David S. Miller [Wed, 3 Jun 2009 09:43:41 +0000 (02:43 -0700)]
Merge branch 'master' of /linux/kernel/git/davem/net-2.6
Conflicts:
drivers/net/forcedeth.c
Yevgeny Petrilin [Mon, 1 Jun 2009 23:29:42 +0000 (23:29 +0000)]
mlx4_en: Updated driver version
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:28:29 +0000 (23:28 +0000)]
mlx4_en: Added vlan_features support
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:24:07 +0000 (23:24 +0000)]
mlx4_en: multiqueue support
By default the driver opens 8 TX queues (defined by MLX4_EN_NUM_TX_RINGS).
If the driver is configured to support Per Priority Flow Control, we open
8 additional TX rings.
dev->real_num_tx_queues is always set to be MLX4_EN_NUM_TX_RINGS.
The mlx4_en_select_queue() function uses standard hashing (skb_tx_hash)
in case that PPFC is not supported or the skb contain a vlan tag,
otherwise the queue is selected according to vlan priority.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:23:13 +0000 (23:23 +0000)]
mlx4_en: Coalescing target is equal for all mtu's
The interrupt moderation should not depend on number of incoming
bytes, but on number of incoming packets.
The previous scheme caused very high interrupts rate for small
messages when big MTU was configured.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:22:18 +0000 (23:22 +0000)]
mlx4_en: Work with part of the ports.
If the initialization of one of the ports failed,
there is no need to fail the other one as well.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:21:20 +0000 (23:21 +0000)]
mlx4_en renamed en_params.c to en_ethtool.c
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 23:20:18 +0000 (23:20 +0000)]
mlx4_en: Moved all module parameters handling to en_main.c
en_params.c file now only handles Ethtool functionality
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 1 Jun 2009 20:27:13 +0000 (20:27 +0000)]
mlx4_en: Giving interface name in debug messages
For each debug message, the message will show interface name in case
that the net device was registered, and PCI bus ID with port number
if we were not registered yet. Messages that are not port/netdev specific
stayed in the old format
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
Minoru Usui [Tue, 2 Jun 2009 09:17:34 +0000 (02:17 -0700)]
net_cls: fix unconfigured struct tcf_proto keeps chaining and avoid kernel panic when we use cls_cgroup
This patch fixes a bug which unconfigured struct tcf_proto keeps
chaining in tc_ctl_tfilter(), and avoids kernel panic in
cls_cgroup_classify() when we use cls_cgroup.
When we execute 'tc filter add', tcf_proto is allocated, initialized
by classifier's init(), and chained. After it's chained,
tc_ctl_tfilter() calls classifier's change(). When classifier's
change() fails, tc_ctl_tfilter() does not free and keeps tcf_proto.
In addition, cls_cgroup is initialized in change() not in init(). It
accesses unconfigured struct tcf_proto which is chained before
change(), then hits Oops.
Signed-off-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: Jamal Hadi Salim <hadi@cyberus.ca>
Tested-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Neil Horman [Tue, 2 Jun 2009 08:29:58 +0000 (01:29 -0700)]
e1000: add missing length check to e1000 receive routine
Patch to fix bad length checking in e1000. E1000 by default does two
things:
1) Spans rx descriptors for packets that don't fit into 1 skb on recieve
2) Strips the crc from a frame by subtracting 4 bytes from the length prior to
doing an skb_put
Since the e1000 driver isn't written to support receiving packets that span
multiple rx buffers, it checks the End of Packet bit of every frame, and
discards it if its not set. This places us in a situation where, if we have a
spanning packet, the first part is discarded, but the second part is not (since
it is the end of packet, and it passes the EOP bit test). If the second part of
the frame is small (4 bytes or less), we subtract 4 from it to remove its crc,
underflow the length, and wind up in skb_over_panic, when we try to skb_put a
huge number of bytes into the skb. This amounts to a remote DOS attack through
careful selection of frame size in relation to interface MTU. The fix for this
is already in the e1000e driver, as well as the e1000 sourceforge driver, but no
one ever pushed it to e1000. This is lifted straight from e1000e, and prevents
small frames from causing the underflow described above
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Tested-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Nivedita Singhvi [Thu, 28 May 2009 07:00:46 +0000 (07:00 +0000)]
ipv4: New multicast-all socket option
After some discussion offline with Christoph Lameter and David Stevens
regarding multicast behaviour in Linux, I'm submitting a slightly
modified patch from the one Christoph submitted earlier.
This patch provides a new socket option IP_MULTICAST_ALL.
In this case, default behaviour is _unchanged_ from the current
Linux standard. The socket option is set by default to provide
original behaviour. Sockets wishing to receive data only from
multicast groups they join explicitly will need to clear this
socket option.
Signed-off-by: Nivedita Singhvi <niv@us.ibm.com>
Signed-off-by: Christoph Lameter<cl@linux.com>
Acked-by: David Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Tue, 2 Jun 2009 07:42:16 +0000 (00:42 -0700)]
net: ipv4/ip_sockglue.c cleanups
Pure cleanups
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Brian Haley [Tue, 2 Jun 2009 07:20:26 +0000 (00:20 -0700)]
IPv6: Print error value when skb allocation fails
Print-out the error value when sock_alloc_send_skb() fails in
the IPv6 neighbor discovery code - can be useful for debugging.
Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ed Swierk [Tue, 2 Jun 2009 07:19:52 +0000 (00:19 -0700)]
forcedeth: add phy_power_down parameter, leave phy powered up by default (v2)
Add a phy_power_down parameter to forcedeth: set to 1 to power down the
phy and disable the link when an interface goes down; set to 0 to always
leave the phy powered up.
The phy power state persists across reboots; Windows, some BIOSes, and
older versions of Linux don't bother to power up the phy again, forcing
users to remove all power to get the interface working (see
http://bugzilla.kernel.org/show_bug.cgi?id=13072). Leaving the phy
powered on is the safest default behavior. Users accustomed to seeing
the link state reflect the interface state and/or wanting to minimize
power consumption can set phy_power_down=1 if compatibility with other
OSes is not an issue.
Signed-off-by: Ed Swierk <eswierk@aristanetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rémi Denis-Courmont [Mon, 1 Jun 2009 01:18:57 +0000 (01:18 +0000)]
usb: gadget: f_phonet: fix memory allocation sizes
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: Felipe Balbi <felipe.balbi@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rémi Denis-Courmont [Mon, 1 Jun 2009 01:18:56 +0000 (01:18 +0000)]
f_phonet: no need to check for carrier - scheduler does it internally
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rémi Denis-Courmont [Mon, 1 Jun 2009 01:18:55 +0000 (01:18 +0000)]
f_phonet: dev_kfree_skb instead of dev_kfree_skb_any in TX callback
Network device TX is never run in IRQ context, and skb is freed outside
of the IRQ-disabling spin lock. So checking for IRQ was a waste of time
here.
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: Felipe Balbi <felipe.balbi@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rémi Denis-Courmont [Mon, 1 Jun 2009 00:35:16 +0000 (00:35 +0000)]
Phonet: fix accounting race between gprs_writeable() and gprs_xmit()
In the unlikely event that gprs_writeable() and gprs_xmit() check for
writeability at the same, we could stop the device queue forever.
Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 1 Jun 2009 21:32:08 +0000 (14:32 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/holtmann/bluetooth-2.6
Joe Perches [Sun, 31 May 2009 20:44:54 +0000 (20:44 +0000)]
atl1c_main.c: add wait_for_idle routine
Slight refactoring of duplicated wait for idle checks
Spelling fix
Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: jie.yang@atheros.com
Signed-off-by: David S. Miller <davem@davemloft.net>
Maciej W. Rozycki [Mon, 1 Jun 2009 10:12:04 +0000 (03:12 -0700)]
3c509: Add missing EISA IDs
Several EISA device IDs for 3c509 family network cards are missing from
the driver, making the cards unusable in their EISA mode. Here's a fix to
add them based on the EISA configuration files distributed by 3Com and our
eisa.ids database.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Brian Haley [Mon, 1 Jun 2009 10:07:33 +0000 (03:07 -0700)]
IPv6: Add 'autoconf' and 'disable_ipv6' module parameters
Add 'autoconf' and 'disable_ipv6' parameters to the IPv6 module.
The first controls if IPv6 addresses are autoconfigured from
prefixes received in Router Advertisements. The IPv6 loopback
(::1) and link-local addresses are still configured.
The second controls if IPv6 addresses are desired at all. No
IPv6 addresses will be added to any interfaces.
Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stefan Weil [Sun, 31 May 2009 10:59:15 +0000 (10:59 +0000)]
cpmac: prevent fatal exception in cpmac_end_xmit
We should not be stopping the subqueues in cpmac_end_xmit
but rather test the status of them. Replace the calls to
netif_subqueue_stop by __netif_subqueue_stopped. This
fixes an unrecoverable exception from happening when
running the driver.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Sun, 31 May 2009 10:57:07 +0000 (10:57 +0000)]
cpmac: remove unused variable.
This patch removes the unused variables in
cpmac_hw_error, cpmac_tx_timeout and cpmac_probe.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wolfgang Grandegger [Sat, 30 May 2009 07:55:50 +0000 (07:55 +0000)]
can: sja1000: generic OF platform bus driver
This patch adds a generic driver for SJA1000 chips on the OpenFirmware
platform bus found on embedded PowerPC systems. You need a SJA1000 node
definition in your flattened device tree source (DTS) file similar to:
can@3,100 {
compatible = "nxp,sja1000";
reg = <3 0x100 0x80>;
interrupts = <2 0>;
interrupt-parent = <&mpic>;
nxp,external-clock-frequency = <
16000000>;
};
See also Documentation/powerpc/dts-bindings/can/sja1000.txt.
CC: devicetree-discuss@ozlabs.org
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wolfgang Grandegger [Sat, 30 May 2009 07:55:49 +0000 (07:55 +0000)]
can: sja1000: stop misusing member base_addr of struct net_device
As discussed on the netdev mailing list, the member "base_addr" of
"struct net_device" should not be (mis)used to store the virtual
address to the SJA1000 register area. According to David Miller,
it's only use is to allow ISA and similar primitive bus devices to
have their I/O ports changed via ifconfig. The virtual address is
now stored in the private data structure of the SJA1000 device and
the callback functions use "struct sja1000_priv" instead of the
unneeded "struct net_device".
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Wolfgang Grandegger [Sat, 30 May 2009 07:55:48 +0000 (07:55 +0000)]
can: some fixes and cleanups to the initial device driver interface
This patch fixes a few errors sneaked into the initial version of the
device driver interface.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Mon, 1 Jun 2009 09:43:17 +0000 (02:43 -0700)]
MAINTAINERS: take maintainership of the cpmac Ethernet driver
This patch adds me as the maintainer of the CPMAC (AR7)
Ethernet driver.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Sun, 31 May 2009 06:04:46 +0000 (23:04 -0700)]
net: unset IFF_XMIT_DST_RELEASE for qeth and ipoib
Last two drivers that need skb->dst in their start_xmit() function
Tell dev_hard_start_xmit() to no release it by unsetting IFF_XMIT_DST_RELEASE
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
André Goddard Rosa [Sat, 30 May 2009 05:13:58 +0000 (22:13 -0700)]
usbnet: ratelimit warning messages invoked from callback handler
Warning messages coming from rtl8150 driver can flood the console
and make a DTV/set-top-box unable to decode video/audio frames.
'Pegasus' driver handles this situation similarly, preventing this
from happening there.
It happens with a low cost BCM MIPS embedded platform, whenever
timeout errors were coming from usbnet device, making platform
unusable for viewer watching.
Signed-off-by: André Goddard Rosa <andre.goddard@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fabio Estevam [Wed, 27 May 2009 03:41:39 +0000 (03:41 +0000)]
fec: Add FEC support for MX35 processor
Add FEC support for MX35 processor.
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tadashi Abe [Mon, 25 May 2009 20:53:27 +0000 (20:53 +0000)]
irda-usb: suspend/resume support
This patch adds power management suspend/resume hooks for irda-usb.
Signed-off-by: Tadashi Abe <tabe@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Finn Thain [Mon, 25 May 2009 16:52:47 +0000 (16:52 +0000)]
mac8390: use printk MAC address format
mac8390: use printk MAC address format
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko [Fri, 22 May 2009 23:22:17 +0000 (23:22 +0000)]
net: convert unicast addr list
This patch converts unicast address list to standard list_head using
previously introduced struct netdev_hw_addr. It also relaxes the
locking. Original spinlock (still used for multicast addresses) is not
needed and is no longer used for a protection of this list. All
reading and writing takes place under rtnl (with no changes).
I also removed a possibility to specify the length of the address
while adding or deleting unicast address. It's always dev->addr_len.
The convertion touched especially e1000 and ixgbe codes when the
change is not so trivial.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
drivers/net/bnx2.c | 13 +--
drivers/net/e1000/e1000_main.c | 24 +++--
drivers/net/ixgbe/ixgbe_common.c | 14 ++--
drivers/net/ixgbe/ixgbe_common.h | 4 +-
drivers/net/ixgbe/ixgbe_main.c | 6 +-
drivers/net/ixgbe/ixgbe_type.h | 4 +-
drivers/net/macvlan.c | 11 +-
drivers/net/mv643xx_eth.c | 11 +-
drivers/net/niu.c | 7 +-
drivers/net/virtio_net.c | 7 +-
drivers/s390/net/qeth_l2_main.c | 6 +-
drivers/scsi/fcoe/fcoe.c | 16 ++--
include/linux/netdevice.h | 18 ++--
net/8021q/vlan.c | 4 +-
net/8021q/vlan_dev.c | 10 +-
net/core/dev.c | 195 +++++++++++++++++++++++++++-----------
net/dsa/slave.c | 10 +-
net/packet/af_packet.c | 4 +-
18 files changed, 227 insertions(+), 137 deletions(-)
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko [Wed, 27 May 2009 05:42:36 +0000 (05:42 +0000)]
bonding: use bond_is_lb() when it's appropriate
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Matt Kraai [Sat, 30 May 2009 05:06:33 +0000 (22:06 -0700)]
net/firmare: Ignore .cis files
Signed-off-by: Matt Kraai <kraai@ftbfs.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Greg Kroah-Hartman [Sat, 30 May 2009 05:04:54 +0000 (22:04 -0700)]
ath1e: add new device id for asus hardware
Gary Lin reports that a new device id needs to be added to the atl1e in
order to get some new Asus hardware to work properly.
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yevgeny Petrilin [Mon, 25 May 2009 20:57:21 +0000 (20:57 +0000)]
mlx4_en: Fix a kernel panic when waking tx queue
When the transmit queue gets full we enable interrupts for TX completions
There was a race that we handled the TX queue both from the interrupt context
and from the transmit function. Using "spin_trylock_irq()" ensures this
doesn't happen.
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Sat, 30 May 2009 04:56:19 +0000 (21:56 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-2.6
Divy Le Ray [Fri, 29 May 2009 12:52:49 +0000 (12:52 +0000)]
cxgb3: add support for the Aquantia 10G-BT phy
Add support for the Aquantia AQ1002 10G-BaseT phy.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Divy Le Ray [Fri, 29 May 2009 12:52:44 +0000 (12:52 +0000)]
cxgb3: Add Aeluros 2020 phy support
Add support for the AEL2020 phy.
Add PCI IDs of the boards using this phy.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Divy Le Ray [Fri, 29 May 2009 12:52:38 +0000 (12:52 +0000)]
cxgb3: move away from LLTX
cxgb3 no longer advertizes LLTX.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Graf Yang [Fri, 29 May 2009 03:41:48 +0000 (03:41 +0000)]
netdev: bfin_mac: add Blackfin MII bus to platform bus to allow DSA access
When we register the MII bus to the platfrom bus, the Distributed Switch
Architecture can hook in transparently.
Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Michael Hennerich [Fri, 29 May 2009 03:41:15 +0000 (03:41 +0000)]
netdev: bfin_mac: fix performance issue found by netperf
- Remove dead long delay
- Use proper defines
- Remove broken implementation of the TX DMA Data Alignment TXDWA feature
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sonic Zhang [Fri, 29 May 2009 03:40:43 +0000 (03:40 +0000)]
netdev: bfin_mac: fix malformed UDP packet transmission when polling with KGDB
Writes to the DMA descriptors may sit in the internal Blackfin data buffers
and not actually be available when the DMA engine goes to fetch them. This
does not typically happen, but when dealing with short/fast packets such as
UDP and polling KGDB, this occurs much more frequently. Same goes for
heavy loads as seen by netperf tests or large scp transfers. So force the
buffers to drain with SSYNC otherwise we get random malformed packets.
Signed-off-by: Sonic Zhang <sonic.zhang@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Michael Hennerich [Fri, 29 May 2009 03:39:45 +0000 (03:39 +0000)]
netdev: bfin_mac: drop useless IRQF_SHARED from Blackfin EMAC interrupt
The IRQ used by the Blackfin EMAC is internal to the peripheral and cannot
be used to generate any other interrupt, so there is no point in marking it
as IRQF_SHARED.
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Fri, 29 May 2009 22:15:30 +0000 (15:15 -0700)]
veth: dont release skb->dst in veth_xmit()
No need to release skb->dst, its now done by core network.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Fri, 29 May 2009 22:02:29 +0000 (15:02 -0700)]
tcp: fix loop in ofo handling code and reduce its complexity
Somewhat luckily, I was looking into these parts with very fine
comb because I've made somewhat similar changes on the same
area (conflicts that arose weren't that lucky though). The loop
was very much overengineered recently in commit
915219441d566
(tcp: Use SKB queue and list helpers instead of doing it
by-hand), while it basically just wants to know if there are
skbs after 'skb'.
Also it got broken because skb1 = skb->next got translated into
skb1 = skb1->next (though abstracted) improperly. Note that
'skb1' is pointing to previous sk_buff than skb or NULL if at
head. Two things went wrong:
- We'll kfree 'skb' on the first iteration instead of the
skbuff following 'skb' (it would require required SACK reneging
to recover I think).
- The list head case where 'skb1' is NULL is checked too early
and the loop won't execute whereas it previously did.
Conclusion, mostly revert the recent changes which makes the
cset very messy looking but using proper accessor in the
previous-like version.
The effective changes against the original can be viewed with:
git-diff
915219441d566f1da0caa0e262be49b666159e17^ \
net/ipv4/tcp_input.c | sed -n -e '57,70 p'
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
John W. Linville [Wed, 27 May 2009 07:21:57 +0000 (09:21 +0200)]
rtl8187: add USB ID for Linksys WUSB54GC-EU v2 USB wifi dongle
http://bugzilla.kernel.org/show_bug.cgi?id=13383
Reported-by: Przemyslaw Kulczycki <azrael@autocom.pl>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
John W. Linville [Thu, 28 May 2009 09:39:02 +0000 (11:39 +0200)]
at76c50x-usb: avoid mutex deadlock in at76_dwork_hw_scan
http://bugzilla.kernel.org/show_bug.cgi?id=13312
at76_dwork_hw_scan holds a mutex while calling ieee80211_scan_completed,
which then calls at76_config which needs the same mutex. This reworks
the ordering to not hold the lock while calling ieee80211_scan_completed.
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Finn Thain [Thu, 28 May 2009 02:05:53 +0000 (02:05 +0000)]
mac8390: fix build with NET_POLL_CONTROLLER
Fix the build for CONFIG_NET_POLL_CONTROLLER that I broke with
217cbfa856dc1cbc2890781626c4032d9e3ec59f ("mac8390: fix regression
caused during net_device_ops conversion").
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Divy Le Ray [Thu, 28 May 2009 11:23:08 +0000 (11:23 +0000)]
cxgb3: link fault fixes
Do not call t3_link_fault() under spinlock, as it calls msleep().
Besides, only the access to pi->link_fault needs to be serialized.
Also initialize local variables before checking the link status,
link state fields might otherwise end up containing garbage.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Divy Le Ray [Thu, 28 May 2009 11:23:02 +0000 (11:23 +0000)]
cxgb3: fix dma mapping regression
Commit
5e68b772e6efd189d6aca76f6872fb75d51ace60
cxgb3: map entire Rx page, feed map+offset to Rx ring.
introduced a regression on platforms defining DECLARE_PCI_UNMAP_ADDR()
and related macros as no-ops.
Rx descriptors are fed with the a page buffer bus address + page chunk offset.
The page buffer bus address is set and retrieved through
pci_unamp_addr_set(), pci_unmap_addr().
These functions being meaningless on x86 (if CONFIG_DMA_API_DEBUG is not set).
The HW ends up with a bogus bus address.
This patch saves the page buffer bus address for all plaftorms.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko [Thu, 28 May 2009 01:05:00 +0000 (01:05 +0000)]
bonding: allow bond in mode balance-alb to work properly in bridge -try4.3
[PATCH net-next] bonding: allow bond in mode balance-alb to work properly in bridge -try4.3
(updated)
changes v4.2 -> v4.3
- memcpy the address always, not just in case it differs from master->dev_addr
- compare_ether_addr_64bits() is not used so there is no direct need to make new
header file (I think it would be good to have bond stuff in separate file
anyway).
changes v4.1 -> v4.2
- use skb->pkt_type == PACKET_HOST compare rather then comparing skb dest addr
against skb->dev->dev_addr
The problem is described in following bugzilla:
https://bugzilla.redhat.com/show_bug.cgi?id=487763
Basically here's what's going on. In every mode, bonding interface uses the same
mac address for all enslaved devices (except fail_over_mac). Only balance-alb
will simultaneously use multiple MAC addresses across different slaves. When you
put this kind of bond device into a bridge it will only add one of mac adresses
into a hash list of mac addresses, say X. This mac address is marked as local.
But this bonding interface also has mac address Y. Now then packet arrives with
destination address Y, this address is not marked as local and the packed looks
like it needs to be forwarded. This packet is then lost which is wrong.
Notice that interfaces can be added and removed from bond while it is in bridge.
***
When the multiple addresses for bridge port approach failed to solve this issue
due to STP I started to think other way to solve this. I returned to previous
solution but tweaked one.
This patch solves the situation in the bonding without touching bridge code.
For every incoming frame to bonding the destination address is compared to
current address of the slave device from which tha packet came. If these two
match destination address is replaced by mac address of the master. This address
is known by bridge so it is delivered properly. Note that the comparsion is not
made directly, it's used skb->pkt_type == PACKET_HOST instead. This is "set"
previously in eth_type_trans().
I experimentally tried that this works as good as searching through the slave
list (v4 of this patch).
Jirka
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Thu, 28 May 2009 00:58:41 +0000 (00:58 +0000)]
korina: periodically poll the media
This patch makes the korina driver poll the media
for link change. This is actually required on
Mikrotik RB532 (not RB532A) for korina to
operate properly.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
FUJITA Tomonori [Wed, 27 May 2009 15:10:42 +0000 (15:10 +0000)]
net: replace dma_sync_single with dma_sync_single_for_cpu
This replaces dma_sync_single() with dma_sync_single_for_cpu() because
dma_sync_single() is an obsolete API; include/linux/dma-mapping.h says:
/* Backwards compat, remove in 2.7.x */
#define dma_sync_single dma_sync_single_for_cpu
#define dma_sync_sg dma_sync_sg_for_cpu
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Thu, 28 May 2009 22:35:10 +0000 (22:35 +0000)]
net: unset IFF_XMIT_DST_RELEASE in ipgre_tunnel_setup()
ipgre_tunnel_xmit() might need skb->dst, so tell dev_hard_start_xmit()
to no release it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Thu, 28 May 2009 11:14:49 +0000 (11:14 +0000)]
atm: unset IFF_XMIT_DST_RELEASE in clip_setup()
clip_start_xmit() needs skb->dst so tell dev_hard_start_xmit()
to no release it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Thu, 28 May 2009 10:44:30 +0000 (10:44 +0000)]
net: unset IFF_XMIT_DST_RELEASE in ipip_tunnel_setup()
ipip_tunnel_xmit() might need skb->dst, so tell dev_hard_start_xmit()
to no release it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Thu, 28 May 2009 00:00:41 +0000 (00:00 +0000)]
net: dont update dev->trans_start
Second round of drivers for Gb cards (and NIU one I forgot in the 10GB round)
Now that core network takes care of trans_start updates, dont do it
in drivers themselves, if possible. Drivers can avoid one cache miss
(on dev->trans_start) in their start_xmit() handler.
Exceptions are NETIF_F_LLTX drivers
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 27 May 2009 19:26:37 +0000 (19:26 +0000)]
net: dont update dev->trans_start in 10GB drivers
Followup of commits
9d21493b4beb8f918ba248032fefa393074a5e2b
and
08baf561083bc27a953aa087dd8a664bb2b88e8e
(net: tx scalability works : trans_start)
(net: txq_trans_update() helper)
Now that core network takes care of trans_start updates, dont do it
in drivers themselves, if possible. Multi queue drivers can
avoid one cache miss (on dev->trans_start) in their start_xmit()
handler.
Exceptions are NETIF_F_LLTX drivers (vxge & tehuti)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 May 2009 08:41:32 +0000 (01:41 -0700)]
Merge branch 'linux-2.6.31.y' of git://git./linux/kernel/git/inaky/wimax
David S. Miller [Fri, 29 May 2009 06:31:56 +0000 (23:31 -0700)]
llc: Kill outdated and incorrect comment.
This comment suggested storing two pieces of state in the
LLC skb control block, and in fact we do. Someone did
the implementation but never killed this todo comment :-)
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 May 2009 06:26:33 +0000 (23:26 -0700)]
irda: Use SKB queue and list helpers instead of doing it by-hand.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Fri, 29 May 2009 04:35:47 +0000 (21:35 -0700)]
tcp: Use SKB queue and list helpers instead of doing it by-hand.
Signed-off-by: David S. Miller <davem@davemloft.net>
Inaky Perez-Gonzalez [Fri, 22 May 2009 07:41:11 +0000 (00:41 -0700)]
wimax: a new API call was added, increment minor protocol version number
As the 'state_get' API call was added, we need to increase the minor
protocol version number so applications that depend on the can check
it's presence.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Fri, 8 May 2009 22:51:44 +0000 (15:51 -0700)]
wimax/i2400m: usb: fix device reset on autosuspend while not yet idle
When the i2400m is connected to a network, the host interface (USB)
cannot be suspended. For that to happen, the device has to have
negotiated with the basestation to put the link on IDLE state.
If the host tries to put the device in standby while it is connected
but not idle, the device resets, as the driver should not do that.
To avoid triggering that, when the USB susbsytem requires the driver
to autosuspend the device, the driver checks if the device is not yet
idle. If it is not, the request is requested (will be retried again
later on after the autosuspend timeout). At some point the device will
enter idle and the request will succeed (unless of course, there is
network traffic, but at that point, there is no idle neither in the
link or the host interface).
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Paulius Zaleckas [Wed, 13 May 2009 13:20:29 +0000 (06:20 -0700)]
wimax: Add netlink interface to get device state
wimax connection manager / daemon has to know what is current
state of the device. Previously it was only possible to get
notification whet state has changed.
Note:
By mistake, the new generic netlink's number for
WIMAX_GNL_OP_STATE_GET was declared inserting into the existing list
of API calls, not appending; thus, it'd break existing API.
Fixed by Inaky Perez-Gonzalez <inaky@linux.intel.com> by moving to
the tail, where we add to the interface, not modify the interface.
Thanks to Stephen Hemminger <shemminger@vyatta.com> for catching this.
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@teltonika.lt>
Inaky Perez-Gonzalez [Thu, 30 Apr 2009 21:39:21 +0000 (14:39 -0700)]
wimax/i2400m: sdio: set the block size before enabling the function
From a fix by Cindy H Kao:
Block size has to be set before sending IOE enable because the
firmware reads the block size register before it reads IOE register.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Wed, 6 May 2009 04:05:16 +0000 (21:05 -0700)]
wimax: document why wimax_msg_*() operations can be used in any state
Funcion documentation for wimax_msg_alloc() and wimax_msg_send() needs
to clarify that they can be used in the very early stages of a
wimax_dev lifecycle.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Sat, 2 May 2009 09:50:03 +0000 (02:50 -0700)]
wimax/i2400m: remove redundant readiness checks from i2400m_report_tlv_*()
Functions i2400m_report_tlv*() are only called from
i2400m_report_hook(), called in a workqueue by
i2400m_report_hook_work(). The scheduler checks for device readiness
before scheduling.
Added an extra check for readiness in i2400m_report_hook_work(), which
makes all the checks down the line redundant.
Obviously the device state could change in the middle, but error
handling would take care of that.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Sat, 25 Apr 2009 07:26:14 +0000 (00:26 -0700)]
wimax/i2400m: factor out 'state report's TLV handling to a function
i2400m_report_state_hook() is going to get messier as we add handling
code.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Mon, 13 Apr 2009 17:26:34 +0000 (10:26 -0700)]
wimax/i2400m: allow kernel commands to device to be logged too
By running 'echo 1 > /sys/kernel/debug/wimax:wmxX/i2400m/trace_msg_from_user',
the driver will echo to user space all the commands being sent to the
device from user space, along with the responses.
However, this only helps with the commands being sent from user space;
with this patch, the trace hook is moved to i2400m_msg_to_dev(), which
is the single access point for running commands to the device (both by
user space and the kernel driver). This allows better debugging by
having a complete stream of commands/acks and reports.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Tue, 31 Mar 2009 00:51:54 +0000 (17:51 -0700)]
wimax/i2400m: trace commands sent from user space on the "echo" pipe
When commands are sent from user space, trace both the command sent
and the answer received over the "echo" pipe instead of over the
"trace" pipe when command tracing is enabled. As well, when the device
sends a reports/indications, send it over the "echo" pipe.
The "trace" pipe is used by the device to send firmware traces;
gets confusing. Another named pipe makes it easier to split debug
information.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Inaky Perez-Gonzalez [Wed, 22 Apr 2009 23:53:08 +0000 (16:53 -0700)]
wimax/i2400m: generate fake source MAC address with random_ether_addr()
The WiMAX i2400m driver needs to generate a fake source MAC address to
fake an ethernet header (for destination, the card's MAC is
used). This is the source of the packet, which is the basestation it
came from. The basestation's mac address is not usable for this, as it
uses its own namespace and it is not always available.
Currently the fake source MAC address was being set to all zeros,
which was causing trouble with bridging.
Use random_ether_addr() to generate a proper one that creates no
trouble.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
David S. Miller [Thu, 28 May 2009 23:46:29 +0000 (16:46 -0700)]
econet: Use SKB queue and list helpers instead of doing it by-hand.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 23:43:52 +0000 (16:43 -0700)]
decnet: Use SKB queue and list helpers instead of doing it by-hand.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 23:36:47 +0000 (16:36 -0700)]
atm: Use SKB queue and list helpers instead of doing it by-hand.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 08:22:57 +0000 (01:22 -0700)]
rtl8187: Remove pointless check in rtl8187_rx_cb().
First of all, it exposes the SKB list implementation.
Second of all it's not needed. If we get called here, we
successfully enqueued the URB with the linked SKB and
such a completion only gets called one time on such an
SKB.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 07:48:16 +0000 (00:48 -0700)]
p54: Use SKB list handling helpers instead of by-hand code.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 00:27:01 +0000 (17:27 -0700)]
wavelan: Remove bogus debugging on skb->next being non-NULL.
This eliminates explicit references to the SKB list handling
implementation.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Thu, 28 May 2009 00:09:44 +0000 (17:09 -0700)]
aoe: Remove superfluous clearing of skb fields in new_skb().
This code uses alloc_skb() which clears them out for us.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 27 May 2009 23:56:47 +0000 (16:56 -0700)]
appletalk: Add proper locking around IPDDP routing table.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Wed, 27 May 2009 22:51:25 +0000 (15:51 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/kaber/nf-2.6
Jiri Pirko [Wed, 27 May 2009 22:48:07 +0000 (15:48 -0700)]
net: correct a comment for the final #endif
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 27 May 2009 11:30:05 +0000 (11:30 +0000)]
net: net/core/sock.c cleanup
Pure style cleanup patch.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 27 May 2009 10:34:50 +0000 (10:34 +0000)]
net: use netdev stats in b44, sundance, via-rhine and via-velocity
Use struct net_device_stats provided in struct net_device instead of
private ones.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet [Wed, 27 May 2009 04:42:37 +0000 (04:42 +0000)]
net: ALIGN/PTR_ALIGN cleanup in alloc_netdev_mq()/netdev_priv()
Use ALIGN() and PTR_ALIGN() macros instead of handcoding them.
Get rid of NETDEV_ALIGN_CONST ugly define
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiri Pirko [Wed, 27 May 2009 00:55:08 +0000 (00:55 +0000)]
bridge: avoid an extra space in br_fdb_update()
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Florian Fainelli [Tue, 26 May 2009 23:38:10 +0000 (23:38 +0000)]
r6040: check and update media status
This patch removes the not-that-useful message in the
r6040_timer which prints the PHY status. Instead
replace it with a call to mii_check_media which will
update the link status and print it on startup.
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pablo Neira Ayuso [Wed, 27 May 2009 13:49:11 +0000 (15:49 +0200)]
netfilter: nfnetlink_log: fix wrong skbuff size calculation
This problem was introduced in
72961ecf84d67d6359a1b30f9b2a8427f13e1e71
since no space was reserved for the new attributes NFULA_HWTYPE,
NFULA_HWLEN and NFULA_HWHEADER.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Jesper Dangaard Brouer [Wed, 27 May 2009 13:45:34 +0000 (15:45 +0200)]
netfilter: xt_hashlimit does a wrong SEQ_SKIP
The function dl_seq_show() returns 1 (equal to SEQ_SKIP) in case
a seq_printf() call return -1. It should return -1.
This SEQ_SKIP behavior brakes processing the proc file e.g. via a
pipe or just through less.
Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Herbert Xu [Tue, 26 May 2009 18:50:34 +0000 (18:50 +0000)]
tcp: Do not check flush when comparing options for GRO
There is no need to repeatedly check flush when comparing TCP
options for GRO as it will be false 99% of the time where it
matters.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:33 +0000 (18:50 +0000)]
gro: Store shinfo in local variable in skb_gro_receive
This patch stores the two shinfo pointers in local variables
because they're used over and over again in skb_gro_receive.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:32 +0000 (18:50 +0000)]
gro: Nasty optimisations for page frags in skb_gro_receive
This patch reverses the direction of the frags array copy in
skb_gro_receive in order simplify the loop conditional. It
also avoids touching the first element of the original frags
array.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:31 +0000 (18:50 +0000)]
gro: Open-code final pskb_may_pull
As we know the only packets which need the final pskb_may_pull
are completely non-linear, and have all the required bits in
frag0, we can perform a straight memcpy instead of going through
pskb_may_pull and doing skb_copy_bits.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:29 +0000 (18:50 +0000)]
ipv4: Use 32-bit loads for ID and length in GRO
This patch optimises the IPv4 GRO code by using 32-bit loads
(instead of 16-bit ones) on the ID and length checks in the receive
function.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:28 +0000 (18:50 +0000)]
gro: Avoid unnecessary comparison after skb_gro_header
For the overwhelming majority of cases, skb_gro_header's return
value cannot be NULL. Yet we must check it because of its current
form. This patch splits it up into multiple functions in order
to avoid this.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Herbert Xu [Tue, 26 May 2009 18:50:27 +0000 (18:50 +0000)]
gro: Optimise length comparison in skb_gro_header
By caching frag0_len, we can avoid checking both frag0 and the
length separately in skb_gro_header. This helps as skb_gro_header
is called four times per packet which amounts to a few million
times at 10Gb/s.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>