conn was never set in struct flush_data passed to flush_cb.
Change-Id: I2a16479bf7660a96f705d1194245da02ee6eeff5
Signed-off-by: Szymon Janc <szymon.janc@tieto.com>
Add BT_GATT_CEP and BT_GATT_LONG_DESCRIPTOR so which can be used by
a task to declare long attributes.
Change-Id: I37a0bb1851b60f3cdfa7e2f00ade70d5c344abbf
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This introduce a new callback called flush to bt_gatt_attr to indicate
when and attribute can cache data, the data is automatically flushed if
regular write request happens and in case of prepare write it checks if
flush is supported otherwise return an error, then upon receiving an
execute write it perform flush over the entire database and respond
with result:
< ACL Data TX: Handle 42 flags 0x00 dlen 6
ATT: Execute Write Request (0x18) len 1
Flags: Immediately write all pending values (0x01)
> ACL Data RX: Handle 42 flags 0x02 dlen 5
ATT: Execute Write Response (0x19) len 0
Change-Id: I2c47f962858b8a741bf40509a648fe2cd6755691
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
When a connection is made, track its role.
Only peripherial/slave role may trigger update of L2CAP connection
parameters.
Change-Id: I72244581650248c7b020c9e5673b73ce46e7399d
Signed-off-by: Arkadiusz Lichwa <arkadiusz.lichwa@tieto.com>
Call attribute write callback taking into account the signature length
so only the actual data is used.
Note: It still don't support validating the signature because that depend
on storing CRSK key.
Change-Id: I93973e2997f57e5715080ef56ad85e0eaf11d0fe
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This reduces the amount of code necessary to check minimum size and
to generate error responses.
Note that not every response can be generate using this method since in
some cases the handle must be set to indicate what attribute has failed.
Change-Id: I0dde7461d97e451407935d5ec3cbfedf6542d5e6
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
The spec says that CCC shall only be persistent accross connection for
bonded devices:
'The Client Characteristic Configuration descriptor value shall be
persistent across connections for bonded devices. The Client Characteristic
Configuration descriptor value shall be set to the default value at each
connection with non-bonded devices'.
To handle this now each configuration has a valid field indicating if the
configuration is valid.
Change-Id: Id74ac54f5e23a7a0b286f90dbc9af4e9ee966dd4
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This fix following compilation error if CONFIG_BLUETOOTH_DEBUG_SMP
is defined:
In file included from net/bluetooth/smp.c:44:0:
net/bluetooth/smp.c: In function 'bt_smp_irk_matches':
net/bluetooth/smp.c:681:31: error: 'val' undeclared
(first use in this function)
BT_DBG("IRK %s bdaddr %s", h(val, 16), bt_addr_str(addr));
Change-Id: I3350620f1f7dc929b949b4483178df66dde55198
Signed-off-by: Szymon Janc <szymon.janc@tieto.com>
This was causing output such as bellow:
bt: bt_gatt_attr_write_ccc: No space to store CCC cfgbt: bt_gatt_attr_write_ccc: No space to store CCC cfgDisconnected
Change-Id: I8aff67b9c4b407e8a2da939c27bea05a597dcf67
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This patch adds callbacks to notify app about connection and disconnection.
Change-Id: If8091254d929fe53d12b153e6f216223a8913f2d
Signed-off-by: Mariusz Skamra <mariusz.skamra@tieto.com>
This checks if connection encrypt flag is set if attribute requires
encryption and in case the connection is already encrypted return
success.
Change-Id: I8af4e9786b63a21f478b3aa9f6df448d7f5760c5
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Respond with an error if attribute read/write return an error.
Change-Id: I9c2f98ac4b73bca5d3714339a47290ca9ae2391b
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
With the introduction of the Number of Completed Packets event
handling this FIFO & fiber aren't anymore strictly HCI command
completion related. Rename them therefore to something that more
appropriately describes their purpose.
Change-Id: Ia53dea8b5ec0098459f23918a5483075757d9202
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
The Number of Completed Packets HCI event belongs to the same category
as the Command Complete/Status events in that it is likely to unlock a
waiter in one of the other fibers. Events which may affect waiters in
this way need to be processed without delay and through a separate
FIFO & fiber to avoid deadlocks.
Change-Id: Ibc2ee2b4f72e2534a8579ad1c36591867e19e6df
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
The PACK_STRUCT macro will be removed soon and __packed is what all
code should use. Convert all the usages in Bluetooth code.
Change-Id: Id5c724566565f67fa8167b61398c508194ece89b
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
We need to make sure we have more event buffers than outgoing ACL
buffers. Currently we allocate 7 + 7 ACL in/out buffers leaving only 7
left for commands/events. Increase the buffer count by two to make
sure we've got 8 command/event buffers.
Change-Id: I4c54aa07019ad9ef262aeecabf151edd60f91b9e
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
When reading a characteristic attribute which contain a 128 bits UUID the
size of the elements were counting only the UUID alone not other fields
that are part of the value causing invalid responses to be sent.
Change-Id: I7474531f586ba19788eb9e0ed59555c5d7672176
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
This checks attribute permission returning an error in case they don't
match.
Change-Id: I56a85ad594cb1ac535b03cdbceb43af84dc6604f
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
If CONFIG_BLUETOOTH_DEBUG_HCI_CORE is enabled it causes the following
error:
hci_core.c:514:12: error: 'struct bt_hci_evt_encrypt_key_refresh_complete' has no member named 'encrypt'
evt->encrypt);
^
Change-Id: I1983653eba40b9aea6b19572b39a285883f00450
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
We can do centralized PDU length checks and more compact code size
with the help of a handler table where the index maps to the SMP op
code.
Change-Id: Iba69ecc14f452411dfb6b4db7c7a5fd19ba3e51f
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Fixes memory leak when buffer is allocated before CCC check, the best
fix is to move CCC check forward.
Change-Id: Ib0492f629a339b576cf811adce47b2f6ab101c33
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
For now we don't support key shortening, so set the minimum value
appropriately to 16 bytes. Once key shortening is implemented this
value can be lowered again.
Change-Id: I172d0bf6d04ba4c5eaa7bbf8ae2accfa373eebb5
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Devices using LE Privacy are identified by their Identity Address
whenever we've paired and received their IRK. We should therefore try
to look up matching IRKs for advertising responses.
Change-Id: I304108da699ff2506e6cc50bd6c5c053a0db024b
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Use correct address to string translation.
Change-Id: Ic766147fd6016894c0f60ec57059984887653a15
Signed-off-by: Andrei Emeltchenko <andrei.emeltchenko@intel.com>
This event comes instead of Encrypt Change when we re-encrypt an
already encrypted connection.
Change-Id: I7fe9545e6c58392a34d72f3dd3962c475ebb7871
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
When a connection occurs it may have happened with an RPA. If so,
check if there's a matching IRK available and update the connection
address with the Identity Address (stored together with the IRK).
Change-Id: I3d995a35171b514ec1902fce617e8e7d6770e785
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Add an API to bt_keys for looking up IRKs based on a given RPA.
Change-Id: If9c0569c8d619f9cb2b6174d77c25dd5358db863
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Add support for the SMP ah() crypto function which is used to match up
IRKs with RPAs.
Change-Id: Iaea8397ea401df800259dfab2ce97187aa5e2df6
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
To avoid spending unnecessary time looking up the same keys multiple
times simply store a pointer to them from the connection context.
Change-Id: I94b550870c863237348a2cab7ba056be2a7f4bc5
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
If we're not actively performing pairing we should ignore any
encryption changes. If we don't do this the code might e.g.
incorrectly clear out existing keys.
Change-Id: Iac49a09abd7386b773e325b9bc939d6b6c741c92
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Add support for handing the Identity Info and Identity Address Info
SMP PDUs.
Change-Id: Ie0ed57e2c1819a1619b9eaa340ed22666c81eea8
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Add support to store IRKs through the bt_keys API.
Change-Id: I01f713f8ad3657605a8bc2738db349d25bb89649
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
We'll need to store many other keys besides slave LTKs. Refactor the
key handling code so that new types can be easily added.
Change-Id: I768db9cb29e8e73e87713e59f7b80af80399f20d
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
In preparation of adding more key support and to not bloat hci_core.c
prepare a new .c file for the key management code.
Change-Id: I5129a7a3a5d495a299f75e4e669253446a276142
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
The input paramter is bt_addr_le_t so the function should be named
accordingly.
Change-Id: I7b796020d6e43ae1aa76a261cbfbaf7cba163029
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
bt_gatt_notify can be used to send ATT notifications:
< ACL Data TX: Handle 3585 flags 0x00 dlen 8
ATT: Handle Value Notification (0x1b) len 3
Handle: 0x0010
Data: 64
Change-Id: I908f10f9e60dc7cef221e4b27c3b6e49e7b0cfc9
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
BT_GATT_CCC uses _bt_gatt_ccc internal struct to handle peer
configurations and special helpers to read and write to minimize the
code necessary to handle CCC in the task.
Note this not yet include handling of connection and disconnection but
the idea is that the core would automatically disable the configurations
while disconnected and re-enable once peer connect back.
Change-Id: I89325b8a074766b9fd2423085565df669f7275e1
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Invalid peer address was being passed since the conn pointer was not
initialized properly.
Change-Id: Ib64584f1eb8bc534389bd4fa057c58dd00ffbd6d
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
To implement CCC properly it is required to know the remote address since
it is per client, futhermore bt_gatt_notify requires an address where to
send the notification.
Change-Id: I9d9fdf72d668831bed76e2aef06816af3620f4b8
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Follow the same style as bt_conn_lookup_handle (i.e. drop the 'by').
Change-Id: I6a31104fc8beac13c357088e46a91bd713fcfdd4
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
To take multiple different ways of looking up connections rename the
existing too generic bt_conn_lookup to bt_conn_lookup_handle.
Change-Id: I1c4965ad76738fa08ccca91b28cda552c9a88373
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
bt_conn_lookup_by_addr looks up for existing BLE connection by address.
Change-Id: If24516511b7d07cd99bbe095a737b38b0adaabad
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Move the usage reports into the function that does the analysis, and
also include a percentage of the usage. To make the results more
relevant give them in terms of the "real" stack size, i.e. the stack
space minus the CSS size in the beginning.
Change-Id: Ibc0418de396e140839c9a810196fbf16df533df5
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
512 is a more accurate description than 1024 of the stack usage impact
of enabling debug logs.
Change-Id: I4b79b4eb9724ba624ff546973a1126c62382d1aa
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
To not have to always do the __noinit, debug log correction and 4-byte
alignement declaration create some helper macros for defining the
stacks.
Change-Id: I2de5068d7044af3f348ed01281498a777d04c97d
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
The CCS always starts at a 4-byte aligned boundary, so if the stack
given to fiber_start doesn't begin in such a place there will be some
unused bytes in the beginning and the 'real' stack will start at a
different offset. Take this into account when calculating the unused
stack portion.
Change-Id: I6c36d0354f4d014ad256b5580b76282d7d4384ce
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
In case there are changes to the way the stacks are defined it's safer
to use sizeof() rather than rely on some specific size define.
Change-Id: Ic9ae090c0dbc0d88995a8883f9b4876be5958f1d
Signed-off-by: Johan Hedberg <johan.hedberg@intel.com>
Szymon Janc