nlmsg_total_size() calculates the length of a netlink message
including header and alignment. nla_total_size() calculates the
space an individual attribute consumes which was meant to be used
in this context.
Also, ensure to account for the attribute header for the
IFLA_INFO_XSTATS attribute as implementations of get_xstats_size()
seem to assume that we do so.
The addition of two message headers minus the missing attribute
header resulted in a calculated message size that was larger than
required. Therefore we never risked running out of skb tailroom.
Signed-off-by: Thomas Graf <tgraf@infradead.org>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
The type of FRAG6_CB(prev)->offset is int, skb->len is *unsigned* int,
and offset is int.
Without this patch, type conversion occurred to this expression, when
(FRAG6_CB(prev)->offset + prev->len) is less than offset.
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
The NF_HOOK_COND returns 0 when it shouldn't due to what I believe to be an
error in the code as the order of operations is not what was intended. C will
evalutate == before =. Which means ret is getting set to the bool result,
rather than the return value of the function call. The code says
if (ret = function() == 1)
when it meant to say:
if ((ret = function()) == 1)
Normally the compiler would warn, but it doesn't notice it because its
a actually complex conditional and so the wrong code is wrapped in an explict
set of () [exactly what the compiler wants you to do if this was intentional].
Fixing this means that errors when netfilter denies a packet get propagated
back up the stack rather than lost.
Problem introduced by commit 2249065f (netfilter: get rid of the grossness
in netfilter.h).
Signed-off-by: Eric Paris <eparis@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Patrick McHardy <kaber@trash.net>
As noted by Steve Chen, since commit
f5fff5dc8a ("tcp: advertise MSS
requested by user") we can end up with a situation where
tcp_select_initial_window() does a divide by a zero (or
even negative) mss value.
The problem is that sometimes we effectively subtract
TCPOLEN_TSTAMP_ALIGNED and/or TCPOLEN_MD5SIG_ALIGNED from the mss.
Fix this by increasing the minimum from 8 to 64.
Reported-by: Steve Chen <schen@mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Robin Holt tried to boot a 16TB machine and found some limits were
reached : sysctl_tcp_mem[2], sysctl_udp_mem[2]
We can switch infrastructure to use long "instead" of "int", now
atomic_long_t primitives are available for free.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Reported-by: Robin Holt <holt@sgi.com>
Reviewed-by: Robin Holt <holt@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
packet_getname_spkt() doesn't initialize all members of sa_data field of
sockaddr struct if strlen(dev->name) < 13. This structure is then copied
to userland. It leads to leaking of contents of kernel stack memory.
We have to fully fill sa_data with strncpy() instead of strlcpy().
The same with packet_getname(): it doesn't initialize sll_pkttype field of
sockaddr_ll. Set it to zero.
Signed-off-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is a possibility malicious users can get limited information about
uninitialized stack mem array. Even if sk_run_filter() result is bound
to packet length (0 .. 65535), we could imagine this can be used by
hostile user.
Initializing mem[] array, like Dan Rosenberg suggested in his patch is
expensive since most filters dont even use this array.
Its hard to make the filter validation in sk_chk_filter(), because of
the jumps. This might be done later.
In this patch, I use a bitmap (a single long var) so that only filters
using mem[] loads/stores pay the price of added security checks.
For other filters, additional cost is a single instruction.
[ Since we access fentry->k a lot now, cache it in a local variable
and mark filter entry pointer as const. -DaveM ]
Reported-by: Dan Rosenberg <drosenberg@vsecurity.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sometimes ax25_getname() doesn't initialize all members of fsa_digipeater
field of fsa struct, also the struct has padding bytes between
sax25_call and sax25_ndigis fields. This structure is then copied to
userland. It leads to leaking of contents of kernel stack memory.
Signed-off-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Followup of commit ef885afbf8 (net: use rcu_barrier() in
rollback_registered_many)
dst_dev_event() scans a garbage dst list that might be feeded by various
network notifiers at device dismantle time.
Its important to call dst_dev_event() after other notifiers, or we might
enter the infamous msleep(250) in netdev_wait_allrefs(), and wait one
second before calling again call_netdevice_notifiers(NETDEV_UNREGISTER,
dev) to properly remove last device references.
Use priority -10 to let dst_dev_notifier be called after other network
notifiers (they have the default 0 priority)
Reported-by: Ben Greear <greearb@candelatech.com>
Reported-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Reported-by: Octavian Purdila <opurdila@ixiacom.com>
Reported-by: Benjamin LaHaise <bcrl@kvack.org>
Tested-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to use build-in debugging macro, pci_dev in priv need to be
assigned first.
This fix iwl3945 driver oopsed at boot with 2.6.37-rc1
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Structure sockaddr_tipc is copied to userland with padding bytes after
"id" field in union field "name" unitialized. It leads to leaking of
contents of kernel stack memory. We have to initialize them to zero.
Signed-off-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As device_set_wakeup_enable can now sleep, move the call to outside
the critical section.
Signed-off-by: Daniel J Blueman <daniel.blueman@gmail.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Andrew Hendry <andrew.hendry@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original patch helps under obscure conditions (no pun) but
some 8168 do not like it. The change needs to be tightened with
a specific 8168 version.
This reverts commit 801e147cde
("r8169: Handle rxfifo errors on 8168 chips").
Regression at https://bugzilla.kernel.org/show_bug.cgi?id=20882
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Tested-by: Andreas Radke <a.radke@arcor.de>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Daniel J Blueman <daniel.blueman@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 8723e1b4ad (inet: RCU changes in inetdev_by_index())
forgot one call site in ip_mc_drop_socket()
We should not decrease idev refcount after inetdev_by_index() call,
since refcount is not increased anymore.
Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Reported-by: Miles Lane <miles.lane@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This cause 'No Bonding' to be used if userspace has not yet been paired
with remote device since the l2cap socket used to create the rfcomm
session does not have any security level set.
Signed-off-by: Luiz Augusto von Dentz <luiz.dentz-von@nokia.com>
Acked-by: Ville Tervo <ville.tervo@nokia.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
Last commit added a wrong endianness conversion. Fixing that.
Reported-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
In function l2cap_get_conf_opt() and l2cap_add_conf_opt() the address of
opt->val sometimes is not at the edge of 2-bytes/4-bytes, so 2-bytes/4 bytes
access will cause data misalignment exeception. Use get_unaligned_le16/32
and put_unaligned_le16/32 function to avoid data misalignment execption.
Signed-off-by: steven miao <realmz6@gmail.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
We've done this for a while in Fedora without any obvious problems other
than some interaction with input devices. Those should be fixed now, so
let's try this in mainline.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
When initiating dedicated bonding a L2CAP raw socket with HIGH security
level is used. The kernel is supposed to trigger the authentication
request in this case but this doesn't happen currently for non-SSP
(pre-2.1) devices. The reason is that the authentication request happens
in the remote extended features callback which never gets called for
non-SSP devices. This patch fixes the issue by requesting also
authentiation in the (normal) remote features callback in the case of
non-SSP devices.
This rule is applied only for HIGH security level which might at first
seem unintuitive since on the server socket side MEDIUM is already
enough for authentication. However, for the clients we really want to
prefer the server side to decide the authentication requrement in most
cases, and since most client sockets use MEDIUM it's better to be
avoided on the kernel side for these sockets. The important socket to
request it for is the dedicated bonding one and that socket uses HIGH
security level.
The patch is based on the initial investigation and patch proposal from
Andrei Emeltchenko <endrei.emeltchenko@nokia.com>.
Signed-off-by: Johan Hedberg <johan.hedberg@nokia.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo F. Padovan <padovan@profusion.mobi>
The allocated externel radio banks have to be freed in
case of ath9k_hw_rf_alloc_ext_banks failure.
Cc: stable@kernel.org
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Since the endpoint descriptors (EP3 & EP4) were changed from Interrupt
to Bulk type by firmware, the urb submission done on Bulk pipes.
And the recent commit "check the endpoint type against the pipe type"
added aditional error checking against pipe types under CONFIG_USB_DEBUG.
So bmAttribute has to be updated for both EP3 & EP4 before submitting
urbs on that pipe. This patch resolves the following failure.
[ 2215.710936] usb 1-1: usb_probe_device
[ 2215.710945] usb 1-1: configuration #1 chosen from 1 choice
[ 2215.711152] usb 1-1: adding 1-1:1.0 (config #1, interface 0)
[ 2215.711252] ath9k_hif_usb 1-1:1.0: usb_probe_interface
[ 2215.711255] ath9k_hif_usb 1-1:1.0: usb_probe_interface - got id
[ 2215.712780] usb 1-1: BOGUS urb xfer, pipe 3 != type 1
[ 2215.713782] usb 1-1: ath9k_htc: Unable to allocate URBs
[ 2215.713801] ath9k_hif_usb: probe of 1-1:1.0 failed with error -22
Reported-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds support for USB dongle with device ID 3346 from IMC Networks.
Signed-off-by: Haitao Zhang <minipanda@linuxrobot.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Bit 22 of AR_WA should be set to fix the situation where chip reset
is asynchronous to clock of analog shift registers, such that when
reset is released, it could mess up the values of analog shift registers
and cause some hw issue on AR9280.
This bit is write only, but the driver does a read-modify-write
on AR_WA without setting bit 22 in ar9002_hw_configpcipowersave()
during radio disable. This causes surprise removal of hw. It can
never recover from this state and the hw will become usable only
after a power on/off cycle, and sometimes only during a cold reboot.
This issue can be triggered by doing frequent roaming with the
simple/test-roam script available from the wifi-test project [1]
when roaming between APs quickly. When roaming there is a is a high
possibility that the device being put into idle (radio disable) state
by mac80211 during AUTH->ASSOC. A device hardware reset would fail
and the kernel would output:
[40251.363799] ath: AWAKE -> FULL-SLEEP
[40251.363815] ieee80211 phy17: device no longer idle - working
[40251.363817] ath: Marking phy17 as not-idle
[40251.363819] ath: FULL-SLEEP -> AWAKE
[40251.415978] pciehp 0000:00:1c.3:pcie04: Card not present on Slot(3)
[40251.419896] ath: ah->misc_mode 0x4
[40251.428138] pciehp 0000:00:1c.3:pcie04: Card present on Slot(3)
[40251.532247] ath: timeout (100000 us) on reg 0x9860: 0xffffffff & 0x00000001 != 0x00000000
[40251.532250] ath: Unable to reset channel (2462 MHz), reset status -5
[40251.532422] ath: Set channel: 5745 MHz
[40251.540639] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.548826] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.557023] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.565211] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.573415] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.581603] ath: Failed to stop TX DMA in 100 msec after killing last frame
[40251.581606] ath: Failed to stop TX DMA. Resetting hardware!
[40251.592679] ath: DMA failed to stop in 10 ms AR_CR=0xffffffff AR_DIAG_SW=0xffffffff
[40251.703330] ath: timeout (100000 us) on reg 0x7000: 0xffffffff & 0x00000003 != 0x00000000
[40251.703333] ath: RTC stuck in MAC reset
[40251.703334] ath: Chip reset failed
[40251.703335] ath: Unable to reset hardware; reset status -22
This is currently only reproducible with some HB92 (Half Mini-PCIE)
cards but the fix applies to all AR9280 cards. This patch fixes this
issue by setting bit 22 during radio disable.
This patch has fixes for all kernels that has ath9k.
[1] http://wireless.kernel.org/en/developers/Testing/wifi-test
Cc: kyungwan.nam@atheros.com
Cc: amod.bodas@atheros.com
Cc: david.quan@atheros.com
Cc: stable@kernel.org
Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There are currently no provisions in place to ensure that the scanning
task has been stopped when the interface is stopped or removed.
This can result in a WARNING at net/wireless/core.c:643 and other badness
when you remove the module while a scan is happening.
Terminate the scanning task during interface stop.
Signed-off-by: Daniel Drake <dsd@laptop.org>
Acked-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
For client STA interfaces, ieee80211_do_stop unsets the relevant
interface's SDATA_STATE_RUNNING state bit prior to cancelling an
interrupted scan. When ieee80211_offchannel_return is invoked as
part of cancelling the scan, it doesn't bother unsetting the
SDATA_STATE_OFFCHANNEL bit because it sees that the interface is
down. Normally this doesn't matter because when the client STA
interface is brought back up, it will probably issue a scan. But
in some cases (e.g., the user changes the interface type while it
is down), the SDATA_STATE_OFFCHANNEL bit will remain set. This
prevents the interface queues from being started. So we
cancel the scan before unsetting the SDATA_STATE_RUNNING bit.
Signed-off-by: Brian Cavagnolo <brian@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
ath9k_ps_wakeup() clears the cycle counters after waking up the
hardware using ath9k_hw_setpower, however if power save is disabled,
then the counters will contain useful data, which then gets discarded.
Fix this by checking the old power mode before discarding any data.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
IS_ERR and PTR_ERR were called with the wrong pointer, leading to a
crash when cfg80211_get_dev_from_ifindex fails.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch includes the following updates:
* add D-Link DWA-130 Rev D
* Netgear has three WNDA3100 versions.
the original WNDA3100 is now called WNDA3100v1.
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Throughput was severely affected in Intel Pinetrail platforms
because of a DMA problem in C3 state. This patch fixes this
issue.
Signed-off-by: Vivek Natarajan <vnatarajan@atheros.com>
CC: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The HW opmode is blindly set to monitor type on monitor mode
change notification. This overrides the opmode when one of the
interfaces is still running as non-monitor iftype. So the monitoring
information needs to be maintained seperately.
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This bug seems to be due to commit 27ae60f8f7 ("ipw2x00: replace
"ieee80211" with "libipw" where appropriate"), where Pavel did this:
- libipw_proc = proc_mkdir(DRV_NAME, init_net.proc_net);
+ libipw_proc = proc_mkdir("ieee80211", init_net.proc_net);
but then the cleanup was kept as
remove_proc_entry(DRV_NAME, init_net.proc_net);
in both places (both in the failure case and in the unload case). The
error string is also total crap, and says
"Unable to create " DRV_NAME " proc directory\n");
Even though it doesn't actually create a proc directory named DRV_NAME at all.
So that patch looks like total and utter crap to me. The commit message says
"Keep /proc/net/ieee80211 under the original name to avoid breaking user
interface."
but the thing is, it really didn't fix anything but that one create
thing. It needs to fix all the other cases too.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Tested-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The sgs allocation error path leaks the allocated message.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
QDIO is running independent from netdevice state. We are not
allowed to schedule NAPI in case the netdevice is not open.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For a certain Hipersockets specific error code in the xmit path, the
qeth driver tries to invoke dev_queue_xmit again.
Commit 79640a4ca6 introduces a busylock
causing locking problems in case of re-invoked dev_queue_xmit by qeth.
This patch removes the attempts to retry packet sending with
dev_queue_xmit from the qeth driver.
Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fix a bug reported by backyes.
Right the first time pktgen's using queue_map that's not been initialized
by set_cur_queue_map(pkt_dev);
Signed-off-by: Junchang Wang <junchangwang@gmail.com>
Signed-off-by: Backyes <backyes@mail.ustc.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
After e6484930d7: net: allocate tx queues in register_netdevice
These calls make net drivers oops at load time, so let's avoid people
git-bisect'ing known problems.
Signed-off-by: Guillaume Chazarain <guichaz@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After e6484930d7: net: allocate tx queues in register_netdevice
It causes an Oops at skge_probe() time.
Signed-off-by: Guillaume Chazarain <guichaz@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The type of FRAG6_CB(prev)->offset is int, skb->len is *unsigned* int,
and offset is int.
Without this patch, type conversion occurred to this expression, when
(FRAG6_CB(prev)->offset + prev->len) is less than offset.
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The basic classifier keeps statistics but does not report it to user space.
This showed up when using basic classifier (with police) as a default catch
all on ingress; no statistics were reported.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The existing 'FirmwareVersion' attribute only covers the DSP firmware as
provided by Conexant; not the overall version of the device firmware. We
do want to be able to see the full version number too.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Presently the b43legacy build fails on an sh randconfig:
In file included from include/net/dst.h:12,
from drivers/net/wireless/b43legacy/xmit.c:32:
include/net/dst_ops.h:28: error: expected ':', ',', ';', '}' or '__attribute__' before '____cacheline_aligned_in_smp'
include/net/dst_ops.h: In function 'dst_entries_get_fast':
include/net/dst_ops.h:33: error: 'struct dst_ops' has no member named 'pcpuc_entries'
include/net/dst_ops.h: In function 'dst_entries_get_slow':
include/net/dst_ops.h:41: error: 'struct dst_ops' has no member named 'pcpuc_entries'
include/net/dst_ops.h: In function 'dst_entries_add':
include/net/dst_ops.h:49: error: 'struct dst_ops' has no member named 'pcpuc_entries'
include/net/dst_ops.h: In function 'dst_entries_init':
include/net/dst_ops.h:55: error: 'struct dst_ops' has no member named 'pcpuc_entries'
include/net/dst_ops.h: In function 'dst_entries_destroy':
include/net/dst_ops.h:60: error: 'struct dst_ops' has no member named 'pcpuc_entries'
make[5]: *** [drivers/net/wireless/b43legacy/xmit.o] Error 1
make[5]: *** Waiting for unfinished jobs....
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This should fix the following warning:
net/core/pktgen.c: In function ‘pktgen_if_write’:
net/core/pktgen.c:890: warning: comparison of distinct pointer types lacks a cast
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Reviewed-by: Nelson Elhage <nelhage@ksplice.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We were using nlmsg_find_attr() to look up the bytecode by attribute when
auditing, but then just using the first attribute when actually running
bytecode. So, if we received a message with two attribute elements, where only
the second had type INET_DIAG_REQ_BYTECODE, we would validate and run different
bytecode strings.
Fix this by consistently using nlmsg_find_attr everywhere.
Signed-off-by: Nelson Elhage <nelhage@ksplice.com>
Signed-off-by: Thomas Graf <tgraf@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This will let us use it on a nlmsghdr stored inside a netlink_callback.
Signed-off-by: Nelson Elhage <nelhage@ksplice.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit ebc0ffae5 (RCU conversion of fib_lookup()),
fib_result_assign() should not change fib refcounts anymore.
Thanks to Michael who did the bisection and bug report.
Reported-by: Michael Ellerman <michael@ellerman.id.au>
Tested-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>