Fix Tx Ack FIFO's first index being advanced beyond a
recorded ack_last value in a node_rx when under race
between ll_rx_get() being pre-empted while executing the
`tx_cmplt_get()` and a call to `ll_rx_put()` in an
interrupt service routine.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
ticker_is_initialized() should only return true when the ticker is
running (triggered regularly). Users like nrf_flash_sync_is_required()
depend on this behavior.
When the bluetooth controller driver is closed, ll_deinit() calls
lll_deinit(), which stops the ticker from being triggered. Also
deinitialize the ticker to ensure that ticker_is_initialized() returns
false.
Signed-off-by: Marco Widmer <marco.widmer@bytesatwork.ch>
Add ll_conn_get() return value check for valid connection
context.
Build command:
cmake -GNinja -DBOARD=nrf52833dk/nrf52833
-DEXTRA_CONF_FILE=overlay-all-bt_ll_sw_split.conf
-DDTC_OVERLAY_FILE=boards/nrf52833dk_nrf52833_df.overlay
-DSNIPPET="bt-ll-sw-split" ../../samples/bluetooth/hci_uart
ninja
Before:
Memory region Used Size Region Size %age Used
FLASH: 283716 B 512 KB 54.11%
RAM: 109752 B 128 KB 83.73%
IDT_LIST: 0 GB 32 KB 0.00%
After:
Memory region Used Size Region Size %age Used
FLASH: 284992 B 512 KB 54.36%
RAM: 109752 B 128 KB 83.73%
IDT_LIST: 0 GB 32 KB 0.00%
After (use of `conn != NULL`):
Memory region Used Size Region Size %age Used
FLASH: 285044 B 512 KB 54.37%
RAM: 109752 B 128 KB 83.73%
IDT_LIST: 0 GB 32 KB 0.00%
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Fix missing removal of redundant ISO Rx size in ACL Rx PDU
allocation size.
Related to commit 7c89f1fe9f ("Bluetooth: controller:
Support for separate ISO RX data path").
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Rename the internal BT_CTLR_ISO_TX_BUFFERS to
BT_CTLR_ISO_TX_PDU_BUFFERS correct represent the number
of actual ISO PDU buffers allocated in the Controller.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Introduces the concept of ll_scan_aux_chain, that allows for
keeping track of several advertising chains with a minimal
overhead. This replaces using several scan aux sets each with
an associated ticker IDs
Only one ticker is used for following chains; This is achieved
by having the scheduled chains in a sorted list and restarting
the ticker when the first in the list expires
EVENT_DONE_EXTRA_TYPE_SCAN_AUX now includes a pointer to the lll
structure to identify which chain to flush (to avoid race
conditions)
New implementation is hidden behind a config for now (defaults
to off): BT_CTLR_SCAN_AUX_USE_CHAINS
Other minor changes:
- rx_incomplete node pointer was moved to ll_sync_set
- data_len removed from ll_sync_set (it uses the chains instead)
Signed-off-by: Troels Nilsson <trnn@demant.com>
Share pause/resume functionality by moving resume ticker handling to
ull_iso. Rename interface function to ull_conn_iso_resume_ticker_start,
and provide handles instead of instance pointer.
Signed-off-by: Morten Priess <mtpr@oticon.com>
Add assertion check to catch infinite loop due to incorrect
node_rx release.
If same node_rx is released twice then it can lead to
infinite looping when releaseing link or node_rx buffers.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
The MFIFO holding TX nodes on the ack path back to the host was only
sized to hold the possible number of ACL TX nodes. However, additional
TX nodes are allocated for LLCP and use the same FIFO.
By adding LLCP_TX_CTRL_BUF_COUNT to the size, the FIFO will be able to
hold the worst-case number of TX nodes.
Signed-off-by: Morten Priess <mtpr@oticon.com>
Utilize a code spell-checking tool to scan for and correct spelling errors
in all files within the subsys/bluetooth/controller directory.
Signed-off-by: Pisit Sawangvonganan <pisit@ndrsolution.com>
In ull_disable, it is imperative that the callback is set up before a
second reference counter check, otherwise it may happen that an LLL done
event has already passed when the disable callback and semaphore is
assigned.
This causes the HCI thread to wait until timeout and assert after
ull_ticker_stop_with_mark.
For certain compilers, due to compiler optimizations, it can be seen
from the assembler code that the callback is assigned after the second
reference counter check.
By adding memory barriers, the code correctly reorders code to the
expected sequence.
Signed-off-by: Morten Priess <mtpr@oticon.com>
Add prepare pipeline assertion checks to detect faults
like stopping preempt timeout without prior start. Also,
check if the preemptor param is correct when preemption
is performed. Check that the prepare pipeline is not
infinitely looping.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Generic node_rx_hdr struct contains data not used for done event, so this
pdu specific data is moved to node_rx_pdu to minimize memory usage for
MFIFO_done alloc.
This however calls for a considerable write-through replacing the generic
node_rx_hdr with node_rx_pdu when relevant and needed.
Signed-off-by: Erik Brockhoff <erbr@oticon.com>
Fix MFIFO_DEFINE to reduce FLASH usage by moving the pool
outside the struct that is static initialized.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Fix missing call to ull_chan_reset(), this fixes
uninitialized channel map being used after LL reset.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Fix ull_prepare_dequeue to not skip events when preempt
does not match the event in the head of the prepare queue.
The head of the prepare queue does not match when
ull_prepare_dequeue has put the head of the queue to the
last of the queue when it is calling lll_resume interface.
This happens when there is already an active event which
needs a preempt timeout to be setup and there is another
non-resume event in the pipeline which now becomes the
head. The fix ensure to restore the order of the events
if a preempt timeout was to be setup.
Currently ull_prepare_dequeue loops through all events
before stopping at the original way the queue was. This is
not efficient and is a scope for improvement in future.
This reverts commit 91781306e9 ("Revert "Bluetooth:
Controller: Fix ull_prepare_dequeue for skipped events"").
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Fix number of completed packets generated when BIS HCI ISO
Data packets sent to Controller are in fragments. Use the
ISOAL stored sdu_fragment count to generate the number of
completed packets count.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Fix a clang warning
ull.c:1235: warning: use of bitwise '|' with boolean operands
The result of ({1/0} &&(int)) is either true(1) or false(0),
not the int and therefore bm is not a bitmask, but just true(1)
or false(0) all together.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Fix ULL_HIGH ticker operations count, it is discovered in
new BabbleSim test implementations that mesh tests using
Extended Advertising and Scanning enqueue upto 4 ticker
operations before ULL_LOW ticker_job could process it.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
These change are reverted as part of fixing regression
failure in LL/DDI/SCN/BV-21-C and LL/DDI/SCN/BV-25-C.
This reverts commit 0d54ca8761.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
In certain scenarios, the semaphore sem_ticker_api_cb may be taken
during attempt to complete a synchronous ticker operation such as start
or stop, but is never released via ull_ticker_status_give. This could
happen if ticker temporarily runs out of ticker operation slots for any
ticker client.
The consequence of ULL stuck waiting in semaphore is anything from
allocation assertions to watchdog timeouts.
This commit also sets a timeout on ull_disable calls, which relies on
the disable callback being invoked from 'done'. Invocation of 'done' is
mandatory, and code shall not allow that not to happen, but to avoid
hiding the real cause, the semaphore now has a timeout which causes
assertion in the location the issue occurred.
Signed-off-by: Morten Priess <mtpr@oticon.com>
The separate done memq was introduced to handle the old LLCP
stalling processing of the rx queue; This is no longer an issue with
the new LLCP, so we can remove it
Signed-off-by: Troels Nilsson <trnn@demant.com>
In order to have clean, self-contained HCI headers that do not have any
dependencies towards the Host or any other part of the system (except
types), reorganize the headers in the following way:
- Split out the macros and structs from hci.h to a new hci_types.h
- Merge the existing hci_err.h into the new hci_types.h
Fixes#58214.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
Fix Central ISO assetion failure on being established due to
regression in commit a53d89450a ("Bluetooth: controller:
fixing CIS establish flow for refactored LLCP") that
required changes due to Create CIS LLCP being refactored.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
When possible re-use the already allocated RX node for notifications.
Store (retain) RX node and Link element on RX if NTF could occur.
Pass link element to LLCP (ull_cp_rx()) together with RX node.
New RX node type RETAIN introduced to signal retention
When no RX node is available allocate one and hold off TX on procedures
until such time that a node is available for NTF.
In case waiting for NTF buffer avail is needed, allocate and store TX
node to use for TX once NTF becomes available.
CIS Established (incl. timeout handling) is now handled entirely as a
specific event driven by ull_conn_iso - ie removal of procedure check
of cis->established and cis->expire, as this is doubling mechanism
in the conn_iso context.
Unit test and helpers updated to handle new node type.
Function ull_cp_release_ntf() was used only in unit test, so moved to
helper context.
Updating release_ntf to handle the fact that with piggy-backing in test
context the node used for NTF can be from two different memory pools
Signed-off-by: Erik Brockhoff <erbr@oticon.com>
Fix ull_prepare_dequeue to not skip events when preempt
does not match the event in the head of the prepare queue.
The head of the prepare queue does not match when
ull_prepare_dequeue has put the head of the queue to the
last of the queue when it is calling lll_resume interface.
This happens when there is already an active event which
needs a preempt timeout to be setup and there is another
non-resume event in the pipeline which now becomes the
head. The fix ensure to restore the order of the events
if a preempt timeout was to be setup.
Currently ull_prepare_dequeue loops through all events
before stopping at the original way the queue was. This is
not efficient and is a scope for improvement in future.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Some recent PRs included conditional compilation for
KCONFIG_BT_LL_SW_LLCP_LEGACY, which must be removed.
Signed-off-by: Andries Kruithof <andries.kruithof@nordicsemi.no>
This commit removes the legacy LLCP code including the
Kconfig option
It also updates the babblesim tests, and removes the
tests for the legacy controller
Signed-off-by: Andries Kruithof <andries.kruithof@nordicsemi.no>
To enable Bluetooth controller coexistence feature, there is
implementation of ticker task, which aborts any ongoing radio events
during assertion of the grant pin. This solves the co-existence issue
in the role of the subordinate transceiver.
Signed-off-by: Tomáš Beneš <tomas@dronetag.cz>
Avoid the use of weak functions and call the respective functions only
if vendor-specific data path is supported. The weak dummy functions
will not implement the desired behavior anyway. This change makes it
explicit that these functions need to be implemented by the vendor.
Signed-off-by: Wolfgang Puffitsch <wopu@demant.com>
When possible re-use the already allocated RX node for notifications.
Store (retain) RX node and Link element on RX if NTF could occur.
Pass link element to LLCP (ull_cp_rx()) together with RX node.
New RX node type RETAIN introduced to signal retention
When no RX node is available allocate one and hold off TX on procedures
until such time that a node is available for NTF.
In case waiting for NTF buffer avail is needed, allocate and store TX
node to use for TX once NTF becomes available.
CIS Established (incl. timeout handling) is now handled entirely as a
specific event driven by ull_conn_iso - ie removal of procedure check
of cis->established and cis->expire, as this is doubling mechanism
in the conn_iso context.
Unit test and helpers updated to handle new node type.
Function ull_cp_release_ntf() was used only in unit test, so moved to
helper context.
Updating release_ntf to handle the fact that with piggy-backing in test
context the node used for NTF can be from two different memory pools
Signed-off-by: Erik Brockhoff <erbr@oticon.com>
Define a internal BT_CTLR_ISO_TX_BUFFERS to allocate ISO Tx
buffers used for Broadcast and Connected ISO.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Make Data, ISO, BIS and CIS PDU structure's octet 3 vendor
specific.
This will allow vendors not supporting the octet 3 or
CTEInfo (8 bits) to save 1 octet in their PDU structures.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
New LLCP will process the CIS establishement in the TX path, meaning it
will only process once every ACL connection interval.
If ACL connection interval is significantly longer than the ISO
interval, CIS packets corresponding to the ACL connection interval will
be lost. This may cause latency in audio rendition.
By processing the CIS establishment in the RX path, notification will be
sent immediately, and audio path is enabled fast to reduce latency.
When a CIS fails to establish, the controller shall complete the
procedure by sending a CIS_ESTABLISHED event containing the error.
Signed-off-by: Morten Priess <mtpr@oticon.com>
The `bluetooth/common/log.h` and `bluetooth/common/log.c` files have been
removed. Files that were using them have been updated to use
`zephyr/logging/log.h` instead.
Those replacement have been done consequently:
- `/BT_DBG/LOG_DBG/`
- `/BT_ERR/LOG_ERR/`
- `/BT_WARN/LOG_WRN/`
- `/BT_INFO/LOG_INF/`
- `/BT_HEXDUMP_DBG/LOG_HEXDUMP_DBG/`
- `/BT_DBG_OBJ_ID/LOG_DBG_OBJ_ID/`
Also, some files were relying on the `common/log.h` include to include
`zephyr/bluetooth/hci.h`, in those cases the include of `hci.h` has
been added.
For files that were including `common/log.h` but not using any logs,
the include has been removed and not replaced.
Signed-off-by: Théo Battrel <theo.battrel@nordicsemi.no>
Fix advertising set's is_enabled flag to be reset when
connection complete is dequeued for connectable advertising
started using legacy advertising enable HCI commands. And
reset the is_enabled flag when advertising set terminated
is dequeued for connectable advertising enabled using
Extended Advertising HCI commands.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
Add long control PDU size support, also the implementation
now allocates control PDU size restricted to the supported
biggest control PDU that is enabled in the application
build. Features that are covered based on the increasing
control PDU sizes are ClearText connections, Encrypted
Connections, Connection Parameter Request support and
Create CIS control Procedure (Connected Central ISO)
support.
Signed-off-by: Vinayak Kariappa Chettimada <vich@nordicsemi.no>
