The ATR code was assuming that it would be able to use tcp_hdr for
every TCP frame that came through.  However this isn't the case as it
is possible for a frame to arrive that is TCP but sent through something
like a raw socket.  As a result the driver was setting up bad filters in
which tcp_hdr was really pointing to the network header so the data was
all invalid.

In order to correct this I have added a bit of parsing logic that will
determine the TCP header location based off of the network header and
either the offset in the case of the IPv4 header, or a walk through the
IPv6 extension headers until it encounters the header that indicates
IPPROTO_TCP.  In addition I have added checks to verify that the lowest
protocol provided is recognized as IPv4 or IPv6 to help mitigate raw
sockets using ETH_P_ALL from having ATR applied to them.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

The VXLAN port number should be stored in network order instead of in host
order as it is accessed from the hot-path in ATR.  This way we can avoid
having to do any byte swaps in order to validate the port number.

I moved the vxlan_port value into a hole in the read-mostly region of the
adapter struct.  This way it should be in a warm cache-line instead of in
some isolated region in memory when it needs to be accessed.

In addition I went through and stripped a bunch of unneeded ifdef flags
since having an extra variable present doesn't really hurt anything and
makes the code easier to read.  I also went through and dropped the
NETIF_F_RXCSUM flag which was being set in hw_encap_features but provides
no value as the flag is not evaluated in the Rx path.
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

commit c9f53e63 ("ixgbe: Refactor MAC address configuration code")
introduced code that doesn't set HW register RAR0 to default mac address
but FF:FF:FF:FF:FF:FF. Due to this, ixgbe HW discards all incoming packets
that doesn't have destination mac address equals to FF:FF:FF:FF:FF:FF.

This commit sets RAR0 correctly to default HW mac address.
Signed-off-by: Tushar Dave <tushar.n.dave@oracle.com>
Tested-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Thomas reports:
"Windows:

00 diagnostics
01 modem
02 at-port
03 nmea
04 nic

Linux:

T:  Bus=02 Lev=01 Prnt=01 Port=03 Cnt=01 Dev#=  4 Spd=480 MxCh= 0
D:  Ver= 2.00 Cls=00(>ifc ) Sub=00 Prot=00 MxPS=64 #Cfgs=  1
P:  Vendor=2001 ProdID=7e19 Rev=02.32
S:  Manufacturer=Mobile Connect
S:  Product=Mobile Connect
S:  SerialNumber=0123456789ABCDEF
C:  #Ifs= 6 Cfg#= 1 Atr=a0 MxPwr=500mA
I:  If#= 0 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=option
I:  If#= 1 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#= 2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#= 3 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#= 4 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=ff Driver=qmi_wwan
I:  If#= 5 Alt= 0 #EPs= 2 Cls=08(stor.) Sub=06 Prot=50 Driver=usb-storage"
Reported-by: Thomas Schäfer <tschaefer@t-online.de>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for you net tree,
they are:

1) There was a race condition between parallel save/swap and delete,
   which resulted a kernel crash due to the increase ref for save, swap,
   wrong ref decrease operations. Reported and fixed by Vishwanath Pai.

2) OVS should call into CT NAT for packets of new expected connections only
   when the conntrack state is persisted with the 'commit' option to the
   OVS CT action. From Jarno Rajahalme.

3) Resolve kconfig dependencies with new OVS NAT support. From Arnd Bergmann.

4) Early validation of entry->target_offset to make sure it doesn't take us
   out from the blob, from Florian Westphal.

5) Again early validation of entry->next_offset to make sure it doesn't take
   out from the blob, also from Florian.

6) Check that entry->target_offset is always of of sizeof(struct xt_entry)
   for unconditional entries, when checking both from check_underflow()
   and when checking for loops in mark_source_chains(), again from
   Florian.

7) Fix inconsistent behaviour in nfnetlink_queue when
   NFQA_CFG_F_FAIL_OPEN is set and netlink_unicast() fails due to buffer
   overrun, we have to reinject the packet as the user expects.

8) Enforce nul-terminated table names from getsockopt GET_ENTRIES
   requests.

9) Don't assume skb->sk is set from nft_bridge_reject and synproxy,
   this fixes a recent update of the code to namespaceify
   ip_default_ttl, patch from Liping Zhang.

This batch comes with four patches to validate x_tables blobs coming
from userspace. CONFIG_USERNS exposes the x_tables interface to
unpriviledged users and to be honest this interface never received the
attention for this move away from the CAP_NET_ADMIN domain. Florian is
working on another round with more patches with more sanity checks, so
expect a bit more Netfilter fixes in this development cycle than usual.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit fa50d974 ("ipv4: Namespaceify ip_default_ttl sysctl knob")
use sock_net(skb->sk) to get the net namespace, but we can't assume
that sk_buff->sk is always exist, so when it is NULL, oops will happen.
Signed-off-by: Liping Zhang <liping.zhang@spreadtrum.com>
Reviewed-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Make sure the table names via getsockopt GET_ENTRIES is nul-terminated
in ebtables and all the x_tables variants and their respective compat
code. Uncovered by KASAN.
Reported-by: Baozeng Ding <sploving1@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

When netlink unicast fails to deliver the message to userspace, we
should also check if the NFQA_CFG_F_FAIL_OPEN flag is set so we reinject
the packet back to the stack.

I think the user expects no packet drops when this flag is set due to
queueing to userspace errors, no matter if related to the internal queue
or when sending the netlink message to userspace.

The userspace application will still get the ENOBUFS error via recvmsg()
so the user still knows that, with the current configuration that is in
place, the userspace application is not consuming the messages at the
pace that the kernel needs.
Reported-by: "Yigal Reiss (yreiss)" <yreiss@cisco.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Tested-by: "Yigal Reiss (yreiss)" <yreiss@cisco.com>

Ben Hawkes says:

 In the mark_source_chains function (net/ipv4/netfilter/ip_tables.c) it
 is possible for a user-supplied ipt_entry structure to have a large
 next_offset field. This field is not bounds checked prior to writing a
 counter value at the supplied offset.

Problem is that mark_source_chains should not have been called --
the rule doesn't have a next entry, so its supposed to return
an absolute verdict of either ACCEPT or DROP.

However, the function conditional() doesn't work as the name implies.
It only checks that the rule is using wildcard address matching.

However, an unconditional rule must also not be using any matches
(no -m args).

The underflow validator only checked the addresses, therefore
passing the 'unconditional absolute verdict' test, while
mark_source_chains also tested for presence of matches, and thus
proceeeded to the next (not-existent) rule.

Unify this so that all the callers have same idea of 'unconditional rule'.
Reported-by: Ben Hawkes <hawkes@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Otherwise this function may read data beyond the ruleset blob.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

We should check that e->target_offset is sane before
mark_source_chains gets called since it will fetch the target entry
for loop detection.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

The openvswitch code has gained support for calling into the
nf-nat-ipv4/ipv6 modules, however those can be loadable modules
in a configuration in which openvswitch is built-in, leading
to link errors:

net/built-in.o: In function `__ovs_ct_lookup':
:(.text+0x2cc2c8): undefined reference to `nf_nat_icmp_reply_translation'
:(.text+0x2cc66c): undefined reference to `nf_nat_icmpv6_reply_translation'

The dependency on (!NF_NAT || NF_NAT) prevents similar issues,
but NF_NAT is set to 'y' if any of the symbols selecting
it are built-in, but the link error happens when any of them
are modular.

A second issue is that even if CONFIG_NF_NAT_IPV6 is built-in,
CONFIG_NF_NAT_IPV4 might be completely disabled. This is unlikely
to be useful in practice, but the driver currently only handles
IPv6 being optional.

This patch improves the Kconfig dependency so that openvswitch
cannot be built-in if either of the two other symbols are set
to 'm', and it replaces the incorrect #ifdef in ovs_ct_nat_execute()
with two "if (IS_ENABLED())" checks that should catch all corner
cases also make the code more readable.

The same #ifdef exists ovs_ct_nat_to_attr(), where it does not
cause a link error, but for consistency I'm changing it the same
way.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 05752523 ("openvswitch: Interface with NAT.")
Acked-by: Joe Stringer <joe@ovn.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

OVS should call into CT NAT for packets of new expected connections only
when the conntrack state is persisted with the 'commit' option to the
OVS CT action.  The test for this condition is doubly wrong, as the CT
status field is ANDed with the bit number (IPS_EXPECTED_BIT) rather
than the mask (IPS_EXPECTED), and due to the wrong assumption that the
expected bit would apply only for the first (i.e., 'new') packet of a
connection, while in fact the expected bit remains on for the lifetime of
an expected connection.  The 'ctinfo' value IP_CT_RELATED derived from
the ct status can be used instead, as it is only ever applicable to
the 'new' packets of the expected connection.

Fixes: 05752523 ('openvswitch: Interface with NAT.')
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

This fix adds a new reference counter (ref_netlink) for the struct ip_set.
The other reference counter (ref) can be swapped out by ip_set_swap and we
need a separate counter to keep track of references for netlink events
like dump. Using the same ref counter for dump causes a race condition
which can be demonstrated by the following script:

ipset create hash_ip1 hash:ip family inet hashsize 1024 maxelem 500000 \
counters
ipset create hash_ip2 hash:ip family inet hashsize 300000 maxelem 500000 \
counters
ipset create hash_ip3 hash:ip family inet hashsize 1024 maxelem 500000 \
counters

ipset save &

ipset swap hash_ip3 hash_ip2
ipset destroy hash_ip3 /* will crash the machine */

Swap will exchange the values of ref so destroy will see ref = 0 instead of
ref = 1. With this fix in place swap will not succeed because ipset save
still has ref_netlink on the set (ip_set_swap doesn't swap ref_netlink).

Both delete and swap will error out if ref_netlink != 0 on the set.

Note: The changes to *_head functions is because previously we would
increment ref whenever we called these functions, we don't do that
anymore.
Reviewed-by: Joshua Hunt <johunt@akamai.com>
Signed-off-by: Vishwanath Pai <vpai@akamai.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

GPIOlib will print warning messages if we call GPIO functions without a
valid GPIO. Change the code to avoid doing so.
Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

It may fail to set coalesce usecs to HW, and Ethtool needs to know if it
is successful to cfg the parameter or not. So it needs return the errno by
dsaf.ko.
Signed-off-by: Lisheng <lisheng011@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The overtime setting and getting REGs in HNS V2 is defferent from HNS V1.
It needs to be distinguished between them if getting or setting the REGs.
Signed-off-by: Lisheng <lisheng011@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For the input parameter count, it's better to use the size
of destination buffer size, as nla_memcpy would take into
account the length of the source netlink attribute when
a data is copied from an attribute.
Signed-off-by: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In ravb_start_xmit dont call skb_tx_timestamp only when hardware
timestamping is requested: in the latter case software timestamps are
suppressed and thus the call of skb_tx_timestamp does not have any effect.

Instead call skb_tx_timestamp unconditionally in ravb_start_xmit, since
the function checks itself if software timestamping is required or should
be skipped due to hardware timestamping.
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Acked-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

We need to use post-decrement to ensure that irq_dispose_mapping is
also called on priv->rxq[0]->irq_no; moreover, if one of the above for
loops failed already at i==0 (so we reach one of these labels with
that value of i), we'll enter an essentially infinite loop of
out-of-bounds accesses.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Reviewed-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix variable assignment inside if statement. It is error-prone and hard to read.
Signed-off-by: Cosmin-Gabriel Samoila <gabrielcsmo@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

For a route with IPv6 encapsulation, the traffic class and hop limit
values are interchanged when returned to userspace by the kernel.
For example, see below.

># ip route add 192.168.0.1 dev eth0.2 encap ip6 dst 0x50 tc 0x50 hoplimit 100 table 1000
># ip route show table 1000
192.168.0.1  encap ip6 id 0 src :: dst fe83::1 hoplimit 80 tc 100 dev eth0.2  scope link
Signed-off-by: Quentin Armitage <quentin@armitage.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Diego Viola <diego.viola@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch removes two BUG_ON() used to notify about RX queue corruptions
on macb (not gem) hardware without actually handling the error.

The new code skips corrupted frames but still processes faultless frames.
Then it resets the RX queue before restarting the reception from a clean
state.

This patch is a rework of an older patch proposed by Neil Armstrong:
http://patchwork.ozlabs.org/patch/371525/Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Just updating version as many fixes got
accumulated over 1.00.00.34
Signed-off-by: Manish Chopra <manish.chopra@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add PHY entries for the Broadcom BCM7346 and BCM7362 chips, these are
40nm generation Ethernet PHY.

Fixes: 815717d1 ("net: phy: bcm7xxx: Remove wildcard entries")
Signed-off-by: Jaedon Shin <jaedon.shin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add map_flags attribute to bpf_map_show_fdinfo(), so that tools like
tc can check for them when loading objects from a pinned entry, e.g.
if user intent wrt allocation (BPF_F_NO_PREALLOC) is different to the
pinned object, it can bail out. Follow-up to 6c905981 ("bpf:
pre-allocate hash map elements"), so that tc can still support this
with v4.6.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

netpoll_setup() does a dev_hold() on np->dev, the netpoll device.  If it
fails, it correctly does a dev_put() but leaves np->dev set.  If we call
netpoll_cleanup() after the failure, np->dev is still set so we do another
dev_put(), which decrements the refcount an extra time.

It's questionable to call netpoll_cleanup() after netpoll_setup() fails,
but it can be difficult to find the problem, and we can easily avoid it in
this case.  The extra decrements can lead to hangs like this:

  unregister_netdevice: waiting for bond0 to become free. Usage count = -3

Set and clear np->dev at the points where we dev_hold() and dev_put() the
device.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

skb_len needs to be skb_headlen(skb) in bcmgenet_xmit_single().

Fragmented skbs can have only Ethernet + IP + TCP headers (14+20+20=54 bytes)
in the linear buffer, followed by the rest in fragments. Bumping skb_len to
ETH_ZLEN would be incorrect for this case, as it would introduce garbage
between TCP header and the fragment data.

This also works with regular/non-fragmented small packets < ETH_ZLEN bytes.
Successfully tested this on GENETv3 with 42-byte ARP frames.

For testing, I used:
ethtool -K eth0 tx-checksum-ipv4 off
ethtool -K eth0 tx-checksum-ipv6 off
echo 0 > /proc/sys/net/ipv4/tcp_timestamps

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: Petri Gynther <pgynther@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1. Add bytes_compl local variable to __bcmgenet_tx_reclaim() to collect
   transmitted bytes. dev->stats updates can then be moved outside the
   while-loop. bytes_compl is also needed for future BQL support.
2. When bcmgenet device uses Tx checksum offload, each transmitted skb
   gets an extra 64-byte header prepended to it. Before this header is
   prepended to the skb, we need to save the skb "wire" length in
   GENET_CB(skb)->bytes_sent, so that proper Tx bytes accounting can
   be done in __bcmgenet_tx_reclaim().
3. skb->len covers the entire length of skb, whether it is linear or
   fragmented. Thus, when we clean the fragments, do not increase
   transmitted bytes.

Fixes: 1c1008c7 ("net: bcmgenet: add main driver file")
Signed-off-by: Petri Gynther <pgynther@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Two minor typo.
Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macb_interrupt() should not use macb_writel(bp, ISR, <value>) but only
queue_writel(queue, ISR, <value>).

There is one IRQ and one set of {ISR, IER, IDR, IMR} [1] registers per
queue on gem hardware, though only queue0 is actually used for now to
receive frames: other queues can already be used to transmit frames.

The queue_readl() and queue_writel() helper macros are designed to access
the relevant IRQ registers.

[1]
ISR: Interrupt Status Register
IER: Interrupt Enable Register
IDR: Interrupt Disable Register
IMR: Interrupt Mask Register
Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Fixes: bfbb92c4 ("net: macb: Handle the RXUBR interrupt on all devices")
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Yisen Zhuang says:

====================
net: hns: fix some bugs in HNS driver

Here are some bug fixed patches for HNS driver.

They are:

>from Kejian, fix for the warning of passing zero to 'PTR_ERR'

>from qianqian, four fixes for inappropriate operation in hns driver

>from Sheng, one fix for optimization of irq proccess in hns driver, and
one fix for hilink status for hns driver.

For more details, please see individual patches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The hilink status reg in HNS V2 is different from HNS v1. In HNS V2, It
distinguishes differnt lane status according to the bit-field of the reg.
As is shown below:
[0:0] ---> lane0
[1:1] ---> lane1
...

But the current driver reads the reg to get the hilink status ONLY
concidering HNS V1 situation. Here is a patch to support both of them.
Signed-off-by: Sheng Li <lisheng011@huawei.com>
Signed-off-by: Daode Huang <huangdaode@hisilicon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

There is a misuse of PTR as shown below:

	ae_node = (void *)of_parse_phandle(dev->of_node,
					   "ae-handle",
					   0);
	if (IS_ERR_OR_NULL(ae_node)) {
		ret = PTR_ERR(ae_node);
		dev_err(dev, "not find ae-handle\n");
		goto out_read_prop_fail;
	}

if the ae_node is NULL, PTR_ERR(ae_node) means it returns success. And the
return value should be -ENODEV.
Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

In hns V1, common_poll should check and clean fbd pkts, because it
can not pend irq to clean them if there is no new pkt comes in.
But hns V2 hw fixes this bug, and will pend irq itself to do this.
So, for hns V2, we set ring_data->fini_process to NULL.
Signed-off-by: Sheng Li <lisheng011@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The variable head in hns_nic_tx_fini_pro has read a value, but it is
obviously no use. The patch will fix it.
And the comment is nothing to do with the routine, so it has to be removed
Signed-off-by: Qianqian Xie <xieqianqian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The return values in the first two functions mdiobus_write()
are ignored. The patch will fix it.
Signed-off-by: Qianqian Xie <xieqianqian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

It misses string format in function snprintf(), as below:
snprintf(buff, ETH_GSTRING_LEN, g_gmac_stats_string[i].desc);

It needs to add "%s" to fix it as below:
snprintf(buff, ETH_GSTRING_LEN, "%s", g_gmac_stats_string[i].desc);
Signed-off-by: Qianqian Xie <xieqianqian@huawei.com>
Signed-off-by: Kejian Yan <yankejian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The cycle index should be varied while the variable j is a fixed value.
The patch will fix this bug.
Signed-off-by: Qianqian Xie <xieqianqian@huawei.com>
Signed-off-by: Yisen Zhuang <Yisen.Zhuang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>