This patch starts the new implementation of feature negotiation:
1. Although it is theoretically possible to perform feature negotiation at any
time (and RFC 4340 supports this), in practice this is prohibitively complex,
as it requires to put traffic on hold for each new negotiation.
2. As a byproduct of restricting feature negotiation to connection setup, the
feature-negotiation retransmit timer is no longer required. This part is now
mapped onto the protocol-level retransmission.
Details indicating why timers are no longer needed can be found on
http://www.erg.abdn.ac.uk/users/gerrit/dccp/notes/feature_negotiation/\
implementation_notes.html
This patch disables anytime negotiation, subsequent patches work out full
feature negotiation support for connection setup.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This inserts the required de-allocation routines for memory allocated by
feature negotiation in the socket destructors, replacing dccp_feat_clean()
in one instance.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
This provides feature-negotiation initialisation for both DCCP sockets and
DCCP request_sockets, to support feature negotiation during connection setup.
It also resolves a FIXME regarding the congestion control initialisation.
Thanks to Wei Yongjun for help with the IPv6 side of this patch.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
This adds list fields and list management functions for the new feature
negotiation implementation. The new code is kept in parallel to the old
code, until removed at the end of the patch set.
Thanks to Arnaldo for suggestions to improve the code.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
A lookup table for feature-negotiation information, extracted from RFC 4340/42,
is provided by this patch. All currently known features can be found in this
table, along with their feature location, their default value, and type.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
This patch prepares for the new and extended feature-negotiation routines.
The following feature-negotiation data structures are provided:
* a container for the various (SP or NN) values,
* symbolic state names to track feature states,
* an entry struct which holds all current information together,
* elementary functions to fill in and process these structures.
Entry structs are arranged as FIFO for the following reason: RFC 4340 specifies
that if multiple options of the same type are present, they are processed in the
order of their appearance in the packet; which means that this order needs to be
preserved in the local data structure (the later insertion code also respects
this order).
The struct list_head has been chosen for the following reasons: the most
frequent operations are
* add new entry at tail (when receiving Change or setting socket options);
* delete entry (when Confirm has been received);
* deep copy of entire list (cloning from listening socket onto request socket).
The NN value has been set to 64 bit, which is a currently sufficient upper limit
(Sequence Window feature has 48 bit).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
The BUG_ON(w_tot == 0) only holds if there is no more than 1 loss interval in
the loss history. If there is only a single loss interval, the calc_i_mean()
routine need in fact not be called (RFC 3448, 6.3.1).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This sets the sysfs permissions so that root can toggle the `debug'
parameter available for nearly every DCCP module. This is useful
since there are various module inter-dependencies. The debug flag
can now be toggled at runtime using
echo 1 > /sys/module/dccp/parameters/dccp_debug
echo 1 > /sys/module/dccp_ccid2/parameters/ccid2_debug
echo 1 > /sys/module/dccp_ccid3/parameters/ccid3_debug
echo 1 > /sys/module/dccp_tfrc_lib/parameters/tfrc_debug
The last is not very useful yet, since no code at the moment calls
the tfrc_debug() macro.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
dccp_disconnect() can be called due to several reasons:
1. when the connection setup failed (inet_stream_connect());
2. when shutting down (inet_shutdown(), inet_csk_listen_stop());
3. when aborting the connection (dccp_close() with 0 linger time).
In case (1) the write queue is empty. This patch empties the write queue,
if in case (2) or (3) it was not yet empty.
This avoids triggering the write-queue BUG_TRAP in sk_stream_kill_queues()
later on.
It also seems natural to do: when breaking an association, to delete all
packets that were originally intended for the soon-disconnected end (compare
with call to tcp_write_queue_purge in tcp_disconnect()).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This updates the use of the `out_invalid_option' label, which produces a
Reset (code 5, "Option Error"), to fill in the Data1...Data3 fields as
specified in RFC 4340, 5.6.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This updates the option-parsing code with regard to RFC 4340, 5.8:
"[..] options with nonsensical lengths (length byte less than two or more
than the remaining space in the options portion of the header) MUST be
ignored, and any option space following an option with nonsensical length
MUST likewise be ignored."
Hence in the following cases erratic options will be ignored:
1. The type byte of a multi-byte option is the last byte of the header
options (i.e. effective option length of 1).
2. The value of the length byte is less than the minimum 2. This has been
changed from previously 3: although no multi-byte option with a length
less than 3 yet exists (cf. table 3 in 5.8), a length of 2 is valid.
(The switch-statement in dccp_parse has further per-option length checks.)
3. The option length exceeds the length of the remaining option space.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
RFC4340 states that if a packet is received with an option error (such as a
Mandatory Option as the last byte of the option list), the endpoint should
repond with a Reset.
In the LISTEN and RESPOND states, the endpoint correctly reponds with Reset,
while in the REQUEST/OPEN states, packets with option errors are just ignored.
The packet sequence is as follows:
Case 1:
Endpoint A Endpoint B
(CLOSED) (CLOSED)
<---------------- REQUEST
RESPONSE -----------------> (*1)
(with invalid option)
<---------------- RESET
(with Reset Code 5, "Option Error")
(*1) currently just ignored, no Reset is sent
Case 2:
Endpoint A Endpoint B
(OPEN) (OPEN)
DATA-ACK -----------------> (*2)
(with invalid option)
<---------------- RESET
(with Reset Code 5, "Option Error")
(*2) currently just ignored, no Reset is sent
This patch fixes the problem, by generating a Reset instead of silently
ignoring option errors.
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
The allocated RX buffer size was 64 bytes bigger than the PCI mapped
size with no good reason. If the packet was actually using the buffer up
to its limit and if the last 64 bytes of the buffer crossed 4KB boundary
then an unmapped PCI page was accessed. The fix is to use only one
parameter for the buffer size - there is no need to differentiate
between the buffer size and the PCI mapping size since the extra 64
bytes can actually be used by the FW to align the Ethernet payload to
64 bytes.
Also updating the driver version and date
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch sets STATUS_EXIT_PENDING on pci_remove. Otherwise
iwl4965_down may fail to uninitialize the driver.
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch calls apm stop on exit and suspend. Without this patch
hardware consumes power even after driver is removed or suspended.
Signed-off-by: Gregory Greenman <gregory.greenman@intel.com>
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch "iwlwifi: do not use GFP_DMA in iwl_tx_queue_init" removes
GFP_DMA from allocation tx command buffers. GFP_DMA allows allocation
only for memory under 16M which causes allocation problems
suspend/resume flows.
Using kmalloc is temporal solution and some consistent/coherent
allocation schema will be more correct. Since iwlwifi hardware
supports 64bit address this solution should work on x86 (32 and
64bit) for now.
This patch fixes memory freeing problem in the previous patch.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Ian Schram <ischram@telenet.be>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes rx_chain computation. The code that adjusts number of
rx chains to number supported by HW was missing. Miss configuration
causes firmware error. Note: iwlwifi supports HW with up to 3 RX
chains (2x2, 2x3, 1x2, and 3x3 MIMO). This patch also simplifies the
whole RX chain computation.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes the wrong use MIMO power save values. Our TX was
configured with our MIMO power save values instead of peer's MIMO power
save values, this may affect connectivity. The peer STA/AP may not sense
our traffic at all as it doesn't have all RX chains opened.
Signed-off-by: Ron Rindjunsky <ron.rindjunsky@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Rx chain might change during power save transitions but it doesn't
require sending Full-ROXN command to the firmware. Full-RXON requires
reconnection to an AP and thus affects user experience. The patch
avoids the Full-RXON by removing the rx_chain modification check in
iwl_full_rxon_required function.
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This enables sending of direct probes on passive channels, as long as
traffic was detected on that channel. This enables connectivity to
hidden/non broadcasting SSIDs APs on passive channels. Note 5000 HW
declares all 5.2 spectrum as passive.
Signed-off-by: Cahill Ben <ben.m.cahill@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Ron Rindjunsky <ron.rindjunsky@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch is a W/A for the TSF sync issue in IBSS merging. HW is not
capable to sync TSF (it's constantly little behind). This creates
constant IBSS merging upon reception of each beacon, adding and removing
station which in turn creates above 50% packet loss and thus dramatically
degrade the throughput. The W/A simply stops the driver from declaring it
has a reliable TSF value and thus eliminates IBSS merging.
Signed-off-by: Assaf Krauss <assaf.krauss@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Remove chipset-specific quirk workaround; the workaround caused
unrecoverable DMA lockups when the driver was loaded following a
PXE boot.
Signed-off-by: Dhananjay Phadke <dhananjay@netxen.com>
Signed-off-by: Michael Brown <mbrown@fensystems.co.uk>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
This commit dropped the setting of the default interrupt throttle rate.
commit 021230d40a
Author: Ayyappan Veeraiyan <ayyappan.veeraiyan@intel.com>
Date: Mon Mar 3 15:03:45 2008 -0800
ixgbe: Introduce MSI-X queue vector code
The following patch adds it back. Without this the default value of 0
causes the performance of this card to be awful. Restoring these to the
default values yields much better performance.
This regression has been around since 2.6.25.
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
CC: stable@kernel.org [2.6.25 and later]
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Make the "pegasus" driver scream less loudly in the face of
problems as it initializes, avoiding hundreds of messages:
- ratelimit some key error messages
- avoid some spurious diagnostics caused by strange codeflow
And fix one instance of goofy indentation.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Andrew Morton reported a build failure on sparc32, because TIPC
uses names like "struct node" and there is a like named data
structure defined in linux/node.h
This just regexp replaces "struct node*" to "struct tipc_node*"
to avoid this and any future similar problems.
Signed-off-by: David S. Miller <davem@davemloft.net>
Ever since commit 4c563f7669
("[XFRM]: Speed up xfrm_policy and xfrm_state walking") it is
illegal to call __xfrm_state_destroy (and thus xfrm_state_put())
with xfrm_state_lock held. If we do, we'll deadlock since we
have the lock already and __xfrm_state_destroy() tries to take
it again.
Fix this by pushing the xfrm_state_put() calls after the lock
is dropped.
Signed-off-by: David S. Miller <davem@davemloft.net>
Re-enable IP when the MTU gets back to a valid size.
This patch just checks if the in_dev is NULL on a NETDEV_CHANGEMTU event
and if MTU is valid (bigger than 68), then re-enable in_dev.
Also a function that checks valid MTU size was created.
Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case of error, the function xfrm_bundle_create returns an ERR
pointer, but never returns a NULL pointer. So a NULL test that comes
after an IS_ERR test should be deleted.
The semantic match that finds this problem is as follows:
(http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@match_bad_null_test@
expression x, E;
statement S1,S2;
@@
x = xfrm_bundle_create(...)
... when != x = E
* if (x != NULL)
S1 else S2
// </smpl>
Signed-off-by: Julien Brunel <brunel@diku.dk>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch addresses an issue with the locking order. ath_rx_flush_tid()
uses spin_lock/unlock_bh when IRQs are disabled in sta_notify by mac80211.
As node clean up is still pending with ath9k and this problematic portion
of the code is expected to change anyway, thinking of a proper fix may not
be worthwhile. So having this interim fix helps the users to get rid of the
kernel warning message.
Pasted the kernel warning message for reference.
kernel: ath0: No ProbeResp from current AP 00:1b:11:60:7a:3d - assume out of range
kernel: ------------[ cut here ]------------
kernel: WARNING: at kernel/softirq.c:136 local_bh_enable+0x3c/0xab()
kernel: Pid: 1029, comm: ath9k Not tainted 2.6.27-rc4-wt-w1fi-wl
kernel:
kernel: Call Trace:
kernel: [<ffffffff802278d8>] warn_on_slowpath+0x51/0x77
kernel: [<ffffffff80224c51>] check_preempt_wakeup+0xf3/0x123
kernel: [<ffffffff80239658>] autoremove_wake_function+0x9/0x2e
kernel: [<ffffffff8022c281>] local_bh_enable+0x3c/0xab
kernel: [<ffffffffa01ab75a>] ath_rx_node_cleanup+0x38/0x6e [ath9k]
kernel: [<ffffffffa01b2280>] ath_node_detach+0x3b/0xb6 [ath9k]
kernel: [<ffffffffa01ab09f>] ath9k_sta_notify+0x12b/0x165 [ath9k]
kernel: [<ffffffff802366cf>] queue_work+0x1d/0x49
kernel: [<ffffffffa018c3fc>] add_todo+0x70/0x99 [mac80211]
kernel: [<ffffffffa017de76>] __sta_info_unlink+0x16b/0x19e [mac80211]
kernel: [<ffffffffa017e6ed>] sta_info_unlink+0x18/0x43 [mac80211]
kernel: [<ffffffffa0182732>] ieee80211_associated+0xaa/0x16d [mac80211]
kernel: [<ffffffffa0184a1a>] ieee80211_sta_work+0x4fb/0x6b4 [mac80211]
kernel: [<ffffffff80469c58>] thread_return+0x30/0xa9
kernel: [<ffffffffa018451f>] ieee80211_sta_work+0x0/0x6b4 [mac80211]
kernel: [<ffffffff802362c2>] run_workqueue+0xb1/0x17a
kernel: [<ffffffff80236be9>] worker_thread+0xd0/0xdb
kernel: [<ffffffff8023964f>] autoremove_wake_function+0x0/0x2e
kernel: [<ffffffff80236b19>] worker_thread+0x0/0xdb
kernel: [<ffffffff8023954a>] kthread+0x47/0x75
kernel: [<ffffffff80223121>] schedule_tail+0x18/0x50
kernel: [<ffffffff8020bc49>] child_rip+0xa/0x11
kernel: [<ffffffff80239503>] kthread+0x0/0x75
kernel: [<ffffffff8020bc3f>] child_rip+0x0/0x11
kernel:
kernel: ---[ end trace e9bb5da661055827 ]---
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Updating sc_keytype multiple times when groupwise and pairwise
ciphers are different results in incorrect pairwise key type
assumed for TX control and normal ping fails. This works fine
for cases where both groupwise and pairwise ciphers are same.
Also use mac80211 provided enums for key length calculation.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
A "Set" to a sign-bit in an "&" operation causes a compiler warning.
Make calculations unsigned.
[ The warning was masked by the old definition of BUILD_BUG_ON() ]
Also remove __builtin_constant_p from FIELD_CHECK since BUILD_BUG_ON
no longer permits non-const values.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
debugfs union in struct ieee80211_sub_if_data is misused by including a
common default_key dentry as a union member. This ends occupying the same
memory area with the first dentry in other union members (structures;
usually drop_unencrypted). Consequently, debugfs operations on
default_key symlinks and drop_unencrypted entry are using the same
dentry pointer even though they are supposed to be separate ones. This
can lead to removing entries incorrectly or potentially leaving
something behind since one of the dentry pointers gets lost.
Fix this by moving the default_key dentry to a new struct
(common_debugfs) that contains dentries (more to be added in future)
that are shared by all vif types. The debugfs union must only be used
for vif type-specific entries to avoid this type of pointer corruption.
Signed-off-by: Jouni Malinen <jouni.malinen@atheros.com>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The leak in if_cs_prog_helper() is obvious.
It looks a bit as if not freeing "fw" in if_cs_prog_real() was done
intentionally, but I'm not seeing why it shouldn't be freed.
Reported-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Holger Schurig <hs4233@mail.mn-solutions.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
When multicasting the driver sets the number of group addresses using
the count from the previous set multicast command. In general this means
you have to set the multicast addresses twice to get the behaviour you
want.
If we were multicasting, and reduce the number of addresses we are
multicasting to, then the driver would write uninitialised data from the
stack into the group addresses to multicast to.
Only write the multicast addresses we have specifically set.
Signed-off-by: David Kilroy <kilroyd@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes loading firmware from memory above 32bit.
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Acked-by: Marcel Holtmann <holtmann@linux.intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The patch fixes CSR_GP_CNTRL_REG_FLAG_INIT_DONE was set instead of
cleared which disabled moving device to D0U state.
Signed-off-by: Mohamed Abbas <mohamed.abbas@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch adds workaround for an interrupt related hardware bug on
some platforms. (Apparently these platforms boot-up w/ INTX_DISABLED
set. -- JWL)
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Zhu Yi <yi.zhu@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
GFP_DMA is not necessary for the iwlwifi hardware and it can cause
allocation failures and/or invoke the OOM killer on lots of systems.
For reference:
https://bugzilla.redhat.com/show_bug.cgi?id=459709
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Current setup with hal and NetworkManager will fail to work
without newest hal version with this config option disabled.
Although this will solve itself by time, at the moment it is
dishonest to say that we don't know any software that uses it,
if there are many many people relying on old hal versions.
Signed-off-by: Florian Mickler <florian@mickler.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Nothing in linux/pim.h should be exported to userspace.
This should fix the XORP build failure reported by
Jose Calhariz, the debain package maintainer.
Nothing originally in linux/mroute.h was exported to userspace
ever, but some of this stuff started to be when it was moved into
this new linux/pim.h, and that was wrong. If we didn't provide these
definitions for 10 years we can reasonably expect that applications
defined this stuff locally or used GLIBC headers providing the
protocol definitions. And as such the only result of this can
be conflict and userland build breakage.
Signed-off-by: David S. Miller <davem@davemloft.net>
Use qdisc_root_sleeping_lock() instead of qdisc_root_lock() where
appropriate. The only difference is while dev is deactivated, when
currently we can use a sleeping qdisc with the lock of noop_qdisc.
This shouldn't be dangerous since after deactivation root lock could
be used only by gen_estimator code, but looks wrong anyway.
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The number of identifiers needs to be checked against the option
length. Also, the identifier index provided needs to be verified
to make sure that it doesn't exceed the bounds of the array.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The bonds check to prevent buffer overlflow was not exactly
right. It still allowed overflow of up to 8 bytes which is
sizeof(struct sctp_authkey).
Since optlen is already checked against the size of that struct,
we are guaranteed not to cause interger overflow either.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
