GitHub/moto-9609/android_kernel_motorola_exynos9610.git
9 years agoamd-xgbe: Checkpatch fixes
Lendacky, Thomas [Fri, 16 Jan 2015 18:46:24 +0000 (12:46 -0600)]
amd-xgbe: Checkpatch fixes

This set of patches resolves some checks reported by the checkpatch
tool.

Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetlink: Fix netlink_insert EADDRINUSE error
Herbert Xu [Fri, 16 Jan 2015 06:23:48 +0000 (17:23 +1100)]
netlink: Fix netlink_insert EADDRINUSE error

The patch c5adde9468b0714a051eac7f9666f23eb10b61f7 ("netlink:
eliminate nl_sk_hash_lock") introduced a bug where the EADDRINUSE
error has been replaced by ENOMEM.  This patch rectifies that
problem.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Fix race in rhashtable_destroy() and use regular work_struct
Ying Xue [Fri, 16 Jan 2015 03:13:09 +0000 (11:13 +0800)]
rhashtable: Fix race in rhashtable_destroy() and use regular work_struct

When we put our declared work task in the global workqueue with
schedule_delayed_work(), its delay parameter is always zero.
Therefore, we should define a regular work in rhashtable structure
instead of a delayed work.

By the way, we add a condition to check whether resizing functions
are NULL before cancelling the work, avoiding to cancel an
uninitialized work.

Lastly, while we wait for all work items we submitted before to run
to completion with cancel_delayed_work(), ht->mutex has been taken in
rhashtable_destroy(). Moreover, cancel_delayed_work() doesn't return
until all work items are accomplished, and when work items are
scheduled, the work's function - rht_deferred_worker() will be called.
However, as rht_deferred_worker() also needs to acquire the lock,
deadlock might happen at the moment as the lock is already held before.
So if the cancel work function is moved out of the lock covered scope,
this will avoid the deadlock.

Fixes: 97defe1 ("rhashtable: Per bucket locks & deferred expansion/shrinking")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Cc: Thomas Graf <tgraf@suug.ch>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'iw_cxgb4-next'
David S. Miller [Fri, 16 Jan 2015 06:07:02 +0000 (01:07 -0500)]
Merge branch 'iw_cxgb4-next'

Hariprasad Shenai says:

====================
Refactor macros to conform to uniform standards

This patch series cleansup macros/register defines, defined in t4.h and
t4fw_ri_api.h and all the affected files.

This patch series is created against net-next tree and includes patches on
iw_cxgb4 tree. Since the patches are dependent on previous cleanup patched we
would line to get this series merged through net-next tree.

We have included all the maintainers of respective drivers. Kindly review the
change and let us know in case of any review comments.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoiw_cxgb4: Cleanup register defines/MACROS defined in t4fw_ri_api.h
Hariprasad Shenai [Fri, 16 Jan 2015 03:54:48 +0000 (09:24 +0530)]
iw_cxgb4: Cleanup register defines/MACROS defined in t4fw_ri_api.h

Cleanup all the MACROS that are defined in t4fw_ri_api.h and affected files

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoiw_cxgb4: Cleanup register defines/MACROS defined in t4.h
Hariprasad Shenai [Fri, 16 Jan 2015 03:54:47 +0000 (09:24 +0530)]
iw_cxgb4: Cleanup register defines/MACROS defined in t4.h

Cleanup all the MACROS defined in t4.h and the affected files

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoiw_cxgb4: Cleanup register defines/MACROS defined in t4fw_ri_api.h
Hariprasad Shenai [Fri, 16 Jan 2015 03:54:48 +0000 (09:24 +0530)]
iw_cxgb4: Cleanup register defines/MACROS defined in t4fw_ri_api.h

Cleanup all the MACROS that are defined in t4fw_ri_api.h and affected files

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoiw_cxgb4: Cleanup register defines/MACROS defined in t4.h
Hariprasad Shenai [Fri, 16 Jan 2015 03:54:47 +0000 (09:24 +0530)]
iw_cxgb4: Cleanup register defines/MACROS defined in t4.h

Cleanup all the MACROS defined in t4.h and the affected files

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: Create gem_ethtool_ops for new statistics functions
Xander Huff [Thu, 15 Jan 2015 21:55:20 +0000 (15:55 -0600)]
net/macb: Create gem_ethtool_ops for new statistics functions

10/100 MACB does not have the same statistics possibilities as GEM. Separate
macb_ethtool_ops to make a new GEM-specific struct with the new statistics
functions included.

Signed-off-by: Xander Huff <xander.huff@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: Add whitespace around arithmetic operators
Xander Huff [Thu, 15 Jan 2015 21:55:19 +0000 (15:55 -0600)]
net/macb: Add whitespace around arithmetic operators

Spaces should surround add, multiply, and bitshift operators.

Signed-off-by: Xander Huff <xander.huff@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: Fix comments to meet style guidelines
Xander Huff [Thu, 15 Jan 2015 21:55:18 +0000 (15:55 -0600)]
net/macb: Fix comments to meet style guidelines

Change comments to not exceed 80 characters per line.
Update block comments in macb.h to start on the line after /*.

Signed-off-by: Xander Huff <xander.huff@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoDriver: Vmxnet3: Fix ethtool -S to return correct rx queue stats
Shrikrishna Khare [Thu, 15 Jan 2015 19:54:30 +0000 (11:54 -0800)]
Driver: Vmxnet3: Fix ethtool -S to return correct rx queue stats

Signed-off-by: Gao Zhenyu <gzhenyu@vmware.com>
Signed-off-by: Shrikrishna Khare <skhare@vmware.com>
Reviewed-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: use MDBA_SET_ENTRY_MAX for maxtype in nlmsg_parse()
Nicolas Dichtel [Thu, 15 Jan 2015 15:29:12 +0000 (16:29 +0100)]
bridge: use MDBA_SET_ENTRY_MAX for maxtype in nlmsg_parse()

This is just a cleanup, because in the current code MDBA_SET_ENTRY_MAX ==
MDBA_SET_ENTRY.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: smc91x: Add Atari EtherNAT support
Michael Schmitz [Thu, 15 Jan 2015 13:06:15 +0000 (14:06 +0100)]
net: smc91x: Add Atari EtherNAT support

Add Atari specific code to the smc91x Ethernet driver. This code is used
on the EtherNAT adapter card for the Atari Falcon extension port.

Signed-off-by: Michael Schmitz <schmitz@debian.org>
Tested-by: Christian Steigies <cts@debian.org>
[geert: Sort Kconfig entries, split in hard and soft dependencies]
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'mac80211-next-for-davem-2015-01-15' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Fri, 16 Jan 2015 00:16:56 +0000 (19:16 -0500)]
Merge tag 'mac80211-next-for-davem-2015-01-15' of git://git./linux/kernel/git/jberg/mac80211-next

Here's a big pile of changes for this round.

We have
 * a lot of regulatory code changes to deal with the
   way newer Intel devices handle this
 * a change to drop packets while disconnecting from
   an AP instead of trying to wait for them
 * a new attempt at improving the tailroom accounting
   to not kick in too much for performance reasons
 * improvements in wireless link statistics
 * many other small improvements and small fixes that
   didn't seem necessary for 3.19 (e.g. in hwsim which
   is testing only code)

Conflicts:
drivers/staging/rtl8723au/os_dep/ioctl_cfg80211.c

Minor overlapping changes.

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'cxgb4-next'
David S. Miller [Thu, 15 Jan 2015 23:59:55 +0000 (18:59 -0500)]
Merge branch 'cxgb4-next'

Anish Bhatt says:

====================
cxgb4/cxgb4i : Update & use ipv6 handling api

This patch series consolidates and updates the ipv6 api, as well as exports
it for use by upper level drivers dependent on cxgb4

v2: Fix formatting issues in clip_tbl.c
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4i : Call into recently added cxgb4 ipv6 api
Anish Bhatt [Wed, 14 Jan 2015 23:17:35 +0000 (15:17 -0800)]
cxgb4i : Call into recently added cxgb4 ipv6 api

Get a reference on every ipv6 address we offload to hardware so that it cannot
be prematurely cleared out before cleanup.

Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4 : Update ipv6 address handling api
Anish Bhatt [Wed, 14 Jan 2015 23:17:34 +0000 (15:17 -0800)]
cxgb4 : Update ipv6 address handling api

This patch improves on previously added support for ipv6 addresses. The code
is consolidated to a single file and adds an api for use by dependent upper
level drivers such as cxgb4i/iw_cxgb4 etc.

Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: Deepak Singh <deepak.s@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv4: per cpu uncached list
Eric Dumazet [Wed, 14 Jan 2015 23:17:06 +0000 (15:17 -0800)]
ipv4: per cpu uncached list

RAW sockets with hdrinc suffer from contention on rt_uncached_lock
spinlock.

One solution is to use percpu lists, since most routes are destroyed
by the cpu that created them.

It is unclear why we even have to put these routes in uncached_list,
as all outgoing packets should be freed when a device is dismantled.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: caacf05e5ad1 ("ipv4: Properly purge netdev references on uncached routes.")
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: sh_eth: Use u32 for 32-bit register data
Geert Uytterhoeven [Thu, 15 Jan 2015 10:54:28 +0000 (11:54 +0100)]
net: sh_eth: Use u32 for 32-bit register data

Replace "unsigned long" by "u32" when handling 32-bit register data.

A.o., this saves 11 * 36 bytes in the static sh_eth_cpu_data structures
when compiling for 64-bit.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodm9000: Add regulator and reset support to dm9000
Zubair Lutfullah Kakakhel [Thu, 15 Jan 2015 10:12:26 +0000 (10:12 +0000)]
dm9000: Add regulator and reset support to dm9000

In boards, the dm9000 chip's power and reset can be controlled by gpio.

It makes sense to add them to the dm9000 driver and let dt be used to
enable power and reset the phy.

Signed-off-by: Zubair Lutfullah Kakakhel <Zubair.Kakakhel@imgtec.com>
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosocket: use iov_length()
Nicolas Dichtel [Thu, 15 Jan 2015 09:35:58 +0000 (10:35 +0100)]
socket: use iov_length()

Better to use available helpers.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf-next
David S. Miller [Thu, 15 Jan 2015 06:50:25 +0000 (01:50 -0500)]
Merge git://git./linux/kernel/git/pablo/nf-next

Pablo Neira Ayuso says:

====================
netfilter updates for net-next

The following patchset contains netfilter updates for net-next, just a
bunch of cleanups and small enhancement to selectively flush conntracks
in ctnetlink, more specifically the patches are:

1) Rise default number of buckets in conntrack from 16384 to 65536 in
   systems with >= 4GBytes, patch from Marcelo Leitner.

2) Small refactor to save one level on indentation in xt_osf, from
   Joe Perches.

3) Remove unnecessary sizeof(char) in nf_log, from Fabian Frederick.

4) Another small cleanup to remove redundant variable in nfnetlink,
   from Duan Jiong.

5) Fix compilation warning in nfnetlink_cthelper on parisc, from
   Chen Gang.

6) Fix wrong format in debugging for ctseqadj, from Gao feng.

7) Selective conntrack flushing through the mark for ctnetlink, patch
   from Kristian Evensen.

8) Remove nf_ct_conntrack_flush_report() exported symbol now that is
   not required anymore after the selective flushing patch, again from
   Kristian.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'vxlan_group_policy_extension'
David S. Miller [Thu, 15 Jan 2015 06:12:13 +0000 (01:12 -0500)]
Merge branch 'vxlan_group_policy_extension'

Thomas Graf says:

====================
VXLAN Group Policy Extension

Implements supports for the Group Policy VXLAN extension [0] to provide
a lightweight and simple security label mechanism across network peers
based on VXLAN. The security context and associated metadata is mapped
to/from skb->mark. This allows further mapping to a SELinux context
using SECMARK, to implement ACLs directly with nftables, iptables, OVS,
tc, etc.

The extension is disabled by default and should be run on a distinct
port in mixed Linux VXLAN VTEP environments. Liberal VXLAN VTEPs
which ignore unknown reserved bits will be able to receive VXLAN-GBP
frames.

Simple usage example:

10.1.1.1:
   # ip link add vxlan0 type vxlan id 10 remote 10.1.1.2 gbp
   # iptables -I OUTPUT -m owner --uid-owner 101 -j MARK --set-mark 0x200

10.1.1.2:
   # ip link add vxlan0 type vxlan id 10 remote 10.1.1.1 gbp
   # iptables -I INPUT -m mark --mark 0x200 -j DROP

iproute2 [1] and OVS [2] support will be provided in separate patches.

[0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
[1] https://github.com/tgraf/iproute2/tree/vxlan-gbp
[2] https://github.com/tgraf/ovs/tree/vxlan-gbp
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoopenvswitch: Support VXLAN Group Policy extension
Thomas Graf [Thu, 15 Jan 2015 02:53:59 +0000 (03:53 +0100)]
openvswitch: Support VXLAN Group Policy extension

Introduces support for the group policy extension to the VXLAN virtual
port. The extension is disabled by default and only enabled if the user
has provided the respective configuration.

  ovs-vsctl add-port br0 vxlan0 -- \
     set Interface vxlan0 type=vxlan options:exts=gbp

The configuration interface to enable the extension is based on a new
attribute OVS_VXLAN_EXT_GBP nested inside OVS_TUNNEL_ATTR_EXTENSION
which can carry additional extensions as needed in the future.

The group policy metadata is stored as binary blob (struct ovs_vxlan_opts)
internally just like Geneve options but transported as nested Netlink
attributes to user space.

Renames the existing TUNNEL_OPTIONS_PRESENT to TUNNEL_GENEVE_OPT with the
binary value kept intact, a new flag TUNNEL_VXLAN_OPT is introduced.

The attributes OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and existing
OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS are implemented mutually exclusive.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoopenvswitch: Allow for any level of nesting in flow attributes
Thomas Graf [Thu, 15 Jan 2015 02:53:58 +0000 (03:53 +0100)]
openvswitch: Allow for any level of nesting in flow attributes

nlattr_set() is currently hardcoded to two levels of nesting. This change
introduces struct ovs_len_tbl to define minimal length requirements plus
next level nesting tables to traverse the key attributes to arbitrary depth.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoopenvswitch: Rename GENEVE_TUN_OPTS() to TUN_METADATA_OPTS()
Thomas Graf [Thu, 15 Jan 2015 02:53:57 +0000 (03:53 +0100)]
openvswitch: Rename GENEVE_TUN_OPTS() to TUN_METADATA_OPTS()

Also factors out Geneve validation code into a new separate function
validate_and_copy_geneve_opts().

A subsequent patch will introduce VXLAN options. Rename the existing
GENEVE_TUN_OPTS() to reflect its extended purpose of carrying generic
tunnel metadata options.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovxlan: Only bind to sockets with compatible flags enabled
Thomas Graf [Thu, 15 Jan 2015 02:53:56 +0000 (03:53 +0100)]
vxlan: Only bind to sockets with compatible flags enabled

A VXLAN net_device looking for an appropriate socket may only consider
a socket which has a matching set of flags/extensions enabled. If
incompatible flags are enabled, return a conflict to have the caller
create a distinct socket with distinct port.

The OVS VXLAN port is kept unaware of extensions at this point.

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovxlan: Group Policy extension
Thomas Graf [Thu, 15 Jan 2015 02:53:55 +0000 (03:53 +0100)]
vxlan: Group Policy extension

Implements supports for the Group Policy VXLAN extension [0] to provide
a lightweight and simple security label mechanism across network peers
based on VXLAN. The security context and associated metadata is mapped
to/from skb->mark. This allows further mapping to a SELinux context
using SECMARK, to implement ACLs directly with nftables, iptables, OVS,
tc, etc.

The group membership is defined by the lower 16 bits of skb->mark, the
upper 16 bits are used for flags.

SELinux allows to manage label to secure local resources. However,
distributed applications require ACLs to implemented across hosts. This
is typically achieved by matching on L2-L4 fields to identify the
original sending host and process on the receiver. On top of that,
netlabel and specifically CIPSO [1] allow to map security contexts to
universal labels.  However, netlabel and CIPSO are relatively complex.
This patch provides a lightweight alternative for overlay network
environments with a trusted underlay. No additional control protocol
is required.

           Host 1:                       Host 2:

      Group A        Group B        Group B     Group A
      +-----+   +-------------+    +-------+   +-----+
      | lxc |   | SELinux CTX |    | httpd |   | VM  |
      +--+--+   +--+----------+    +---+---+   +--+--+
  \---+---/                     \----+---/
      |                              |
  +---+---+                      +---+---+
  | vxlan |                      | vxlan |
  +---+---+                      +---+---+
      +------------------------------+

Backwards compatibility:
A VXLAN-GBP socket can receive standard VXLAN frames and will assign
the default group 0x0000 to such frames. A Linux VXLAN socket will
drop VXLAN-GBP  frames. The extension is therefore disabled by default
and needs to be specifically enabled:

   ip link add [...] type vxlan [...] gbp

In a mixed environment with VXLAN and VXLAN-GBP sockets, the GBP socket
must run on a separate port number.

Examples:
 iptables:
  host1# iptables -I OUTPUT -m owner --uid-owner 101 -j MARK --set-mark 0x200
  host2# iptables -I INPUT -m mark --mark 0x200 -j DROP

 OVS:
  # ovs-ofctl add-flow br0 'in_port=1,actions=load:0x200->NXM_NX_TUN_GBP_ID[],NORMAL'
  # ovs-ofctl add-flow br0 'in_port=2,tun_gbp_id=0x200,actions=drop'

[0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
[1] http://lwn.net/Articles/204905/

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Thu, 15 Jan 2015 05:53:17 +0000 (00:53 -0500)]
Merge git://git./linux/kernel/git/davem/net

Conflicts:
drivers/net/xen-netfront.c

Minor overlapping changes in xen-netfront.c, mostly to do
with some buffer management changes alongside the split
of stats into TX and RX.

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Wed, 14 Jan 2015 22:17:37 +0000 (11:17 +1300)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Don't use uninitialized data in IPVS, from Dan Carpenter.

 2) conntrack race fixes from Pablo Neira Ayuso.

 3) Fix TX hangs with i40e, from Jesse Brandeburg.

 4) Fix budget return from poll calls in dnet and alx, from Eric
    Dumazet.

 5) Fix bugus "if (unlikely(x) < 0)" test in AF_PACKET, from Christoph
    Jaeger.

 6) Fix bug introduced by conversion to list_head in TIPC retransmit
    code, from Jon Paul Maloy.

 7) Don't use GFP_NOIO under spinlock in USB kaweth driver, from Alexey
    Khoroshilov.

 8) Fix bridge build with INET disabled, from Arnd Bergmann.

 9) Fix netlink array overrun for PROBE attributes in openvswitch, from
    Thomas Graf.

10) Don't hold spinlock across synchronize_irq() in tg3 driver, from
    Prashant Sreedharan.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (44 commits)
  tg3: Release tp->lock before invoking synchronize_irq()
  tg3: tg3_reset_task() needs to use rtnl_lock to synchronize
  tg3: tg3_timer() should grab tp->lock before checking for tp->irq_sync
  team: avoid possible underflow of count_pending value for notify_peers and mcast_rejoin
  openvswitch: packet messages need their own probe attribtue
  i40e: adds FCoE configure option
  cxgb4vf: Fix queue allocation for 40G adapter
  netdevice: Add missing parentheses in macro
  bridge: only provide proxy ARP when CONFIG_INET is enabled
  neighbour: fix base_reachable_time(_ms) not effective immediatly when changed
  net: fec: fix MDIO bus assignement for dual fec SoC's
  xen-netfront: use different locks for Rx and Tx stats
  drivers: net: cpsw: fix multicast flush in dual emac mode
  cxgb4vf: Initialize mdio_addr before using it
  net: Corrected the comment describing the ndo operations to reflect the actual prototype for couple of operations
  usb/kaweth: use GFP_ATOMIC under spin_lock in usb_start_wait_urb()
  MAINTAINERS: add me as ibmveth maintainer
  tipc: fix bug in broadcast retransmit code
  update ip-sysctl.txt documentation (v2)
  net/at91_ether: prepare and unprepare clock
  ...

9 years agoMerge branch 'tg3-net'
David S. Miller [Wed, 14 Jan 2015 22:05:55 +0000 (17:05 -0500)]
Merge branch 'tg3-net'

Prashant Sreedharan says:

====================
tg3: synchronize_irq() should be called without taking locks

v2: Added Reported-by, Tested-by fields and reference to the thread that
    reported the problem

This series addresses the problem reported by Peter Hurley in mail thread
https://lkml.org/lkml/2015/1/12/1082
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: Release tp->lock before invoking synchronize_irq()
Prashant Sreedharan [Wed, 14 Jan 2015 19:34:44 +0000 (11:34 -0800)]
tg3: Release tp->lock before invoking synchronize_irq()

synchronize_irq() can sleep waiting, for pending IRQ handlers so driver
should release the tp->lock spin lock before invoking synchronize_irq()

Reported-by: Peter Hurley <peter@hurleysoftware.com>
Tested-by: Peter Hurley <peter@hurleysoftware.com>
Signed-off-by: Prashant Sreedharan <prashant@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: tg3_reset_task() needs to use rtnl_lock to synchronize
Prashant Sreedharan [Wed, 14 Jan 2015 19:34:17 +0000 (11:34 -0800)]
tg3: tg3_reset_task() needs to use rtnl_lock to synchronize

Currently tg3_reset_task() uses only tp->lock for synchronizing with code
paths like tg3_open() etc. But since tp->lock is released before doing
synchronize_irq(), rtnl_lock should be taken in tg3_reset_task() to
synchronize it with other code paths.

Reported-by: Peter Hurley <peter@hurleysoftware.com>
Tested-by: Peter Hurley <peter@hurleysoftware.com>
Signed-off-by: Prashant Sreedharan <prashant@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotg3: tg3_timer() should grab tp->lock before checking for tp->irq_sync
Prashant Sreedharan [Wed, 14 Jan 2015 19:33:49 +0000 (11:33 -0800)]
tg3: tg3_timer() should grab tp->lock before checking for tp->irq_sync

This is to avoid the race between tg3_timer() and the execution paths
which does not invoke tg3_timer_stop() and releases tp->lock before
calling synchronize_irq()

Reported-by: Peter Hurley <peter@hurleysoftware.com>
Tested-by: Peter Hurley <peter@hurleysoftware.com>
Signed-off-by: Prashant Sreedharan <prashant@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Wed, 14 Jan 2015 21:54:30 +0000 (10:54 +1300)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "Two bugfixes for arm64.  I will have another pull request next week,
  but otherwise things are calm"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  arm64: KVM: Fix HCR setting for 32bit guests
  arm64: KVM: Fix TLB invalidation by IPA/VMID

9 years agoteam: avoid possible underflow of count_pending value for notify_peers and mcast_rejoin
Jiri Pirko [Wed, 14 Jan 2015 17:15:30 +0000 (18:15 +0100)]
team: avoid possible underflow of count_pending value for notify_peers and mcast_rejoin

This patch is fixing a race condition that may cause setting
count_pending to -1, which results in unwanted big bulk of arp messages
(in case of "notify peers").

Consider following scenario:

count_pending == 2
   CPU0                                           CPU1
team_notify_peers_work
  atomic_dec_and_test (dec count_pending to 1)
  schedule_delayed_work
 team_notify_peers
   atomic_add (adding 1 to count_pending)
team_notify_peers_work
  atomic_dec_and_test (dec count_pending to 1)
  schedule_delayed_work
team_notify_peers_work
  atomic_dec_and_test (dec count_pending to 0)
   schedule_delayed_work
team_notify_peers_work
  atomic_dec_and_test (dec count_pending to -1)

Fix this race by using atomic_dec_if_positive - that will prevent
count_pending running under 0.

Fixes: fc423ff00df3a1955441 ("team: add peer notification")
Fixes: 492b200efdd20b8fcfd  ("team: add support for sending multicast rejoins")
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Wed, 14 Jan 2015 21:50:29 +0000 (10:50 +1300)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:
 "Two small performance tweaks, the plumbing for the execveat system
  call and a couple of bug fixes"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/uprobes: fix user space PER events
  s390/bpf: Fix JMP_JGE_X (A > X) and JMP_JGT_X (A >= X)
  s390/bpf: Fix ALU_NEG (A = -A)
  s390/mm: avoid using pmd_to_page for !USE_SPLIT_PMD_PTLOCKS
  s390/timex: fix get_tod_clock_ext() inline assembly
  s390: wire up execveat syscall
  s390/kernel: use stnsm 255 instead of stosm 0
  s390/vtime: Get rid of redundant WARN_ON
  s390/zcrypt: kernel oops at insmod of the z90crypt device driver

9 years agoopenvswitch: packet messages need their own probe attribtue
Thomas Graf [Wed, 14 Jan 2015 13:56:19 +0000 (13:56 +0000)]
openvswitch: packet messages need their own probe attribtue

User space is currently sending a OVS_FLOW_ATTR_PROBE for both flow
and packet messages. This leads to an out-of-bounds access in
ovs_packet_cmd_execute() because OVS_FLOW_ATTR_PROBE >
OVS_PACKET_ATTR_MAX.

Introduce a new OVS_PACKET_ATTR_PROBE with the same numeric value
as OVS_FLOW_ATTR_PROBE to grow the range of accepted packet attributes
while maintaining to be binary compatible with existing OVS binaries.

Fixes: 05da589 ("openvswitch: Add support for OVS_FLOW_ATTR_PROBE.")
Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
Tracked-down-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Reviewed-by: Jesse Gross <jesse@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40e: adds FCoE configure option
Vasu Dev [Wed, 14 Jan 2015 13:14:07 +0000 (05:14 -0800)]
i40e: adds FCoE configure option

Adds FCoE config option I40E_FCOE, so that FCoE can be enabled
as needed but otherwise have it disabled by default.

This also eliminate multiple FCoE config checks, instead now just
one config check for CONFIG_I40E_FCOE.

The I40E FCoE was added with 3.17 kernel and therefore this patch
shall be applied to stable 3.17 kernel also.

CC: <stable@vger.kernel.org>
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Tested-by: Jim Young <jamesx.m.young@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4vf: Fix queue allocation for 40G adapter
Hariprasad Shenai [Wed, 14 Jan 2015 12:28:10 +0000 (17:58 +0530)]
cxgb4vf: Fix queue allocation for 40G adapter

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge tag 'locks-v3.19-1' of git://git.samba.org/jlayton/linux
Linus Torvalds [Wed, 14 Jan 2015 21:38:07 +0000 (10:38 +1300)]
Merge tag 'locks-v3.19-1' of git://git.samba.org/jlayton/linux

Pull file locking fix from Jeff Layton:
 "Just a simple bugfix for a regression that I introduced into v3.18
  with the internal lease API overhaul -- mea culpa.  Kudos to Linda and
  Neil for tracking this down and fixing it"

* tag 'locks-v3.19-1' of git://git.samba.org/jlayton/linux:
  locks: fix NULL-deref in generic_delete_lease

9 years agoipv6:icmp:remove unnecessary brackets
zhuyj [Wed, 14 Jan 2015 09:23:59 +0000 (17:23 +0800)]
ipv6:icmp:remove unnecessary brackets

There are too many brackets. Maybe only one bracket is enough.

Signed-off-by: Zhu Yanjun <Yanjun.Zhu@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonetdevice: Add missing parentheses in macro
Benjamin Poirier [Wed, 14 Jan 2015 07:52:35 +0000 (16:52 +0900)]
netdevice: Add missing parentheses in macro

For example, one could conceivably call
for_each_netdev_in_bond_rcu(condition ? bond1 : bond2, slave)
and get an unexpected result.

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoopenvswitch: Introduce ovs_tunnel_route_lookup
Fan Du [Wed, 14 Jan 2015 05:10:35 +0000 (13:10 +0800)]
openvswitch: Introduce ovs_tunnel_route_lookup

Introduce ovs_tunnel_route_lookup to consolidate route lookup
shared by vxlan, gre, and geneve ports.

Signed-off-by: Fan Du <fan.du@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Wed, 14 Jan 2015 21:27:56 +0000 (10:27 +1300)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block layer fixes from Jens Axboe:
 "The major part is an update to the NVMe driver, fixing various issues
  around surprise removal and hung controllers.  Most of that is from
  Keith, and parts are simple blk-mq fixes or exports/additions of minor
  functions to aid this effort, and parts are changes directly to the
  NVMe driver.

  Apart from the above, this contains:

   - Small blk-mq change from me, killing an unused member of the
     hardware queue structure.

   - Small fix from Ming Lei, fixing up a few drivers that didn't
     properly check for ERR_PTR() returns from blk_mq_init_queue()"

* 'for-linus' of git://git.kernel.dk/linux-block:
  NVMe: Fix locking on abort handling
  NVMe: Start and stop h/w queues on reset
  NVMe: Command abort handling fixes
  NVMe: Admin queue removal handling
  NVMe: Reference count admin queue usage
  NVMe: Start all requests
  blk-mq: End unstarted requests on a dying queue
  blk-mq: Allow requests to never expire
  blk-mq: Add helper to abort requeued requests
  blk-mq: Let drivers cancel requeue_work
  blk-mq: Export if requests were started
  blk-mq: Wake tasks entering queue on dying
  blk-mq: get rid of ->cmd_size in the hardware queue
  block: fix checking return value of blk_mq_init_queue
  block: wake up waiters when a queue is marked dying
  NVMe: Fix double free irq
  blk-mq: Export freeze/unfreeze functions
  blk-mq: Exit queue on alloc failure

9 years agoMerge branch 'vxlan_rco'
David S. Miller [Wed, 14 Jan 2015 20:20:11 +0000 (15:20 -0500)]
Merge branch 'vxlan_rco'

Tom Herbert says:

====================
net: Remote checksum offload for VXLAN

This patch set adds support for remote checksum offload in VXLAN.

The remote checksum offload is generalized by creating a common
function (remcsum_adjust) that does the work of modifying the
checksum in remote checksum offload. This function can be called
from normal or GRO path. GUE was modified to use this function.

To support RCO is VXLAN we use the 9th bit in the reserved
flags to indicated remote checksum offload. The start and offset
values are encoded n a compressed form in the low order (reserved)
byte of the vni field.

Remote checksum offload is described in
https://tools.ietf.org/html/draft-herbert-remotecsumoffload-01

Changes in v2:
  - Add udp_offload_callbacks which has GRO functions that take a
    udp_offload pointer argument. This argument can be used to retrieve
    a per port structure of the encapsulation for use in gro processing
    (mostly by doing container_of on the structure).
  - Use the 10th bit in VXLAN flags for RCO which does not seem to
    conflict with other proposals at this time (ie. VXLAN-GPE and
    VXLAN-GPB)
  - Require that RCO must be explicitly enabled on the receiver
    as well as the sender.

Tested by running 200 TCP_STREAM connections with VXLAN (over IPv4).

With UDP checksums and Remote Checksum Offload
  IPv4
      Client
        11.84% CPU utilization
      Server
        12.96% CPU utilization
      9197 Mbps
  IPv6
      Client
        12.46% CPU utilization
      Server
        14.48% CPU utilization
      8963 Mbps

With UDP checksums, no remote checksum offload
  IPv4
      Client
        15.67% CPU utilization
      Server
        14.83% CPU utilization
      9094 Mbps
  IPv6
      Client
        16.21% CPU utilization
      Server
        14.32% CPU utilization
      9058 Mbps

No UDP checksums
  IPv4
      Client
        15.03% CPU utilization
      Server
        23.09% CPU utilization
      9089 Mbps
  IPv6
      Client
        16.18% CPU utilization
      Server
        26.57% CPU utilization
       8954 Mbps
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agovxlan: Remote checksum offload
Tom Herbert [Tue, 13 Jan 2015 01:00:38 +0000 (17:00 -0800)]
vxlan: Remote checksum offload

Add support for remote checksum offload in VXLAN. This uses a
reserved bit to indicate that RCO is being done, and uses the low order
reserved eight bits of the VNI to hold the start and offset values in a
compressed manner.

Start is encoded in the low order seven bits of VNI. This is start >> 1
so that the checksum start offset is 0-254 using even values only.
Checksum offset (transport checksum field) is indicated in the high
order bit in the low order byte of the VNI. If the bit is set, the
checksum field is for UDP (so offset = start + 6), else checksum
field is for TCP (so offset = start + 16). Only TCP and UDP are
supported in this implementation.

Remote checksum offload for VXLAN is described in:

https://tools.ietf.org/html/draft-herbert-vxlan-rco-00

Tested by running 200 TCP_STREAM connections with VXLAN (over IPv4).

With UDP checksums and Remote Checksum Offload
  IPv4
      Client
        11.84% CPU utilization
      Server
        12.96% CPU utilization
      9197 Mbps
  IPv6
      Client
        12.46% CPU utilization
      Server
        14.48% CPU utilization
      8963 Mbps

With UDP checksums, no remote checksum offload
  IPv4
      Client
        15.67% CPU utilization
      Server
        14.83% CPU utilization
      9094 Mbps
  IPv6
      Client
        16.21% CPU utilization
      Server
        14.32% CPU utilization
      9058 Mbps

No UDP checksums
  IPv4
      Client
        15.03% CPU utilization
      Server
        23.09% CPU utilization
      9089 Mbps
  IPv6
      Client
        16.18% CPU utilization
      Server
        26.57% CPU utilization
       8954 Mbps

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoudp: pass udp_offload struct to UDP gro callbacks
Tom Herbert [Tue, 13 Jan 2015 01:00:37 +0000 (17:00 -0800)]
udp: pass udp_offload struct to UDP gro callbacks

This patch introduces udp_offload_callbacks which has the same
GRO functions (but not a GSO function) as offload_callbacks,
except there is an argument to a udp_offload struct passed to
gro_receive and gro_complete functions. This additional argument
can be used to retrieve the per port structure of the encapsulation
for use in gro processing (mostly by doing container_of on the
structure).

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: only provide proxy ARP when CONFIG_INET is enabled
Arnd Bergmann [Tue, 13 Jan 2015 14:10:27 +0000 (15:10 +0100)]
bridge: only provide proxy ARP when CONFIG_INET is enabled

When IPV4 support is disabled, we cannot call arp_send from
the bridge code, which would result in a kernel link error:

net/built-in.o: In function `br_handle_frame_finish':
:(.text+0x59914): undefined reference to `arp_send'
:(.text+0x59a50): undefined reference to `arp_tbl'

This makes the newly added proxy ARP support in the bridge
code depend on the CONFIG_INET symbol and lets the compiler
optimize the code out to avoid the link error.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 958501163ddd ("bridge: Add support for IEEE 802.11 Proxy ARP")
Cc: Kyeyoon Park <kyeyoonp@codeaurora.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agor8152: replace tasklet with NAPI
hayeswang [Mon, 12 Jan 2015 04:06:23 +0000 (12:06 +0800)]
r8152: replace tasklet with NAPI

Replace tasklet with NAPI.

Add rx_queue to queue the remaining rx packets if the number of the
rx packets is more than the request from poll().

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocfg80211: docs: remove station_info_flags
Johannes Berg [Wed, 14 Jan 2015 12:57:36 +0000 (13:57 +0100)]
cfg80211: docs: remove station_info_flags

I removed the enum earlier, but forgot to remove it from the
documentation - do that now.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agonl80211: send netdetect configuration info in NL80211_CMD_GET_WOWLAN
Luciano Coelho [Fri, 9 Jan 2015 12:06:37 +0000 (14:06 +0200)]
nl80211: send netdetect configuration info in NL80211_CMD_GET_WOWLAN

Send the netdetect configuration information in the response to
NL8021_CMD_GET_WOWLAN commands.  This includes the scan interval,
SSIDs to match and frequencies to scan.

Additionally, add the NL80211_WOWLAN_TRIG_NET_DETECT with
NL80211_ATTR_WOWLAN_TRIGGERS_SUPPORTED.

Signed-off-by: Luciano Coelho <luciano.coelho@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: avoid reg-hints in self-managed only systems
Arik Nemtsov [Wed, 7 Jan 2015 14:47:20 +0000 (16:47 +0200)]
cfg80211: avoid reg-hints in self-managed only systems

When a system contains only self-managed regulatory devices all hints
from the regulatory core are ignored. Stop hint processing early in this
case. These systems usually don't have CRDA deployed, which results in
endless (irrelevent) logs of the form:
cfg80211: Calling CRDA to update world regulatory domain

Make sure there's at least one self-managed device before discarding a
hint, in order to prevent initial hints from disappearing on CRDA
managed systems.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agocfg80211: introduce sync regdom set API for self-managed
Arik Nemtsov [Wed, 7 Jan 2015 14:47:19 +0000 (16:47 +0200)]
cfg80211: introduce sync regdom set API for self-managed

A self-managed device will sometimes need to set its regdomain synchronously.
Notably it should be set before usermode has a chance to query it. Expose
a new API to accomplish this which requires the RTNL.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Ilan Peer <ilan.peer@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: don't defer scans in case of radar detection
Eliad Peller [Wed, 7 Jan 2015 15:50:11 +0000 (17:50 +0200)]
mac80211: don't defer scans in case of radar detection

Radar detection can last indefinite time. There is no
point in deferring a scan request in this case - simply
return -EBUSY.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: consider only relevant vifs for radar_required calculation
Eliad Peller [Wed, 7 Jan 2015 15:50:10 +0000 (17:50 +0200)]
mac80211: consider only relevant vifs for radar_required calculation

ctx->conf.radar_enabled should reflect whether radar
detection is enabled for the channel context.

When calculating it, make it consider only the vifs
that have this context assigned (instead of all the
vifs).

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: remove local->radar_detect_enabled
Eliad Peller [Wed, 7 Jan 2015 15:50:09 +0000 (17:50 +0200)]
mac80211: remove local->radar_detect_enabled

local->radar_detect_enabled should tell whether
radar_detect is enabled on any interface belonging
to local.

However, it's not getting updated correctly
in many cases (actually, when testing with hwsim
it's never been set, even when the dfs master
is beaconing).

Instead of handling all the corner cases
(e.g. channel switch), simply check whether
radar detection is enabled only when needed,
instead of caching the result.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: add TDLS supported channels correctly
Arik Nemtsov [Wed, 7 Jan 2015 14:45:07 +0000 (16:45 +0200)]
mac80211: add TDLS supported channels correctly

The function adding the supported channels IE during a TDLS connection had
several issues:
1. If the entire subband is usable, the function exitted the loop without
   adding it
2. The function only checked chandef_usable, ignoring flags like RADAR
   which would prevent TDLS off-channel communcation.
3. HT20 was explicitly required in the chandef, while not a requirement
   for TDLS off-channel.

Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agomac80211: let flush() drop packets when possible
Emmanuel Grumbach [Wed, 7 Jan 2015 13:42:39 +0000 (15:42 +0200)]
mac80211: let flush() drop packets when possible

When roaming / suspending, it makes no sense to wait until
the transmit queues of the device are empty. In extreme
condition they can be starved (VO saturating the air), but
even in regular cases, it is pointless to delay the roaming
because the low level driver is trying to send packets to
an AP which is far away. We'd rather drop these packets and
let TCP retransmit if needed. This will allow to speed up
the roaming.

For suspend, the explanation is even more trivial.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoMerge branch 'hip04'
David S. Miller [Wed, 14 Jan 2015 06:52:51 +0000 (01:52 -0500)]
Merge branch 'hip04'

Ding Tianhong says:

====================
add hisilicon hip04 ethernet driver

v13:
- Fix the problem of alignment parameters for function and checkpatch warming.

v12:
- According Alex's suggestion, modify the changelog and add MODULE_DEVICE_TABLE
  for hip04 ethernet.

v11:
- Add ethtool support for tx coalecse getting and setting, the xmit_more
  is not supported for this patch, but I think it could work for hip04,
  will support it later after some tests for performance better.

  Here are some performance test results by ping and iperf(add tx_coalesce_frames/users),
  it looks that the performance and latency is more better by tx_coalesce_frames/usecs.

  - Before:
    $ ping 192.168.1.1 ...
    === 192.168.1.1 ping statistics ===
    24 packets transmitted, 24 received, 0% packet loss, time 22999ms
    rtt min/avg/max/mdev = 0.180/0.202/0.403/0.043 ms

    $ iperf -c 192.168.1.1 ...
    [ ID] Interval       Transfer     Bandwidth
    [  3]  0.0- 1.0 sec   115 MBytes   945 Mbits/sec

  - After:
    $ ping 192.168.1.1 ...
    === 192.168.1.1 ping statistics ===
    24 packets transmitted, 24 received, 0% packet loss, time 22999ms
    rtt min/avg/max/mdev = 0.178/0.190/0.380/0.041 ms

    $ iperf -c 192.168.1.1 ...
    [ ID] Interval       Transfer     Bandwidth
    [  3]  0.0- 1.0 sec   115 MBytes   965 Mbits/sec

v10:
- According Arnd's suggestion, remove the skb_orphan and use the hrtimer
  for the cleanup of the TX queue and add some modification for the hip04
  drivers.
  1) drop the broken skb_orphan call
  2) drop the workqueue
  3) batch cleanup based on tx_coalesce_frames/usecs for better throughput
  4) use a reasonable default tx timeout (200us, could be shorted
     based on measurements) with a range timer
  5) fix napi poll function return value
  6) use a lockless queue for cleanup

v9:
- There is no tx completion interrupts to free DMAd Tx packets, it means taht
  we rely on new tx packets arriving to run the destructors of completed packets,
  which open up space in their sockets's send queues. Sometimes we don't get such
  new packets causing Tx to stall, a single UDP transmitter is a good example of
  this situation, so we need a clean up workqueue to reclaims completed packets,
  the workqueue will only free the last packets which is already stay for several jiffies.
  Also fix some format cleanups.

v8:
- Use poll to reclaim xmitted buffer as workaround since no tx done interrupt

v7:
- Remove select NET_CORE in 0002

v6:
- Suggest by Russell: Use netdev_sent_queue & netdev_completed_queue to solve latency issue
  Also shorten the period of timer, which is used to wakeup the queue since no
  tx completed interrupt.

v5:
- no big change, fix typo

v4:
- Modify accoringly to the suggetion from Arnd, Florian, Eric, David
  Use of_parse_phandle_with_fixed_args & syscon_node_to_regmap get ppe info
  Add skb_orphan() and tx_timer for reclaim since no tx_finished interrupt
  Update timeout, and move of_phy_connect to probe to reuse open/stop

v3:
- Suggest from Arnd, use syscon & regmap_write/read to replace static void __iomem *ppebase.
  Modify hisilicon-hip04-net.txt accrordingly to suggestion from Florian and Sergei.

v2:
- Got many suggestions from Russell, Arnd, Florian, Mark and Sergei
  Remove memcpy, use dma_map/unmap_single, use dma_alloc_coherent rather than dma_pool, etc.
  Refer property in ethernet.txt, change ppe description, etc.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: hisilicon: new hip04 ethernet driver
dingtianhong [Wed, 14 Jan 2015 06:34:14 +0000 (14:34 +0800)]
net: hisilicon: new hip04 ethernet driver

Support Hisilicon hip04 ethernet driver, including 100M / 1000M controller.
The controller has no tx done interrupt, reclaim xmitted buffer in the poll.

v13: Fix the problem of alignment parameters for function and checkpatch warming.

v12: According Alex's suggestion, modify the changelog and add MODULE_DEVICE_TABLE
     for hip04 ethernet.

v11: Add ethtool support for tx coalecse getting and setting, the xmit_more
     is not supported for this patch, but I think it could work for hip04,
     will support it later after some tests for performance better.

     Here are some performance test results by ping and iperf(add tx_coalesce_frames/users),
     it looks that the performance and latency is more better by tx_coalesce_frames/usecs.

     - Before:
     $ ping 192.168.1.1 ...
     === 192.168.1.1 ping statistics ===
     24 packets transmitted, 24 received, 0% packet loss, time 22999ms
     rtt min/avg/max/mdev = 0.180/0.202/0.403/0.043 ms

     $ iperf -c 192.168.1.1 ...
     [ ID] Interval       Transfer     Bandwidth
     [  3]  0.0- 1.0 sec   115 MBytes   945 Mbits/sec

     - After:
     $ ping 192.168.1.1 ...
     === 192.168.1.1 ping statistics ===
     24 packets transmitted, 24 received, 0% packet loss, time 22999ms
     rtt min/avg/max/mdev = 0.178/0.190/0.380/0.041 ms

     $ iperf -c 192.168.1.1 ...
     [ ID] Interval       Transfer     Bandwidth
     [  3]  0.0- 1.0 sec   115 MBytes   965 Mbits/sec

v10: According David Miller and Arnd Bergmann's suggestion, add some modification
     for v9 version
     - drop the workqueue
     - batch cleanup based on tx_coalesce_frames/usecs for better throughput
     - use a reasonable default tx timeout (200us, could be shorted
       based on measurements) with a range timer
     - fix napi poll function return value
     - use a lockless queue for cleanup

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: hisilicon: new hip04 MDIO driver
Zhangfei Gao [Wed, 14 Jan 2015 06:34:13 +0000 (14:34 +0800)]
net: hisilicon: new hip04 MDIO driver

Hisilicon hip04 platform mdio driver
Reuse Marvell phy drivers/net/phy/marvell.c

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoDocumentation: add Device tree bindings for Hisilicon hip04 ethernet
Zhangfei Gao [Wed, 14 Jan 2015 06:34:12 +0000 (14:34 +0800)]
Documentation: add Device tree bindings for Hisilicon hip04 ethernet

This patch adds the Device Tree bindings for the Hisilicon hip04
Ethernet controller, including 100M / 1000M controller.

Signed-off-by: Zhangfei Gao <zhangfei.gao@linaro.org>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoneighbour: fix base_reachable_time(_ms) not effective immediatly when changed
Jean-Francois Remy [Wed, 14 Jan 2015 03:22:39 +0000 (04:22 +0100)]
neighbour: fix base_reachable_time(_ms) not effective immediatly when changed

When setting base_reachable_time or base_reachable_time_ms on a
specific interface through sysctl or netlink, the reachable_time
value is not updated.

This means that neighbour entries will continue to be updated using the
old value until it is recomputed in neigh_period_work (which
    recomputes the value every 300*HZ).
On systems with HZ equal to 1000 for instance, it means 5mins before
the change is effective.

This patch changes this behavior by recomputing reachable_time after
each set on base_reachable_time or base_reachable_time_ms.
The new value will become effective the next time the neighbour's timer
is triggered.

Changes are made in two places: the netlink code for set and the sysctl
handling code. For sysctl, I use a proc_handler. The ipv6 network
code does provide its own handler but it already refreshes
reachable_time correctly so it's not an issue.
Any other user of neighbour which provide its own handlers must
refresh reachable_time.

Signed-off-by: Jean-Francois Remy <jeff@melix.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: fec: fix MDIO bus assignement for dual fec SoC's
Stefan Agner [Tue, 13 Jan 2015 23:20:21 +0000 (00:20 +0100)]
net: fec: fix MDIO bus assignement for dual fec SoC's

On i.MX28, the MDIO bus is shared between the two FEC instances.
The driver makes sure that the second FEC uses the MDIO bus of the
first FEC. This is done conditionally if FEC_QUIRK_ENET_MAC is set.
However, in newer designs, such as Vybrid or i.MX6SX, each FEC MAC
has its own MDIO bus. Simply removing the quirk FEC_QUIRK_ENET_MAC
is not an option since other logic, triggered by this quirk, is
still needed.

Furthermore, there are board designs which use the same MDIO bus
for both PHY's even though the second bus would be available on the
SoC side. Such layout are popular since it saves pins on SoC side.
Due to the above quirk, those boards currently do work fine. The
boards in the mainline tree with such a layout are:
- Freescale Vybrid Tower with TWR-SER2 (vf610-twr.dts)
- Freescale i.MX6 SoloX SDB Board (imx6sx-sdb.dts)

This patch adds a new quirk FEC_QUIRK_SINGLE_MDIO for i.MX28, which
makes sure that the MDIO bus of the first FEC is used in any case.

However, the boards above do have a SoC with a MDIO bus for each FEC
instance. But the PHY's are not connected in a 1:1 configuration. A
proper device tree description is needed to allow the driver to
figure out where to find its PHY. This patch fixes that shortcoming
by adding a MDIO bus child node to the first FEC instance, along
with the two PHY's on that bus, and making use of the phy-handle
property to add a reference to the PHY's.

Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: improved ethtool statistics support
Xander Huff [Tue, 13 Jan 2015 22:15:51 +0000 (16:15 -0600)]
net/macb: improved ethtool statistics support

Currently `ethtool -S` simply returns "no stats available". It
would be more useful to see what the various ethtool statistics
registers' values are. This change implements get_ethtool_stats,
get_strings, and get_sset_count functions to accomplish this.

Read all GEM statistics registers and sum them into
macb.ethtool_stats. Add the necessary infrastructure to make this
accessible via `ethtool -S`.

Update gem_update_stats to utilize ethtool_stats.

Signed-off-by: Xander Huff <xander.huff@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/macb: Adding comments to various #defs to make interpretation easier
Xander Huff [Tue, 13 Jan 2015 22:15:50 +0000 (16:15 -0600)]
net/macb: Adding comments to various #defs to make interpretation easier

This change is to help improve at-a-glace knowledge of the purpose of the
various Cadence MACB/GEM registers. Comments are more helpful for human
readability than short acronyms.

Describe various #define varibles Cadence MACB/GEM registers as documented
in Xilinix's "Zynq-7000 All Programmable SoC TechnicalReference Manual, v1.9.1
(UG-585)"

Signed-off-by: Xander Huff <xander.huff@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'xen-netfront-next'
David S. Miller [Wed, 14 Jan 2015 05:22:12 +0000 (00:22 -0500)]
Merge branch 'xen-netfront-next'

David Vrabel says:

====================
xen-netfront: refactor making Tx requests

As netfront as evolved to handle different sorts of skbs the code to
fill a Tx requests has been copy and pasted several times.  The series
refactors this and a few other areas.

The first patch is to a Xen header but this can be merged via
net-next.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen-netfront: refactor making Tx requests
David Vrabel [Tue, 13 Jan 2015 17:16:44 +0000 (17:16 +0000)]
xen-netfront: refactor making Tx requests

Eliminate all the duplicate code for making Tx requests by
consolidating them into a single xennet_make_one_txreq() function.

xennet_make_one_txreq() and xennet_make_txreqs() work with pages and
offsets so it will be easier to make netfront handle highmem frags in
the future.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen-netfront: refactor skb slot counting
David Vrabel [Tue, 13 Jan 2015 17:16:43 +0000 (17:16 +0000)]
xen-netfront: refactor skb slot counting

A function to count the number of slots an skb needs is more useful
than one that counts the slots needed for only the frags.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen: add page_to_mfn()
David Vrabel [Tue, 13 Jan 2015 17:16:42 +0000 (17:16 +0000)]
xen: add page_to_mfn()

pfn_to_mfn(page_to_pfn(p)) is a common use case so add a generic
helper for it.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Add MAINTAINERS entry
Thomas Graf [Tue, 13 Jan 2015 00:01:24 +0000 (01:01 +0100)]
rhashtable: Add MAINTAINERS entry

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorhashtable: Lower/upper bucket may map to same lock while shrinking
Thomas Graf [Mon, 12 Jan 2015 23:58:21 +0000 (23:58 +0000)]
rhashtable: Lower/upper bucket may map to same lock while shrinking

Each per bucket lock covers a configurable number of buckets. While
shrinking, two buckets in the old table contain entries for a single
bucket in the new table. We need to lock down both while linking.
Check if they are protected by different locks to avoid a recursive
lock.

Fixes: 97defe1e ("rhashtable: Per bucket locks & deferred expansion/shrinking")
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'leds-fixes-for-3.19' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 13 Jan 2015 23:22:56 +0000 (12:22 +1300)]
Merge branch 'leds-fixes-for-3.19' of git://git./linux/kernel/git/cooloney/linux-leds

Pull LED fix from Bryan Wu.

* 'leds-fixes-for-3.19' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/linux-leds:
  leds: netxbig: fix oops at probe time

9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/borntraeger...
Linus Torvalds [Tue, 13 Jan 2015 22:54:12 +0000 (11:54 +1300)]
Merge tag 'for-linus' of git://git./linux/kernel/git/borntraeger/linux

Pull WRITE_ONCE argument order change from Christian Borntraeger:
 "As discussed on LKML[1] it was agreed that WRITE_ONCE(x, val) is
  better than ASSIGN_ONCE(val, x)

  Lets change that for 3.19 as 3.19 has no user yet, but the first users
  will hit linux-next soon"

[1] http://marc.info/?l=linux-kernel&m=142081181707596

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/borntraeger/linux:
  kernel: Change ASSIGN_ONCE(val, x) to WRITE_ONCE(x, val)

9 years agonet: rename vlan_tx_* helpers since "tx" is misleading there
Jiri Pirko [Tue, 13 Jan 2015 16:13:44 +0000 (17:13 +0100)]
net: rename vlan_tx_* helpers since "tx" is misleading there

The same macros are used for rx as well. So rename it.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet: sched: fix skb->protocol use in case of accelerated vlan path
Jiri Pirko [Tue, 13 Jan 2015 16:13:43 +0000 (17:13 +0100)]
net: sched: fix skb->protocol use in case of accelerated vlan path

tc code implicitly considers skb->protocol even in case of accelerated
vlan paths and expects vlan protocol type here. However, on rx path,
if the vlan header was already stripped, skb->protocol contains value
of next header. Similar situation is on tx path.

So for skbs that use skb->vlan_tci for tagging, use skb->vlan_proto instead.

Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoatm: horizon: Remove some unused functions
Rickard Strandqvist [Tue, 13 Jan 2015 18:50:21 +0000 (19:50 +0100)]
atm: horizon: Remove some unused functions

Removes some functions that are not used anywhere:
channel_to_vpivci() query_tx_channel_config() rx_disabled_handler()

This was partially found by using a static code analysis program called cppcheck.

Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoatm: lanai: Remove unused function
Rickard Strandqvist [Tue, 13 Jan 2015 18:45:19 +0000 (19:45 +0100)]
atm: lanai: Remove unused function

Remove the function aal5_spacefor() that is not used anywhere.

This was partially found by using a static code analysis program called cppcheck.

Signed-off-by: Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agotipc: correctly handle releasing a not fully initialized sock
Sasha Levin [Tue, 13 Jan 2015 17:46:41 +0000 (12:46 -0500)]
tipc: correctly handle releasing a not fully initialized sock

Commit f2f9800d4955 "tipc: make tipc node table aware of net
namespace" has added a dereference of sock->sk before making sure it's
not NULL, which makes releasing a tipc socket NULL pointer dereference
for sockets that are not fully initialized.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agosunvnet: fix rx packet length check to allow for TSO
David L Stevens [Tue, 13 Jan 2015 17:45:05 +0000 (12:45 -0500)]
sunvnet: fix rx packet length check to allow for TSO

This patch fixes the rx packet length check in the sunvnet driver to allow
for a TSO max packet length greater than the LDC channel negotiated MTU.
These are negotiated separately and there is no requirement that
port->tsolen be less than port->rmtu, but if it isn't, it'll drop packets
with rx length errors.

Signed-off-by: David L Stevens <david.stevens@oracle.com>
Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoxen-netfront: use different locks for Rx and Tx stats
David Vrabel [Tue, 13 Jan 2015 16:42:42 +0000 (16:42 +0000)]
xen-netfront: use different locks for Rx and Tx stats

In netfront the Rx and Tx path are independent and use different
locks.  The Tx lock is held with hard irqs disabled, but Rx lock is
held with only BH disabled.  Since both sides use the same stats lock,
a deadlock may occur.

  [ INFO: possible irq lock inversion dependency detected ]
  3.16.2 #16 Not tainted
  ---------------------------------------------------------
  swapper/0/0 just changed the state of lock:
   (&(&queue->tx_lock)->rlock){-.....}, at: [<c03adec8>]
  xennet_tx_interrupt+0x14/0x34
  but this lock took another, HARDIRQ-unsafe lock in the past:
   (&stat->syncp.seq#2){+.-...}
  and interrupts could create inverse lock ordering between them.
  other info that might help us debug this:
   Possible interrupt unsafe locking scenario:

         CPU0                    CPU1
         ----                    ----
    lock(&stat->syncp.seq#2);
                                 local_irq_disable();
                                 lock(&(&queue->tx_lock)->rlock);
                                 lock(&stat->syncp.seq#2);
    <Interrupt>
      lock(&(&queue->tx_lock)->rlock);

Using separate locks for the Rx and Tx stats fixes this deadlock.

Reported-by: Dmitry Piotrovsky <piotrovskydmitry@gmail.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomISDN: avoid arch specific __builtin_return_address call
Arnd Bergmann [Tue, 13 Jan 2015 16:10:58 +0000 (17:10 +0100)]
mISDN: avoid arch specific __builtin_return_address call

Not all architectures are able to call __builtin_return_address().
On ARM, the mISDN code produces this warning:

hardware/mISDN/w6692.c: In function 'w6692_dctrl':
hardware/mISDN/w6692.c:1181:75: warning: unsupported argument to '__builtin_return_address'
  pr_debug("%s: %s dev(%d) open from %p\n", card->name, __func__,
                                                                           ^
hardware/mISDN/mISDNipac.c: In function 'open_dchannel':
hardware/mISDN/mISDNipac.c:759:75: warning: unsupported argument to '__builtin_return_address'
  pr_debug("%s: %s dev(%d) open from %p\n", isac->name, __func__,
                                                                           ^

In a lot of cases, this is relatively easy to work around by
passing the value of __builtin_return_address(0) from the
callers into the functions that want it. One exception is
the indirect 'open' function call in struct isac_hw. While it
would be possible to fix this as well, this patch only addresses
the other callers properly and lets this one return the direct
parent function, which should be good enough.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoinfiniband: mlx5: avoid a compile-time warning
Arnd Bergmann [Tue, 13 Jan 2015 16:09:43 +0000 (17:09 +0100)]
infiniband: mlx5: avoid a compile-time warning

The return type of find_first_bit() is architecture specific,
on ARM it is 'unsigned int', while the asm-generic code used
on x86 and a lot of other architectures returns 'unsigned long'.

When building the mlx5 driver on ARM, we get a warning about
this:

infiniband/hw/mlx5/mem.c: In function 'mlx5_ib_cont_pages':
infiniband/hw/mlx5/mem.c:84:143: warning: comparison of distinct pointer types lacks a cast
     m = min(m, find_first_bit(&tmp, sizeof(tmp)));

This patch changes the driver to use min_t to make it behave
the same way on all architectures.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agomlx5: avoid build warnings on 32-bit
Arnd Bergmann [Tue, 13 Jan 2015 16:08:06 +0000 (17:08 +0100)]
mlx5: avoid build warnings on 32-bit

The mlx5 driver passes a string pointer in through a 'u64' variable,
which on 32-bit machines causes a build warning:

drivers/net/ethernet/mellanox/mlx5/core/debugfs.c: In function 'qp_read_field':
drivers/net/ethernet/mellanox/mlx5/core/debugfs.c:303:11: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

The code is in fact safe, so we can shut up the warning by adding
extra type casts.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agorocker: fix harmless warning on 32-bit machines
Arnd Bergmann [Tue, 13 Jan 2015 14:23:52 +0000 (15:23 +0100)]
rocker: fix harmless warning on 32-bit machines

The rocker driver tries to assign a pointer to a 64-bit integer
and then back to a pointer. This is safe on all architectures,
but causes a compiler warning when pointers are shorter than
64-bit:

rocker/rocker.c: In function 'rocker_desc_cookie_ptr_get':
rocker/rocker.c:809:9: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
  return (void *) desc_info->desc->cookie;
         ^

This adds another cast to uintptr_t to tell the compiler
that it's safe.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agodrivers: net: cpsw: fix multicast flush in dual emac mode
Mugunthan V N [Tue, 13 Jan 2015 12:05:49 +0000 (17:35 +0530)]
drivers: net: cpsw: fix multicast flush in dual emac mode

Since ALE table is a common resource for both the interfaces in Dual EMAC
mode and while bringing up the second interface in cpsw_ndo_set_rx_mode()
all the multicast entries added by the first interface is flushed out and
only second interface multicast addresses are added. Fixing this by
flushing multicast addresses based on dual EMAC port vlans which will not
affect the other emac port multicast addresses.

Fixes: d9ba8f9 (driver: net: ethernet: cpsw: dual emac interface implementation)
Cc: <stable@vger.kernel.org> # v3.9+
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4: Ripping out old hard-wired initialization code in driver
Hariprasad Shenai [Tue, 13 Jan 2015 09:49:25 +0000 (15:19 +0530)]
cxgb4: Ripping out old hard-wired initialization code in driver

Removing old hard-wired initialization code in the driver, which is no longer
used. Also deprecating few module parameters.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoleds: netxbig: fix oops at probe time
Simon Guinot [Tue, 2 Dec 2014 15:32:10 +0000 (07:32 -0800)]
leds: netxbig: fix oops at probe time

This patch fixes a NULL pointer dereference on led_dat->mode_val. Due to
this bug, a kernel oops can be observed at probe time on the LaCie 2Big
and 5Big v2 boards:

Unable to handle kernel NULL pointer dereference at virtual address 00000008
[...]
[<c03f244c>] (netxbig_led_probe) from [<c02c8c6c>] (platform_drv_probe+0x4c/0x9c)
[<c02c8c6c>] (platform_drv_probe) from [<c02c72d0>] (driver_probe_device+0x98/0x25c)
[<c02c72d0>] (driver_probe_device) from [<c02c7520>] (__driver_attach+0x8c/0x90)
[<c02c7520>] (__driver_attach) from [<c02c5c24>] (bus_for_each_dev+0x68/0x94)
[<c02c5c24>] (bus_for_each_dev) from [<c02c6408>] (bus_add_driver+0x124/0x1dc)
[<c02c6408>] (bus_add_driver) from [<c02c7ac0>] (driver_register+0x78/0xf8)
[<c02c7ac0>] (driver_register) from [<c000888c>] (do_one_initcall+0x80/0x1cc)
[<c000888c>] (do_one_initcall) from [<c0733618>] (kernel_init_freeable+0xe4/0x1b4)
[<c0733618>] (kernel_init_freeable) from [<c058db9c>] (kernel_init+0xc/0xec)
[<c058db9c>] (kernel_init) from [<c0009850>] (ret_from_fork+0x14/0x24)
[...]

This bug was introduced by commit 588a6a99286ae30afb1339d8bc2163517b1b7dd1
("leds: netxbig: fix attribute-creation race").

Signed-off-by: Simon Guinot <simon.guinot@sequanux.org>
Cc: <stable@vger.kernel.org> # 3.17+
Acked-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Bryan Wu <cooloney@gmail.com>
9 years agotipc: remove redundant timer defined in tipc_sock struct
Ying Xue [Tue, 13 Jan 2015 09:07:48 +0000 (17:07 +0800)]
tipc: remove redundant timer defined in tipc_sock struct

Remove the redundant timer defined in tipc_sock structure, instead we
can directly reuse the sk_timer defined in sock structure.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/fsl: replace (1 << x) with BIT(x) for bit definitions in xgmac_mdio
Shaohui Xie [Tue, 13 Jan 2015 02:30:59 +0000 (10:30 +0800)]
net/fsl: replace (1 << x) with BIT(x) for bit definitions in xgmac_mdio

Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agonet/fsl: fix a bug in xgmac_mdio
Shaohui Xie [Tue, 13 Jan 2015 02:30:31 +0000 (10:30 +0800)]
net/fsl: fix a bug in xgmac_mdio

There is a bug in xgmac_mdio_read when clear the bit MDIO_STAT_ENC,
which '&' is missed in 'mdio_stat &= ~MDIO_STAT_ENC'.

Signed-off-by: Shaohui Xie <Shaohui.Xie@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agobridge: fix uninitialized variable warning
Roopa Prabhu [Tue, 13 Jan 2015 00:25:28 +0000 (16:25 -0800)]
bridge: fix uninitialized variable warning

net/bridge/br_netlink.c: In function ‘br_fill_ifinfo’:
net/bridge/br_netlink.c:146:32: warning: ‘vid_range_flags’ may be used uninitialized in this function [-Wmaybe-uninitialized]
  err = br_fill_ifvlaninfo_range(skb, vid_range_start,
                                ^
net/bridge/br_netlink.c:108:6: note: ‘vid_range_flags’ was declared here
  u16 vid_range_flags;

Reported-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoipv6: directly include libc-compat.h in ipv6.h
Willem de Bruijn [Mon, 12 Jan 2015 19:29:34 +0000 (14:29 -0500)]
ipv6: directly include libc-compat.h in ipv6.h

Patch 3b50d9029809 ("ipv6: fix redefinition of in6_pktinfo ...")
fixed a libc compatibility issue in ipv6 structure definitions
as described in include/uapi/linux/libc-compat.h.

It relies on including linux/in6.h to include libc-compat.h itself.
Include that file directly to clearly communicate the dependency
(libc-compat.h: "This include must be as early as possible").

Signed-off-by: Willem de Bruijn <willemb@google.com>
----

As discussed in http://patchwork.ozlabs.org/patch/427384/
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agocxgb4vf: Initialize mdio_addr before using it
Hariprasad Shenai [Mon, 12 Jan 2015 16:36:16 +0000 (22:06 +0530)]
cxgb4vf: Initialize mdio_addr before using it

In commit 5ad24def21b205a8 ("cxgb4vf: Fix ethtool get_settings for VF driver")
mdio_addr of port_info structure was used unininitialzed. Fixing it.

Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Tue, 13 Jan 2015 21:06:07 +0000 (16:06 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2015-01-13

This series contains updates to i40e and i40evf.

Mitch provides a fix for i40e to move the call to pci_disable_sriov() so
that it is called earlier to ensure that the PF driver won't free VF
resources before the VF remove routine can complete.  Also cleans up
redundant and duplicate code in the i40evf.  Refactors the i40evf
shutdown code and let the watchdog take care of shutting things down.
Fix a possible memory leak, if we are using VLANs and the communication
with the PF fail during shutdown.  On some versions of the firmware, the
VF admin send queue may become stalled.  In this case, the easiest
solution is to place another descriptor on the queue and the firmware
will then process both requests.

Greg adds a warning when the NPAR enabled partitions detected a link speed
less than 10 Gpbs.

Vasu removes redundant VN2VN MAC address which were already added by
the FCoE stack.

Shannon adds code to find how many partitions there are per port and
what is the current partition_id when in NPAR mode.  In multifunction
mode, make sure we only allow SR/IOV on the master PF of a port and
only allow partition 1 to set WoL, speed and flow control.

Kamil adds code to read the PBA block from shadow RAM and returns
the part number in a string format.

Catherine provides a fix to check if link state and link speed has
changed before exiting link event

v2: remove un-needed {} in patch #3 of the series based on feedback from
    Sergei Shtylyov
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agoi40e: limit sriov to partition 1 of NPAR configurations
Shannon Nelson [Thu, 11 Dec 2014 07:06:34 +0000 (07:06 +0000)]
i40e: limit sriov to partition 1 of NPAR configurations

Make sure we only allow SR/IOV on the master PF of a port in multifunction
mode.  This should be the case anyway based on the num_vfs configured in
the NVM, but this will help make sure there's no question.  If we're not
in multifunction mode the partition_id will always be 1.

Change-ID: I8b2592366fe6782f15301bde2ebd1d4da240109d
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Tested-by: Jim Young <james.m.young@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: Don't exit link event early if link speed has changed
Catherine Sullivan [Thu, 11 Dec 2014 07:06:33 +0000 (07:06 +0000)]
i40e: Don't exit link event early if link speed has changed

Previously we were only checking if the link up state had changed,
and if it hadn't exiting the link event routine early. We should
also check if speed has changed, and if it has, stay and finish
processing the link event.

Change-ID: I9c8e0991b3f0279108a7858898c3c5ce0a9856b8
Signed-off-by: Catherine Sullivan <catherine.sullivan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
9 years agoi40e: limit WoL and link settings to partition 1
Shannon Nelson [Thu, 11 Dec 2014 07:06:32 +0000 (07:06 +0000)]
i40e: limit WoL and link settings to partition 1

When in multi-function mode, e.g. Dell's NPAR, only partition 1
of each MAC is allowed to set WoL, speed, and flow control.

Change-ID: I87a9debc7479361c55a71f0120294ea319f23588
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>