload skb header pointers after calling pskb_may_pull
   in _decode_session4 as this may change the pointers into
   the packet.

5) Add a missing statistic counter on inner mode errors.

Please pull or let me know if there are problems.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'wireless-drivers-next-for-davem-2015-10-29' of git://git.kernel.org/pub...
David S. Miller [Fri, 30 Oct 2015 11:38:24 +0000 (20:38 +0900)]
Merge tag 'wireless-drivers-next-for-davem-2015-10-29' of git://git.kernel.org/pub/scm/linux/kernel/git/kvalo/wireless-drivers-next

Kalle Valo says:

====================
iwlwifi

* bug fix for TDLS
* fixes and cleanups in scan
* support of several scan plans
* improvements in FTM
* fixes in FW API
* improvements in the failure paths when the bus is dead
* other various small things here and there

ath10k

* add QCA9377 support
* fw_stats support for 10.4 firmware

ath6kl

* report antenna configuration to user space
* implement ethtool stats

ssb

* add Kconfig SSB_HOST_SOC for compiling SoC related code
* move functions specific to SoC hosted bus to separated file
* pick PCMCIA host code support from b43 driver
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: fix: pass correct obj size when deferring obj add
Scott Feldman [Thu, 29 Oct 2015 06:17:31 +0000 (23:17 -0700)]
switchdev: fix: pass correct obj size when deferring obj add

Fixes: 4d429c5dd ("switchdev: introduce possibility to defer obj_add/del")
Signed-off-by: Scott Feldman <sfeldma@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: fix: erasing too much of vlan obj when handling multiple vlan specs
Scott Feldman [Thu, 29 Oct 2015 06:17:30 +0000 (23:17 -0700)]
switchdev: fix: erasing too much of vlan obj when handling multiple vlan specs

When adding vlans with multiple IFLA_BRIDGE_VLAN_INFO attrs set in AFSPEC,
we would wipe the vlan obj struct after the first IFLA_BRIDGE_VLAN_INFO.
Fix this by only clearing what's necessary on each IFLA_BRIDGE_VLAN_INFO
iteration.

Fixes: 9e8f4a54 ("switchdev: push object ID back to object structure")
Signed-off-by: Scott Feldman <sfeldma@gmail.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge tag 'nfc-next-4.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo...
David S. Miller [Fri, 30 Oct 2015 11:19:43 +0000 (20:19 +0900)]
Merge tag 'nfc-next-4.4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/nfc-next

Samuel Ortiz says:

====================
NFC 4.4 pull request

This is the NFC pull request for 4.4.

It's a bit bigger than usual, the 3 main culprits being:

- A new driver for Intel's Fields Peak NCI chipset. In order to
  support this chipset we had to export a few NCI routines and
  extend the driver NCI ops to not only support proprietary
  commands but also core ones.

- Support for vendor commands for both STM drivers, st-nci
  and st21nfca. Those vendor commands allow to run factory tests
  through the NFC netlink interface.

- New i2c and SPI support for the Marvell driver, together with
  firmware download support for this driver's core.

Besides that we also have:

- A few file renames in the STM drivers, to keep the naming
  consistent between drivers.

- Some improvements and fixes on the NCI HCI layer, mostly to
  properly reach a secure element over a legacy HCI link.

- A few fixes for the s3fwrn5 and trf7970a drivers.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Fri, 30 Oct 2015 10:41:10 +0000 (19:41 +0900)]
Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next

Johan Hedberg says:

====================
pull request: bluetooth-next 2015-10-28

Here are a some more Bluetooth patches for 4.4 which collected up during
the past week. The most important ones are from Kuba Pawlak for fixing
locking issues with SCO sockets. There's also a fix from Alexander Aring
for 6lowpan, a memleak fix from Julia Lawall for the btmrvl driver and
some cleanup patches from Marcel.

Please let me know if there are any issues pulling. Thanks.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoipv6: recreate ipv6 link-local addresses when increasing MTU over IPV6_MIN_MTU
Alexander Duyck [Mon, 26 Oct 2015 18:06:33 +0000 (11:06 -0700)]
ipv6: recreate ipv6 link-local addresses when increasing MTU over IPV6_MIN_MTU

This change makes it so that we reinitialize the interface if the MTU is
increased back above IPV6_MIN_MTU and the interface is up.

Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'mlxsw-flooding-and-cosmetics'
David S. Miller [Fri, 30 Oct 2015 03:27:07 +0000 (12:27 +0900)]
Merge branch 'mlxsw-flooding-and-cosmetics'

Jiri Pirko says:

====================
mlxsw: driver update

This driver update mainly brings support for user to be able to setup
flooding on specified port, via bridge flag. Also, there is a fix in ageing
time conversion. The rest is just cosmetics.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Make mlxsw_sp_port_switchdev_ops static
Jiri Pirko [Wed, 28 Oct 2015 09:17:05 +0000 (10:17 +0100)]
mlxsw: spectrum: Make mlxsw_sp_port_switchdev_ops static

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: Put braces on all arms of branch statement
Or Gerlitz [Wed, 28 Oct 2015 09:17:04 +0000 (10:17 +0100)]
mlxsw: Put braces on all arms of branch statement

Fix a place where checkpatch complains that braces should be used
on all arms of this statement.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: Put constant on the right side of comparisons
Or Gerlitz [Wed, 28 Oct 2015 09:17:03 +0000 (10:17 +0100)]
mlxsw: Put constant on the right side of comparisons

Fixes those places where checkpatch complains that comparisons
should place the constant on the right side of the test.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Fix ageing time value
Jiri Pirko [Wed, 28 Oct 2015 09:17:02 +0000 (10:17 +0100)]
mlxsw: spectrum: Fix ageing time value

The value passed through switchdev attr set is not in jiffies, but in
clock_t, so fix the convert.

Reported-by: Sagi Rotem <sagir@mellanox.com>
Fixes: 56ade8fe3f ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: reg: Avoid unnecessary line wrap for mlxsw_reg_sfd_uc_unpack
Jiri Pirko [Wed, 28 Oct 2015 09:17:01 +0000 (10:17 +0100)]
mlxsw: reg: Avoid unnecessary line wrap for mlxsw_reg_sfd_uc_unpack

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: reg: Fix desription typos of couple of SFN items
Jiri Pirko [Wed, 28 Oct 2015 09:17:00 +0000 (10:17 +0100)]
mlxsw: reg: Fix desription typos of couple of SFN items

Fix copy-paste errors.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: reg: Fix description for reg_sfd_uc_sub_port
Jiri Pirko [Wed, 28 Oct 2015 09:16:59 +0000 (10:16 +0100)]
mlxsw: reg: Fix description for reg_sfd_uc_sub_port

The original description was for LAG, so fix it.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Add support for flood control
Ido Schimmel [Wed, 28 Oct 2015 09:16:58 +0000 (10:16 +0100)]
mlxsw: spectrum: Add support for flood control

Add or remove a bridged port from the flooding domain of unknown unicast
packets according to user configuration.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: Add support for VLAN ranges in flooding configuration
Ido Schimmel [Wed, 28 Oct 2015 09:16:57 +0000 (10:16 +0100)]
mlxsw: spectrum: Add support for VLAN ranges in flooding configuration

When enabling a range of VLANs on a bridged port we can configure
flooding for these VLANs by one register access instead of calling the
same register for each VLAN. This is accomplished by using the 'range'
field of the Switch Flooding Table Register (SFTR).

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agomlxsw: spectrum: move "bridged" bool to u8 flags
Jiri Pirko [Wed, 28 Oct 2015 09:16:56 +0000 (10:16 +0100)]
mlxsw: spectrum: move "bridged" bool to u8 flags

It is a flag anyway, so move it to existing u8 flag and don't waste mem.
Fix the flags to be in single u8 on the way.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: Make flood to CPU optional
Ido Schimmel [Wed, 28 Oct 2015 09:16:55 +0000 (10:16 +0100)]
switchdev: Make flood to CPU optional

In certain use cases it is not always desirable for the switch device to
flood traffic to CPU port. Instead, only certain packet types (e.g.
STP, LACP) should be trapped to it.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoswitchdev: Add support for flood control
Ido Schimmel [Wed, 28 Oct 2015 09:16:54 +0000 (10:16 +0100)]
switchdev: Add support for flood control

Allow devices supporting this feature to control the flooding of unknown
unicast traffic, by making switchdev infrastructure propagate this setting
to the switch driver.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge branch 'xgene_txrx_delay'
David S. Miller [Fri, 30 Oct 2015 03:21:57 +0000 (12:21 +0900)]
Merge branch 'xgene_txrx_delay'

Iyappan Subramanian says:

====================
drivers: xgene: Add support RGMII TX/RX delay configuration

X-Gene RGMII ethernet controller has a RGMII bridge that performs the
task of converting the RGMII signal {RX_CLK,RX_CTL, RX_DATA[3:0]} from
PHY to GMII signal {RX_DV,RX_ER,RX_DATA[7:0]} and vice versa.  This
RGMII bridge has a provision to internally delay the input RX_CLK and
the output TX_CLK using configuration registers. This will help in
maintain the CLK-CTL delay relationship in various operating
conditions.

This patch adds support RGMII TX/RX delay configuration.
====================

Signed-off-by: Iyappan Subramanian <isubramanian@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoDocumentation: dts: xgene: Add TX/RX delay field
Iyappan Subramanian [Mon, 26 Oct 2015 22:25:16 +0000 (15:25 -0700)]
Documentation: dts: xgene: Add TX/RX delay field

Signed-off-by: Iyappan Subramanian <isubramanian@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agodrivers: net: xgene: Add support RGMII TX/RX delay configuration
Iyappan Subramanian [Mon, 26 Oct 2015 22:25:15 +0000 (15:25 -0700)]
drivers: net: xgene: Add support RGMII TX/RX delay configuration

Add RGMII TX/RX delay configuration support. RGMII standard requires 2ns
delay to help the RGMII bridge receiver to sample data correctly. If the
default value does not provide proper centering of the data sample, the
TX/RX delay parameters can be used to adjust accordingly.

Signed-off-by: Iyappan Subramanian <isubramanian@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agobridge: set is_local and is_static before fdb entry is added to the fdb hashtable
Roopa Prabhu [Tue, 27 Oct 2015 14:52:56 +0000 (07:52 -0700)]
bridge: set is_local and is_static before fdb entry is added to the fdb hashtable

Problem Description:
We can add fdbs pointing to the bridge with NULL ->dst but that has a
few race conditions because br_fdb_insert() is used which first creates
the fdb and then, after the fdb has been published/linked, sets
"is_local" to 1 and in that time frame if a packet arrives for that fdb
it may see it as non-local and either do a NULL ptr dereference in
br_forward() or attach the fdb to the port where it arrived, and later
br_fdb_insert() will make it local thus getting a wrong fdb entry.
Call chain br_handle_frame_finish() -> br_forward():
But in br_handle_frame_finish() in order to call br_forward() the dst
should not be local i.e. skb != NULL, whenever the dst is
found to be local skb is set to NULL so we can't forward it,
and here comes the problem since it's running only
with RCU when forwarding packets it can see the entry before "is_local"
is set to 1 and actually try to dereference NULL.
The main issue is that if someone sends a packet to the switch while
it's adding the entry which points to the bridge device, it may
dereference NULL ptr. This is needed now after we can add fdbs
pointing to the bridge.  This poses a problem for
br_fdb_update() as well, while someone's adding a bridge fdb, but
before it has is_local == 1, it might get moved to a port if it comes
as a source mac and then it may get its "is_local" set to 1

This patch changes fdb_create to take is_local and is_static as
arguments to set these values in the fdb entry before it is added to the
hash. Also adds null check for port in br_forward.

Fixes: 3741873b4f73 ("bridge: allow adding of fdb entries pointing to the bridge device")
Reported-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Reviewed-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogeneve: add IPv6 bits to geneve_fill_metadata_dst
John W. Linville [Tue, 27 Oct 2015 13:49:00 +0000 (09:49 -0400)]
geneve: add IPv6 bits to geneve_fill_metadata_dst

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogeneve: handle ipv6 priority like ipv4 tos
John W. Linville [Mon, 26 Oct 2015 21:01:45 +0000 (17:01 -0400)]
geneve: handle ipv6 priority like ipv4 tos

Other callers of udp_tunnel6_xmit_skb just pass 0 for the prio
argument.  Jesse Gross <jesse@nicira.com> suggested that prio is really
the same as IPv4's tos and should be handled the same, so this is my
interpretation of that suggestion.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Reported-by: Jesse Gross <jesse@nicira.com>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agogeneve: implement support for IPv6-based tunnels
John W. Linville [Mon, 26 Oct 2015 21:01:44 +0000 (17:01 -0400)]
geneve: implement support for IPv6-based tunnels

NOTE: Link-local IPv6 addresses for remote endpoints are not supported,
since the driver currently has no capacity for binding a geneve
interface to a specific link.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoarm64/efi: do not assume DRAM base is aligned to 2 MB
Ard Biesheuvel [Thu, 29 Oct 2015 14:07:25 +0000 (15:07 +0100)]
arm64/efi: do not assume DRAM base is aligned to 2 MB

The current arm64 Image relocation code in the UEFI stub assumes that
the dram_base argument it receives is always a multiple of 2 MB. In
reality, it is simply the lowest start address of all RAM entries in
the UEFI memory map, which means it could be any multiple of 4 KB.

Since the arm64 kernel Image needs to reside TEXT_OFFSET bytes beyond
a 2 MB aligned base, or it will fail to boot, make sure we round dram_base
to 2 MB before using it to calculate the relocation address.

Fixes: e38457c361b30c5a ("arm64: efi: prefer AllocatePages() over efi_low_alloc() for vmlinux")
Reported-by: Timur Tabi <timur@codeaurora.org>
Tested-by: Timur Tabi <timur@codeaurora.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
8 years agoipv6: protect mtu calculation of wrap-around and infinite loop by rounding issues
Hannes Frederic Sowa [Wed, 28 Oct 2015 12:21:04 +0000 (13:21 +0100)]
ipv6: protect mtu calculation of wrap-around and infinite loop by rounding issues

Raw sockets with hdrincl enabled can insert ipv6 extension headers
right into the data stream. In case we need to fragment those packets,
we reparse the options header to find the place where we can insert
the fragment header. If the extension headers exceed the link's MTU we
actually cannot make progress in such a case.

Instead of ending up in broken arithmetic or rounding towards 0 and
entering an endless loop in ip6_fragment, just prevent those cases by
aborting early and signal -EMSGSIZE to user space.

This is the second version of the patch which doesn't use the
overflow_usub function, which got reverted for now.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoRevert "Merge branch 'ipv6-overflow-arith'"
Hannes Frederic Sowa [Wed, 28 Oct 2015 12:21:03 +0000 (13:21 +0100)]
Revert "Merge branch 'ipv6-overflow-arith'"

Linus dislikes these changes. To not hold up the net-merge let's revert
it for now and fix the bug like Linus suggested.

This reverts commit ec3661b42257d9a06cf0d318175623ac7a660113, reversing
changes made to c80dbe04612986fd6104b4a1be21681b113b5ac9.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
8 years agoMerge ath-next from ath.git
Kalle Valo [Thu, 29 Oct 2015 13:34:27 +0000 (15:34 +0200)]
Merge ath-next from ath.git

Major changes:

ath10k:

* add QCA9377 support
* fw_stats support for 10.4 firmware

ath6kl:

* report antenna configuration to user space
* implement ethtool stats

8 years agoath6kl: implement ethtool stats
Ben Greear [Thu, 22 Oct 2015 16:07:39 +0000 (09:07 -0700)]
ath6kl: implement ethtool stats

This supports a way to get target stats through normal
ethtool stats API.

For instance:
# ethtool -S wlan1
NIC statistics:
     tx_pkts_nic: 353
     tx_bytes_nic: 25142
     rx_pkts_nic: 6
     rx_bytes_nic: 996
     d_tx_ucast_pkts: 89
     d_tx_bcast_pkts: 264
     d_tx_ucast_bytes: 3020
     d_tx_bcast_bytes: 22122
...

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agoath6kl: break stats gathering code into separate method
Ben Greear [Thu, 22 Oct 2015 16:07:38 +0000 (09:07 -0700)]
ath6kl: break stats gathering code into separate method

This will allow us to call it from elsewhere when implementing
ethtool stats.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agoath6kl: fix firmware version assignment
Ben Greear [Wed, 21 Oct 2015 21:53:21 +0000 (14:53 -0700)]
ath6kl: fix firmware version assignment

Improper use of strlcpy caused garbage to be appended to the
firmware version string.  Fix this by paying attention to the
ie_lenth.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agoath6kl: add error message to explain lack of HT
Ben Greear [Wed, 21 Oct 2015 21:53:20 +0000 (14:53 -0700)]
ath6kl: add error message to explain lack of HT

It can take a user a while to understand why their
NIC that advertises 802.11n support cannot actually
do 802.11n.  Print out a warning in the logs to save
the next poor person to use this NIC some trouble.

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agoath6kl: report antenna configuration
Ben Greear [Wed, 21 Oct 2015 21:53:19 +0000 (14:53 -0700)]
ath6kl: report antenna configuration

This lets 'iw phy phy0 info' report antennas for
the radio device:

...
Available Antennas: TX 0x2 RX 0x2
Configured Antennas: TX 0x2 RX 0x2
...

Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agowil6210: handle failure in Tx vring config
Vladimir Kondratiev [Sun, 25 Oct 2015 13:59:24 +0000 (15:59 +0200)]
wil6210: handle failure in Tx vring config

When configuring Tx vring for new connection,
WMI call to the firmware may fail. In this case, need to
clean up properly. In particular, need to call
cfg80211_del_sta() in case of AP like interface.

Perform full "disconnect" procedure for proper clean up

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agowil6210: fix device ready detection
Vladimir Kondratiev [Sun, 25 Oct 2015 13:59:23 +0000 (15:59 +0200)]
wil6210: fix device ready detection

Adjust driver behavior during FW boot. Proper sequence of
events after reset and FW download, is as following:

- FW prepares mailbox structure and reports IRQ "FW_READY"
- driver caches mailbox registers, marks mailbox readiness
- FW sends WMI_FW_READY event, ignore it
- FW sends WMI_READY event with some data
- driver stores relevant data marks FW is operational

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agowil6210: Fix TSO overflow handling
Hamad Kadmany [Sun, 25 Oct 2015 13:59:22 +0000 (15:59 +0200)]
wil6210: Fix TSO overflow handling

When Tx ring full is encountered with TSO,
printout of "DMA error" was wrongly printed.

In addition, in case of Tx ring full return
proper error code so that NETDEV_TX_BUSY is
returned to network stack in order not to
drop the packets and retry transmission of the
packets when ring is emptied.

Signed-off-by: Hamad Kadmany <qca_hkadmany@qca.qualcomm.com>
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agowil6210: ignore selected WMI events
Vladimir Kondratiev [Sun, 25 Oct 2015 13:59:21 +0000 (15:59 +0200)]
wil6210: ignore selected WMI events

Some events are ignored for purpose; such events should not
be treated as "unhandled events". Replace info message
saying "unhandled" with debug one saying "ignore", to reduce
dmesg pollution

Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>
8 years agoath10k: add QCA9377 chipset support
Bartosz Markowski [Wed, 28 Oct 2015 14:09:53 +0000 (15:09 +0100)]
ath10k: add QCA9377 chipset support

Add the hardware name, revision and update the pci_id table.

Currently there're two HW ref. designs available I'm aware of,
with 1.0.2 and 1.1 chip revisions. I've access and been using
the first one so far and this patch cover only it.

QCA9377 inherits most of the stuff (e.g. fw interfaces)
from QCA61x4 design, so the integration was pretty straightforward.

Signed-off-by: Bartosz Markowski <bartosz.markowski@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>