virtio-net: move sg off stack

Move sg structure off stack and into virtnet_info structure.
This helps remove extra sg_init_table calls as well as reduce
stack usage.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

benet: fix the misusage of zero dma address

benet driver wrongly assumes that zero is an invalid dma address
(calls dma_unmap_page for only non zero dma addresses). Zero is a
valid dma address on some architectures. The dma length can be used
here.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

benet: use the dma state API instead of the pci equivalents

The DMA API is preferred.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Sathya Perla <sathyap@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3c503: Fix IRQ probing

The driver attempts to select an IRQ for the NIC automatically by
testing which of the supported IRQs are available and then probing
each available IRQ with probe_irq_{on,off}().  There are obvious race
conditions here, besides which:
1. The test for availability is done by passing a NULL handler, which
   now always returns -EINVAL, thus the device cannot be opened:
   <http://bugs.debian.org/566522>
2. probe_irq_off() will report only the first ISA IRQ handled,
   potentially leading to a false negative.

There was another bug that meant it ignored all error codes from
request_irq() except -EBUSY, so it would 'succeed' despite this
(possibly causing conflicts with other ISA devices).  This was fixed
by ab08999d6029bb2c79c16be5405d63d2bedbdfea 'WARNING: some
request_irq() failures ignored in el2_open()', which exposed bug 1.

This patch:
1. Replaces the use of probe_irq_{on,off}() with a real interrupt handler
2. Adds a delay before checking the interrupt-seen flag
3. Disables interrupts on all failure paths
4. Distinguishes error codes from the second request_irq() call,
   consistently with the first

Compile-tested only.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

caif: tty's are kref objects so take a reference

I don't think this can be abused in this case but do things properly.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

caif: check write operations

write is optional for a tty device. Check that we have a write op rather
than calling NULL.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: include linux/proc_fs.h in dev_addr_lists.c

As pointed by Randy Dunlap, we must include linux/proc_fs.h in
net/core/dev_addr_lists.c, regardless of CONFIG_PROC_FS

Reported-by: Randy Dunlap <randy.dunlap@oracle.com>,
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix definition of netdev_for_each_mc_addr()

The first argument should be called ha, not mclist.  All callers use the
name "ha", but if they used a different name, there would be a compile
error.

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

flow: delayed deletion of flow cache entries

Speed up lookups by freeing flow cache entries later. After
virtualizing flow cache entry operations, the flow cache may now
end up calling policy or bundle destructor which can be slowish.

As gc_list is more effective with double linked list, the flow cache
is converted to use common hlist and list macroes where appropriate.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm: remove policy garbage collection

Policies are now properly reference counted and destroyed from
all code paths. The delayed gc is just an overhead now and can
be removed.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm: cache bundles instead of policies for outgoing flows

__xfrm_lookup() is called for each packet transmitted out of
system. The xfrm_find_bundle() does a linear search which can
kill system performance depending on how many bundles are
required per policy.

This modifies __xfrm_lookup() to store bundles directly in
the flow cache. If we did not get a hit, we just create a new
bundle instead of doing slow search. This means that we can now
get multiple xfrm_dst's for same flow (on per-cpu basis).

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

flow: virtualize flow cache entry methods

This allows to validate the cached object before returning it.
It also allows to destruct object properly, if the last reference
was held in flow cache. This is also a prepartion for caching
bundles in the flow cache.

In return for virtualizing the methods, we save on:
- not having to regenerate the whole flow cache on policy removal:
  each flow matching a killed policy gets refreshed as the getter
  function notices it smartly.
- we do not have to call flow_cache_flush from policy gc, since the
  flow cache now properly deletes the object if it had any references

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: bump to 1.7

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: Use (dev|netdev|netif)_<level> macro helpers

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: implement set/get_msglevel

{set,get}_msglevel is required to use netif_{err,dbg} macros.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: fix checkpatch errors.

This patch fixes multiple errors reported by checkpatch:
- else not on the ending brace of an if { }
- multiple occurences of for( instead of for (
- c99 comments
- assignment and tests on the same line
- test and statements on the same line
- macro with complex value not between parenthesis
- static variable with initialization value

No functionnal change.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: prefix all functions with au1000_

In order to avoid namespace clashes, prefix all internal driver functions
with au1000_.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: set MODULE_VERSION

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

au1000-eth: allow driver to be compiled as a module

This patch allows the au1000-eth driver to be compiled as a module.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of /linux/kernel/git/davem/net-2.6

Conflicts:
drivers/net/bonding/bond_main.c
drivers/net/via-velocity.c
drivers/net/wireless/iwlwifi/iwl-agn.c

NET: usb: Adding URB_ZERO_PACKET flag to usbnet.c

This patch adds setting of the urb transfer flag URB_ZERO_PACKET  before
submitting an urb for drivers that have requested it (by advertising flag
FLAG_SEND_ZLP).
The modification is in usbnet.c function usbnet_start_xmit().
This patch only adds the zero length flag.
A subsequent patch will address the buggy code we found when devices do not
advertise FLAG_SEND_ZLP in which case there is a possibility of transferring
packets with non-deterministic length.

This patch has been tested on kernel-2.6.34-rc3.
This patch has been checked against net-2.6 tree.
Signed-off-by: Elina Pasheva <epasheva@sierrawireless.com>
Signed-off-by: Rory Filer <rfiler@sierrawireless.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/irda: Add SuperH IrDA driver support

This is very simple driver for SuperH Mobile IrDA
which support SIR/MIR/FIR.
This patch add only SIR support for now.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/irda: sh_sir: Modify iounmap wrong execution

On sh_sir_probe function, there was a possibility that
iounmap is executed even though self->membase was NULL when error case.
This patch modify it.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/irda: sh_sir: fixup err return value on sh_sir_open

On sh_sir_open function, there was a possibility that
err variable didn't have value even though it is return value.
This patch modify it.

Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

socket: remove duplicate declaration of struct timespec

struct timespec ts was alreay defined. Reuse the previously
defined one and reduce the memory footprint on the stack by
16 bytes.

Signed-off-by: Hagen Paul Pfeifer <hagen@jauu.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

TIPC: Updated topology subscription protocol according to latest spec

This patch makes it explicit in the API that all fields in subscriptions and events exchanged with the Topology Server must be in
network byte order.
It also ensures that all fields of a subscription are compared when cancelling a subscription, in order to avoid inadvertent
cancelling of the wrong subscription.
Finally, the tipc module version is updated to 2.0.0, to reflect the API change.

Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add non-Virtex5 support for LL TEMAC driver

This patch adds support for using the LL TEMAC Ethernet driver on
non-Virtex 5 platforms by adding support for accessing the Soft DMA
registers as if they were memory mapped instead of solely through the
DCR's (available on the Virtex 5).

The patch also updates the driver so that it runs on the MicroBlaze.
The changes were tested on the PowerPC 440, PowerPC 405, and the
MicroBlaze platforms.

Signed-off-by: John Tyner <jtyner@cs.ucr.edu>
Signed-off-by: John Linn <john.linn@xilinx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

via-velocity: remove private #define

Registers and their bits from mii.h. Courtesy from ed.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

NET: sb1250: Fix compile warning in driver

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Update version to 3.109

This patch updates the tg3 version to 3.109.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Remove tg3_dump_state()

The tg3_dump_state() function is commented out and doesn't get used much
when debugging efforts are underway.  The chip architecture has also
changed enough that any debugging effort that might use it would have to
tailor it to the capabilities of the hardware.  This patch just removes
the function entirely.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Cleanup if codestyle

This patch cleans up the code style as it pertains to if statements.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: The case of switches

This patch fixes checkpatch errors related to switch statements.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Whitespace, constant, and comment updates

This patch fixes whitespace errors, preprocessor definition placement
oddities and updates comments.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Use VPD fw version when present

This patch extracts the firmware version from the VPD area and changes
the code to prefer that version version over the manually extracted
version.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Prepare FW version code for VPD versioning

The code that extracts the firmware version from the device's NVRAM
assumes the firmware version member is a clean slate.  The following
patch will add code to extract the firmware version from the VPD area
of NVRAM, so this assumption will no longer be true.  This patch adjusts
the versioning code to respect the VPD version if it exists.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Fix message 80 char violations

Commit 05dbe005386e7521153dce6c5ad95c98b73b80c7, "Use (pr|netdev)_<level>
macro helpers" extended some of the message lines longer than 80 chars.
This patch reels the line lengths back in.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: netdev_err() => dev_err()

Some uses of netdev_err() happen before the device has been registered.
This will cause the messages to be prepended with "(unregistered
net_device)" strings.  We can clean up the message by using dev_err()
instead.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Replace pr_err with sensible alternatives

All the instances of pr_err() in the tg3 driver have better
alternatives.  This patch replaces all pr_err() calls.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Restore likely() check in tg3_poll_msix()

When creating the new tg3_poll_msix() function, the likely() compiler
hint was dropped.  This patch reintroduces it.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Fix MII_TG3_DSP_EXP8 offset

The offset to the MII_TG3_DSP_EXP8 register is incorrect.  Fix it.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mac80211: Ensure initializing private mc_list in prepare_multicast().

Fix kernel panic by NULL pointer dereference in the context of
ieee80211_ops->prepare_multicast().

This bug was introduced by commit 22bedad3c.. ("net: convert
multicast list to list_head").

Call __hw_addr_init() in ieee80211_alloc_hw() to initialize
list_head of private device multicast list, like we do in
bond_init().

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Reviewed-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Add a missing local_irq_enable()

As noticed by Changli Gao, we must call local_irq_enable() after
rps_unlock()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

rps: fixed missed rps_unlock

Fix spin_unlock_irq which needs to be rps_unlock.

Signed-off-by: Tom Herbert <therbert@google.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: unmanaged L2TPv3 tunnels fixes

Followup to commit 789a4a2c
(l2tp: Add support for static unmanaged L2TPv3 tunnels)

One missing init in l2tp_tunnel_sock_create() could access random kernel
memory, and a bit field should be unsigned.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

SCTP: Change to use ipv6_addr_copy()

Change SCTP IPv6 code to use ipv6_addr_copy()

Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

icmp: Account for ICMP out errors

When ip_append() fails because of socket limit or memory shortage,
increment ICMP_MIB_OUTERRORS counter, so that "netstat -s" can report
these errors.

LANG=C netstat -s | grep "ICMP messages failed"
    0 ICMP messages failed

For IPV6, implement ICMP6_MIB_OUTERRORS counter as well.

# grep Icmp6OutErrors /proc/net/dev_snmp6/*
/proc/net/dev_snmp6/eth0:Icmp6OutErrors                    0
/proc/net/dev_snmp6/lo:Icmp6OutErrors                    0

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

smc91c92_cs: fix the problem of "Unable to find hardware address"

smc91c92_cs:
 *cvt_ascii_address returns 0, if success.
 *call free_netdev, if we can't find hardware address.

Signed-off-by: Ken Kawasaki <ken_kawasaki@spring.nifty.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Fix L2TP_DEBUGFS ifdef tests.

We have to check CONFIG_L2TP_DEBUGFS_MODULE as well as
CONFIG_L2TP_DEBUGFS.

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add missing semicolon to MODULE_ALIAS() in l2tp_netlink.c

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Update documentation

This patch adds documentation about the L2TPv3 functionality.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add support for static unmanaged L2TPv3 tunnels

This patch adds support for static (unmanaged) L2TPv3 tunnels, where
the tunnel socket is created by the kernel rather than being created
by userspace. This means L2TP tunnels and sessions can be created
manually, without needing an L2TP control protocol implemented in
userspace. This might be useful where the user wants a simple ethernet
over IP tunnel.

A patch to iproute2 adds a new command set under "ip l2tp" to make use
of this feature. This will be submitted separately.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add debugfs files for dumping l2tp debug info

The existing pppol2tp driver exports debug info to
/proc/net/pppol2tp. Rather than adding info to that file for the new
functionality added in this patch series, we add new files in debugfs,
leaving the old /proc file for backwards compatibility (L2TPv2 only).

Currently only one file is provided: l2tp/tunnels, which lists
internal debug info for all l2tp tunnels and sessions. More files may
be added later. The info is for debug and problem analysis only -
userspace apps should use netlink to obtain status about l2tp tunnels
and sessions.

Although debugfs does not support net namespaces, the tunnels and
sessions dumped in l2tp/tunnels are only those in the net namespace of
the process reading the file.

Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add L2TP ethernet pseudowire support

This driver presents a regular net_device for each L2TP ethernet
pseudowire instance. These interfaces are named l2tpethN by default,
though userspace can specify an alternative name when the L2TP
session is created, if preferred. When the pseudowire is established,
regular Linux networking utilities may be used to configure the
interface, i.e. give it IP address info or add it to a bridge. Any
data passed over the interface is carried over an L2TP tunnel.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Convert rwlock to RCU

Reader/write locks are discouraged because they are slower than spin
locks. So this patch converts the rwlocks used in the per_net structs
to rcu.

Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add netlink control API for L2TP

In L2TPv3, we need to create/delete/modify/query L2TP tunnel and
session contexts. The number of parameters is significant. So let's
use netlink. Userspace uses this API to control L2TP tunnel/session
contexts in the kernel.

The previous pppol2tp driver was managed using [gs]etsockopt(). This
API is retained for backwards compatibility. Unlike L2TPv2 which
carries only PPP frames, L2TPv3 can carry raw ethernet frames or other
frame types and these do not always have an associated socket
family. Therefore, we need a way to use L2TP sessions that doesn't
require a socket type for each supported frame type. Hence netlink is
used.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

netlink: Export genl_lock() API for use by modules

This lets kernel modules which use genl netlink APIs serialize netlink
processing.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add L2TPv3 IP encapsulation (no UDP) support

This patch adds a new L2TPIP socket family and modifies the core to
handle the case where there is no UDP header in the L2TP
packet. L2TP/IP uses IP protocol 115. Since L2TP/UDP and L2TP/IP
packets differ in layout, the datapath packet handling code needs
changes too. Userspace uses an L2TPIP socket instead of a UDP socket
when IP encapsulation is required.

We can't use raw sockets for this because the semantics of raw sockets
don't lend themselves to the socket-per-tunnel model - we need to

Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Update PPP-over-L2TP driver to work over L2TPv3

This patch makes changes to the L2TP PPP code for L2TPv3.

The existing code has some assumptions about the L2TP header which are
broken by L2TPv3. Also the sockaddr_pppol2tp structure of the original
code is too small to support the increased size of the L2TPv3 tunnel
and session id, so a new sockaddr_pppol2tpv3 structure is needed. In
the socket calls, the size of this structure is used to tell if the
operation is for L2TPv2 or L2TPv3.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add L2TPv3 protocol support

The L2TPv3 protocol changes the layout of the L2TP packet
header. Tunnel and session ids change from 16-bit to 32-bit values,
data sequence numbers change from 16-bit to 24-bit values and PPP-specific
fields are moved into protocol-specific subheaders.

Although this patch introduces L2TPv3 protocol support, there are no
userspace interfaces to create L2TPv3 sessions yet.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Add ppp device name to L2TP ppp session data

When dumping L2TP PPP sessions using /proc/net/pppol2tp, get the
assigned PPP device name from PPP using ppp_dev_name().

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ppp: Add ppp_dev_name() exported function

ppp_dev_name() gives PPP users visibility of a ppp channel's device
name. This can be used by L2TP drivers to dump the assigned PPP
interface name.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Split pppol2tp patch into separate l2tp and ppp parts

This patch splits the pppol2tp driver into separate L2TP and PPP parts
to prepare for L2TPv3 support. In L2TPv3, protocols other than PPP can
be carried, so this split creates a common L2TP core that will handle
the common L2TP bits which protocol support modules such as PPP will
use.

Note that the existing pppol2tp module is split into l2tp_core and
l2tp_ppp by this change.

There are no feature changes here. Internally, however, there are
significant changes, mostly to handle the separation of PPP-specific
data from the L2TP session and to provide hooks in the core for
modules like PPP to access.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

l2tp: Relocate pppol2tp driver to new net/l2tp directory

This patch moves the existing pppol2tp driver from drivers/net into a
new net/l2tp directory, which is where the upcoming L2TPv3 code will
live. The existing CONFIG_PPPOL2TP config option is left in its
current place to avoid "make oldconfig" issues when an existing
pppol2tp user takes this change. (This is the same approach used for
the pppoatm driver, which moved to net/atm.)

There are no code changes. The existing drivers/net/pppol2tp.c is
simply moved to net/l2tp.

Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: convert multicast list to list_head

Converts the list and the core manipulating with it to be the same as uc_list.

+uses two functions for adding/removing mc address (normal and "global"
 variant) instead of a function parameter.
+removes dev_mcast.c completely.
+exposes netdev_hw_addr_list_* macros along with __hw_addr_* functions for
 manipulation with lists on a sandbox (used in bonding and 80211 drivers)

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: move address list functions to a separate file

+little renaming of unicast functions to be smooth with multicast ones

Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: update version to 5.0.1

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix endianness in fw validation

cpu_to_le32 was missing and used improperly.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix interface attach sequence

Interface should be visible even if resource allocation fails.
netif_device_attach should be called for every netif_device_detach.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: add driver debug support

Add debug print in driver, can be tuned by ethtool msg level
callback.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: use IDC defined timeout value

o USE/Read IDC defined timeout value from ROM.
o While resetting chip, don't wait for other pci-func to respond,
  more than reset_ack_timeo seconds,

Signed-off-by: Sucheta Chakraborty <sucheta.chakraborty@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix onchip memory access

Fix incorrect offset calculation and remove unnecessary remap
of the region in bar 0 to access onchip memory.

This was leading to read incorrect values by debug tools.

Signed-off-by: Dhananjay Phadke <dhananjay.phadke@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: update oncard memory size check

All QLogic converged NICs have 128-bit 128MB on card memory.
Fix the limit check from 64MB to 128MB and remove unnecessary
64-bit read/write checks.

Signed-off-by: Dhananjay Phadke <dhananjay.phadke@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: handle queue manager access

Check the access by tools for hardware queue engine and handle it
separately than other block registers, otherwise incorrect data
is returned.

Signed-off-by: Dhananjay Phadke <dhananjay.phadke@qlogic.com>
Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: fix fw load from file

Rarely: Fw file size can be unaligned to 8.

Signed-off-by: Amit Kumar Salecha <amit.salecha@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Add module table to all existing phy drivers

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

phylib: Support phy module autoloading

We don't use the normal hotplug mechanism because it doesn't work. It will
load the module some time after the device appears, but that's not good
enough for us -- we need the driver loaded _immediately_ because otherwise
the NIC driver may just abort and then the phy 'device' goes away.

[bwh: s/phy/mdio/ in module alias, kerneldoc for struct mdio_device_id]

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Acked-by: Andy Fleming <afleming@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: illegal_highdma() fix

Followup to commit 5acbbd428db47b12f137a8a2aa96b3c0a96b744e
(net: change illegal_highdma to use dma_mask)

If dev->dev.parent is NULL, we should not try to dereference it.

Dont force inline illegal_highdma() as its pretty big now.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: Adding PCI SRIOV support

- Patch adds support to enable PCI SRIOV in the driver and changes to handle initialization of PCI virtual functions.
- Function handler to change mac addresses for VF from its corresponding PF.

Signed-off-by: Sarveshwar Bandi <sarveshwarb@serverengines.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/pcmcia/3c589_cs: using netdev_info and friends where appropriate

Signed-off-by: Alexander Kurz <linux@kbdbabel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

netdev/fec.c: add phylib supporting to enable carrier detection (v2)

BugLink: http://bugs.launchpad.net/bugs/457878
v2:
 - remove duplicated phy_speed caculation
 - fix the phy_speed caculation according to the DataSheet

v1:
 - removed old MII phy control code
 - add phylib supporting
 - add ethtool interface to make user space NetworkManager works

Tested on Freescale i.MX51 Babbage board.

This patch is based on a patch from Frederic Rodo <fred.rodo@gmail.com>

Cc: Frederic Rodo <fred.rodo@gmail.com>
Signed-off-by: Bryan Wu <bryan.wu@canonical.com>
Acked-by: Amit Kucheria <amit.kucheria@canonical.com>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: change illegal_highdma to use dma_mask

Robert Hancock pointed out two problems about NETIF_F_HIGHDMA:

-Many drivers only set the flag when they detect they can use 64-bit DMA,
since otherwise they could receive DMA addresses that they can't handle
(which on platforms without IOMMU/SWIOTLB support is fatal). This means that if
64-bit support isn't available, even buffers located below 4GB will get copied
unnecessarily.

-Some drivers set the flag even though they can't actually handle 64-bit DMA,
which would mean that on platforms without IOMMU/SWIOTLB they would get a DMA
mapping error if the memory they received happened to be located above 4GB.

http://lkml.org/lkml/2010/3/3/530

We can use the dma_mask if we need bouncing or not here. Then we can
safely fix drivers that misuse NETIF_F_HIGHDMA.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: Added GRO support

Adding GRO support on top of the HW LRO (TPA) support –
there is no measurable performance drawback of adding GRO
on top of it, and it allows better performance when LRO (TPA)
is turned off for virtualization or bridging.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

flow: structurize flow cache

Group all per-cpu data to one structure instead of having many
globals. Also prepare the internals so that we can have multiple
instances of the flow cache if needed.

Only the kmem_cache is left as a global as all flow caches share
the same element size, and benefit from using a common cache.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm: remove policy lock when accessing policy->walk.dead

All of the code considers ->dead as a hint that the cached policy
needs to get refreshed. The read side can just drop the read lock
without any side effects.

The write side needs to make sure that it's written only exactly
once. Only possible race is at xfrm_policy_kill(). This is fixed
by checking result of __xfrm_policy_unlink() when needed. It will
always succeed if the policy object is looked up from the hash
list (so some checks are removed), but it needs to be checked if
we are trying to unlink policy via a reference (appropriate
checks added).

Since policy->walk.dead is written exactly once, it no longer
needs to be protected with a write lock.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm_user: verify policy direction at XFRM_MSG_POLEXPIRE handler

Add missing check for policy direction verification. This is
especially important since without this xfrm_user may end up
deleting per-socket policy which is not allowed.

Signed-off-by: Timo Teras <timo.teras@iki.fi>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

xfrm: Remove xfrm_state_genid

The xfrm state genid only needs to be matched against the copy
saved in xfrm_dst.  So we don't need a global genid at all.  In
fact, we don't even need to initialise it.

Based on observation by Timo Teräs.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

r8169: clean up my printk uglyness

Fix formatting on r8169 printk

Brandon Philips noted that I had a spacing issue in my printk for the
last r8169 patch that made it quite ugly.  Fix that up and add the PFX
macro to it as well so it looks like the other r8169 printks

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: Hook up cxgb4 to Kconfig and Makefile

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: Add main driver file and driver Makefile

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: Add remaining driver headers and L2T management

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: Add packet queues and packet DMA code

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: Add HW and FW support code

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4: Add register, message, and FW definitions

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6 fib: Make rt6_info{} more cache-line aware.

The head element of rt6_info{} is dst_entry{}, and
IPv6 specific elements follow.

Because elements at the end of dst_entry{} are frequently
updated, it is not good to put frequently-used static
elements, such as rt6i_idev, rt6i_dst or rt6i_flags in the
same cache line.

On the other hand, fib6_table, rt6i_node or rt6i_gateway are
rarely used, so it is okay to stay in the same cache line.

Let's rearrange rt6_info{}.

Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

rps: keep the old behavior on SMP without rps

keep the old behavior on SMP without rps

RPS introduces a lock operation to per cpu variable input_pkt_queue on
SMP whenever rps is enabled or not. On SMP without RPS, this lock isn't
needed at all.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
net/core/dev.c | 42 ++++++++++++++++++++++++++++--------------
1 file changed, 28 insertions(+), 14 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>

r8169: Fix rtl8169_rx_interrupt()

In case a reset is performed, rtl8169_rx_interrupt() is called from
process context instead of softirq context. Special care must be taken
to call appropriate network core services (netif_rx() instead of
netif_receive_skb()). VLAN handling also corrected.

Reported-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Tested-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Diagnosed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

gen_estimator: deadlock fix

One of my test machine got a deadlock during "tc" sessions,
adding/deleting classes & filters, using traffic estimators.

After some analysis, I believe we have a potential use after free case
in est_timer() :

spin_lock(e->stats_lock); << HERE >>
read_lock(&est_lock);
if (e->bstats == NULL)   << TEST >>
goto skip;

Test is done a bit late, because after estimator is killed, and before
rcu grace period elapsed, we might already have freed/reuse memory where
e->stats_locks points to (some qdisc->q.lock)

A possible fix is to respect a rcu grace period at Qdisc dismantle time.

On 64bit, sizeof(struct Qdisc) is exactly 192 bytes. Adding 16 bytes to
it (for struct rcu_head) is a problem because it might change
performance, given QDISC_ALIGNTO is 32 bytes.

This is why I also change QDISC_ALIGNTO to 64 bytes, to satisfy most
current alignment requirements.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: remove redundant verification code

The check if error signaling is wanted (inet->recverr != 0) is done by
the caller: raw.c:raw_err() and udp.c:__udp4_lib_err(), so there is no
need to check this condition again.

Signed-off-by: Hagen Paul Pfeifer <hagen@jauu.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

acenic: use the dma state API instead of the pci equivalents

The DMA API is preferred.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>