dev_queue_xmit() needs to dirty fields "state", "q", "bstats" and "qstats"

On x86_64 arch, they currently span three cache lines, involving more
cache line ping pongs than necessary, making longer holding of queue spinlock.

We can reduce this to one cache line, by grouping all read-mostly fields
at the beginning of structure. (Or should I say, all highly modified fields
at the end :) )

Before patch :

offsetof(struct Qdisc, state)=0x38
offsetof(struct Qdisc, q)=0x48
offsetof(struct Qdisc, bstats)=0x80
offsetof(struct Qdisc, qstats)=0x90
sizeof(struct Qdisc)=0xc8

After patch :

offsetof(struct Qdisc, state)=0x80
offsetof(struct Qdisc, q)=0x88
offsetof(struct Qdisc, bstats)=0xa0
offsetof(struct Qdisc, qstats)=0xac
sizeof(struct Qdisc)=0xc0
Signed-off-by: NEric Dumazet <dada1@cosmosbay.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We do this so we can later inform userspace who set the
regulatory domain and provide details of the request.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This is not used as we can always just assume the first
regulatory domain set will _always_ be a static regulatory
domain. REGDOM_SET_BY_CORE will be the first request from
cfg80211 for a regdomain and that then populates the first
regulatory request.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This patch skips the delivery of conntrack events if the packet
was drop due to a race condition in the conntrack insertion.
Signed-off-by: NPablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: NPatrick McHardy <kaber@trash.net>

There's very little need for most of the callsites to get
tp->xmit_goal_size updated. That will cost us divide as is,
so slice the function in two. Also, the only users of the
tp->xmit_goal_size are directly behind tcp_current_mss(),
so there's no need to store that variable into tcp_sock
at all! The drop of xmit_goal_size currently leaves 16-bit
hole and some reorganization would again be necessary to
change that (but I'm aiming to fill that hole with u16
xmit_goal_size_segs to cache the results of the remaining
divide to get that tso on regression).

Bring xmit_goal_size parts into tcp.c
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Cc: Evgeniy Polyakov <zbr@ioremap.net>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Wow, it was quite tricky to merge that stream of negations
but I think I finally got it right:

check & replace_ts_recent:
(s32)(rcv_tsval - ts_recent) >= 0                  => 0
(s32)(ts_recent - rcv_tsval) <= 0                  => 0

discard:
(s32)(ts_recent - rcv_tsval)  > TCP_PAWS_WINDOW    => 1
(s32)(ts_recent - rcv_tsval) <= TCP_PAWS_WINDOW    => 0

I toggled the return values of tcp_paws_check around since
the old encoding added yet-another negation making tracking
of truth-values really complicated.
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It turns out that net_alive is unnecessary, and the original problem
that led to it being added was simply that the icmp code thought
it was a network device and wound up being unable to handle packets
while there were still packets in the network namespace.

Now that icmp and tcp have been fixed to properly register themselves
this problem is no longer present and we have a stronger guarantee
that packets will not arrive in a network namespace then that provided
by net_alive in netif_receive_skb.  So remove net_alive allowing
packet reception run a little faster.

Additionally document the strong reason why network namespace cleanup
is safe so that if something happens again someone else will have
a chance of figuring it out.
Signed-off-by: NEric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Commit faee47cd
(sctp: Fix the RTO-doubling on idle-link heartbeats)
broke the RTO doubling for data retransmits.  If the
heartbeat was sent before the data T3-rtx time, the
the RTO will not double upon the T3-rtx expiration.
Distingish between the operations by passing an argument
to the function.

Additionally, Wei Youngjun pointed out that our treatment
of requested HEARTBEATS and timer HEARTBEATS is the same
wrt resetting congestion window.  That needs to be separated,
since user requested HEARTBEATS should not treat the link
as idle.
Signed-off-by: NVlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The above functions from include/net/tcp.h have been defined with an
argument that they never use. The argument is 'u32 ack' which is never
used inside the function body, and thus it can be removed. The rest of
the patch involves the necessary changes to the function callers of the
above two functions.
Signed-off-by: NHantzis Fotis <xantzis@ceid.upatras.gr>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Also fixes insignificant bug that would cause sending of stale
SACK block (would occur in some corner cases).
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It seems that implementation in yeah was inconsistent to what
other did as it would increase cwnd one ack earlier than the
others do.

Size benefits:

  bictcp_cong_avoid |  -36
  tcp_cong_avoid_ai |  +52
  bictcp_cong_avoid |  -34
  tcp_scalable_cong_avoid |  -36
  tcp_veno_cong_avoid |  -12
  tcp_yeah_cong_avoid |  -38

= -104 bytes total
Signed-off-by: NIlpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When devices are world roaming they cannot beacon or do active scan
on 5 GHz or on channels 12, 13 and 14 on the 2 GHz band. Although
we have a good regulatory API some cards may _always_ world roam, this
is also true when a system does not have CRDA present. Devices doing world
roaming can still passive scan, if they find a beacon from an AP on
one of the world roaming frequencies we make the assumption we can do
the same and we also remove the passive scan requirement.

This adds support for providing beacon regulatory hints based on scans.
This works for devices that do either hardware or software scanning.
If a channel has not yet been marked as having had a beacon present
on it we queue the beacon hint processing into the workqueue.

All wireless devices will benefit from beacon regulatory hints from
any wireless device on a system including new devices connected to
the system at a later time.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

All regulatory hints (core, driver, userspace and 11d) are now processed in
a workqueue.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

We do this so later on we can move the pending requests onto a
workqueue. By using the wiphy_idx instead of the wiphy we can
later easily check if the wiphy has disappeared or not.
Signed-off-by: NLuis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This adds optional notifier functions for software scan.
Signed-off-by: NMichael Buesch <mb@bu3sch.de>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The previous patch made cfg80211 generally aware of the signal
type a given hardware will give, so now it can implement
SIOCGIWRANGE itself, removing more wext stuff from mac80211.
Might need to be a little more parametrized once we have
more hardware using cfg80211 and new hardware capabilities.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

It wasn't a good idea to make the signal type a per-BSS option,
although then it is closer to the actual value. Move it to be
a per-wiphy setting, update mac80211 to match.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The TX/RX packet counters are needed to fill in RADIUS Accounting
attributes Acct-Output-Packets and Acct-Input-Packets. We already
collect the needed information, but only the TX/RX bytes were
previously exposed through nl80211. Allow applications to fetch the
packet counters, too, to provide more complete support for accounting.
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Acked-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This extends the NL80211_CMD_TRIGGER_SCAN command to allow applications
to specify a set of information element(s) to be added into Probe
Request frames with NL80211_ATTR_IE. This provides support for the
MLME-SCAN.request primitive parameter VendorSpecificInfo and can be
used, e.g., to implement WPS scanning.
Signed-off-by: NJouni Malinen <jouni.malinen@atheros.com>
Acked-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The AP can switch dynamically between 20/40 Mhz channel width,
in which case we switch the local operating channel, but the
rate control algorithm is not notified. This patch adds a new callback
to indicate such changes to the RC algorithm.

Currently, HT channel width change is notified, but this callback
can be used to indicate any new requirements that might come up later on.
Signed-off-by: NSujith <Sujith.Manoharan@atheros.com>
Acked-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Hardware with AMPDU queues currently has broken aggregation.

This patch fixes it by making all A-MPDUs go over the regular AC queues,
but keeping track of the hardware queues in mac80211. As a first rough
version, it actually stops the AC queue for extended periods of time,
which can be removed by adding buffering internal to mac80211, but is
currently not a huge problem because people rarely use multiple TIDs
that are in the same AC (and iwlwifi currently doesn't operate as AP).

This is a short-term fix, my current medium-term plan, which I hope to
execute soon as well, but am not sure can finish before .30, looks like
this:
 1) rework the internal queuing layer in mac80211 that we use for
    fragments if the driver stopped queue in the middle of a fragmented
    frame to be able to queue more frames at once (rather than just a
    single frame with its fragments)
 2) instead of stopping the entire AC queue, queue up the frames in a
    per-station/per-TID queue during aggregation session initiation,
    when the session has come up take all those frames and put them
    onto the queue from 1)
 3) push the ampdu queue layer abstraction this patch introduces in
    mac80211 into the driver, and remove the virtual queue stuff from
    mac80211 again

This plan will probably also affect ath9k in that mac80211 queues the
frames instead of passing them down, even when there are no ampdu queues.
Signed-off-by: NJohannes Berg <johannes@sipsolutions.net>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Only ipw2x00 now uses it.  Reduce confusion.  Profit!
Signed-off-by: NDan Williams <dcbw@redhat.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Impact: Attribute function with __releases(...)

Fix this sparse warning:
  net/ipv4/inet_fragment.c:276:35: warning: context imbalance in 'inet_frag_find' - unexpected unlock
Signed-off-by: NHannes Eder <hannes@hanneseder.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Some of the qualification tests demand that in case of failures in L2CAP
the HCI disconnect should indicate a reason why L2CAP fails. This is a
bluntly layer violation since multiple L2CAP connections could be using
the same ACL and thus forcing a disconnect reason is not a good idea.

To comply with the Bluetooth test specification, the disconnect reason
is now stored in the L2CAP connection structure and every time a new
L2CAP channel is added it will set back to its default. So only in the
case where the L2CAP channel with the disconnect reason is really the
last one, it will propagated to the HCI layer.

The HCI layer has been extended with a disconnect indication that allows
it to ask upper layers for a disconnect reason. The upper layer must not
support this callback and in that case it will nicely default to the
existing behavior. If an upper layer like L2CAP can provide a disconnect
reason that one will be used to disconnect the ACL or SCO link.

No modification to the ACL disconnect timeout have been made. So in case
of Linux to Linux connection the initiator will disconnect the ACL link
before the acceptor side can signal the specific disconnect reason. That
is perfectly fine since Linux doesn't make use of this value anyway. The
L2CAP layer has a perfect valid error code for rejecting connection due
to a security violation. It is unclear why the Bluetooth specification
insists on having specific HCI disconnect reason.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

In preparation for L2CAP fixed channel support, the CID value of a
L2CAP connection needs to be accessible via the socket interface. The
CID is the connection identifier and exists as source and destination
value. So extend the L2CAP socket address structure with this field and
change getsockname() and getpeername() to fill it in.

The bind() and connect() functions have been modified to handle L2CAP
socket address structures of variable sizes. This makes them future
proof if additional fields need to be added.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

If the extended features mask indicates support for fixed channels,
request the list of available fixed channels. This also enables the
fixed channel features bit so remote implementations can request
information about it. Currently only the signal channel will be
listed.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

The recommendation for the L2CAP PSM 1 (SDP) is to not use any kind
of authentication or encryption. So don't trigger authentication
for incoming and outgoing SDP connections.

For L2CAP PSM 3 (RFCOMM) there is no clear requirement, but with
Bluetooth 2.1 the initiator is required to enable authentication
and encryption first and this gets enforced. So there is no need
to trigger an additional authentication step. The RFCOMM service
security will make sure that a secure enough link key is present.

When the encryption gets enabled after the SDP connection setup,
then switch the security level from SDP to low security.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

If the remote L2CAP server uses authentication pending stage and
encryption is enabled it can happen that a L2CAP connection request is
sent twice due to a race condition in the connection state machine.

When the remote side indicates any kind of connection pending, then
track this state and skip sending of L2CAP commands for this period.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

When two L2CAP connections are requested quickly after the ACL link has
been established there exists a window for a race condition where a
connection request is sent before the information response has been
received. Any connection request should only be sent after an exchange
of the extended features mask has been finished.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

When receiving incoming connection to specific services, always use
general bonding. This ensures that the link key gets stored and can be
used for further authentications.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

When attempting to setup eSCO connections it can happen that some link
manager implementations fail to properly negotiate the eSCO parameters
and thus fail the eSCO setup. Normally the link manager is responsible
for the negotiation of the parameters and actually fallback to SCO if
no agreement can be reached. In cases where the link manager is just too
stupid, then at least try to establish a SCO link if eSCO fails.

For the Bluetooth devices with EDR support this includes handling packet
types of EDR basebands. This is particular tricky since for the EDR the
logic of enabling/disabling one specific packet type is turned around.
This fix contains an extra bitmask to disable eSCO EDR packet when
trying to fallback to a SCO connection.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

A role switch with devices following the Bluetooth pre-2.1 standards
or without Encryption Pause and Resume support is not possible if
encryption is enabled. Most newer headsets require the role switch,
but also require that the connection is encrypted.

For connections with a high security mode setting, the link will be
immediately dropped. When the connection uses medium security mode
setting, then a grace period is introduced where the TX is halted and
the remote device gets a change to re-enable encryption after the
role switch. If not re-enabled the link will be dropped.

Based on initial work by Ville Tervo <ville.tervo@nokia.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Change the RFCOMM internals to use the new security levels and remove
the link mode details.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Change the L2CAP internals to use the new security levels and remove
the link mode details.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

The current security model is based around the flags AUTH, ENCRYPT and
SECURE. Starting with support for the Bluetooth 2.1 specification this is
no longer sufficient. The different security levels are now defined as
SDP, LOW, MEDIUM and SECURE.

Previously it was possible to set each security independently, but this
actually doesn't make a lot of sense. For Bluetooth the encryption depends
on a previous successful authentication. Also you can only update your
existing link key if you successfully created at least one before. And of
course the update of link keys without having proper encryption in place
is a security issue.

The new security levels from the Bluetooth 2.1 specification are now
used internally. All old settings are mapped to the new values and this
way it ensures that old applications still work. The only limitation
is that it is no longer possible to set authentication without also
enabling encryption. No application should have done this anyway since
this is actually a security issue. Without encryption the integrity of
the authentication can't be guaranteed.

As default for a new L2CAP or RFCOMM connection, the LOW security level
is used. The only exception here are the service discovery sessions on
PSM 1 where SDP level is used. To have similar security strength as with
a Bluetooth 2.0 and before combination key, the MEDIUM level should be
used. This is according to the Bluetooth specification. The MEDIUM level
will not require any kind of man-in-the-middle (MITM) protection. Only
the HIGH security level will require this.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

In order to decide if listening RFCOMM sockets should be accept()ed
the BD_ADDR of the remote device needs to be known. This patch adds
a socket option which defines a timeout for deferring the actual
connection setup.

The connection setup is done after reading from the socket for the
first time. Until then writing to the socket returns ENOTCONN.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

The L2CAP and RFCOMM applications require support for authorization
and the ability of rejecting incoming connection requests. The socket
interface is not really able to support this.

This patch does the ground work for a socket option to defer connection
setup. Setting this option allows calling of accept() and then the
first read() will trigger the final connection setup. Calling close()
would reject the connection.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

It turns out that net_alive is unnecessary, and the original problem
that led to it being added was simply that the icmp code thought
it was a network device and wound up being unable to handle packets
while there were still packets in the network namespace.

Now that icmp and tcp have been fixed to properly register themselves
this problem is no longer present and we have a stronger guarantee
that packets will not arrive in a network namespace then that provided
by net_alive in netif_receive_skb.  So remove net_alive allowing
packet reception run a little faster.

Additionally document the strong reason why network namespace cleanup
is safe so that if something happens again someone else will have
a chance of figuring it out.
Signed-off-by: NEric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The untracked conntrack actually does usually have events marked for
delivery as its not special-cased in that part of the code. Skip the
actual delivery since it impacts performance noticeably.
Signed-off-by: NPatrick McHardy <kaber@trash.net>

A long time ago we had bugs, primarily in TCP, where we would modify
skb->truesize (for TSO queue collapsing) in ways which would corrupt
the socket memory accounting.

skb_truesize_check() was added in order to try and catch this error
more systematically.

However this debugging check has morphed into a Frankenstein of sorts
and these days it does nothing other than catch false-positives.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>