With the %pK format specifier we hide the kernel addresses
with the help of kptr_restrict sysctl.
In this patch, %p is changed to %pK in the driver code.

The sysctl is documented in Documentation/sysctl/kernel.txt.
Signed-off-by: Maharaja Kennadyrajan <c_mkenna@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

WMI_SERVICE_PERIODIC_CHAN_STAT_SUPPORT service has missed in
the commit 7e247a9e ("ath10k: add dynamic tx mode switch
config support for qca4019"). This patch adds the service to
avoid mismatch between host and target.

Fixes: 7e247a9e ("ath10k: add dynamic tx mode switch config support for qca4019")
Signed-off-by: Tamizh chelvam <c_traja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Currently the FW version is taken from the sw_version field
of the FW ready event. This version is based on internal
version control revision and it is difficult to map to actual
FW version.
Fix this by using the actual FW version stored in the FW file
header record.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

When driver is loaded, extract a capabilities record
from the FW file. This record contains bits indicating
which optional features are supported by this FW.
The driver can use this information to determine
which functionality to support and/or expose to user
space.
The extraction is done before wiphy structure is
registered, because the capabilities can affect
information published by the this structure.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Align to latest version of the auto generated wmi file
describing the interface with FW.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Currently the places that check wil->scan_request and
call cfg80211_scan_done are not consistently protected,
so there is a risk that cfg80211_scan_done will be called
with NULL scan_request, causing a kernel crash.
Fix this by using p2p_wdev_mutex in few other places
that access scan_request. This makes sense since
scan_request may point to p2p_wdev, and it is not worth
the extra complexity of adding a new mutex.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

HW Rx filters and masks are not configured
properly by firmware during boot sequences. The
MAC_PCU_ADDR1 is set to 0s instead of 1s which
allows the HW to ACK any frame that passes through
MAC_PCU_RX_FILTER. The MAC_PCU_RX_FILTER itself
is misconfigured on boot as well.

The combination of these bugs ended up with the
following manifestations:
 - "no channel configured; ignoring frame(s)!"
   warnings in the driver
 - spurious ACKs (transmission) on the air during
   firmware bootup sequences

The former was a long standing and known bug
originally though mostly harmless.

However Marek recently discovered that this
problem also involves ACKing *all* frames the HW
receives (including beacons ;). Such frames
are delivered to host and generate the former
warning as well.

This could be a problem with regulatory compliance
in some rare cases (e.g. Taiwan which forbids
transmissions on channel 36 which is the default
bootup channel on 5Ghz band cards). The good news
is that it'd require someone else to violate
regulatory first to coerce our device to generate
and transmit an ACK.

The problem could be reproduced in a rather busy
environment that has a lot of APs. The likelihood
could be increased by injecting an msleep() of
5000 or longer immediately after
ath10k_htt_setup() in ath10k_core_start().

The reason why the former warnings were only
showing up seldom is because the device was either
quickly reset again (i.e. during firmware probing)
or wmi vdev was created (which fixes hw and fw
states).

It is technically possible for host driver to
override adequate hw registers however this can't
work reliably because the bug root cause lies in
incorrect firmware state on boot (internal
structure used to program MAC_PCU_ADDR1 is not
properly initialized) and only vdev create/delete
events can fix it. This is why the patch takes
dummy vdev approach.

This could be fixed in firmware as well but having
this fixed in driver is more robust, most notably
when thinking of users of older firmware such as
999.999.0.636.
Reported-by: Marek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

This allows placing command barriers for explicit
serializing and synchronizing state.

Useful for future driver development.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Will be useful for implementing command barriers.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Will be useful for implementing command barriers.
Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

This patch adds testmode support for 10.4 based chipsets
and added code swap support for UTF binary.
Signed-off-by: Tamizh chelvam <c_traja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Currently testmode events for 10.x firmware are processed from
10.1 wmi event processing. This patch is used to handle testmode
events in 10.2 and 10.4 based firmware.
Signed-off-by: Tamizh chelvam <c_traja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Preparation to make use of firmware_swap_code_seg_info for UTF binary.
Signed-off-by: Tamizh chelvam <c_traja@qti.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Currently wiphy structure is initialized and registered
in wil_if_alloc, before some information is available such
as MAC address and capabilities. As a result there is a
small chance user space will get incorrect information
from calls such as NL80211_CMD_GET_WIPHY.
Fix this by seperating the registration and moving it
to wil_if_add which is executed later, after all
relevant information is known.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

txdata->enabled is used in order to determine if the TX vring
is valid. As the data transmit is handled in a different context,
in case txdata->enabled is set before vring->hwtail is updated,
an old or corrupted vring->hwtail can be used.
Protect setting of txdata->enabled and vring->hwtail to prevent a
case where TX vring start handling TX packets before setting
vring->hwtail.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Stations disconnection is executed as part of wil_reset so no
need to do it in wil_down.
Removal of the disconnect operation will also preserve the lock
of wil->mutex during the whole reset flow and prevent handling of
connect event while resetting.
Set wil_status_resetting in earlier stage in the flow to prevent
double resetting call in case communication with FW fails while
bringing the interface down.
Signed-off-by: Lazar Alexei <qca_ailizaro@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

fix stop p2p device handling to identify between search
and listen and update the upper layers with the appropriate
notification.
The stop of p2p radio operations also needs to be performed
in __wil_down.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Change the logging category of HALP functions from
MISC to IRQ, since the HALP mechanism is closely
related to interrupts. Both HALP and IRQ create
a heavy load of logging messages when enabled,
so their logging is typically disabled during normal debug
scenarios. Having them in the same logging category
will make it easier to disable logging for both in one go.
Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

According to the spec the PN should be calculated per TID.
In the current implementation, the PN and key_set were set
only for TID 0, therefore only traffic for TID 0 was supported.
In order to support all TIDs, the key_set and PN should be set
for all the TIDs.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

In case HALP vote times out, we need to mask the HALP IRQ, as done
in case the interrupt is received, as this interrupt should be
set until completion of the low latency operation.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Align to latest version of the auto generated wmi file
describing the interface with FW.
Signed-off-by: Maya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

just to comply with current ath9k_hw_nvram_read to return value, hence
behaving reacting accordingly.
Signed-off-by: Eduardo Abinader <eduardo.abinader@riverbed.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Fixes smatch warning:

ath9k_vif_iter_set_beacon() warn if statement not indented
Signed-off-by: Bob Copeland <me@bobcopeland.com>
Reviewed-by: Julian Calaby <julian.calaby@gmail.com>
Signed-off-by: Kalle Valo <kvalo@qca.qualcomm.com>

Minor overlapping changes for both merge conflicts.

Resolution work done by Stephen Rothwell was used
as a reference.
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull networking fixes from David Miller:

 1) Buffers powersave frame test is reversed in cfg80211, fix from Felix
    Fietkau.

 2) Remove bogus WARN_ON in openvswitch, from Jarno Rajahalme.

 3) Fix some tg3 ethtool logic bugs, and one that would cause no
    interrupts to be generated when rx-coalescing is set to 0.  From
    Satish Baddipadige and Siva Reddy Kallam.

 4) QLCNIC mailbox corruption and napi budget handling fix from Manish
    Chopra.

 5) Fix fib_trie logic when walking the trie during /proc/net/route
    output than can access a stale node pointer.  From David Forster.

 6) Several sctp_diag fixes from Phil Sutter.

 7) PAUSE frame handling fixes in mlxsw driver from Ido Schimmel.

 8) Checksum fixup fixes in bpf from Daniel Borkmann.

 9) Memork leaks in nfnetlink, from Liping Zhang.

10) Use after free in rxrpc, from David Howells.

11) Use after free in new skb_array code of macvtap driver, from Jason
    Wang.

12) Calipso resource leak, from Colin Ian King.

13) mediatek bug fixes (missing stats sync init, etc.) from Sean Wang.

14) Fix bpf non-linear packet write helpers, from Daniel Borkmann.

15) Fix lockdep splats in macsec, from Sabrina Dubroca.

16) hv_netvsc bug fixes from Vitaly Kuznetsov, mostly to do with VF
    handling.

17) Various tc-action bug fixes, from CONG Wang.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (116 commits)
  net_sched: allow flushing tc police actions
  net_sched: unify the init logic for act_police
  net_sched: convert tcf_exts from list to pointer array
  net_sched: move tc offload macros to pkt_cls.h
  net_sched: fix a typo in tc_for_each_action()
  net_sched: remove an unnecessary list_del()
  net_sched: remove the leftover cleanup_a()
  mlxsw: spectrum: Allow packets to be trapped from any PG
  mlxsw: spectrum: Unmap 802.1Q FID before destroying it
  mlxsw: spectrum: Add missing rollbacks in error path
  mlxsw: reg: Fix missing op field fill-up
  mlxsw: spectrum: Trap loop-backed packets
  mlxsw: spectrum: Add missing packet traps
  mlxsw: spectrum: Mark port as active before registering it
  mlxsw: spectrum: Create PVID vPort before registering netdevice
  mlxsw: spectrum: Remove redundant errors from the code
  mlxsw: spectrum: Don't return upon error in removal path
  i40e: check for and deal with non-contiguous TCs
  ixgbe: Re-enable ability to toggle VLAN filtering
  ixgbe: Force VLNCTRL.VFE to be set in all VMDq paths
  ...

Tom Herbert says:

====================
strp: Stream parser for messages

This patch set introduces a utility for parsing application layer
protocol messages in a TCP stream. This is a generalization of the
mechanism implemented of Kernel Connection Multiplexor.

This patch set adapts KCM to use the strparser. We expect that kTLS
can use this mechanism also. RDS would probably be another candidate
to use a common stream parsing mechanism.

The API includes a context structure, a set of callbacks, utility
functions, and a data ready function. The callbacks include
a parse_msg function that is called to perform parsing (e.g.
BPF parsing in case of KCM), and a rcv_msg function that is called
when a full message has been completed.

For strparser we specify the return codes from the parser to allow
the backend to indicate that control of the socket should be
transferred back to userspace to handle some exceptions in the
stream: The return values are:

      >0 : indicates length of successfully parsed message
       0  : indicates more data must be received to parse the message
       -ESTRPIPE : current message should not be processed by the
          kernel, return control of the socket to userspace which
          can proceed to read the messages itself
       other < 0 : Error is parsing, give control back to userspace
          assuming that synchronization is lost and the stream
          is unrecoverable (application expected to close TCP socket)

There is one issue I haven't been able to fully resolve. If parse_msg
returns ESTRPIPE (wants control back to userspace) the parser may
already have consumed some bytes of the message. There is no way to
put bytes back into the TCP receive queue and tcp_read_sock does not
allow an easy way to peek messages. In lieu of a better solution, we
return ENODATA on the socket to indicate that the data stream is
unrecoverable (application needs to close socket). This condition
should only happen if an application layer message header is split
across two skbuffs and parsing just the first skbuff wasn't sufficient
to determine the that transfer to userspace is needed.

This patch set contains:

  - strparser implementation
  - changes to kcm to use strparser
  - strparser.txt documentation

v2:
  - Add copyright notice to C files
  - Remove GPL module license from strparser.c
  - Add report of rxpause

v3:
  - Restore GPL module license
  - Use EXPORT_SYMBOL_GPL

v4:
  - Removed unused function, changed another to be static as suggested
    by davem
  - Rewoked data_ready to be called from upper layer, no longer requires
    taking over socket data_ready callback as suggested by Lance Chao

Tested:
  - Ran a KCM thrash test for 24 hours. No behavioral or performance
    differences observed.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Adapt KCM to use the stream parser. This mostly involves removing
the RX handling and setting up the strparser using the interface.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch introduces a utility for parsing application layer protocol
messages in a TCP stream. This is a generalization of the mechanism
implemented of Kernel Connection Multiplexor.

The API includes a context structure, a set of callbacks, utility
functions, and a data ready function.

A stream parser instance is defined by a strparse structure that
is bound to a TCP socket. The function to initialize the structure
is:

int strp_init(struct strparser *strp, struct sock *csk,
              struct strp_callbacks *cb);

csk is the TCP socket being bound to and cb are the parser callbacks.

The upper layer calls strp_tcp_data_ready when data is ready on the lower
socket for strparser to process. This should be called from a data_ready
callback that is set on the socket:

void strp_tcp_data_ready(struct strparser *strp);

A parser is bound to a TCP socket by setting data_ready function to
strp_tcp_data_ready so that all receive indications on the socket
go through the parser. This is assumes that sk_user_data is set to
the strparser structure.

There are four callbacks.
 - parse_msg is called to parse the message (returns length or error).
 - rcv_msg is called when a complete message has been received
 - read_sock_done is called when data_ready function exits
 - abort_parser is called to abort the parser

The input to parse_msg is an skbuff which contains next message under
construction. The backend processing of parse_msg will parse the
application layer protocol headers to determine the length of
the message in the stream. The possible return values are:

   >0 : indicates length of successfully parsed message
   0  : indicates more data must be received to parse the message
   -ESTRPIPE : current message should not be processed by the
      kernel, return control of the socket to userspace which
      can proceed to read the messages itself
   other < 0 : Error is parsing, give control back to userspace
      assuming that synchronzation is lost and the stream
      is unrecoverable (application expected to close TCP socket)

In the case of error return (< 0) strparse will stop the parser
and report and error to userspace. The application must deal
with the error. To handle the error the strparser is unbound
from the TCP socket. If the error indicates that the stream
TCP socket is at recoverable point (ESTRPIPE) then the application
can read the TCP socket to process the stream. Once the application
has dealt with the exceptions in the stream, it may again bind the
socket to a strparser to continue data operations.

Note that ENODATA may be returned to the application. In this case
parse_msg returned -ESTRPIPE, however strparser was unable to maintain
synchronization of the stream (i.e. some of the message in question
was already read by the parser).

strp_pause and strp_unpause are used to provide flow control. For
instance, if rcv_msg is called but the upper layer can't immediately
consume the message it can hold the message and pause strparser.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

While commit 9c706a49 ("net: ipconfig: fix use after free") avoids
the use after free, the resulting code still ends up calling both the
ic_setup_if() and ic_setup_routes() after calling ic_close_devs(), and
access to the device is still required.

Move the call to ic_close_devs() to the very end of the function.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Cong Wang says:

====================
net_sched: tc action fixes and updates

This patchset fixes a few regressions caused by the previous
code refactor and more. Thanks to Jamal for catching them!

Note, patch 3/7 and 4/7 are not strictly necessary for this patchset,
I just want to carry them together.

---
v4: adjust an indention for Jamal
    add two more patches

v3: avoid list for fast path, suggested by Jamal

v2: replace flex_array with regular dynamic array
    keep tcf_action_stats_update() in act_api.h
    fix macro typos found by Amir
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

The act_police uses its own code to walk the
action hashtable, which leads to that we could
not flush standalone tc police actions, so just
switch to tcf_generic_walker() like other actions.

(Joint work from Roman and Cong.)
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Jamal reported a crash when we create a police action
with a specific index, this is because the init logic
is not correct, we should always create one for this
case. Just unify the logic with other tc actions.

Fixes: a03e6fe5 ("act_police: fix a crash during removal")
Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

As pointed out by Jamal, an action could be shared by
multiple filters, so we can't use list to chain them
any more after we get rid of the original tc_action.
Instead, we could just save pointers to these actions
in tcf_exts, since they are refcount'ed, so convert
the list to an array of pointers.

The "ugly" part is the action API still accepts list
as a parameter, I just introduce a helper function to
convert the array of pointers to a list, instead of
relying on the C99 feature to iterate the array.

Fixes: a85a970a ("net_sched: move tc_action into tcf_common")
Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

struct tcf_exts belongs to filters, should not be visible
to plain tc actions.

Cc: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

It is harmless because all users pass 'a' to this macro.

Fixes: 00175aec ("net/sched: Macro instead of CONFIG_NET_CLS_ACT ifdef")
Cc: Amir Vadai <amir@vadai.me>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This list_del() for tc action is not needed actually,
because we only use this list to chain bulk operations,
therefore should not be carried for latter operations.

Fixes: ec0595cc ("net_sched: get rid of struct tcf_common")
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

After refactoring tc_action into tcf_common, we no
longer need to cleanup temporary "actions" in list,
they are permanently stored in the hashtable.

Fixes: a85a970a ("net_sched: move tc_action into tcf_common")
Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Simon Wunderlich says:

====================
pull request for net-next: batman-adv 2016-08-16

This feature patchset is all about adding netlink support, which should
supersede our debugfs configuration interface in the long run. It is
especially necessary when batman-adv should be used in different
namespaces, since debugfs can not differentiate between those.

More specifically, the following changes are included:

 - Two fixes for namespace handling by Andrew Lunn, checking also the
   namespaces for parent interfaces, and supress debugfs entries
   for non-default netns

 - Implement various netlink commands for the new interface, by
   Matthias Schiffer, Andrew Lunn, Sven Eckelmann and Simon Wunderlich
   (13 patches):
    * routing algorithm list
    * hardif list
    * translation tables (local and global)
    * TTVN for the translation tables
    * originator and neighbor tables for B.A.T.M.A.N. IV
      and B.A.T.M.A.N. V
    * gateway dump functionality for B.A.T.M.A.N. IV
      and B.A.T.M.A.N. V
    * Bridge Loop Avoidance claims, and corresponding BLA group
    * Bridge Loop Avoidance backbone tables

 - Finally, mark batman-adv as netns compatible, by Andrew Lunn (1 patch)
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Jeff Kirsher says:

====================
Intel Wired LAN Driver Updates 2016-08-16

This series contains fixes to e1000e, igb, ixgbe and i40e.

Kshitiz Gupta provides a fix for igb to resolve the PHY delay compensation
math in several functions.

Jarod Wilson provides a fix for e1000e which had to broken up into 2
patches, first is prepares the driver for expanding the list of NICs
that have occasional ~10 hour clock jumps when being used for PTP.
Second patch actually fixes i218 silicon which has been experiencing
the clock jumps while using PTP.

Alex provides 2 patches for ixgbe now that he is back at Intel.  First
fixes setting VLNCTRL.VFE bit, which was left unchanged in earlier patches
which resulted in disabling VLAN filtering for all the VFs.  Second
corrects the support for disabling the VLAN tag filtering via the
feature bit.

Lastly, David fixes i40e which was causing a kernel panic when
non-contiguous traffic classes or traffic classes not starting with TC0,
were configured on a link partner switch.  To fix this, changed the
logic when determining the total number of TCs enabled.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>