When building fragmented skb's skb->len keeps track of the size of head
plus all fragments combined, however when queueing the skb for sending we
need to report the head size instead of the total size, so we just set
skb->len to skb_headlen().

This bug appeared when implementing MSG_MORE support for L2CAP sockets, it
never showed up before because l2cap_skbuff_fromiovec() never accounted skb
size correctly. A following patch will fix this.
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>
Reviewed-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

If alloc() fails we let the frags linked list with garbage value (the
err ptr value) in its last element.
Reported-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

Commit 105bdf9e introduced a
regression in L2CAP streaming mode due to rearranged initialization
code that is shared between ERTM and streaming mode.  This change
makes sure the transmit queue is initialized in both modes.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Acked-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Commit 94122bbe introduced a problem
where tx_send_head was not set to point to the first skb in the ERTM
transmit queue, which stalled data transmission.  This change sets
that pointer when transmission is not already in progress.
Reported-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

The mgmt_ev_device_connected signal must be sent before any event
indications happen for sockets associated with the connection. Otherwise
e.g. device authorization for the sockets will fail with ENOTCONN as
user space things that there is no baseband link.

This patch fixes the issue by ensuring that the device_connected event
if sent (if it hasn't been so already) as soon as the first ACL data
packet arrives from the remote device.
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

It fixes L2CAP socket based security level elevation during a
connection. The HID profile needs this (for keyboards) and it is the only
way to achieve the security level elevation when using the management
interface to talk to the kernel (hence the management enabling patch
being the one that exposes this issue).

It enables the userspace a security level change when the socket is
already connected and create a way to notify the socket the result of the
request. At the moment of the request the socket is made non writable, if
the request fails the connections closes, otherwise the socket is made
writable again, POLL_OUT is emmited.
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

In commit b0302aba, the rtlwifi family of drivers was converted to use
asynchronous firmware loading. Unfortumately, the implementation was
racy, and the ieee80211 routines could be started before rtl_init_core()
was called to setup the data.

This patch fixes the bug noted in https://bugzilla.kernel.org/show_bug.cgi?id=43187.
Reported-by: NJoshua Roys <Joshua.Roys@gtri.gatech.edu>
Tested-by: NNeptune Ning <frostyplanet@gmail.com>
Signed-off-by: NLarry Finger <Larry.Finger@lwfinger.net>
Cc: Stable <stable@vger.kernel.org>    [3.3]
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Fixed some checkpatch warnings in mgmt.c.
Signed-off-by: NCristian Chilipirea <cristian.chilipirea@gmail.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

The dot1q ethertype number (0x8100) is embedded in the code, although
it is already defined in included headers.
Signed-off-by: NEldad Zack <eldad@fogrefinery.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Use more common code for ERTM and streaming mode segmentation and
transmission, and begin using skb control block data for delaying
extended or enhanced header generation until just before the packet is
transmitted.  This code is also better suited for resegmentation,
which is needed when L2CAP links are reconfigured after an AMP channel
move.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Reviewed-by: NUlisses Furquim <ulisses@profusion.mobi>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Remove magic number unneeded calculation since
hlen = L2CAP_HDR_SIZE + L2CAP_PSMLEN_SIZE
Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

hlen has a fixed size of L2CAP_HDR_SIZE, use this instead.
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

The Bluetooth Low Energy support so far was disabled by default via
a module parameter. With this change the module parameter will be removed
and Low Energy is enabled by default.
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

No one is using hci_le_ltk_neg_reply() in bluetooth subsystem.
Signed-off-by: NSyam Sidhardhan <s.syam@samsung.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Host sleep is activated using already configured host sleep
parameters in suspend handler and it is cancelled in resume
handler.
Signed-off-by: NAmitkumar Karwar <akarwar@marvell.com>
Signed-off-by: NBing Zhao <bzhao@marvell.com>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

hlen - L2CAP_HDR_SIZE = 0, so we don't need to add them in the
calculation.
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>
Signed-off-by: NJohan Hedberg <johan.hedberg@intel.com>

In this API, we were using sizeof operator for an array
given as function argument, which is invalid.
However this API is not used anywhere.
Signed-off-by: NSyam Sidhardhan <s.syam@samsung.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Adding Code Aurora Forum copyright information due to significant
additions of code.
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

L2CAP sockets contain a pointer to l2cap_chan that needs to be
reference counted in order to prevent a possible dangling pointer when
the channel is freed.

There were a few other cases where an l2cap_chan pointer on the stack
was dereferenced after a call to l2cap_chan_del. Those pointers are
also now reference counted.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

l2cap_get_chan_by_ident was not used, but didn't generate a compiler
warning because it was an inline function.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Structure members used by ERTM or streaming mode need to be
initialized when an ERTM or streaming mode link is configured.  Some
duplicate code is also eliminated by moving in to the ERTM init
function.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Moving these functions simplifies future patches by eliminating
forward declarations, makes future patches easier to review, and
better preserves 'git blame' information.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

These values are now in the nested l2cap_ctrl struct.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Instead of using modular division, the offset can be calculated using
only addition and subtraction.  The previous calculation did not work
as intended and was more difficult to understand, involving unsigned
integer underflow and a check for a negative value where one was not
possible.
Signed-off-by: NMat Martineau <mathewm@codeaurora.org>
Acked-by: NMarcel Holtmann <marcel@holtmann.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

Currently debugfs commands "hscfgcmd" and "gpiogap" are provided
for host sleep configuration. But if user doesn't configure host
sleep parameters using these commands, host sleep activation is
failed during suspend (support for suspend and resume handlers is
added in next patch).

Default host sleep configuration is done during driver initialisation
in this patch.
Signed-off-by: NAmitkumar Karwar <akarwar@marvell.com>
Signed-off-by: NBing Zhao <bzhao@marvell.com>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

User-space pass the remote device address type to kernel through
struct sockaddr_l2 what makes the advertising useless. This patch
removes all advertising cache code.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

In order to establish a LE connection we need the address type
information. User-space already pass this information to kernel
through struct sockaddr_l2.

This patch adds the dst_type parameter to l2cap_chan_connect so we
are able to pass the address type info from user-space down to
hci_conn layer.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This patch adds the dst_type parameter to hci_connect function.
Instead of searching the address type in advertising cache, we
use the dst_type parameter to establish LE connections.

The dst_type is ignored for BR/EDR connection establishment.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This patch moves the helper function bdaddr_to_le to hci_core, so it
can be used in mgmt.c and hci_conn.c.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Since address type macros are not only related to Management
Interface anymore, it makes sense to rename the helper function
mgmt_to_le to bdaddr_to_le.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This patch adds the address type info to struct sockaddr_l2 so
user-space can inform the remote device address type required
to establish LE connections.

Soon, instead of looking the advertising cache up to discover the
address type, we'll use this address type info to establish LE
connections.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Since address type macros are not only related to Management
Interface anymore, it makes sense to rename the helper function
link_to_mgmt to link_to_bdaddr.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This patch moves address type macros to bluetooth.h since they will be
used by management interface and Bluetooth socket interface. It also
replaces the macro prefix MGMT_ADDR_ by BDADDR_.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Acked-by: NJohan Hedberg <johan.hedberg@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Signed-off-by: NAndrei Emeltchenko <andrei.emeltchenko@intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This patch removes unneeded variable assignments in hci_connect.
'sec_level' is already assigned to BT_SECURITY_LOW in hci_le_connect
and 'pending_sec_level' and 'auth_type' are assigned right after
if statement.
Signed-off-by: NAndre Guedes <andre.guedes@openbossa.org>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

As most LE devices leave advertising mode when they enter the connected
state, we may want to "pass" that connection to other users.

The first user will be the pairing procedure, the connection is
established without an associated socket, after the pairing is
complete, userspace may want to discover via GATT what services the
newly bonded device has.

If userspace establishes the connection while the timeout still
hasn't expired, the connection will be re-used.
Signed-off-by: NVinicius Costa Gomes <vinicius.gomes@openbossa.org>
Tested-by: NJoão Paulo Rechi Vita <jprvita@openbossa.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

The cid or psm and the source address might not be enough to uniquely
identify a global channel, especially when the source address is our
own.

For instance, when trying to communicate with two LE devices in master
mode, data received from the both devices is sent to the same socket.

Fix this by taking the destination address into account when choosing
the socket.
Signed-off-by: NIdo Yariv <ido@wizery.com>
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Tested-by: NJoão Paulo Rechi Vita <jprvita@openbossa.org>
Signed-off-by: NGustavo Padovan <gustavo@padovan.org>

We allocate memory with kzalloc() so there is no need to call
memset(..., 0, ...) or similar.
Signed-off-by: NDavid Herrmann <dh.herrmann@googlemail.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

We currently initialize locks, lists, works, etc. in hci_register_dev()
(hci_alloc_dev() was added later) which is bogus because an hdev is in an
invalid state if it is not registered.
This patch moves all memory initialization to hci_alloc_dev(). Device
registering and registration of sub-modules is still left in
hci_register_dev() as it belongs there.

The benefit is (despite cleaning up the code-base) we can now always be
sure that an hdev is a valid object and can be locked and worked on even
though it may not be registered.

This patch also reorders the initialization to be easier to understand.
First the memory is initialized, then all generic structures and as last
step the sub-init functions are called. This guarantees that all
dependencies are initialized in the right order and makes it also easier
to find a specific line. We previously initialized it in the same order as
the "struct hci_dev" is declared which seems pretty random.
Signed-off-by: NDavid Herrmann <dh.herrmann@googlemail.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>