tcp_adv_win_scale default value is 2, meaning we expect a good citizen
skb to have skb->len / skb->truesize ratio of 75% (3/4)
In 2.6 kernels we (mis)accounted for typical MSS=1460 frame :
1536 + 64 + 256 = 1856 'estimated truesize', and 1856 * 3/4 = 1392.
So these skbs were considered as not bloated.
With recent truesize fixes, a typical MSS=1460 frame truesize is now the
more precise :
2048 + 256 = 2304. But 2304 * 3/4 = 1728.
So these skb are not good citizen anymore, because 1460 < 1728
(GRO can escape this problem because it build skbs with a too low
truesize.)
This also means tcp advertises a too optimistic window for a given
allocated rcvspace : When receiving frames, sk_rmem_alloc can hit
sk_rcvbuf limit and we call tcp_prune_queue()/tcp_collapse() too often,
especially when application is slow to drain its receive queue or in
case of losses (netperf is fast, scp is slow). This is a major latency
source.
We should adjust the len/truesize ratio to 50% instead of 75%
This patch :
1) changes tcp_adv_win_scale default to 1 instead of 2
2) increase tcp_rmem[2] limit from 4MB to 6MB to take into account
better truesize tracking and to allow autotuning tcp receive window to
reach same value than before. Note that same amount of kernel memory is
consumed compared to 2.6 kernels.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
l2tp_ip_sendmsg could return without releasing socket lock, making it all the
way to userspace, and generating the following warning:
[ 130.891594] ================================================
[ 130.894569] [ BUG: lock held when returning to user space! ]
[ 130.897257] 3.4.0-rc5-next-20120501-sasha #104 Tainted: G W
[ 130.900336] ------------------------------------------------
[ 130.902996] trinity/8384 is leaving the kernel with locks still held!
[ 130.906106] 1 lock held by trinity/8384:
[ 130.907924] #0: (sk_lock-AF_INET){+.+.+.}, at: [<ffffffff82b9503f>] l2tp_ip_sendmsg+0x2f/0x550
Introduced by commit 2f16270 ("l2tp: Fix locking in l2tp_ip.c").
Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I just noticed after some recent updates, that the init path for the drop
monitor protocol has a minor error. drop monitor maintains a per cpu structure,
that gets initalized from a single cpu. Normally this is fine, as the protocol
isn't in use yet, but I recently made a change that causes a failed skb
allocation to reschedule itself . Given the current code, the implication is
that this workqueue reschedule will take place on the wrong cpu. If drop
monitor is used early during the boot process, its possible that two cpus will
access a single per-cpu structure in parallel, possibly leading to data
corruption.
This patch fixes the situation, by storing the cpu number that a given instance
of this per-cpu data should be accessed from. In the case of a need for a
reschedule, the cpu stored in the struct is assigned the rescheule, rather than
the currently executing cpu
Tested successfully by myself.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
If register_netdev returns failure, the dev->interrupt and
its transfer buffer should be released, so just fix it.
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The transfer buffer of dev->interrupt is allocated in .probe path,
but not freed in .disconnet path, so mark the interrupt URB as
URB_FREE_BUFFER to free the buffer when the URB is destroyed.
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Acked-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Creating a VLAN interface on top of ucc_geth adds 4 bytes
to the frame and the HW controller is not prepared to
TX a frame bigger than 1518 bytes which is 4 bytes too
small for a full VLAN frame. Add 16 bytes which will handle
the a simple VLAN and leaves 12 bytes for future expansion.
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
In a busy network we see ucc_geth is dropping RX pkgs every now
and then. Increase the RX queues HW descriptors from
16 to 32 to deal with this.
Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_checksum_help(skb) can return an error, we must free skb in this
case. qdisc_drop(skb, sch) can also be feeded with a NULL skb (if
skb_unshare() failed), so lets use this generic helper.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Bug: The VLAN bit of the MAC RX Status Word is unreliable in several older
supported chips. Sometimes the VLAN bit is not set for valid VLAN packets
and also sometimes the VLAN bit is set for non-VLAN packets that came after
a VLAN packet. This results in a receive length error when VLAN hardware
tagging is enabled.
Fix: Variation on original fix proposed by Mirko.
The VLAN information is decoded in the status loop, and can be
applied to the received SKB there. This eliminates the need for the
separate tag field in the interface data structure. The tag has to
be copied and cleared if packet is copied. This version checked out
with vlan and normal traffic.
Note: vlan_tx_tag_present should be renamed vlan_tag_present, but that
is outside scope of this.
Reported-by: Mirko Lindner <mlindner@marvell.com>
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a small packet is received, the driver copies it to a new skb to allow
reusing the full size Rx buffer. The copy was propogating the checksum offload
but not the receive hash information. The bug is impact was mostly harmless
and therefore not observed until reviewing this area of code.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver calls cxgb_vlan_mode() from init_one(). This calls into
synchronize_rx(), which locks all the q locks, but the q locks are not
initialized until cxgb_up() -> setup_sge_qsets(). So move the call to
cxgb_vlan_mode() into cxgb_up(), after the call to setup_sge_qsets().
We also move the body of these functions up higher to avoid having to
a forward declaration.
This was found because of the lockdep warning:
INFO: trying to register non-static key.
the code is fine but needs lockdep annotation.
turning off the locking correctness validator.
Pid: 323, comm: work_for_cpu Not tainted 3.4.0-rc5 #28
Call Trace:
[<ffffffff8106e767>] register_lock_class+0x108/0x2d0
[<ffffffff8106ff42>] __lock_acquire+0xd3/0xd06
[<ffffffff81070fd0>] lock_acquire+0xbf/0xfe
[<ffffffff813862a6>] _raw_spin_lock_irq+0x36/0x45
[<ffffffffa01e71aa>] cxgb_vlan_mode+0x96/0xcb [cxgb3]
[<ffffffffa01f90eb>] init_one+0x8c4/0x980 [cxgb3]
[<ffffffff811fcbf0>] local_pci_probe+0x3f/0x70
[<ffffffff81042206>] do_work_for_cpu+0x10/0x22
[<ffffffff810482de>] kthread+0xa1/0xa9
[<ffffffff8138e234>] kernel_thread_helper+0x4/0x10
Contrary to what lockdep says, the code is not fine: we are locking an
uninitialized spinlock.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the support to bounce buffer added, the skb is coming as nonlinear in the
case of non-DDPed data frames for FCoE, which is mostly ok as the FCoE stack
would take care of that. However, for target mode, we have to set the FC CRC
and FC EOF field to allow the protocol stack to not drop the frame for the last
data frame of that sequence. So fix this by linearizing the skb first before
doing skb_put().
Signed-off-by: Yi Zou <yi.zou@intel.com>
Tested-by: Marcus Dennis <marcusx.e.dennis@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The driver was freeing memory in shutdown instead of remove. As a result
we were leaking memory if IEEE DCB was enabled and we loaded/unloaded the
driver. This change moves the freeing of the memory into the remove
routine where it belongs.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Maybe it's a typo, but it cause that igbvf can't be initialized successfully.
Set perm_addr value using valid dev_addr, although which is equal to hw.mac.addr.
Signed-off-by: Samuel Liao <samuelliao@tencent.com>
Signed-off-by: Shan Wei <davidshan@tencent.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch sets the automatic speed and duplex detection bits
in MAC_CR to enable the mac to determine its speed automatically
from the phy.
Note this must be done BEFORE the receiver or transmitter is
enabled.
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
also explicitly set the phy to advertise 1000 speeds
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
smsc75xx phy interrupt acknowledge needs an mdio_write to clear
PHY_INT_SRC instead of just a read like in smsc95xx.
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
fix bug in phy_init loop that was ignoring BMCR reset bit, akin to smsc95xx's d946092000
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
smsc75xx needs MII_ACCESS_BUSY to be set to correctly trigger mdio I/O. Note smsc75xx is different from smsc95xx in this regard.
Signed-off-by: Stephane Fillod <fillods@users.sf.net>
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes the same issue as reported on smsc95xx, where the
usb device is connected with no ethernet cable plugged-in.
Without this patch sysfs reports the cable as present
flag@flag-desktop:~$ cat /sys/class/net/eth0/carrier
1
while it's not:
flag@flag-desktop:~$ sudo mii-tool eth0
eth0: no link
Signed-off-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
EAP frames for stations in an AP VLAN are sent on the main AP interface
to avoid race conditions wrt. moving stations.
For that to work properly, sta_info_get_bss must be used instead of
sta_info_get when sending EAP packets.
Previously this was only done for cooked monitor injected packets, so
this patch adds a check for tx->skb->protocol to the same place.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: stable@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When the cwnd reduction is done, ssthresh may be infinite
if TCP enters CWR via ECN or F-RTO. If cwnd is not undone, i.e.,
undo_marker is set, tcp_complete_cwr() falsely set cwnd to the
infinite ssthresh value. The correct operation is to keep cwnd
intact because it has been updated in ECN or F-RTO.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now the helper function from filter.c for negative offsets is exported,
it can be used it in the jit to handle negative offsets.
First modify the asm load helper functions to handle:
- know positive offsets
- know negative offsets
- any offset
then the compiler can be modified to explicitly use these helper
when appropriate.
This fixes the case of a negative X register and allows to lift
the restriction that bpf programs with negative offsets can't
be jited.
Tested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Jan Seiffert <kaffeemonster@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current checking always succeeded. We have to check the first
character of the string to check that it's empty, thus, skipping
the timeout path.
This fixes the use of the CT target without the timeout option.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Change order of init so netns init is ready
when register ioctl and netlink.
Ver2
Whitespace fixes and __init added.
Reported-by: "Ryan O'Hara" <rohara@redhat.com>
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
ip_vs_create_timeout_table() can return NULL
All functions protocol init_netns is affected of this patch.
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Avoid crash when registering shedulers after
the IPVS core initialization for netns fails. Do this by
checking for present core (net->ipvs).
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Eric Dumazet pointed out to me that the drop_monitor protocol has some holes in
its smp protections. Specifically, its possible to replace data->skb while its
being written. This patch corrects that by making data->skb an rcu protected
variable. That will prevent it from being overwritten while a tracepoint is
modifying it.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: David Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet pointed out this warning in the drop_monitor protocol to me:
[ 38.352571] BUG: sleeping function called from invalid context at kernel/mutex.c:85
[ 38.352576] in_atomic(): 1, irqs_disabled(): 0, pid: 4415, name: dropwatch
[ 38.352580] Pid: 4415, comm: dropwatch Not tainted 3.4.0-rc2+ #71
[ 38.352582] Call Trace:
[ 38.352592] [<ffffffff8153aaf0>] ? trace_napi_poll_hit+0xd0/0xd0
[ 38.352599] [<ffffffff81063f2a>] __might_sleep+0xca/0xf0
[ 38.352606] [<ffffffff81655b16>] mutex_lock+0x26/0x50
[ 38.352610] [<ffffffff8153aaf0>] ? trace_napi_poll_hit+0xd0/0xd0
[ 38.352616] [<ffffffff810b72d9>] tracepoint_probe_register+0x29/0x90
[ 38.352621] [<ffffffff8153a585>] set_all_monitor_traces+0x105/0x170
[ 38.352625] [<ffffffff8153a8ca>] net_dm_cmd_trace+0x2a/0x40
[ 38.352630] [<ffffffff8154a81a>] genl_rcv_msg+0x21a/0x2b0
[ 38.352636] [<ffffffff810f8029>] ? zone_statistics+0x99/0xc0
[ 38.352640] [<ffffffff8154a600>] ? genl_rcv+0x30/0x30
[ 38.352645] [<ffffffff8154a059>] netlink_rcv_skb+0xa9/0xd0
[ 38.352649] [<ffffffff8154a5f0>] genl_rcv+0x20/0x30
[ 38.352653] [<ffffffff81549a7e>] netlink_unicast+0x1ae/0x1f0
[ 38.352658] [<ffffffff81549d76>] netlink_sendmsg+0x2b6/0x310
[ 38.352663] [<ffffffff8150824f>] sock_sendmsg+0x10f/0x130
[ 38.352668] [<ffffffff8150abe0>] ? move_addr_to_kernel+0x60/0xb0
[ 38.352673] [<ffffffff81515f04>] ? verify_iovec+0x64/0xe0
[ 38.352677] [<ffffffff81509c46>] __sys_sendmsg+0x386/0x390
[ 38.352682] [<ffffffff810ffaf9>] ? handle_mm_fault+0x139/0x210
[ 38.352687] [<ffffffff8165b5bc>] ? do_page_fault+0x1ec/0x4f0
[ 38.352693] [<ffffffff8106ba4d>] ? set_next_entity+0x9d/0xb0
[ 38.352699] [<ffffffff81310b49>] ? tty_ldisc_deref+0x9/0x10
[ 38.352703] [<ffffffff8106d363>] ? pick_next_task_fair+0x63/0x140
[ 38.352708] [<ffffffff8150b8d4>] sys_sendmsg+0x44/0x80
[ 38.352713] [<ffffffff8165f8e2>] system_call_fastpath+0x16/0x1b
It stems from holding a spinlock (trace_state_lock) while attempting to register
or unregister tracepoint hooks, making in_atomic() true in this context, leading
to the warning when the tracepoint calls might_sleep() while its taking a mutex.
Since we only use the trace_state_lock to prevent trace protocol state races, as
well as hardware stat list updates on an rcu write side, we can just convert the
spinlock to a mutex to avoid this problem.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: David Miller <davem@davemloft.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This driver currently leaves elp_work behind when stopping, which
occasionally results in data corruption because work function ends
up accessing freed memory, typical symptoms of this are various
worker_thread crashes. Fix it by cancelling elp_work.
Cc: stable@vger.kernel.org # v2.6.37+
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Currently SDIO glue frees it's own structure before calling
wl1251_free_hw(), which in turn calls ieee80211_unregister_hw().
The later call may result in a need to communicate with the chip
to stop it (as it happens now if the interface is still up before
rmmod), which means calls are made back to the glue, resulting in
freed memory access.
Fix this by freeing glue data last.
Cc: stable@vger.kernel.org # v2.6.37+
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Commit 2a19032 (b43: reload phy and bss settings after core restarts)
introduced an unconditional call to b43_op_config() at the end of
b43_op_start(). When firmware fails to load this can wedge the system.
There's no need to reload the configuration after a failed
initialization anyway, so only make the call if initialization was
successful.
BugLink: http://bugs.launchpad.net/bugs/950295
Cc: Felix Fietkau <nbd@openwrt.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
"iwlwifi: use correct released ucode version" change
the ucode api ok from 6000G2 to 6000G2B, but it shall belong
to 6030 device series, not the 6005 device series. Fix it
Cc: stable@vger.kernel.org #3.3+
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When our driver device is removed on the AHB bus, our IO memory is never unmapped.
Signed-off-by: Jonathan Bither <jonbither@gmail.com>
Reviewed-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
dpc takes care of all data packets transmissions for sdio function
2. It is possible that it misses some completion events when the
traffic is heavy or it's running on a slow cpu. A linked list is
introduced to make sure dpc is invoked whenever needed.
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
SDIO stack doesn't have a structure for function 0. The structure
pointer stored in card->sdio_func[0] is actually for function 1.
With current implementation the register read/write is applied to
function 1. This pathch fixes the issue.
Reviewed-by: Pieter-Paul Giesberts <pieterpg@broadcom.com>
Reviewed-by: Arend van Spriel <arend@broadcom.com>
Reviewed-by: Florian Fainelli <florian@openwrt.org>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Franky Lin <frankyl@broadcom.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Clean up a reference to jiffies in tcp_rcv_rtt_measure() that should
instead reference tcp_time_stamp. Since the result of the subtraction
is passed into a function taking u32, this should not change any
behavior (and indeed the generated assembly does not change on
x86_64). However, it seems worth cleaning this up for consistency and
clarity (and perhaps to avoid bugs if this is copied and pasted
somewhere else).
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Without this patch sysfs reports the cable as present
flag@flag-desktop:~$ cat /sys/class/net/eth0/carrier
1
while it's not:
flag@flag-desktop:~$ sudo mii-tool eth0
eth0: no link
Tested on my Beagle XM.
v2: added mantainer to the list of recipient
Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
Acked-by: Steve Glendinning <steve.glendinning@shawell.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The dl2k driver's rio_ioctl call has a few issues:
- No permissions checking
- Implements SIOCGMIIREG and SIOCGMIIREG using the SIOCDEVPRIVATE numbers
- Has a few ioctls that may have been used for debugging at one point
but have no place in the kernel proper.
This patch removes all but the MII ioctls, renumbers them to use the
standard ones, and adds the proper permission check for SIOCSMIIREG.
We can also get rid of the dl2k-specific struct mii_data in favor of
the generic struct mii_ioctl_data.
Since we have the phyid on hand, we can add the SIOCGMIIPHY ioctl too.
Most of the MII code for the driver could probably be converted to use
the generic MII library but I don't have a device to test the results.
Reported-by: Stephan Mueller <stephan.mueller@atsec.com>
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add missing spin_lock_init() for frames list lock.
Signed-off-by: Alexander Smirnov <alex.bluesman.smirnov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Clean all the pending fragments and relative timers if 6lowpan link
is going to be deleted.
Signed-off-by: Alexander Smirnov <alex.bluesman.smirnov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add nescesary mlme callbacks to satisfy "iz list" request from user space.
Due to 6lowpan device doesn't have its own phy, mlme implemented as a pipe
to a real phy to which 6lowpan is attached.
Signed-off-by: Alexander Smirnov <alex.bluesman.smirnov@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
