sdk/emulator/emulator-kernel.git
10 years agonet: Verify UDP checksum before handoff to encap
Tom Herbert [Wed, 7 May 2014 23:52:39 +0000 (16:52 -0700)]
net: Verify UDP checksum before handoff to encap

Moving validation of UDP checksum to be done in UDP not encap layer.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoicmp6: Call skb_checksum_validate
Tom Herbert [Wed, 7 May 2014 23:52:29 +0000 (16:52 -0700)]
icmp6: Call skb_checksum_validate

Use skb_checksum_validate to verify checksum.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoicmp: Call skb_checksum_simple_validate
Tom Herbert [Wed, 7 May 2014 23:52:21 +0000 (16:52 -0700)]
icmp: Call skb_checksum_simple_validate

Use skb_checksum_simple_validate to verify checksum.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoigmp: Call skb_checksum_simple_validate
Tom Herbert [Wed, 7 May 2014 23:52:10 +0000 (16:52 -0700)]
igmp: Call skb_checksum_simple_validate

Use skb_checksum_simple_validate to verify checksum.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agogre6: Call skb_checksum_simple_validate
Tom Herbert [Wed, 7 May 2014 23:51:57 +0000 (16:51 -0700)]
gre6: Call skb_checksum_simple_validate

Use skb_checksum_simple_validate to verify checksum.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agogre: Call skb_checksum_simple_validate
Tom Herbert [Wed, 7 May 2014 23:51:46 +0000 (16:51 -0700)]
gre: Call skb_checksum_simple_validate

Use skb_checksum_simple_validate to verify checksum.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agomellanox: Logging message cleanups
Joe Perches [Wed, 7 May 2014 19:52:57 +0000 (12:52 -0700)]
mellanox: Logging message cleanups

Use a more current logging style.

o Coalesce formats
o Add missing spaces for coalesced formats
o Align arguments for modified formats
o Add missing newlines for some logging messages
o Use DRV_NAME as part of format instead of %s, DRV_NAME to
  reduce overall text.
o Use ..., ##__VA_ARGS__ instead of args... in macros
o Correct a few format typos
o Use a single line message where appropriate

Signed-off-by: Joe Perches <joe@perches.com>
Acked-By: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: remove the unused macro
dingtianhong [Wed, 7 May 2014 14:10:21 +0000 (22:10 +0800)]
bonding: remove the unused macro

Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: simplify the slave_do_arp_validate_only()
dingtianhong [Wed, 7 May 2014 14:10:20 +0000 (22:10 +0800)]
bonding: simplify the slave_do_arp_validate_only()

The argument slave is not used for slave_do_arp_validate_only(), so no need
to keep it, make the function more simple.

Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobonding: remove the unnecessary struct bond_net
dingtianhong [Wed, 7 May 2014 14:10:19 +0000 (22:10 +0800)]
bonding: remove the unnecessary struct bond_net

Move the structure bond_net forward, and remove the unnecessary structure declaration.

Cc: Jay Vosburgh <j.vosburgh@gmail.com>
Cc: Veaceslav Falico <vfalico@gmail.com>
Cc: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv4: remove inet_addr_hash_lock in devinet.c
WANG Cong [Tue, 6 May 2014 18:15:56 +0000 (11:15 -0700)]
ipv4: remove inet_addr_hash_lock in devinet.c

All the callers hold RTNL lock, so there is no need to use inet_addr_hash_lock
to protect the hash list.

Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqlcnic: Fix Kconfig dependency on HWMON
Harish Patil [Wed, 7 May 2014 06:06:15 +0000 (02:06 -0400)]
qlcnic: Fix Kconfig dependency on HWMON

Commit 1f0f467b670e "qlcnic: Add hwmon interface to export board
temperature" introduced a randconfig build error in the case
when the hwmon framework is built as a module and the qlcnic
driver itself is built-in:

drivers/built-in.o: In function `qlcnic_register_hwmon_dev':
drivers/net/ethernet/qlogic/qlcnic/qlcnic_sysfs.c:1301:
undefined reference to `hwmon_device_register_with_groups'
drivers/built-in.o: In function `qlcnic_unregister_hwmon_dev':
drivers/net/ethernet/qlogic/qlcnic/qlcnic_sysfs.c:1309:
undefined reference to `hwmon_device_unregister'.

This changes the Kconfig logic to enforce that the qlcnic hwmon
support can only be enabled if it is possible to successfully
build it.

Signed-off-by: Harish Patil <harish.patil@qlogic.com>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'micrel_ksz9031'
David S. Miller [Wed, 7 May 2014 20:51:22 +0000 (16:51 -0400)]
Merge branch 'micrel_ksz9031'

Hubert Chaumette says:

====================
net/phy: micrel: Add DT configuration support and documentation for KSZ9031

 - Adds DT configuration support for ksz9031
 - Renames micrel-ksz9021.txt to micrel-ksz90x1.txt and adds ksz9031 binding
   documentation

Changes since v3:
 - Rebased on net-next
Changes since v2:
 - Merged together ksz9031_load_{clk,data,ctrl}_skew_values()
 - Added field length and number of fields prameter to account for registers
  specificities
 - Added binding documentation
Changes since v1:
 - Removed ksz9021 and ksz9031 fixup deletions from arch/arm/mach-imx/mach-imx6q.c

Hubert Chaumette (2):
  Update Micrel KSZ90x1 binding documentation
  ARM: i.MX6: Add OF configuration support for ksz9031
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoARM: i.MX6: Add OF configuration support for ksz9031
Hubert Chaumette [Tue, 6 May 2014 07:40:17 +0000 (09:40 +0200)]
ARM: i.MX6: Add OF configuration support for ksz9031

Adds support for ksz9031 PAD skew configuration over devicetree.

Signed-off-by: Hubert Chaumette <hchaumette@adeneo-embedded.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoUpdate Micrel KSZ90x1 binding documentation
Hubert Chaumette [Tue, 6 May 2014 07:40:16 +0000 (09:40 +0200)]
Update Micrel KSZ90x1 binding documentation

Renames micrel-ksz9021.txt to micrel-ksz90x1.txt and adds documentation for
the KSZ9031 binding from patch 1. Also adds step increment information, and
note about phy fixups.

Signed-off-by: Hubert Chaumette <hchaumette@adeneo-embedded.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: clean up snmp stats code
WANG Cong [Mon, 5 May 2014 22:55:55 +0000 (15:55 -0700)]
net: clean up snmp stats code

commit 8f0ea0fe3a036a47767f9c80e (snmp: reduce percpu needs by 50%)
reduced snmp array size to 1, so technically it doesn't have to be
an array any more. What's more, after the following commit:

commit 933393f58fef9963eac61db8093689544e29a600
Date:   Thu Dec 22 11:58:51 2011 -0600

    percpu: Remove irqsafe_cpu_xxx variants

    We simply say that regular this_cpu use must be safe regardless of
    preemption and interrupt state.  That has no material change for x86
    and s390 implementations of this_cpu operations.  However, arches that
    do not provide their own implementation for this_cpu operations will
    now get code generated that disables interrupts instead of preemption.

probably no arch wants to have SNMP_ARRAY_SZ == 2. At least after
almost 3 years, no one complains.

So, just convert the array to a single pointer and remove snmp_mib_init()
and snmp_mib_free() as well.

Cc: Christoph Lameter <cl@linux.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoisdn: hisax: remove some dead code
Dan Carpenter [Mon, 5 May 2014 08:53:05 +0000 (11:53 +0300)]
isdn: hisax: remove some dead code

The HISAX_HFC4S8S_PCIMEM code hasn't been able to compile since before
the start of git history.  I have deleted it.

There are also a few indenting mistakes where one side of the ifdef
wasn't indented correctly which I fixed as well.

Reported-by: Walter Harms <wharms@bfs.de>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotunnel: fix RFC number in comment for INET_ECN_decapsulate()
Neal Cardwell [Mon, 5 May 2014 00:55:39 +0000 (20:55 -0400)]
tunnel: fix RFC number in comment for INET_ECN_decapsulate()

The quoted text and figure are from RFC 6040 ("Tunnelling of Explicit
Congestion Notification").

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agow83977af_ir: remove two faux Kconfig macros
Paul Bolle [Sun, 4 May 2014 11:03:12 +0000 (13:03 +0200)]
w83977af_ir: remove two faux Kconfig macros

Ever since v2.3.5 the driver for "Winbond W83977AF (IR)" contains two
unneeded preprocessor macros.

CONFIG_NETWINDER_TX_DMA_PROBLEMS is never defined and can safely be
removed. And CONFIG_NETWINDER_RX_DMA_PROBLEMS is just an alias for
CONFIG_ARCH_NETWINDER, so that (valid) Kconfig macro can be used
instead.

Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: Need to sock_put on csum error
Tom Herbert [Mon, 5 May 2014 23:43:58 +0000 (16:43 -0700)]
ipv6: Need to sock_put on csum error

Commit 4068579e1e098fa81d48db9ba4432ab664c58561 ("net: Implmement
RFC 6936 (zero RX csums for UDP/IPv6)") introduced zero checksums
being allowed for IPv6, but in the case that a socket disallows a
zero checksum on RX we need to sock_put.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'tipc-next'
David S. Miller [Mon, 5 May 2014 21:26:54 +0000 (17:26 -0400)]
Merge branch 'tipc-next'

Ying Xue says:

====================
tipc: purge signal handler infrastructure

When we delay some actions to be executed in asynchronous contexts,
these usually add unnecessary code complexities, and make their
behaviours unpredictable and indeterministic. Moreover, as the signal
handler infrastructure is first stopped when tipc module is removed,
this may cause some potential risks for us. For instance, although
signal handler is already stopped, some tipc components still submit
signal requests to signal handler infrastructure, which may lead to
some resources not to be released or freed correctly.

So the series aims to convert all actions being performed in tasklet
context asynchronously with interface provided by signal handler
infrastructure to be executed synchronously, thereby deleting the
whole infrastructure of signal handler.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: purge signal handler infrastructure
Ying Xue [Mon, 5 May 2014 00:56:18 +0000 (08:56 +0800)]
tipc: purge signal handler infrastructure

In the previous commits of this series, we removed all asynchronous
actions which were based on the tasklet handler - "tipc_k_signal()".

So the moment has now come when we can completely remove the tasklet
handler infrastructure. That is done with this commit.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: avoid to asynchronously reset all links
Ying Xue [Mon, 5 May 2014 00:56:17 +0000 (08:56 +0800)]
tipc: avoid to asynchronously reset all links

Postpone the actions of resetting all links until after bclink
lock is released, avoiding to asynchronously reset all links.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: convert allocations of global variables associated with bclink
Ying Xue [Mon, 5 May 2014 00:56:16 +0000 (08:56 +0800)]
tipc: convert allocations of global variables associated with bclink

Convert allocations of global variables associated with bclink from
static way to dynamical way for the convenience of bclink instance
initialisation. Meanwhile, this also helps TIPC support name space
in the future easily.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: define new functions to operate bc_lock
Ying Xue [Mon, 5 May 2014 00:56:15 +0000 (08:56 +0800)]
tipc: define new functions to operate bc_lock

As we are going to do more jobs when bc_lock is released, the two
operations of holding/releasing the lock should be encapsulated with
functions. In addition, we move bc_lock spin lock into tipc_bclink
structure avoiding to define the global variable.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: avoid to asynchronously deliver name tables to peer node
Ying Xue [Mon, 5 May 2014 00:56:14 +0000 (08:56 +0800)]
tipc: avoid to asynchronously deliver name tables to peer node

Postpone the actions of delivering name tables until after node
lock is released, avoiding to do it under asynchronous context.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: remove TIPC_NAMES_GONE node flag
Ying Xue [Mon, 5 May 2014 00:56:13 +0000 (08:56 +0800)]
tipc: remove TIPC_NAMES_GONE node flag

Since previously what all publications pertaining to the lost node
were removed from name table was finished in tasklet context
asynchronously, we need to TIPC_NAMES_GONE flag indicating whether
the node cleanup work is finished or not. But now as the cleanup work
has been finished when node lock is released, the flag becomes
meaningless for us.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: avoid to asynchronously notify subscriptions
Ying Xue [Mon, 5 May 2014 00:56:12 +0000 (08:56 +0800)]
tipc: avoid to asynchronously notify subscriptions

Postpone the actions of notifying subscriptions until after node lock
is released, avoiding to asynchronously execute registered handlers
when node is lost.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: rename setup_blocked variable of node struct to flags
Ying Xue [Mon, 5 May 2014 00:56:11 +0000 (08:56 +0800)]
tipc: rename setup_blocked variable of node struct to flags

Rename setup_blocked variable of node struct to a more common
name called "flags", which will be used to represent kinds of
node states.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: adjust order of variables in tipc_node structure
Ying Xue [Mon, 5 May 2014 00:56:10 +0000 (08:56 +0800)]
tipc: adjust order of variables in tipc_node structure

Move more frequently used variables up to the head of tipc_node
structure, hopefully improving a bit performance.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: always use tipc_node_lock() to hold node lock
Ying Xue [Mon, 5 May 2014 00:56:09 +0000 (08:56 +0800)]
tipc: always use tipc_node_lock() to hold node lock

Although we obtain node lock with tipc_node_lock() in most time, there
are still places where we directly use native spin lock interface
to grab node lock. But as we will do more jobs in the future when node
lock is released, we should ensure that tipc_node_lock() is always
called when node lock is taken.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reviewed-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: via-rhine: Convert #ifdef USE_MMIO to a runtime flag
Alexey Charkov [Sat, 3 May 2014 12:40:53 +0000 (16:40 +0400)]
net: via-rhine: Convert #ifdef USE_MMIO to a runtime flag

This introduces another flag in 'quirks' to replace the preprocessor
define (USE_MMIO) used to indicate whether the device needs a
separate enable routine to operate in MMIO mode.

All of the currently known platform Rhine cores operate in MMIO
mode by default, and on PCI it is preferred over PIO for performance
reasons. However, a comment in code suggests that some (?) early
Rhine cores only work in PIO mode, so they should not be switched
to MMIO.

Enabling MMIO on PCI is still triggered by the same Kconfig option
to avoid breaking user configs needlessly, but this can be changed
going forward towards automatic runtime detection in case a list of
PIO-only Rhine revisions can be compiled.

This also fixes a couple of compiler warnings detected by Fengguang
Wu's test bot (!USE_MMIO case):

   drivers/net/ethernet/via/via-rhine.c: In function 'rhine_init_one_pci':
   drivers/net/ethernet/via/via-rhine.c:1108:1: warning: label 'err_out_unmap' defined but not used [-Wunused-label]
    err_out_unmap:
    ^
   drivers/net/ethernet/via/via-rhine.c:1022:6: warning: unused variable 'i' [-Wunused-variable]
     int i, rc;
         ^
   drivers/net/ethernet/via/via-rhine.c:916:22: warning: 'quirks' may be used uninitialized in this function [-Wmaybe-uninitialized]
     phy_id = rp->quirks & rqIntPHY ? 1 : 0;
                         ^
   drivers/net/ethernet/via/via-rhine.c:1026:6: note: 'quirks' was declared here
     u32 quirks;
         ^

Signed-off-by: Alexey Charkov <alchark@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: remove unused function ipv6_inherit_linklocal()
WANG Cong [Sat, 3 May 2014 00:21:01 +0000 (17:21 -0700)]
ipv6: remove unused function ipv6_inherit_linklocal()

It is no longer used after commit e837735ec406a347756e
(ip6_tunnel: ensure to always have a link local address).

Cc: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'inet_csums'
David S. Miller [Mon, 5 May 2014 19:28:52 +0000 (15:28 -0400)]
Merge branch 'inet_csums'

Tom Herbert says:

====================
net: Checksum offload changes

I am working on overhauling RX checksum offload. Goals of this effort
are:

- Specify what exactly it means when driver returns CHECKSUM_UNNECESSARY
- Preserve CHECKSUM_COMPLETE through encapsulation layers
- Don't do skb_checksum more than once per packet
- Unify GRO and non-GRO csum verification as much as possible
- Unify the checksum functions (checksum_init)
- Simply code

What is in this first patch set:

- Create a common "checksum_init" function which is called from
  TCPv{4,6} and UDPv{4,6}
- Add some for RFC6936, UDP/IPv6 zero checksums
- Add architecture support for csum_add and provide implementations
  for x86_64 and Sparc 32 and 64 bit (please test the latter)

Please review carefully and test if possible, mucking with basic
checksum functions is always a little precarious :-)
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Implmement RFC 6936 (zero RX csums for UDP/IPv6)
Tom Herbert [Fri, 2 May 2014 23:29:58 +0000 (16:29 -0700)]
net: Implmement RFC 6936 (zero RX csums for UDP/IPv6)

RFC 6936 relaxes the requirement of RFC 2460 that UDP/IPv6 packets which
are received with a zero UDP checksum value must be dropped. RFC 6936
allows zero checksums to support tunnels over UDP.

When sk_no_check is set we allow on a socket we allow a zero IPv6
UDP checksum. This is for both sending zero checksum and accepting
a zero checksum on receive.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Call skb_checksum_init in IPv6
Tom Herbert [Fri, 2 May 2014 23:29:51 +0000 (16:29 -0700)]
net: Call skb_checksum_init in IPv6

Call skb_checksum_init instead of private functions.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Call skb_checksum_init in IPv4
Tom Herbert [Fri, 2 May 2014 23:29:38 +0000 (16:29 -0700)]
net: Call skb_checksum_init in IPv4

Call skb_checksum_init instead of private functions.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Generalize checksum_init functions
Tom Herbert [Fri, 2 May 2014 23:29:18 +0000 (16:29 -0700)]
net: Generalize checksum_init functions

Create a general __skb_checksum_validate function (actually a
macro) to subsume the various checksum_init functions. This
function can either init the checksum, or do the full validation
(logically checksum_init+skb_check_complete)-- a flag specifies
if full vaidation is performed. Also, there is a flag to the function
to indicate that zero checksums are allowed (to support optional
UDP checksums).

Added several stub functions for calling __skb_checksum_validate.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agosparc: csum_add for Sparc
Tom Herbert [Fri, 2 May 2014 23:28:54 +0000 (16:28 -0700)]
sparc: csum_add for Sparc

versions.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Change x86_64 add32_with_carry to allow memory operand
Tom Herbert [Fri, 2 May 2014 23:28:40 +0000 (16:28 -0700)]
net: Change x86_64 add32_with_carry to allow memory operand

Note add32_with_carry(a, b) is suboptimal, as it forces
a and b in registers.

b could be a memory or a register operand.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agox86_64: csum_add for x86_64
Tom Herbert [Fri, 2 May 2014 23:28:15 +0000 (16:28 -0700)]
x86_64: csum_add for x86_64

Add csum_add function for x86_64.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Allow csum_add to be provided in arch
Tom Herbert [Fri, 2 May 2014 23:28:03 +0000 (16:28 -0700)]
net: Allow csum_add to be provided in arch

csum_add is really nothing more then add-with-carry which
can be implemented efficiently in some architectures.
Allow architecture to define this protected by HAVE_ARCH_CSUM_ADD.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Mon, 5 May 2014 17:36:26 +0000 (13:36 -0400)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

John W. Linville says:

====================
pull request: wireless-next 2014-05-02

Please pull this batch of updates intended for the 3.16 stream...

For the mac80211 bits, Johannes says:

"In this round we have a large number of small features and
improvements from people too numerous to list here. The only really
bit thing is Michał and Luca's CSA work (including changing how
interface combination verification is done)."

For the Bluetooth bits, Gustavo says:

"Here goes some patches for the -next release. There is nothing
really special for this pull request, just a bunch of refactors,
fixes and clean ups."

For the ath10k/ath6kl bits, Kalle says:

"For ath6kl Kalle fixed a bunch of checkpatch warnings.

In ath10k we had more changes, major ones being:

* fix memory allocation failures after a firmware crash (Michal)

* some rework of DFS configuration to enable it correctly in all cases
  (Michal)

* add a new firmware crash option to make it possible to crash 10.1
  firmware for testing purposes (Marek P)

* fix RTS/CTS protection in certain cases (Marek K)

* fix wrong RSSI and rate reporting in some cases (Janusz)

* fix firmware stats reporting (Chun, Ben & Bartosz)"

For the iwlwifi bits, Emmanuel says:

"I have here a bunch of unrelated things. I disabled support for
-7.ucode which means that I can removed a lot of code. Eliad has
a brand new feature: we reduce the Tx power when the link allows -
this reduces our power consumption. The regular changes in power and
scan area. One interesting thing though is the patches from Johannes,
we have now GRO which allows to increase our throughput in TCP Rx. The
main advantage is that it reduces the number of TCP Acks - these TCP
Acks are completely useless when we are using A-MPDU since the first
packet of the A-MPDU generates a TCP Ack which is made obsolete by
the next packets."

Along with that, there are a variety of updates to b43, mwifiex,
rtl8180 and wil6210 drivers and a handful of other updates here
and there.

Please let me know if there are problems!
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'am437x'
David S. Miller [Mon, 5 May 2014 17:19:00 +0000 (13:19 -0400)]
Merge branch 'am437x'

George Cherian says:

====================
The series adds CPTS support for AM4372.

Patch 1 - DT changes w.r.t clock changes for AM33xx.
Patch 2 - CPTS clock name harcoding in the driver is removed.
  Easier to pass the clock name from dt rather than hardcoding in driver.
  Also in prepration for DRA7x CPTS support.
Patch 3 - Enable the CPTS support for both DRA7x and AM4372 in the driver.
Patch 4 - Enable the Annexe F for L2 PTP for AM437x and DRA7x.
Patch 5 - Change the default clocksource to dpll_core_m5
Patch 6 - DT changes for AM4372.

v1 -> v2
Patch 1 and 2 Re-ordering.
Seperate TS_BITS define for Hw version V2 and V3
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoARM: dts: am4372: Add clock names for cpsw and cpts
George Cherian [Fri, 2 May 2014 06:32:04 +0000 (12:02 +0530)]
ARM: dts: am4372: Add clock names for cpsw and cpts

Add CPSW fck and CPTS clock and clock names for AM4372

Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoARM: AM43xx: clk: Change the cpts ref clock source to dpll_core_m5 clk
George Cherian [Fri, 2 May 2014 06:32:03 +0000 (12:02 +0530)]
ARM: AM43xx: clk: Change the cpts ref clock source to dpll_core_m5 clk

cpsw_cpts_rft_clk has got the choice of 3 clocksources
 -dpll_core_m4_ck
 -dpll_core_m5_ck
 -dpll_disp_m2_ck

By default dpll_core_m4_ck is selected, witn this as clock
source the CPTS doesnot work properly. It gives clockcheck errors
while running PTP.

 clockcheck: clock jumped backward or running slower than expected!

By selecting dpll_core_m5_ck as the clocksource fixes this issue.
In AM335x dpll_core_m5_ck is the default clocksource.

Signed-off-by: George Cherian <george.cherian@ti.com>
Acked-by: Tero Kristo <t-kristo@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers: net: cpsw: Enable Annexe F Time sync
George Cherian [Fri, 2 May 2014 06:32:02 +0000 (12:02 +0530)]
drivers: net: cpsw: Enable Annexe F Time sync

Enable the Annex F Time Sync explicitly for DRA7x and AM4372.
With this enabled the L2 PTP is working.

while at that rename TS_BIT8 to TS_TTL_NONZERO

Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers: net: cpsw: Enable CPTS for DRA7xx and AM4372
George Cherian [Fri, 2 May 2014 06:32:01 +0000 (12:02 +0530)]
drivers: net: cpsw: Enable CPTS for DRA7xx and AM4372

Enable cpts hardware time stamping for Dra7xx and AM4372.
This enables PTPv2 for DRA7xx and AM4372.

Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agodrivers: net: cpts: Remove hardcoded clock name for CPTS
George Cherian [Fri, 2 May 2014 06:32:00 +0000 (12:02 +0530)]
drivers: net: cpts: Remove hardcoded clock name for CPTS

CPTS refclk name is hardcoded, which makes it fail in case of DRA7x
Remove the hardcoded clock name for CPTS refclk and get the same from DT.

Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoARM: dts: am33xx: Add clock names for cpsw and cpts
George Cherian [Fri, 2 May 2014 06:31:59 +0000 (12:01 +0530)]
ARM: dts: am33xx: Add clock names for cpsw and cpts

Add CPSW fck and CPTS clock and clock names

Signed-off-by: George Cherian <george.cherian@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agounregister_netdevice : move RTM_DELLINK to until after ndo_uninit
Roopa Prabhu [Thu, 1 May 2014 18:40:30 +0000 (11:40 -0700)]
unregister_netdevice : move RTM_DELLINK to until after ndo_uninit

This patch fixes ordering of rtnl notifications during unregister_netdevice
by moving RTM_DELLINK notification to until after ndo_uninit.

The problem was seen with unregistering bond netdevices.

bond ndo_uninit callback generates a few RTM_NEWLINK notifications for
NETDEV_CHANGEADDR and NETDEV_FEAT_CHANGE. This is seen mostly when the
bond is deleted with slaves still enslaved to the bond.

During unregister netdevice (rollback_registered_many to be specific)
bond ndo_uninit is called after RTM_DELLINK notification goes out.
This results in userspace seeing RTM_DELLINK followed by a couple of
RTM_NEWLINK's.

In userspace problem was seen with libnl. libnl cache deletes the bond
when it sees RTM_DELLINK and re-adds the bond with the following
RTM_NEWLINK. Resulting in a stale bond entry in libnl cache when the kernel
has already deleted the bond.

This patch has been tested for bond, bridges and vlan devices.

Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'filter-cleanups'
David S. Miller [Sun, 4 May 2014 23:46:59 +0000 (19:46 -0400)]
Merge branch 'filter-cleanups'

Daniel Borkmann says:

====================
BPF cleanups

v3->v4:
 - Sorry, noticed and fixed a typo in patch 3, rest as is
v2->v3:
 - Included Dave's feedback for unsigned long type in patch 3
 - Patch 1 and patch 2 unchanged since v1, dropped other
   two for now
v1->v2:
 - Only changed patch 5 as to suggestion from Alexei
 - Rest is the same
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: misc/various cleanups
Daniel Borkmann [Thu, 1 May 2014 16:34:20 +0000 (18:34 +0200)]
net: filter: misc/various cleanups

This contains only some minor misc cleanpus. We can spare us the
extra variable declaration in __skb_get_pay_offset(), the cast in
__get_random_u32() is rather unnecessary and in __sk_migrate_realloc()
we can remove the memcpy() and do a direct assignment of the structs.
Latter was suggested by Fengguang Wu found with coccinelle. Also,
remaining pointer casts of long should be unsigned long instead.

Suggested-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: make register naming more comprehensible
Daniel Borkmann [Thu, 1 May 2014 16:34:19 +0000 (18:34 +0200)]
net: filter: make register naming more comprehensible

The current code is a bit hard to parse on which registers can be used,
how they are mapped and all play together. It makes much more sense to
define this a bit more clearly so that the code is a bit more intuitive.
This patch cleans this up, and makes naming a bit more consistent among
the code. This also allows for moving some of the defines into the header
file. Clearing of A and X registers in __sk_run_filter() do not get a
particular register name assigned as they have not an 'official' function,
but rather just result from the concrete initial mapping of old BPF
programs. Since for BPF helper functions for BPF_CALL we already use
small letters, so be consistent here as well. No functional changes.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: simplify label names from jump-table
Daniel Borkmann [Thu, 1 May 2014 16:34:18 +0000 (18:34 +0200)]
net: filter: simplify label names from jump-table

This patch simplifies label naming for the BPF jump-table.
When we define labels via DL(), we just concatenate/textify
the combination of instruction opcode which consists of the
class, subclass, word size, target register and so on. Each
time we leave BPF_ prefix intact, so that e.g. the preprocessor
generates a label BPF_ALU_BPF_ADD_BPF_X for DL(BPF_ALU, BPF_ADD,
BPF_X) whereas a label name of ALU_ADD_X is much more easy
to grasp. Pure cleanup only.

Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: filter: doc: expand and improve BPF documentation
Alexei Starovoitov [Thu, 1 May 2014 15:16:03 +0000 (08:16 -0700)]
net: filter: doc: expand and improve BPF documentation

In particular, this patch tries to clarify internal BPF calling
convention and adds internal BPF examples, JIT guide, use cases.

Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: remove in_flight parameter from cong_avoid() methods
Eric Dumazet [Sat, 3 May 2014 04:18:05 +0000 (21:18 -0700)]
tcp: remove in_flight parameter from cong_avoid() methods

Commit e114a710aa505 ("tcp: fix cwnd limited checking to improve
congestion control") obsoleted in_flight parameter from
tcp_is_cwnd_limited() and its callers.

This patch does the removal as promised.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotcp: fix cwnd limited checking to improve congestion control
Eric Dumazet [Wed, 30 Apr 2014 18:58:13 +0000 (11:58 -0700)]
tcp: fix cwnd limited checking to improve congestion control

Yuchung discovered tcp_is_cwnd_limited() was returning false in
slow start phase even if the application filled the socket write queue.

All congestion modules take into account tcp_is_cwnd_limited()
before increasing cwnd, so this behavior limits slow start from
probing the bandwidth at full speed.

The problem is that even if write queue is full (aka we are _not_
application limited), cwnd can be under utilized if TSO should auto
defer or TCP Small queues decided to hold packets.

So the in_flight can be kept to smaller value, and we can get to the
point tcp_is_cwnd_limited() returns false.

With TCP Small Queues and FQ/pacing, this issue is more visible.

We fix this by having tcp_cwnd_validate(), which is supposed to track
such things, take into account unsent_segs, the number of segs that we
are not sending at the moment due to TSO or TSQ, but intend to send
real soon. Then when we are cwnd-limited, remember this fact while we
are processing the window of ACKs that comes back.

For example, suppose we have a brand new connection with cwnd=10; we
are in slow start, and we send a flight of 9 packets. By the time we
have received ACKs for all 9 packets we want our cwnd to be 18.
We implement this by setting tp->lsnd_pending to 9, and
considering ourselves to be cwnd-limited while cwnd is less than
twice tp->lsnd_pending (2*9 -> 18).

This makes tcp_is_cwnd_limited() more understandable, by removing
the GSO/TSO kludge, that tried to work around the issue.

Note the in_flight parameter can be removed in a followup cleanup
patch.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: Allow tc changes in user namespaces
Stéphane Graber [Wed, 30 Apr 2014 15:25:43 +0000 (11:25 -0400)]
net: Allow tc changes in user namespaces

This switches a few remaining capable(CAP_NET_ADMIN) to ns_capable so
that root in a user namespace may set tc rules inside that namespace.

Signed-off-by: Stéphane Graber <stgraber@ubuntu.com>
Acked-by: Serge E. Hallyn <serge.hallyn@ubuntu.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: "David S. Miller" <davem@davemloft.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'davinci_mdio'
David S. Miller [Fri, 2 May 2014 20:17:18 +0000 (16:17 -0400)]
Merge branch 'davinci_mdio'

Grygorii Strashko says:

====================
introduce devm_mdiobus_alloc/free and clean up davinci mdio

Introduce a resource managed devm_mdiobus_alloc[_size]()/devm_mdiobus_free()
to automatically clean up MDIO bus alocations made by MDIO drivers,
thus leading to simplified MDIO drivers code.

Clean up Davinci MDIO driver and use new devm API.

Changes in v3:
- added devm_mdiobus_alloc_size() and
  devm_mdiobus_alloc() converted to be just a simple wrapper now.

Changes in v2:
- minor comments taken into account
- additional patches added for cleaning up Davinci MDIO driver
====================

Acked-by: Santosh Shilimkar<santosh.shilimkar@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: davinci_mdio: simplify IO memory mapping
Grygorii Strashko [Wed, 30 Apr 2014 12:23:36 +0000 (15:23 +0300)]
net: davinci_mdio: simplify IO memory mapping

Simplify IO memory mapping by using devm_ioremap_resource()
which will do all errors handling and reporting for us.

Acked-and-tested-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: davinci_mdio: drop pinctrl_pm_select_default_state from probe
Grygorii Strashko [Wed, 30 Apr 2014 12:23:35 +0000 (15:23 +0300)]
net: davinci_mdio: drop pinctrl_pm_select_default_state from probe

The "default" pinctrl state is set by Drivers core now before
calling the driver's probe.
Hence, it's safe to drop pinctrl_pm_select_default_state() call
from Davinci mdio driver probe.

Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Acked-and-tested-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: davinci_mdio: use devm_* api
Grygorii Strashko [Wed, 30 Apr 2014 12:23:34 +0000 (15:23 +0300)]
net: davinci_mdio: use devm_* api

Use devm_* API for memory allocation and to get device's clock
to simplify driver's code.

Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-and-tested-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agomdio_bus: implement devm_mdiobus_alloc/devm_mdiobus_free
Grygorii Strashko [Wed, 30 Apr 2014 12:23:33 +0000 (15:23 +0300)]
mdio_bus: implement devm_mdiobus_alloc/devm_mdiobus_free

Add a resource managed devm_mdiobus_alloc[_size]()/devm_mdiobus_free()
to automatically clean up MDIO bus alocations made by MDIO drivers,
thus leading to simplified MDIO drivers code.

Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Acked-and-tested-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: via-rhine: Drop revision property, use quirks instead
Alexey Charkov [Wed, 30 Apr 2014 18:21:09 +0000 (22:21 +0400)]
net: via-rhine: Drop revision property, use quirks instead

This adds two new flags to quirks and thus removes the need to carry
revision in rhine_private. As a result, the init logic is simplified
a bit.

This also fixes a compiler warning in OF code on 64bit due to pointer
casting:

        drivers/net/ethernet/via/via-rhine.c: In function ‘rhine_init_one_platform’:
        drivers/net/ethernet/via/via-rhine.c:1132:13: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
          revision = (u32)match->data;
                     ^

That code was added in commit 2d283862dc62daead9db0dc89cd0d0351e91f765
("net: via-rhine: add OF bus binding").

Tested in platform configuration on a VIA WM8950 APC Rock board.

Reported-by: Jan Moskyto Matejka <mq@suse.cz>
Signed-off-by: Alexey Charkov <alchark@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Fri, 2 May 2014 17:47:50 +0000 (13:47 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

10 years agohyperv: Enable sendbuf mechanism on the send path
KY Srinivasan [Wed, 30 Apr 2014 17:14:31 +0000 (10:14 -0700)]
hyperv: Enable sendbuf mechanism on the send path

We send packets using a copy-free mechanism (this is the Guest to Host transport
via VMBUS). While this is obviously optimal for large packets,
it may not be optimal for small packets. Hyper-V host supports
a second mechanism for sending packets that is "copy based". We implement that
mechanism in this patch.

In this version of the patch I have addressed a comment from David Miller.

With this patch (and all of the other offload and VRSS patches), we are now able
to almost saturate a 10G interface between Linux VMs on Hyper-V
on different hosts - close to  9 Gbps as measured via iperf.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: stmmac: set phy to use polling by default
Dinh Nguyen [Tue, 29 Apr 2014 17:56:21 +0000 (12:56 -0500)]
net: stmmac: set phy to use polling by default

mii_irq[] array is never initialized anywhere in the driver, thus mii_irq[]
will always equate to zero. So, for the case where the PHY does not have an
irq, we should use PHY_POLL for that situation.

Signed-off-by: Dinh Nguyen <dinguyen@altera.com>
Tested-by: Vince Bridgers <vbridger@altera.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovirtio-net: Set needed_headroom for virtio-net when VIRTIO_F_ANY_LAYOUT is true
Zhangjie \(HZ\) [Tue, 29 Apr 2014 10:43:22 +0000 (18:43 +0800)]
virtio-net: Set needed_headroom for virtio-net when VIRTIO_F_ANY_LAYOUT is true

This is a small supplement for commit e7428e95a06fb516fac1308bd0e176e27c0b9287
("virtio-net: put virtio-net header inline with data"). TCP packages have
enough room to put virtio-net header in, but UDP packages do not. By
setting dev->needed_headroom for virtio-net device, UDP packages could have
enough room.

For UDP packages, sk_buff is alloced in fun __ip_append_data. The size is
"alloclen + hh_len + 15", and "hh_len = LL_RESERVED_SPACE(rt-dst.dev);".
The Macro is defined as follows:
#define LL_RESERVED_SPACE(dev) \
     ((((dev)->hard_header_len+(dev)->needed_headroom)\
     &~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
By default, for UDP packages, after skb is allocated, only 16 bytes
reserved. And 2 bytes remained after mac header is set. That is not enough
to put virtio-net header in. If we set dev->needed_headroom to 12 or 10
(according to mergeable_rx_bufs is on or off ), more room can be reserved.
Then there is enough room for UDP packages to put the header in.

test result list as below:
guest and host: suse11sp3, netperf, intel 2.4GHz
+-------+---------+---------+---------+---------+
|       |   old             |   new             |
+-------+---------+---------+---------+---------+
| UDP   |  Gbit/s | pps     |  Gbit/s | pps     |
| 64    |  0.57   | 692232  |  0.61   | 742420  |
| 256   |  1.60   | 686860  |  1.71   | 733331  |
| 512   |  2.92   | 674576  |  3.07   | 710446  |
| 1024  |  4.99   | 598977  |  5.17   | 620821  |
| 1460  |  5.68   | 483757  |  7.16   | 610519  |
| 4096  |  6.98   | 637468  |  7.21   | 658471  |
+-------+---------+---------+---------+---------+

Signed-off-by: Zhang Jie <zhangjie14@huawei.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: fix memory leak of publications
Ying Xue [Tue, 29 Apr 2014 03:12:18 +0000 (11:12 +0800)]
tipc: fix memory leak of publications

Commit 1bb8dce57f4d15233688c68990852a10eb1cd79f ("tipc: fix memory
leak during module removal") introduced a memory leak issue: when
name table is stopped, it's forgotten that publication instances are
freed properly. Additionally the useless "continue" statement in
tipc_nametbl_stop() is removed as well.

Reported-by: Jason <huzhijiang@gmail.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Acked-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ipv6: Introduce ip6_sk_dst_hoplimit.
Lorenzo Colitti [Tue, 29 Apr 2014 02:57:34 +0000 (11:57 +0900)]
net: ipv6: Introduce ip6_sk_dst_hoplimit.

This replaces 6 identical code snippets with a call to a new
static inline function.

Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: dsa: update DSA drivers to use ds_to_priv
Florian Fainelli [Mon, 28 Apr 2014 18:14:28 +0000 (11:14 -0700)]
net: dsa: update DSA drivers to use ds_to_priv

Use the helper function to retrieve the driver private context instead of
using (void *)(ds + 1).

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: dsa: add ds_to_priv
Florian Fainelli [Mon, 28 Apr 2014 18:14:27 +0000 (11:14 -0700)]
net: dsa: add ds_to_priv

DSA drivers have a trick which consists in allocating "priv_size" more
bytes to account for the DSA driver private context. Add a helper
function to access that private context instead of open-coding it in
drivers with (void *)(ds + 1).

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agorsi: Changing opcode for sta mode according to changes in firmware
Jahnavi Meher [Mon, 28 Apr 2014 19:33:53 +0000 (01:03 +0530)]
rsi: Changing opcode for sta mode according to changes in firmware

Signed-off-by: Jahnavi Meher <jahnavi.meher@redpinesignals.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: fix worse EVM for 11b rates
Rajkumar Manoharan [Mon, 28 Apr 2014 15:47:08 +0000 (21:17 +0530)]
ath9k_hw: fix worse EVM for 11b rates

Adjust FIR filter co-efficients to improve EVM for 11b rates.

Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath9k_hw: update ar9300 initvals
Rajkumar Manoharan [Mon, 28 Apr 2014 15:47:07 +0000 (21:17 +0530)]
ath9k_hw: update ar9300 initvals

 * rfsat gainchange hysteresis of rf_gain stuck with large
   interference present.

Signed-off-by: Rajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agob43: N-PHY: complete radio 0x2056 setup
Rafał Miłecki [Sat, 26 Apr 2014 18:56:05 +0000 (20:56 +0200)]
b43: N-PHY: complete radio 0x2056 setup

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoath5k: Fix AR5K_PHY_TXPOWER_RATE_MAX register value setting.
Nickolay Ledovskikh [Fri, 25 Apr 2014 18:53:34 +0000 (22:53 +0400)]
ath5k: Fix AR5K_PHY_TXPOWER_RATE_MAX register value setting.

I was reading ath5k power setting code and
noticed typing error in ath5k_hw_txpower function.
Invalid value was written to AR5K_PHY_TXPOWER_RATE_MAX
register.

Signed-off-by: Nikolay Ledovskikh <nledovskikh@gmail.com>
Acked-by: Nick Kossifidis <mickflemm@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoiwlegacy: Convert /n to \n
Joe Perches [Fri, 25 Apr 2014 01:51:00 +0000 (18:51 -0700)]
iwlegacy: Convert /n to \n

Use a newline character appropriately.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agowlcore: fix usage of platform_device_add_data()
Christian Engelmayer [Thu, 24 Apr 2014 20:35:56 +0000 (22:35 +0200)]
wlcore: fix usage of platform_device_add_data()

Coverity CID 986698 reports leakage of struct wlcore_platdev_data in the
probe functions of both the SPI/SDIO interfaces. The structure passed to
platform_device_add_data() is dynamically allocated and only freed in the
error paths, however, platform_device_add_data() adds a copy of the platform
specific data to the device. Move the temporary struct that is kmemdup'ed
to the stack. This issue exists since afb43e6d (wlcore: remove if_ops from
platform_data).

Signed-off-by: Christian Engelmayer <cengelma@gmx.at>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agortlwifi: rtl8723be: disable MSI interrupts mode
Adam Lee [Thu, 24 Apr 2014 03:08:44 +0000 (11:08 +0800)]
rtlwifi: rtl8723be: disable MSI interrupts mode

94010fa0dd07e8b904e7c6b6589f15573008ab15 ("rtlwifi: add MSI interrupts
mode support") introduced MSI interrupts mode support, which seemed
safe enough with RTL8188EE and RTL8723BE as RealTek's testing results,
but some users reported their RTL8188EE modules could not connect to
any wireless network after the MSI mode was enabled by Ubuntu 14.04.

So, let's fallback to pin-based mode until rtlwifi's MSI support get
good compatibility.

BugLink: https://bugs.launchpad.net/bugs/1310512
Signed-off-by: Adam Lee <adam.lee@canonical.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoRevert "rtlwifi: rtl8188ee: enable MSI interrupts mode"
Adam Lee [Thu, 24 Apr 2014 03:08:43 +0000 (11:08 +0800)]
Revert "rtlwifi: rtl8188ee: enable MSI interrupts mode"

This reverts commit 2a54eb5e1476426ee639bbfbe179b52342a0d82c
("rtlwifi: rtl8188ee: enable MSI interrupts mode").

94010fa0dd07e8b904e7c6b6589f15573008ab15 ("rtlwifi: add MSI interrupts
mode support") introduced MSI interrupts mode support, which seemed
safe enough with RTL8188EE and RTL8723BE as RealTek's testing results,
but some users reported their RTL8188EE modules could not connect to
any wireless network after the MSI mode was enabled by Ubuntu 14.04.

So, let's fallback to pin-based mode until rtlwifi's MSI support get
good compatibility.

BugLink: https://bugs.launchpad.net/bugs/1310512
Signed-off-by: Adam Lee <adam.lee@canonical.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agob43: don't pre-set radio_on variable to true
Rafał Miłecki [Tue, 22 Apr 2014 11:54:37 +0000 (13:54 +0200)]
b43: don't pre-set radio_on variable to true

Setting it to true during init doesn't seem to be any workaround while
it can cause problems (not enabling radio due to belief it's enabled).

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agob43: use b43_software_rfkill helper
Rafał Miłecki [Tue, 22 Apr 2014 11:54:36 +0000 (13:54 +0200)]
b43: use b43_software_rfkill helper

This removes dealing with pointers directly and allows tracking radio
state with radio_on variable.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agob43: bcma: respect GMODE (band choice) during core reset
Rafał Miłecki [Tue, 22 Apr 2014 11:54:35 +0000 (13:54 +0200)]
b43: bcma: respect GMODE (band choice) during core reset

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
10 years agoMerge branch 'for-john' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac802...
John W. Linville [Wed, 30 Apr 2014 16:04:27 +0000 (12:04 -0400)]
Merge branch 'for-john' of git://git./linux/kernel/git/jberg/mac80211-next

Conflicts:
net/mac80211/chan.c

10 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
John W. Linville [Wed, 30 Apr 2014 15:56:43 +0000 (11:56 -0400)]
Merge branch 'for-upstream' of git://git./linux/kernel/git/bluetooth/bluetooth-next

10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Mon, 28 Apr 2014 18:53:00 +0000 (14:53 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/jkirsher/net-next

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates

This series contains updates to i40e and i40evf.

Jacob provides a i40e patch to remove ptp_tx_work() because it conflicts
with the interrupt-based handler and will cause spurious warning messages
to be displayed in the kernel log.  Since the hardware can properly trigger
an interrupt, we do not need to poll for an available timestamp in a work
queue any more.

Greg provides two patches for i40e, first to retain the MAC filter list
when changing a port VLAN.  Second fixes an issue where the function to
set the VF MAC address was not taking the port VLAN filter into account
when setting/clearing/resetting the VF's host administered MAC address.

Mitch provides three patches for i40evf, first adds support for getting
and setting RSS hashing options, RSS LUT entries, and getting the number
of device channels from ethtool.  Fixes a panic that woudl occur in the VF
if the PF driver failed or was removed from the host kernel.

Shannon adds driver version string to the driver version command so that
can be passed down through the firmware to low-level NC-SI functions.

Neerav provides several patches, first redistributes queue vectors after
DCB reconfiguration because the total number of queues that are configured
for a VSI may change based on a change in the total number of TCs enabled
as a result of a change in the DCB configuration based on DCBX.  Second,
allows the driver probe to continue when querying DCB config fails.
Lastly, provides changes to retrieve and store missing link config
information that was not being cached in the struct i40e_link_status.

Kevin provides a couple of patches, first to change a int variable type
to u16 to avoid a typecheck failure.  Second to update a function
parameters to remove an argument which is no longer used.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: move the delivery of named messages out of nametbl lock
Ying Xue [Mon, 28 Apr 2014 10:00:10 +0000 (18:00 +0800)]
tipc: move the delivery of named messages out of nametbl lock

Commit a89778d8baf19cd7e728d81121a294a06cedaad1 ("tipc: add support
for link state subscriptions") introduced below possible deadlock
scenario:

       CPU0                          CPU1
T0:   tipc_publish()                 link_timeout()
T1:   tipc_nametbl_publish()         [grab node lock]*
T2:   [grab nametbl write lock]*     link_state_event()
T3:   named_cluster_distribute()     link_activate()
T4:   [grab node lock]*              tipc_node_link_up()
T5:                                  tipc_nametbl_publish()
T6:                                  [grab nametble write lock]*

The opposite order of holding nametbl write lock and node lock on
above two different paths may result in a deadlock. If we move the
the delivery of named messages via link out of name nametbl lock,
the reverse order of holding locks will be eliminated, as a result,
the deadlock will be killed as well.

Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotipc: fix out of bounds indexing
Erik Hugne [Mon, 28 Apr 2014 06:20:09 +0000 (08:20 +0200)]
tipc: fix out of bounds indexing

Commit 78acb1f9b898e85fa2c1e28e700b54b66b288e8d ("tipc: add
ioctl to fetch link names") introduced a buffer overflow bug where
specially crafted ioctl requests could cause out-of-bounds indexing
of the node->links array. This was caused by an incorrect check vs
MAX_BEARERS, and the static code checker complaint is:
net/tipc/node.c:459 tipc_node_get_linkname() error: buffer overflow 'node->links' 2 <= 2

Signed-off-by: Erik Hugne <erik.hugne@ericsson.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoctc: replace PTR_RET with PTR_ERR_OR_ZERO
Frank Blaschka [Mon, 28 Apr 2014 08:05:11 +0000 (10:05 +0200)]
ctc: replace PTR_RET with PTR_ERR_OR_ZERO

PTR_RET is deprecated. Do not recommend its usage anymore.
Use PTR_ERR_OR_ZERO instead.

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoclaw: replace PTR_RET with PTR_ERR_OR_ZERO
Frank Blaschka [Mon, 28 Apr 2014 08:05:10 +0000 (10:05 +0200)]
claw: replace PTR_RET with PTR_ERR_OR_ZERO

PTR_RET is deprecated. Do not recommend its usage anymore.
Use PTR_ERR_OR_ZERO instead.

Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqeth: Add new priority queueing options
Stefan Raspl [Mon, 28 Apr 2014 08:05:09 +0000 (10:05 +0200)]
qeth: Add new priority queueing options

The existing options for priority queueing within OSA devices were based on
the now partially defunct TOS field. This patch adds two new options as
follows:
   - prio_queueing_skb: bases priority queueing on skb-priority
   - prio_queueing_vlan: uses the priority code point in the 802.1Q header

Signed-off-by: Stefan Raspl <raspl@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Reviewed-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqeth: Extend priority queueing to IPv6
Stefan Raspl [Mon, 28 Apr 2014 08:05:08 +0000 (10:05 +0200)]
qeth: Extend priority queueing to IPv6

Make the current priority queueing logic apply to IPv6 traffic.

Signed-off-by: Stefan Raspl <raspl@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Reviewed-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqeth: replace PTR_RET with PTR_ERR_OR_ZERO
Duan Jiong [Mon, 28 Apr 2014 08:05:07 +0000 (10:05 +0200)]
qeth: replace PTR_RET with PTR_ERR_OR_ZERO

PTR_RET is deprecated. Do not recommend its usage anymore.
Use PTR_ERR_OR_ZERO instead.

Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agolcs: replace PTR_RET with PTR_ERR_OR_ZERO
Duan Jiong [Mon, 28 Apr 2014 08:05:06 +0000 (10:05 +0200)]
lcs: replace PTR_RET with PTR_ERR_OR_ZERO

PTR_RET is deprecated. Do not recommend its usage anymore.
Use PTR_ERR_OR_ZERO instead.

Signed-off-by: Duan Jiong <duanj.fnst@cn.fujitsu.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoqeth: Fix default queue setting in priority queueing
Stefan Raspl [Mon, 28 Apr 2014 08:05:05 +0000 (10:05 +0200)]
qeth: Fix default queue setting in priority queueing

The default queue for queueing modes prio_queueing_prec and
prio_queueing_tos is supposed to be QETH_DEFAULT_QUEUE. However, neither
mode will reset the default default_out_queue value when enabled, leaving
a previously set value activate as the default.
E.g. enabling one of the fixed queueing modes, e.g. no_prio_queueing:0,
and successively switching to any of the prio_queueing_* modes will
leave default_out_queue at the previous (wrong) value 0.

Signed-off-by: Stefan Raspl <raspl@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoptp: validate the requested frequency adjustment.
Richard Cochran [Sun, 27 Apr 2014 13:01:27 +0000 (15:01 +0200)]
ptp: validate the requested frequency adjustment.

PTP Hardware Clock drivers specify a maximum frequency adjustment that
their clocks can accommodate. Normally, user space programs will want to
respect the advertised limits. However, no kernel or driver code checks
that the dialed frequency offset is within the bounds, and out of range
values can lead to surprising results.

This patch fixes the issue by rejecting bad values.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv4: Use predefined value for readability
Hisao Tanabe [Sun, 27 Apr 2014 10:03:45 +0000 (19:03 +0900)]
ipv4: Use predefined value for readability

Signed-off-by: Hisao Tanabe <xtanabe@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoethtool: exit the loop when invalid index occurs
Jean Sacren [Sun, 27 Apr 2014 08:20:38 +0000 (02:20 -0600)]
ethtool: exit the loop when invalid index occurs

The commit 3de0b592394d ("ethtool: Support for configurable RSS hash
key") introduced a new function ethtool_copy_validate_indir() with
full iteration of the loop to validate the ring indices, which could
be an overkill. To minimize the impact, we ought to exit the loop as
soon as the invalid index occurs for the very first time. The
remaining loop simply doesn't serve any more purpose.

Signed-off-by: Jean Sacren <sakiwit@gmail.com>
Cc: Venkata Duvvuru <VenkatKumar.Duvvuru@Emulex.Com>
Signed-off-by: David S. Miller <davem@davemloft.net>