cnic: Fix rtnl deadlock

When cnic_stop_hw() -> cnic_cm_stop_bnx2x_hw() is called under rtnl_lock()
from NETDEV_DOWN event, it waits for cnic_delete_task() to complete.
It will deadlock when cnic_delete_task() takes rtnl_lock() before
calling cnic_ulp_stop_one().

We fix it by removing the rtnl_lock() in cnic_delete_task().
cnic_ulp_stop_one() has mutex and atomic bit ops to prevent important
operations from being done more than once, so it is not necessary to take
rtnl_lock().

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: atl*: convert to hw_features

Things left as they were:
 - atl1: is RX checksum really enabled?
 - atl2: copy-paste from atl1, with-errors-on-modify I presume
 - atl1c: there's a bug: MTU can't be changed if device is not up

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: tg3: convert to hw_features

Cleanup hint: Some features are calculated in tg3_get_invariants() and
the rest in its caller --- tg3_init_one(). This is not changed here.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Acked-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: enic: convert to hw_features

As the driver uses GRO and not LRO, LRO settings are ignored anyway
and are removed here to avoid confusion.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: benet: convert to hw_features

Simple conversion.

This also fixes a bug in TX checksum toggling --- driver was changing
NETIF_F_HW_CSUM instead of NETIF_F_IP_CSUM+NETIF_F_IPV6_CSUM.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

skge: convert to hw_features

just IP_CSUM.  This needs testing and so is not changed here.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: call FLR after setup wol in be_shutdown

Calling setup_wol after a reset is inconsequential.
The WOL setting should be programmed before FLR.
And yes, FLR does not erase wol information.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: dynamically allocate adapter->vf_cfg

Instead of a fixed sized array for vf_cfg, allocate the size dynamically
depending on number of VFs the device supports.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: fix to get max VFs supported from adapter

The user supplied num_vfs value need not be compared
against a static BE_MAX_VF, but can be checked against
the actual VFs that the device can support.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: use common method to check for sriov function type

Lancer and BE can both use SLI_INTF_REG to check a VF or a PF.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: add rxhash support

Add rxhash support,
Based on initial work by Eric Dumazet.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

Conflicts:
drivers/net/benet/be_main.c

ipv4: Fix "Set rt->rt_iif more sanely on output routes."

Commit 1018b5c01636c7c6bda31a719bda34fc631db29a ("Set rt->rt_iif more
sanely on output routes.")  breaks rt_is_{output,input}_route.

This became the cause to return "IP_PKTINFO's ->ipi_ifindex == 0".

To fix it, this does:

1) Add "int rt_route_iif;" to struct rtable

2) For input routes, always set rt_route_iif to same value as rt_iif

3) For output routes, always set rt_route_iif to zero.  Set rt_iif
   as it is done currently.

4) Change rt_is_{output,input}_route() to test rt_route_iif

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix skb_add_data_nocache() to calc csum correctly

commit c6e1a0d12ca7b4f22c58e55a16beacfb7d3d8462 broken the calc
 (net: Allow no-cache copy from user on transmit)
of checksum, which may cause some tcp packets be dropped because
incorrect checksum. ssh does not work under today's net-next-2.6
tree.

Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ewrk3: convert to set_phys_id

Use ethtool infrastructure for blinking, which is now does
locking at higher level.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: convert to set_phys_id

Convert driver to use new ethtool set_phys_id.
Not completely sure that this is correct for all cases of device
up/down and doing operation. Compile tested only.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: netxen: convert to hw_features

Rather simple conversion to hw_features.

Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

capi: Perform scheduled capifs removal

udev fully replaces this special file system that only contains CAPI
NCCI TTY device nodes. User space (pppdcapiplugin) works without
noticing the difference.

Signed-off-by: Jan Kiszka <jan.kiszka@web.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed bit field; int have_hotplug_status_watch:1

Fixes error from sparse:
  CHECK   drivers/net/xen-netback/xenbus.c
drivers/net/xen-netback/xenbus.c:29:40: error: dubious one-bit signed bitfield

        int have_hotplug_status_watch:1;

Reported-by: Dr. David Alan Gilbert <linux@treblig.org>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

MAINTAINERS: add entry for Xen network backend

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: Fix suspend/resume operation

eq_next_idx is not getting reset to zero during suspend.
This causes resume to fail. Added the fix.

Signed-off-by: Sarveswara Rao Mygapula <sarveswararao.mygapula@emulex.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: Rename some struct members for clarity

Renamed msix_vec_idx to eq_idx in be_eq_obj struct.
Renamed msix_vec_next_idx to eq_next_idx in be_adapter structure.
These members are used in INTX mode also.

Signed-off-by: Sarveswara Rao Mygapula <sarveswararao.mygapula@emulex.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s2io: convert to set_phys_id (v2)

Convert to new ethtool set physical id model. Remove no longer used
timer, and fix docbook comment.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

niu: convert to new ethtool set_phys_id

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pcnet32: convert to set_phys_id

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

benet: convert to set_phys_id

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2x: convert to set_phys_id

Also cleanup error codes to no lie about things that driver doesn't
support. If device is down report -EAGAIN (same as Broadcom), and
if port doesn't blink then error as well.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bnx2: convert to set_phys_id

In this case, need to add element to device private to hold
original led state.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

vxge: convert to set_phys_id

Also fix up incorrect docbook comment

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb3: implement set_phys_id

Implement new ethtool set_phys_id on Chelsio cxgb3 board.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: implement ethtool set_phys_id

Implement control of LED via set_phys_id.
Note: since PHY is powered off if device is down, this board
only allows blinking if device is up.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

skge: implement set_phys_id

Implement set_phys_id led control on SysKonnect board.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

sky2: support ethtool set_phys_id

Use ethtool set_phys_id to control LED. Fixes issues with RTNL being
held for extended periods.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pppoe: drop PPPOX_ZOMBIEs in pppoe_flush_dev

otherwise we loop forever if a PPPoE socket was set
to PPPOX_ZOMBIE state by a PADT message when the
ethernet device is going down afterwards.

Signed-off-by: Ulrich Weber <uweber@astaro.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6

dsa/mv88e6131: add support for mv88e6085 switch

The mv88e6085 is identical to the mv88e6095, except that all ports are
10/100 Mb/s, so use the existing setup code except for the cpu/dsa speed
selection in _setup_port().

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Acked-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: Enable RFS sk_rxhash tracking for ipv6 sockets (v2)

properly record sk_rxhash in ipv6 sockets (v2)

Noticed while working on another project that flows to sockets which I had open
on a test systems weren't getting steered properly when I had RFS enabled.
Looking more closely I found that:

1) The affected sockets were all ipv6
2) They weren't getting steered because sk->sk_rxhash was never set from the
incomming skbs on that socket.

This was occuring because there are several points in the IPv4 tcp and udp code
which save the rxhash value when a new connection is established.  Those calls
to sock_rps_save_rxhash were never added to the corresponding ipv6 code paths.
This patch adds those calls.  Tested by myself to properly enable RFS
functionalty on ipv6.

Change notes:
v2:
Filtered UDP to only arm RFS on bound sockets (Eric Dumazet)

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

qlcnic: Update version number to 5.0.16

Bumped up version number to 5.0.16

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Fix LRO disable

o In dev->open LRO was enabled by default, enable it depending
  upon netdev->features , kernel may have disabled it.
o Configure LRO when interface is up.

Signed-off-by: Rajesh Borundia <Rajesh.Borundia@qlogic.com>
Signed-off-by: Anirban Chakraborty <Anirban.Chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Use flt method to determine flash fw region

Use flash layout table to get flash fw starting address and its size.
If that fails, use legacy method.

Signed-off-by: Sritej Velaga <sritej.velaga@qlogic.com>
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Remove unused code

Cleaned up unused codes for interrupt coalescence settings

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Changes to VLAN code

Made changes to VLAN code comply with new VLAN infrastructure in kernel

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Code optimization patch

Optimized code resulted in achieving lower CPU utilization on transmit path
and higher throughput for small packet sizes (64 bytes).

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Cleanup patch

 1. Changed adapter structure to move away from embedding hardware and
    receive context structs and use pointers to those objects
 2. Packed all the structs that interface with FW
 3. Removed unused code and structs

Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Memory leak fix

Fix a memory leak in error path of pci info.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

qlcnic: Make PCI info available in all modes

Before this fix, PCI info was available only when multiple NIC functions
are present on the same port.

Signed-off-by: Sony Chacko <sony.chacko@qlogic.com>
Signed-off-by: Anirban Chakraborty <anirban.chakraborty@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

be2net: Fix a potential crash during shutdown.

adapter could remain uninitialized if probe fails for some reason.
A null pointer access could cause a crash if be_shutdown
is called after that.

Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bna: Fix for handling firmware heartbeat failure

This patch contains a fix for gracefully handling firmware heartbeat
failure instead of forcing panic.

Signed-off-by: Debashis Dutt <ddutt@brocade.com>
Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: convert protocol handling to RCU

This patch removes spin_locks at CAN socket creation time by using RCU.

Inspired by the discussion with Kurt van Dijck and Eric Dumazet the RCU code
was partly derived from af_phonet.c

Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-davem' of git://git./linux/kernel/git/bwh/sfc-next-2.6

can: mcp251x: Allow pass IRQ flags through platform data.

When an interrupt occurs, the INT pin is driven low by the
MCP251x controller (falling edge) but in some cases the INT
pin can be connected to the MPU through a transistor or level
translator which inverts this signal. In this case interrupt
should be configured in rising edge.

This patch adds support to pass the IRQ flags via
mcp251x_platform_data.

Signed-off-by: Enric Balletbo i Serra <eballetbo@iseebcn.com>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Acked-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

smsc911x: fix mac_lock acquision before calling smsc911x_mac_read

When SMSC911X_SAVE_MAC_ADDRESS flag is enabled the driver calls
smsc911x_mac_read and smsc911x_mac_read function without acquiring mac_lock
spinlock

This patch fixes following warning

smsc911x: Driver version 2008-10-21.
------------[ cut here ]------------
WARNING: at drivers/net/smsc911x.c:261 smsc911x_mac_read+0x24/0x220()
Modules linked in:
[<c0060858>] (unwind_backtrace+0x0/0xe0) from [<c009322c>] (warn_slowpath_common+0x4c/0x64)
[<c009322c>] (warn_slowpath_common+0x4c/0x64) from [<c009325c>] (warn_slowpath_null+0x18/0x1c)
[<c009325c>] (warn_slowpath_null+0x18/0x1c) from [<c0324bec>] (smsc911x_mac_read+0x24/0x220)
[<c0324bec>] (smsc911x_mac_read+0x24/0x220) from [<c0434788>] (smsc911x_read_mac_address+0x18/0x6c)
[<c0434788>] (smsc911x_read_mac_address+0x18/0x6c) from [<c0434c74>] (smsc911x_drv_probe+0x498/0x1788)
[<c0434c74>] (smsc911x_drv_probe+0x498/0x1788) from [<c02d3e54>] (platform_drv_probe+0x14/0x18)
[<c02d3e54>] (platform_drv_probe+0x14/0x18) from [<c02d2d60>] (driver_probe_device+0xc8/0x184)
[<c02d2d60>] (driver_probe_device+0xc8/0x184) from [<c02d2e84>] (__driver_attach+0x68/0x8c)
[<c02d2e84>] (__driver_attach+0x68/0x8c) from [<c02d1fc8>] (bus_for_each_dev+0x48/0x74)
[<c02d1fc8>] (bus_for_each_dev+0x48/0x74) from [<c02d2660>] (bus_add_driver+0x9c/0x228)
[<c02d2660>] (bus_add_driver+0x9c/0x228) from [<c02d3598>] (driver_register+0xa0/0x124)
[<c02d3598>] (driver_register+0xa0/0x124) from [<c0050668>] (do_one_initcall+0x94/0x168)
[<c0050668>] (do_one_initcall+0x94/0x168) from [<c0008984>] (kernel_init+0xa0/0x154)
[<c0008984>] (kernel_init+0xa0/0x154) from [<c005afac>] (kernel_thread_exit+0x0/0x8)
---[ end trace 2c931a35b7885770 ]---

Signed-off-by: Enric Balletbo i Serra <eballetbo@iseebcn.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Support 4mb flash sizes for 5717 and 5719

If a 5717 or 5719 NVRAM part is manually strapped and is 2mb in size,
the driver needs to look at the NVRAM size field rather than infer it
from the strapping itself.

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

tg3: Enable 5720 support

This patch adds the 5720 device ID to the PCI table, thus enabling 5720
support.

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

tg3: Add 5720 PHY ID

This patch adds the 5720 PHY ID.

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

tg3: Add 5720 H2BMC support

This patch adds support for the new Host to BMC feature.

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

tg3: Add 5720 NVRAM decoding

The 5720 implements its own NVRAM pin strapping scheme.  This patch adds
the required support.

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

tg3: Add 5720 ASIC rev

This patch adds support for the 5720 ASIC rev.

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

tg3: Reintroduce 5717_PLUS

This patch reintroduces the TG3_FLG3_5717_PLUS to identify 5717 and
later devices.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: 5717_PLUS => 57765_PLUS

The 57765 arrived before the 5717 and has a subset of the features
supported by the 5717.  This patch renames the 5717_PLUS flag so that it
can be reintroduced to designate only 5717 and later devices.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tg3: Cleanup extended rx ring size code

Hardcoded values are used in multiple places to describe the maximum rx
ring sizes.  This patch replaces those values with preprocessor
constants.  This patch also introduces a new TG3_FLG3_LRG_PROD_RING_CAP
to determine if the device is capable of supporting larger ring sizes.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-2.6

sfc: Implement ethtool_ops::set_phys_id instead of ethtool_ops::phys_id

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

ethtool: Change ETHTOOL_PHYS_ID implementation to allow dropping RTNL

The ethtool ETHTOOL_PHYS_ID command runs for an arbitrarily long
period of time, holding the RTNL lock.  This blocks routing updates,
device enumeration, and various important operations that one might
want to keep running while hunting for the flashing LED.

We need to drop the RTNL lock during this operation, but currently the
core implementation is a thin wrapper around a driver operation and
drivers may well depend upon holding the lock.

Define a new driver operation 'set_phys_id' with an argument that sets
the ID indicator on/off/inactive/active (the last optional, for any
driver or firmware that prefers to handle blinking asynchronously).
When this is defined, the ethtool core drops the lock while waiting
and only acquires it around calls to this operation.

Deprecate the 'phys_id' operation in favour of this.  It can be
removed once all in-tree drivers are converted.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

ethtool: Fill out and update comment for struct ethtool_ops

Briefly document all operations (except get_rx_ntuple), including
whether they may return an error code and whether they are deprecated.
Also mention some things that should be handled by the ethtool core
rather than by drivers.

Briefly document general requirements for callers.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

ethtool: Convert struct ethtool_ops comment to kernel-doc format

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

sfc: Implement generic features interface

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

sfc: Enable all TSO features on VLANs

The TSO code already supports IPv6 on VLAN, so enable it.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>

net: Allow no-cache copy from user on transmit

This patch uses __copy_from_user_nocache on transmit to bypass data
cache for a performance improvement.  skb_add_data_nocache and
skb_copy_to_page_nocache can be called by sendmsg functions to use
this feature, initial support is in tcp_sendmsg.  This functionality is
configurable per device using ethtool.

Presumably, this feature would only be useful when the driver does
not touch the data.  The feature is turned on by default if a device
indicates that it does some form of checksum offload; it is off by
default for devices that do no checksum offload or indicate no checksum
is necessary.  For the former case copy-checksum is probably done
anyway, in the latter case the device is likely loopback in which case
the no cache copy is probably not beneficial.

This patch was tested using 200 instances of netperf TCP_RR with
1400 byte request and one byte reply.  Platform is 16 core AMD x86.

No-cache copy disabled:
   672703 tps, 97.13% utilization
   50/90/99% latency:244.31 484.205 1028.41

No-cache copy enabled:
   702113 tps, 96.16% utilization,
   50/90/99% latency 238.56 467.56 956.955

Using 14000 byte request and response sizes demonstrate the
effects more dramatically:

No-cache copy disabled:
   79571 tps, 34.34 %utlization
   50/90/95% latency 1584.46 2319.59 5001.76

No-cache copy enabled:
   83856 tps, 34.81% utilization
   50/90/95% latency 2508.42 2622.62 2735.88

Note especially the effect on latency tail (95th percentile).

This seems to provide a nice performance improvement and is
consistent in the tests I ran.  Presumably, this would provide
the greatest benfits in the presence of an application workload
stressing the cache and a lot of transmit data happening.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: range check STP parameters

Apply restrictions on STP parameters based 802.1D 1998 standard.
   * Fixes missing locking in set path cost ioctl
   * Uses common code for both ioctl and sysfs

This is based on an earlier patch Sasikanth V but with overhaul.

Note:
1. It does NOT enforce the restriction on the relationship max_age and
   forward delay or hello time because in existing implementation these are
   set as independant operations.

2. If STP is disabled, there is no restriction on forward delay

3. No restriction on holding time because users use Linux code to act
   as hub or be sticky.

4. Although standard allow 0-255, Linux only allows 0-63 for port priority
   because more bits are reserved for port number.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: allow creating bridge devices with netlink

Add netlink device ops to allow creating bridge device via netlink.
This works in a manner similar to vlan, macvlan and bonding.

Example:
  # ip link add link dev br0 type bridge
  # ip link del dev br0

The change required rearranging initializtion code to deal with
being called by create link. Most of the initialization happens
in br_dev_setup, but allocation of stats is done in ndo_init callback
to deal with allocation failure. Sysfs setup has to wait until
after the network device kobject is registered.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: allow creating/deleting fdb entries via netlink

Use RTM_NEWNEIGH and RTM_DELNEIGH to allow updating of entries
in bridge forwarding table. This allows manipulating static entries
which is not possible with existing tools.

Example (using bridge extensions to iproute2)
   # br fdb add 00:02:03:04:05:06 dev eth0

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: add netlink notification on forward entry changes

This allows applications to query and monitor bridge forwarding
table in the same method used for neighbor table. The forward table
entries are returned in same structure format as used by the ioctl.
If more information is desired in future, the netlink method is
extensible.

Example (using bridge extensions to iproute2)
  # br monitor

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: split rcu and no-rcu cases of fdb lookup

In some cases, look up of forward database entry is done with RCU;
and for others no RCU is needed because of locking. Split the two
cases into two differnt loops (and take off inline).

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: track last used time in forwarding table

Adds tracking the last used time in forwarding table.
Rename ageing_timer to updated to better describe it.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: change arguments to fdb_create

Later patch provides ability to create non-local static entry.
To make this easier move the updating of the flag values to
after the code that creates entry.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: Don't pass invalid dst_entry pointer to dst_release().

Make sure dst_release() is not called with error pointer. This is
similar to commit 4910ac6c526d2868adcb5893e0c428473de862b5 ("ipv4:
Don't ip_rt_put() an error pointer in RAW sockets.").

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

iwlwifi: accept EEPROM version 0x423 for iwl6000

A number of these devices have appeared "in the wild", and apparently
the Windows driver is perfectly happy to support this EEPROM version.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cc: stable@kernel.org
Acked-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>

rt2x00: fix cancelling uninitialized work

{rx,tx}done_work's are only initialized for usb devices.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: stable@kernel.org
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

rtlwifi: Fix some warnings/bugs

Some compiler/architecture combinations generate some warnings that are
not seen on my main system. Two of the "warnings" about unitialized variables
are really bugs.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

p54usb: IDs for two new devices

Cc: <stable@kernel.org>
Reported-by: Mark Davis [via p54/devices wiki]
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wl12xx: fix potential buffer overflow in testmode nvs push

We were allocating the size of the NVS file struct and not checking
whether the length of the buffer passed was correct before copying it
into the allocated memory.  This is a security hole because buffer
overflows can occur if the userspace passes a bigger file than what is
expected.

With this patch, we check if the size of the data passed from
userspace matches the size required.

This bug was introduced in 2.6.36.

Cc: stable@kernel.org
Reported-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: reset rx idle timer from tasklet

2.6.38 added WARN_ON(in_irq) in del_timer_sync that triggers on zd1211rw when
reseting rx idle timer in urb completion handler.

Move timer reseting to tasklet.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

zd1211rw: remove URB_SHORT_NOT_OK flag in zd_usb_iowrite16v_async()

Patch removes the bogus flag introduced by upstream commit
eefdbec1ea8b7093d2c09d1825f68438701723cf. Old code had buffer length check
that new code tried to handle with URB_SHORT_NOT_OK flag. With USB debugging
enabled bogus flag caused usb_submit_urb fail.

Remove URB_SHORT_NOT_OK flag and add buffer length check to urb completion
handler.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=32092
Reported-by: Jonathan Callen <abcd@gentoo.org>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: Fix duplicate frames on cooked monitor

Cleaning the ieee80211_rx_data.flags field here is wrong, instead the
flags should be valid accross processing the frame on different
interfaces. Fix this by removing the incorrect flags=0 assignment.

Introduced in commit 554891e63a29af35cc6bb403ef34e319518114d0
(mac80211: move packet flags into packet).

Signed-off-by: Helmut Schaa <helmut.schaa@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

mac80211: fix comment regarding aggregation buf_size

The description for buf_size was misleading and
just said you couldn't TX larger aggregates, but
of course you can't TX aggregates in a way that
would exceed the window either, which is possible
even if the aggregates are shorter than that.

Expand the description, thanks to Emmanuel for
explaining this to me.

Cc: Emmanuel Grumbach <egrumbach@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

ath9k: Fix phy info print message with AR9485 chipset.

The phy information print during driver init time doesn't show
the numeric part of the chip name properly for AR9485. This patch
addresses this issue by adding the string to the respective array.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

wl12xx: fix module author's email address in the spi and sdio modules

The MODULE_AUTHOR() macro in the main module (wl12xx) has been updated
to reflect one of the author's new email address, but the wl12xx_spi
and wl12xx_sdio modules haven't been updated.  This patches updates
them.

Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

b43: allocate receive buffers big enough for max frame len + offset

Otherwise, skb_put inside of dma_rx can fail...

https://bugzilla.kernel.org/show_bug.cgi?id=32042

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: stable@kernel.org

pkt_sched: QFQ - quick fair queue scheduler

This is an implementation of the Quick Fair Queue scheduler developed
by Fabio Checconi. The same algorithm is already implemented in ipfw
in FreeBSD. Fabio had an earlier version developed on Linux, I just
cleaned it up.  Thanks to Eric Dumazet for testing this under load.

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

xen: netfront: assume all hw features are available until backend connection setup

We need to assume that all features will be available when registering the
netdev otherwise they are ommitted from the initial set of
dev->wanted_features. When we connect to the backed we reduce the set as
necessary due to the call to netdev_update_features() in xennet_connect().

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

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

netfilter: xt_conntrack: fix inverted conntrack direction test

--ctdir ORIGINAL matches REPLY packets, and vv:

userspace sets "invert_flags &= ~XT_CONNTRACK_DIRECTION" in ORIGINAL
case.

Thus: (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) ^
      !!(info->invert_flags & XT_CONNTRACK_DIRECTION))

yields "1 ^ 0", which is true -> returns false.

Reproducer:
iptables -I OUTPUT 1 -p tcp --syn -m conntrack --ctdir ORIGINAL

Signed-off-by: Florian Westphal <fwestphal@astaro.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: xt_addrtype: replace rt6_lookup with nf_afinfo->route

This avoids pulling in the ipv6 module when using (ipv4-only) iptables
-m addrtype.

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: af_info: add 'strict' parameter to limit lookup to .oif

ipv6 fib lookup can set RT6_LOOKUP_F_IFACE flag to restrict search
to an interface, but this flag cannot be set via struct flowi.

Also, it cannot be set via ip6_route_output: this function uses the
passed sock struct to determine if this flag is required
(by testing for nonzero sk_bound_dev_if).

Work around this by passing in an artificial struct sk in case
'strict' argument is true.

This is required to replace the rt6_lookup call in xt_addrtype.c with
nf_afinfo->route().

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: af_info: add network namespace parameter to route hook

This is required to eventually replace the rt6_lookup call in
xt_addrtype.c with nf_afinfo->route().

Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Patrick McHardy <kaber@trash.net>

IPVS: fix NULL ptr dereference in ip_vs_ctl.c ip_vs_genl_dump_daemons()

ipvsadm -ln --daemon will trigger a Null pointer exception because
ip_vs_genl_dump_daemons() uses skb_net() instead of skb_sknet().

To prevent others from NULL ptr a check is made in ip_vs.h skb_net().

Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: h323: bug in parsing of ASN1 SEQOF field

Static analyzer of clang found a dead store which appears to be a bug in
reading count of items in SEQOF field, only the lower byte of word is
stored. This may lead to corrupted read and communication shutdown.

The bug has been in the module since it's first inclusion into linux
kernel.

[Patrick: the bug is real, but without practical consequence since the
 largest amount of sequence-of members we parse is 30.]

Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: ipset: references are protected by rwlock instead of mutex

The timeout variant of the list:set type must reference the member sets.
However, its garbage collector runs at timer interrupt so the mutex
protection of the references is a no go. Therefore the reference protection
is converted to rwlock.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>

netfilter: ipset: list:set timeout variant fixes

- the timeout value was actually not set
- the garbage collector was broken

The variant is fixed, the tests to the ipset testsuite are added.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Patrick McHardy <kaber@trash.net>