This change addresses an L2CAP ERTM throughput problem when a remote
device does not fully utilize the available transmit window.

The L2CAP ERTM transmit window size determines the maximum number of
unacked frames that may be outstanding at any time. It is configured
separately for each direction of an ERTM connection. Each side sends a
configuration request with a tx_win field indicating how many unacked
frames it is capable of receiving before sending an ack. The
configuration response's tx_win field shows how many frames the
transmitter will actually send before waiting for an ack.

It's important to trace both the actual transmit window (to check for
validity of incoming frames) and the number of frames that the
transmitter will send before waiting (to send acks at the appropriate
time). Now there are separate tx_win and ack_win values. ack_win is
updated based on configuration responses, and is used to determine
when acks are sent.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Correct places where CID and PSM were printed as int. For CID: 0x%4.4x
is used and for PSM: 0x%2.2x.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Avoid unneeded type conversion by correcting type specifiers in debug
statements for L2CAP.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

If remote device sends bogus RFC option with invalid length,
undefined options values are used. Fix this by using defaults when
remote misbehaves.

This also fixes the following warning reported by gcc 4.7.0:

net/bluetooth/l2cap_core.c: In function 'l2cap_config_rsp':
net/bluetooth/l2cap_core.c:3302:13: warning: 'rfc.max_pdu_size' may be used uninitialized in this function [-Wmaybe-uninitialized]
net/bluetooth/l2cap_core.c:3266:24: note: 'rfc.max_pdu_size' was declared here
net/bluetooth/l2cap_core.c:3298:25: warning: 'rfc.monitor_timeout' may be used uninitialized in this function [-Wmaybe-uninitialized]
net/bluetooth/l2cap_core.c:3266:24: note: 'rfc.monitor_timeout' was declared here
net/bluetooth/l2cap_core.c:3297:25: warning: 'rfc.retrans_timeout' may be used uninitialized in this function [-Wmaybe-uninitialized]
net/bluetooth/l2cap_core.c:3266:24: note: 'rfc.retrans_timeout' was declared here
net/bluetooth/l2cap_core.c:3295:2: warning: 'rfc.mode' may be used uninitialized in this function [-Wmaybe-uninitialized]
net/bluetooth/l2cap_core.c:3266:24: note: 'rfc.mode' was declared here
Signed-off-by: NSzymon Janc <szymon.janc@tieto.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

The l2cap_conn_del function tries to cancel_sync the security timer, but
when it's called from the timeout function itself a deadlock occurs.
Subsequently the "hcon->l2cap_data = NULL" that's supposed to protect
multiple calls to l2cap_conn_del never gets cleared and when the
connection finally drops we double free's etc which will crash the
kernel.

This patch fixes the issue by using the HCI_CONN_LE_SMP_PEND for
protecting against this. The same flag is also used for the same purpose
in other places in the SMP code.
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

This patch changes the default MTU value for L2CAP ATT fixed channel
to L2CAP_DEFAULT_MTU (672 octets).

Differently from others L2CAP channels, in L2CAP ATT fixed channel
there is no MTU negotiation. The MTU value for that channel is up to
the L2CAP implementation. The only restriction in L2CAP spec is the
MTU value must not be less than 23 octets.

At ATT protocol level (on top of L2CAP), we have the ATT_MTU which
defines the maximum size of any ATT message sent between client and
server. GATT profile defines ATT_MTU default value to 23 octets. If
a GATT based profile wants to use ATT_MTU greater than 23 octets
(e.g. HID over GATT profile), it should negotiate a new value by
executing the GATT Exchange MTU sub-procedure.

Thus, in order to support any value of ATT_MTU negotiated at ATT
protocol level, our L2CAP implementation should have L2CAP ATT
fixed channel MTU equal or greater than ATT_MAX_MTU (512 octets).
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Trivial fix, let the code cleaner.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

Simplify the handling of different ERTM header size. We were the same
check in some places of the code, and more is expected to come, so just
replace them with a function.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

l2cap_att_channel always returns 0 which is not used.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

l2cap_conless_channel always return 0 which is not used.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

l2cap_data_channel always return 0 which is not used anywhere,
make it void function.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Make A2MP channel special channel which do not reference hci_conn.
This prevents from keeping ACL connection open when all L2CAP
channels got closed.

hci_conn_hold and hci_conn_put are not reference counts on the
hci_conn structure in the typical way.  They are reference counts for
the ACL.  When you do the last hci_conn_put, the ACL is disconnected
after a timeout.
Reported-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Functions will be always defined and in case not implemented as
dummy __no_func.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Handle incoming A2MP connection by creating AMP manager and
processing A2MP messages.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

A2MP fixed channel do not have sk
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Create and initialize fixed A2MP channel
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Move this check to before the channel time creation simplifies the code
and avoid memory allocation if the channel already exist.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Remove socket specific code from l2cap_core.c
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

These vars are kept in sync so we can use chan->state here.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This move socket specific code to l2cap_sock.c.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This remove a bit more of socket code from l2cap core, this calls set the
SOCK_ZAPPED and do some clean up depending on the socket state.
Reported-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Use chan instead of void * makes more sense here.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This replace code in l2cap_le_conn_ready() by a similar code in
l2cap_chan_ready().
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This is already performed inside l2cap_chan_ready(), so we don't need it
here again.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Acked-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Both FCS and Extended header might be or might not be present in a ERTM
channel.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

__constant_cpu_to_le16() is the right go here.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Define assigned Protocol and Service Multiplexor (PSM) identifiers
and use them instead of magic numbers.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Sometimes an ACL link must be raised to a higher security level after
an L2CAP connection is requested, but before a connection response is
sent.  In this case, a connection response sent by L2CAP was not
immediately followed by a configuration request.  Other code paths do
send this configuration request right away.  It was possible for the
connection to stall while L2CAP waited for the remote device (like
PTS) to trigger the configuration process.

Here is an abbreviated hcidump of the failure case with PTS:

1337806446.051982 > ACL data: handle 43 flags 0x02 dlen 10
    L2CAP(s): Info req: type 2
1337806446.052050 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Info rsp: type 2 result 0
      Extended feature mask 0x000000b8
1337806446.595320 > ACL data: handle 43 flags 0x02 dlen 12
    L2CAP(s): Connect req: psm 4097 scid 0x0041
1337806446.595673 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 1 status 0
1337806446.595679 < ACL data: handle 43 flags 0x00 dlen 10
    L2CAP(s): Info req: type 2
1337806446.669835 > ACL data: handle 43 flags 0x02 dlen 16
    L2CAP(s): Info rsp: type 2 result 0
      Extended feature mask 0x00000028
1337806446.669899 < HCI Command: Authentication Requested (0x01|0x0011) plen 2
1337806446.669906 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 1 status 1
<security setup here>
1337806446.769888 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 0 status 0

At this point, the connection stalls and no further messages are sent
on the L2CAP signaling channel.  No data is received either.

If we immediately send a configuration request after a successful connect
response, the connection completes:

1337724090.041162 > ACL data: handle 43 flags 0x02 dlen 10
    L2CAP(s): Info req: type 2
1337724090.041236 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Info rsp: type 2 result 0
      Extended feature mask 0x000000b8
1337724090.597128 > ACL data: handle 43 flags 0x02 dlen 12
    L2CAP(s): Connect req: psm 4097 scid 0x0041
1337724090.597236 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 1 status 0
1337724090.597244 < ACL data: handle 43 flags 0x00 dlen 10
    L2CAP(s): Info req: type 2
1337724090.660842 > ACL data: handle 43 flags 0x02 dlen 16
    L2CAP(s): Info rsp: type 2 result 0
      Extended feature mask 0x00000028
1337724090.660926 < HCI Command: Authentication Requested (0x01|0x0011) plen 2
1337724090.660934 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 1 status 1
<security setup here>
1337724090.755162 < ACL data: handle 43 flags 0x00 dlen 16
    L2CAP(s): Connect rsp: dcid 0x0040 scid 0x0041 result 0 status 0
1337724090.755171 < ACL data: handle 43 flags 0x00 dlen 23
    L2CAP(s): Config req: dcid 0x0041 flags 0x00 clen 11
1337724091.361847 > ACL data: handle 43 flags 0x02 dlen 29
    L2CAP(s): Config rsp: scid 0x0040 flags 0x00 result 0 clen 15
1337724091.863808 > ACL data: handle 43 flags 0x02 dlen 23
    L2CAP(s): Config req: dcid 0x0040 flags 0x00 clen 11
1337724091.863882 < ACL data: handle 43 flags 0x00 dlen 29
    L2CAP(s): Config rsp: scid 0x0041 flags 0x00 result 0 clen 15
1337724092.683745 > ACL data: handle 43 flags 0x02 dlen 12
    L2CAP(d): cid 0x0040 len 8 [psm 4097]
      0000: 00 00 11 22 33 44 34 2f                           ..."3D4/
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Define Continuation flag which the only flag used from Flags field
in L2CAP Configuration Request and Response.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Most of the include were unnecessary or already included by some other
header.
Replace module.h by export.h where possible.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Previous callers of l2cap_build_conf_rsp in l2cap_config_req use
flags instead of continuation flag hardcoded value. It does not change
logic and preserve future possible flags.
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

These functions were returning always 0, we just make then void.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

The same check is done just before call l2cap_streaming_send()
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Silence warnings below:

net/bluetooth/l2cap_core.c:1662:24: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:1662:27: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:1683:24: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:1683:27: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:2260:46: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:2574:33: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:2581:33: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:4556:24: warning: Using plain integer
as NULL pointer
net/bluetooth/l2cap_core.c:4556:27: warning: Using plain integer
as NULL pointer
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

chan->tx_q is only initialized if we use ERTM or Streaming mode.
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This enables the new receive and transmit state machines.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

The transmit window values must be configured for streaming mode, even
though streaming mode does not have a window.  This enables use of
extended headers when the transmit window socket option is set to 64
or larger.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>

Let the compiler decide if inlining is appropriate.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo.padovan@collabora.co.uk>