This patch adds missing memset which are required to initialize
flow key member. For example for IP flow we need to initialize
ip.frag for all cases.

Found by inspection.

This bug is introduced by commit 07148121
("openvswitch: Eliminate memset() from flow_extract").
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

replace asm/uaccess.h by linux/uaccess.h
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit b4d2394d ("dsa: Replace mii_bus with a generic host device")
replaces mii_bus with a generic host_dev, and introduces
dsa_host_dev_to_mii_bus() to support conversion from host_dev to mii_bus.
However, in some cases it uses to_mii_bus to perform that conversion.
Since host_dev is not the phy bus device but typically a platform device,
this fails and results in a crash with the affected drivers.

BUG: unable to handle kernel NULL pointer dereference at           (null)
IP: [<ffffffff81781d35>] __mutex_lock_slowpath+0x75/0x100
PGD 406783067 PUD 406784067 PMD 0
Oops: 0002 [#1] SMP
...
Call Trace:
[<ffffffff810a538b>] ? pick_next_task_fair+0x61b/0x880
[<ffffffff81781de3>] mutex_lock+0x23/0x37
[<ffffffff81533244>] mdiobus_read+0x34/0x60
[<ffffffff8153b95a>] __mv88e6xxx_reg_read+0x8a/0xa0
[<ffffffff8153b9bc>] mv88e6xxx_reg_read+0x4c/0xa0

Fixes: b4d2394d ("dsa: Replace mii_bus with a generic host device")
Cc: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In commit ec8a2e56 ("tipc: same receive
code path for connection protocol and data messages") we omitted the
the possiblilty that an arriving message extracted from a bundle buffer
may be a multicast message. Such messages need to be to be delivered to
the socket via a separate function, tipc_sk_mcast_rcv(). As a result,
small multicast messages arriving as members of a bundle buffer will be
silently dropped.

This commit corrects the error by considering this case in the function
tipc_link_bundle_rcv().
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 971f10ec ("tcp: better TCP_SKB_CB layout to reduce cache line
misses") added a regression for SO_BINDTODEVICE on IPv6.

This is because we still use inet6_iif() which expects that IP6 control
block is still at the beginning of skb->cb[]

This patch adds tcp_v6_iif() helper and uses it where necessary.

Because __inet6_lookup_skb() is used by TCP and DCCP, we add an iif
parameter to it.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Fixes: 971f10ec ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
Acked-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Don't ring the doorbell, and don't do PIO.  This will also prevent
 TX Push, because there will be more than one buffer waiting when
 the doorbell is rung.
Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Remove calling cancel_delayed_work_sync() for runtime suspend,
because it would cause dead lock. Instead, return -EBUSY to
avoid the device enters suspending if the net is running and
the delayed work is pending or running. The delayed work would
try to wake up the device later, so the suspending is not
necessary.
Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pskb_may_pull() maybe change skb->data and make uh pointer oboslete,
so reload uh and guehdr

Fixes: 37dd0247 ("gue: Receive side for Generic UDP Encapsulation")
Cc: Tom Herbert <therbert@google.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pskb_may_pull() maybe change skb->data and make eth pointer oboslete,
so set eth after pskb_may_pull()

Fixes:3d7b46cd("ip_tunnel: push generic protocol handling to ip_tunnel module")
Cc: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In case that the IP header has optional field at the end, this patch will
get the port numbers after that field, and compute the hash. The general
parser skb_flow_dissect() is used here.
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

If megaflows are disabled, the userspace does not send the netlink attribute
OVS_FLOW_ATTR_MASK, and the kernel must create an exact match mask.

sw_flow_mask_set() sets every bytes (in 'range') of the mask to 0xff, even the
bytes that represent padding for struct sw_flow, or the bytes that represent
fields that may not be set during ovs_flow_extract().
This is a problem, because when we extract a flow from a packet,
we do not memset() anymore the struct sw_flow to 0.

This commit gets rid of sw_flow_mask_set() and introduces mask_set_nlattr(),
which operates on the netlink attributes rather than on the mask key. Using
this approach we are sure that only the bytes that the user provided in the
flow are matched.

Also, if the parse_flow_mask_nlattrs() for the mask ENCAP attribute fails, we
now return with an error.

This bug is introduced by commit 07148121
("openvswitch: Eliminate memset() from flow_extract").
Reported-by: Alex Wang <alexw@nicira.com>
Signed-off-by: Daniele Di Proietto <ddiproietto@vmware.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pskb_may_pull maybe change skb->data and make eth pointer oboslete,
so eth needs to reload

Fixes: 91269e39 ("vxlan: using pskb_may_pull as early as possible")
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pskb_may_pull() called by arphdr_ok can change skb->data, so put the arp
setting after arphdr_ok to avoid the use the freed memory

Fixes: 07148121 ("openvswitch: Eliminate memset() from flow_extract.")
Cc: Jesse Gross <jesse@nicira.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Acked-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ip_setup_cork() called inside ip_append_data() steals dst entry from rt to cork
and in case errors in __ip_append_data() nobody frees stolen dst entry

Fixes: 2e77d89b ("net: avoid a pair of dst_hold()/dst_release() in ip_append_data()")
Signed-off-by: Vasily Averin <vvs@parallels.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We can retrieve opt from skb, no need to pass it as a parameter.
And opt should always be non-NULL, no need to check.

Cc: Krzysztof Kolasa <kkolasa@winsoft.pl>
Cc: Eric Dumazet <edumazet@google.com>
Tested-by: Krzysztof Kolasa <kkolasa@winsoft.pl>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cookie_v4_check() allocates ip_options_rcu in the same way
with tcp_v4_save_options(), we can just make it a helper function.

Cc: Krzysztof Kolasa <kkolasa@winsoft.pl>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 971f10ec ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
missed that cookie_v4_check() still calls ip_options_echo() which uses
IPCB(). It should use TCPCB() at TCP layer, so call __ip_options_echo()
instead.

Fixes: commit 971f10ec ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
Cc: Krzysztof Kolasa <kkolasa@winsoft.pl>
Cc: Eric Dumazet <edumazet@google.com>
Reported-by: Krzysztof Kolasa <kkolasa@winsoft.pl>
Tested-by: Krzysztof Kolasa <kkolasa@winsoft.pl>
Signed-off-by: Cong Wang <cwang@twopensource.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This simplifies the lanai.c driver by using
the module_pci_driver() macro, at the expense
of losing only debugging messages.
Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Avoid confusion between pid and portid.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2014-10-16

This series contains updates to fm10k and ixgbe.

Matthew provides two fixes for fm10k, first sets the flag to fetch the
host state before kicking off the service task that reads the host
state when bringing the interface up.  The second makes sure that we
release the mailbox lock after detecting an error and before we return
the error code.

Andy Zhou provides a compile fix for fm10k, when the driver is compiled
into the kernel and the VXLAN driver is compiled as a module.

Emil provides a fix for ixgbe to prevent against a panic by trying
to dereference a NULL pointer in ixgbe_ndo_set_vf_spoofchk().
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The check for vfinfo is not sufficient because it does not protect
against specifying vf that is outside of sriov_num_vfs range.
All of the ndo functions have a check for it except for
ixgbevf_ndo_set_spoofcheck().

The following patch is all we need to protect against this panic:

ip link set p96p1 vf 0 spoofchk off
BUG: unable to handle kernel NULL pointer dereference at 0000000000000052
IP: [<ffffffffa044a1c1>]
ixgbe_ndo_set_vf_spoofchk+0x51/0x150 [ixgbe]
Reported-by: Thierry Herbelot <thierry.herbelot@6wind.com>
Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Acked-by: Thierry Herbelot <thierry.herbelot@6wind.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Compiling with CONFIG_FM10K=y and VXLAN=m resulting in linking error:

   drivers/built-in.o: In function `fm10k_open':
   (.text+0x1f9d7a): undefined reference to `vxlan_get_rx_port'
   make: *** [vmlinux] Error 1

The fix follows the same strategy as I40E.
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

After grabbing the mailbox lock and detecting an error, the lock must be
released before the error code can be returned.
Signed-off-by: Matthew Vick <matthew.vick@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Set the flag to fetch the host state before kicking off the service task
that reads the host state when bringing the interface back up.
Signed-off-by: Matthew Vick <matthew.vick@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

pskb_may_pull should be used to check if skb->data has enough space,
skb->len can not ensure that.

Cc: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

when netif_rx() is done, the netif_rx handled skb maybe be freed,
and should not be used.
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

All functions used struct vport *vport except
ovs_vport_find_upcall_portid.

This fixes 1 kerneldoc warning
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

s/sock/gs
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For each Rx frame the eTSEC writes its FCS (Frame Check Sequence)
to the Rx buffer.

The eTSEC h/w manual states in the "Receive Buffer Descriptor Field
Descriptions" table:
"Data length is the number of octets written by the eTSEC into this BD's
data buffer if L is cleared (the value is equal to MRBLR), or, if L is
set, the length of the frame including *CRC*, FCB (if RCTRL[PRSDEP > 00),
preamble (if MACCFG2[PreAmRxEn]=1), time stamp (if RCTRL[TS] = 1) and
any padding (RCTRL[PAL])."

Though the FCS bytes are removed by the driver before passing the skb
to the net stack, the Rx buffer size computation does not currently
take into account the FCS bytes (4 bytes).
Because the Rx buffer size is multiple of 512 bytes, leaving out the
FCS is not a problem for the default MTU of 1500, as the Rx buffer size
is 1536 in this case.  However, for custom MTUs, where the difference
between the MTU size and the Rx buffer size is less, this can be a
problem as the computed Rx buffer size won't be enough to accomodate
the FCS for a received frame that is big enough (close to MTU size).
In such case the received frame is considered to be incomplete (L flag
not set in the RxBD status) and silently dropped.

Note that the driver does not currently support S/G on Rx, so it has to
compute its Rx buffer size based on the MTU of the device.
Reported-by: Kristian Otnes <kotnes@cisco.com>
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

commit 0b725a2c
    net: Remove ndo_xmit_flush netdev operation, use signalling instead.

added code that looks at skb->xmit_more after the skb has
been put in TX VQ. Since some paths process the ring and free the skb
immediately, this can cause use after free.

Fix by storing xmit_more in a local variable.

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

iMX6SX IEEE 1588 module has one hw issue in capturing the ATVR register.
The current SW flow is:
		ENET0->ATCR |= ENET_ATCR_CAPTURE_MASK;
		ts_counter_ns = ENET0->ATVR;
The ATVR value is not expected value that cause LinuxPTP stack cannot be convergent.

ENET Block Guide/ Chapter for the iMX6SX (PELE) address the issue:
After set ENET_ATCR[Capture], there need some time cycles before the counter
value is capture in the register clock domain. The wait-time-cycles is at least
6 clock cycles of the slower clock between the register clock and the 1588 clock.
So need something like:
		ENET0->ATCR |= ENET_ATCR_CAPTURE_MASK;
		wait();
		ts_counter_ns = ENET0->ATVR;

For iMX6SX, the 1588 ts_clk is fixed to 25Mhz, register clock is 66Mhz, so the
wait-time-cycles must be greater than 240ns (40ns * 6). The patch add 1us delay
before cpu read ATVR register.

Changes V2:
Modify the commit/comments log to describe the issue clearly.
Signed-off-by: Fugang Duan <B38611@freescale.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Identified by kbuild test robot. csk family is always set to be AF_INET or
AF_INET6, so skb will always be initialized to some value but there is no harm
in silencing the warning anyways.
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Fixes : f42bb57c ('cxgb4i : Fix -Wunused-function warning')
Signed-off-by: David S. Miller <davem@davemloft.net>

Add ndo_gso_check which a device can define to indicate whether is
is capable of doing GSO on a packet. This funciton would be called from
the stack to determine whether software GSO is needed to be done. A
driver should populate this function if it advertises GSO types for
which there are combinations that it wouldn't be able to handle. For
instance a device that performs UDP tunneling might only implement
support for transparent Ethernet bridging type of inner packets
or might have limitations on lengths of inner headers.
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

this patch is to fix the stmmac data compatibilities for
all the SoCs inside the platform file.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Anish Bhatt says:

====================
ipv6 and related cleanup for cxgb4/cxgb4i

This patch set removes some duplicated/extraneous code from cxgb4i, guards
cxgb4 against compilation failure based on ipv6 tristate, make ipv6 related
code no longer be enabled by default irrespective of ipv6 tristate and fixes
a refcnt issue.
-Anish

v2 : Provide more detailed commit messages, make subject more concise as
recommended by Dave Miller.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

There is an extra call to dst_neigh_lookup() leftover in cxgb4i that can cause
an unreleased refcnt issue. Remove extraneous call.
Signed-off-by: Anish Bhatt <anish@chelsio.com>

Fixes : 759a0cc5 ('cxgb4i: Add ipv6 code to driver, call into libcxgbi ipv6 api')
Signed-off-by: David S. Miller <davem@davemloft.net>

A bunch of ipv6 related code is left on by default. While this causes no
compilation issues, there is no need to have this enabled by default. Guard
with an ipv6 check, which also takes care of a -Wunused-function warning.
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4 ipv6 does not guard against ipv6 being disabled, or the standard
ipv6 module vs inbuilt tri-state issue. This was fixed for cxgb4i & iw_cxgb4
but missed for cxgb4.
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

cxgb4 already handles CLIP updates from a previous changeset for iw_cxgb4,
there is no need to have this functionality in cxgb4i. Remove duplicated code
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

TCP Small queues tries to keep number of packets in qdisc
as small as possible, and depends on a tasklet to feed following
packets at TX completion time.
Choice of tasklet was driven by latencies requirements.

Then, TCP stack tries to avoid reorders, by locking flows with
outstanding packets in qdisc in a given TX queue.

What can happen is that many flows get attracted by a low performing
TX queue, and cpu servicing TX completion has to feed packets for all of
them, making this cpu 100% busy in softirq mode.

This became particularly visible with latest skb->xmit_more support

Strategy adopted in this patch is to detect when tcp_wfree() is called
from ksoftirqd and let the outstanding queue for this flow being drained
before feeding additional packets, so that skb->ooo_okay can be set
to allow select_queue() to select the optimal queue :

Incoming ACKS are normally handled by different cpus, so this patch
gives more chance for these cpus to take over the burden of feeding
qdisc with future packets.

Tested:

lpaa23:~# ./super_netperf 1400 --google-pacing-rate 3028000 -H lpaa24 -l 3600 &

lpaa23:~# sar -n DEV 1 10 | grep eth1
06:16:18 AM      eth1 595448.00 1190564.00  38381.09 1760253.12      0.00      0.00      1.00
06:16:19 AM      eth1 594858.00 1189686e.00  38340.76 1758952.72      0.00      0.00      0.00
06:16:20 AM      eth1 597017.00 1194019.00  38480.79 1765370.29      0.00      0.00      1.00
06:16:21 AM      eth1 595450.00 1190936.00  38380.19 1760805.05      0.00      0.00      0.00
06:16:22 AM      eth1 596385.00 1193096.00  38442.56 1763976.29      0.00      0.00      1.00
06:16:23 AM      eth1 598155.00 1195978.00  38552.97 1768264.60      0.00      0.00      0.00
06:16:24 AM      eth1 594405.00 1188643.00  38312.57 1757414.89      0.00      0.00      1.00
06:16:25 AM      eth1 593366.00 1187154.00  38252.16 1755195.83      0.00      0.00      0.00
06:16:26 AM      eth1 593188.00 1186118.00  38232.88 1753682.57      0.00      0.00      1.00
06:16:27 AM      eth1 596301.00 1192241.00  38440.94 1762733.09      0.00      0.00      0.00
Average:         eth1 595457.30 1190843.50  38381.69 1760664.84      0.00      0.00      0.50
lpaa23:~# ./tc -s -d qd sh dev eth1 | grep backlog
 backlog 7606336b 2513p requeues 167982
 backlog 224072b 74p requeues 566
 backlog 581376b 192p requeues 5598
 backlog 181680b 60p requeues 1070
 backlog 5305056b 1753p requeues 110166    // Here, this TX queue is attracting flows
 backlog 157456b 52p requeues 1758
 backlog 672216b 222p requeues 3025
 backlog 60560b 20p requeues 24541
 backlog 448144b 148p requeues 21258

lpaa23:~# echo 1 >/proc/sys/net/ipv4/tcp_tsq_enable_tcp_wfree_ksoftirqd_detect

Immediate jump to full bandwidth, and traffic is properly
shard on all tx queues.

lpaa23:~# sar -n DEV 1 10 | grep eth1
06:16:46 AM      eth1 1397632.00 2795397.00  90081.87 4133031.26      0.00      0.00      1.00
06:16:47 AM      eth1 1396874.00 2793614.00  90032.99 4130385.46      0.00      0.00      0.00
06:16:48 AM      eth1 1395842.00 2791600.00  89966.46 4127409.67      0.00      0.00      1.00
06:16:49 AM      eth1 1395528.00 2791017.00  89946.17 4126551.24      0.00      0.00      0.00
06:16:50 AM      eth1 1397891.00 2795716.00  90098.74 4133497.39      0.00      0.00      1.00
06:16:51 AM      eth1 1394951.00 2789984.00  89908.96 4125022.51      0.00      0.00      0.00
06:16:52 AM      eth1 1394608.00 2789190.00  89886.90 4123851.36      0.00      0.00      1.00
06:16:53 AM      eth1 1395314.00 2790653.00  89934.33 4125983.09      0.00      0.00      0.00
06:16:54 AM      eth1 1396115.00 2792276.00  89984.25 4128411.21      0.00      0.00      1.00
06:16:55 AM      eth1 1396829.00 2793523.00  90030.19 4130250.28      0.00      0.00      0.00
Average:         eth1 1396158.40 2792297.00  89987.09 4128439.35      0.00      0.00      0.50

lpaa23:~# tc -s -d qd sh dev eth1 | grep backlog
 backlog 7900052b 2609p requeues 173287
 backlog 878120b 290p requeues 589
 backlog 1068884b 354p requeues 5621
 backlog 996212b 329p requeues 1088
 backlog 984100b 325p requeues 115316
 backlog 956848b 316p requeues 1781
 backlog 1080996b 357p requeues 3047
 backlog 975016b 322p requeues 24571
 backlog 990156b 327p requeues 21274

(All 8 TX queues get a fair share of the traffic)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>