Fix the retransmit timer start logic to account for remote busy state
in L2CAP BR/EDR flow control.
The previous implementation would start the retransmit timer when there
are outstanding unacknowledged I-frames, but it didn't check if the
remote side is in a busy state (RNR received). Starting the timer while
the remote is busy is incorrect since the remote cannot process I-frames
until it sends an RR to clear the busy condition.
Add a check for L2CAP_FLAG_REMOTE_BUSY before starting the retransmit
timer to ensure the timer is only started when the remote side is ready
to receive I-frames.
This prevents unnecessary timer expirations and retransmissions when the
remote peer has signaled it cannot accept additional I-frames.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Refactor the I-frame retransmission mechanism to improve clarity and
correctness in L2CAP BR/EDR flow control.
The previous implementation used multiple flags (L2CAP_FLAG_PDU_RETRANS
and L2CAP_FLAG_REQ_SEQ_UPDATED) to track retransmission state, which
made the logic complex and error-prone. This change introduces a
per-window retransmit flag and dedicated helper functions to manage
retransmission of all unacknowledged I-frames.
Key changes:
- Replace L2CAP_FLAG_PDU_RETRANS with L2CAP_FLAG_RET_I_FRAME for
timeout-triggered retransmission of the first unacked I-frame
- Replace L2CAP_FLAG_REQ_SEQ_UPDATED with L2CAP_FLAG_RET_I_FRAMES for
retransmission of all unacked I-frames
- Add `retransmit` flag to `bt_l2cap_br_window` structure to track
which I-frames need retransmission
- Add `l2cap_br_retransmit_i_frames()` to mark all outstanding I-frames
for retransmission
- Add `l2cap_br_stop_retransmit_i_frames()` to clear retransmission
state
- Add `l2cap_br_get_ret_win()` to retrieve the next I-frame marked
for retransmission
- Update `l2cap_br_ret_fc_data_pull()` to handle both single I-frame
timeout retransmission and bulk retransmission separately
- Fix window memory management to only free newly allocated windows on
error, not retransmitted ones
- Update S-frame handlers (RR, REJ, RNR) to use new retransmission
helpers instead of setting flags directly
- Update comment for L2CAP_FLAG_RET_I_FRAME to clarify it handles
timeout retransmission
These changes make the retransmission logic more explicit and easier
to maintain while ensuring proper flow control behavior.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
There is a corner case that the S-frame (RR) needs to be sent. And
there is an I-frame (and only one I-frame) is in pending. But the
I-frame is pending for waiting for ack instead of sending. In this
case, the S-frame (RR) will not be preformed.
The previous implementation incorrectly checked for pending data
before sending S-frames, which could prevent timely acknowledgments.
Fix the S-frame sending logic to ensure proper flow control in L2CAP
BR/EDR channels.
Simplify the S-frame sending mechanism by removing the dedicated
S-frame buffer allocation and transmission path. Instead, rely on
the existing data ready mechanism to trigger S-frame transmission
when needed.
Key changes:
- Remove S-frame specific macros and flags for identifying S-frames
- Change `l2cap_br_send_s_frame()` to only set the pending flag
instead of allocating and sending buffers directly
- Remove error handling paths that would disconnect on S-frame send
failures since S-frames are now queued through the normal path
- Simplify `l2cap_br_get_next_sdu()` and `l2cap_br_ret_fc_data_pull()`
by removing S-frame specific handling
- Update `bt_l2cap_br_chan_recv_complete()` to always return 0
These changes ensure S-frames are sent promptly for flow control
regardless of pending I-frames in the transmission queue.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Fix buffer pool allocation size for BR/EDR L2CAP non-basic modes by
using the appropriate macro that accounts for SDU segmentation
overhead.
The previous implementation used BT_L2CAP_BUF_SIZE() which is designed
for basic mode and does not account for the additional overhead
required when segmenting SDUs in retransmission or flow control modes
(I-frame control field, FCS, and SDU length field).
- Replace BT_L2CAP_BUF_SIZE() with BT_L2CAP_RT_FC_MAX_SDU_BUF_SIZE()
- Ensures sufficient buffer space for segmented I-frames
- Aligns buffer allocation with recent MPS and MTU calculation fixes
This prevents potential buffer overruns when transmitting segmented
SDUs in non-basic modes.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Simplify the PDU length calculation in get_pdu_len() by using the
configured MPS value directly without manual adjustments for overhead.
The previous implementation attempted to manually calculate available
payload space by subtracting header/tail sizes and then adding back
the SDU length field size, which was unnecessarily complex and
error-prone.
- Remove manual calculation of actual_mps with header/tail overhead
- Use br_chan->tx.mps directly as the MPS limit
- Remove special handling for start segments (SDU length field
adjustment)
- Simplify logic: return pdu_len if within MPS, otherwise return MPS
The MPS value already represents the maximum PDU size that can be
transmitted, so no additional adjustments are needed. This aligns with
the recent fixes to MPS validation and MTU calculation.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Fix incorrect MPS validation for segmented I-frames by extracting the
SDU length field before checking payload size against MPS.
When an I-frame has SAR (Segmentation and Reassembly) set to START,
it contains a 2-byte SDU length field that precedes the actual payload.
The previous implementation validated buf->len against MPS before
removing this header, causing false positives when the SDU length field
pushed the total size over MPS.
- Extract SAR field from control header for all L2CAP modes
- Initialize SAR to UNSEG by default
- Pull SDU length (2 bytes) when SAR indicates START segment
- Perform MPS validation after SDU length extraction
- Move net_buf_simple_save/restore to encompass SDU length handling
This ensures MPS validation checks only the actual payload size,
preventing spurious disconnections for legitimate segmented transfers.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Fix incorrect RX MTU calculation when retransmission/flow control modes
are enabled. The previous implementation did not account for I-frame
overhead (control field and FCS), which could cause MTU to exceed the
maximum I-frame payload.
- Add l2cap_br_get_rx_mtu() helper to calculate proper RX MTU
- Discount extended control field size (4 bytes) when enhanced
retransmission or streaming mode is enabled
- Discount standard control field size (2 bytes) for basic mode
- Always discount FCS size (2 bytes) as it may be required by peer
- Add assertion to validate MTU meets minimum requirements
- Update l2cap_br_check_chan_config() to use new MTU calculation
This ensures the negotiated MTU fits within the I-frame payload limits
and prevents buffer overruns in non-basic L2CAP modes.
Signed-off-by: Lyle Zhu <lyle.zhu@nxp.com>
Add a new field to the `sntp_time` struct which is an estimate of the
delay between the SNTP server sending the response and the SNTP client
(the application) receiving it. This information can then be used by
the caller to correct for the systemic bias introduced by non-zero
network latencies.
Includes internal variable naming to make it clearer which timestamps
are which.
Signed-off-by: Jordan Yates <jordan@embeint.com>
Change ctr_drb random implementation to use PSA Crypto API.
Since name convention is very different and PSA abstracts the
algorithm used to generate CSPRNG, file and Kconfig options were
changed. Current symbols were deprecated and just select the new
one.
Signed-off-by: Flavio Ceolin <flavio@hubblenetwork.com>
Currently, the CID for dynamic channels are not cleared on reuse.
l2cap_chan_alloc_cid() sees the non-zero CID and skips allocation.
If another dynamic channel (e.g. EATT) has claimed that CID in the
meantime, the reused CID collides and the remote rejects with
BT_L2CAP_LE_ERR_SCID_IN_USE.
This commit adds clearing of bt_l2cap_le_chan.rx.cid to
`l2cap_chan_rx_init()`, to match the pattern for how state is cleared
for bt_l2cap_le_chan.tx in `l2cap_chan_tx_init()`. As a consequence of
this, the function calling the CID-allocating function on the server
side (`l2cap_chan_add()`) is moved after the init functions.
Signed-off-by: Håvard Reierstad <haavard.reierstad@nordicsemi.no>
Replace the manual linked list (using _next pointer) for RFCOMM
servers with sys_slist_t. This enables use of slist utility functions
like sys_slist_find_and_remove(), which is needed to support server
unregistration in a subsequent commit.
Signed-off-by: Kai Cheng <chengkai@xiaomi.com>
Move the avdtp_tx_raise() call from avdtp_tx_processor() into
avdtp_tx_frags() to ensure the tx is not raised if there is no idle
net_buf.
The issue is: the `avdtp_tx_raise` is called always to trigger the
worker to run again, it will keeps occupying the cpu if there is pending
tx, but there is no idle net_buf. After moving the calling to
`avdtp_tx_frags`, it is only called when there is still idle net_buf.
After the previous avdtp sending net_buf is released, the `avdtp_tx_cb`
is called, and the `avdtp_tx_raise` will be triggered again.
Signed-off-by: Mark Wang <yichang.wang@nxp.com>
Fix the qsort comparison function that was hitting an integer underflow.
Some video sources such as video-sw-generator have UINT32_MAX for maximum
frame interval (arbitrarily slow), which made the comparison underflow,
which in turn lead to frame interval being listed in the wrong order,
breaking Windows support.
Fixes#101259.
Signed-off-by: Josuah Demangeon <josuah.demangeon@nordicsemi.no>
Reduce the log level to debug for a number of internal status logs within
the USB device_next stack.
Signed-off-by: Henrik Brix Andersen <henrik@brixandersen.dk>
Move AVRCP SDP record registration and AVCTP server register from
bt_avrcp_init() to callback registration functions to align with
other Bluetooth Classic profiles.
Signed-off-by: Make Shi <make.shi@nxp.com>
Adds a Kconfig option to detect if checksum offloading
is supported by any driver and one to activate it.
Main benefit is that `net_if_need_calc_rx_checksum()` and
`net_if_need_calc_tx_checksum()` are now inline and therefore the
compiler can optimize more, when checksum offloading is not supported.
Signed-off-by: Fin Maaß <f.maass@vogl-electronic.com>
Add calls to sys_trace_idle_exit before leaving idle state
to track CPU load.
Extend CPU_LOAD to CPU_AARCH32_CORTEX_R and CPU_AARCH32_CORTEX_A, thus
we can support CPU_LOAD for all CPU_CORTEX.
Signed-off-by: Jisheng Zhang <jszhang@kernel.org>
Remove the unconditional `conn_mgr_if_used` call on data reception.
Receiving data ends up being a poor metric for "interface in use" in
many situations. This is especially the case when connected to networks
with broadcast packets that are not necessarily relevant for us, such as
ARP discovery requests.
Signed-off-by: Jordan Yates <jordan@embeint.com>
Commit adds checking that the PB-ADV bearer has been
closed before opening the PB-Remote link for NPPI
interface. NPPI interface callbacks are NULL if
PB-ADV server bearer is still open.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
Inside the init function, it was only checked if the
entropy device is ready and if it is not ready a error code
is returned. This is useless as the return codes of SYS_INIT()
functions are not used or saved, in difference to device inits.
Signed-off-by: Fin Maaß <f.maass@vogl-electronic.com>
When comparing time sources for `CONFIG_SNTP_UNCERTAINTY` to validate
the server responses, `dest_ts_us` and `orig_ts_us` need to be
constructed from the same time sources and have the same offsets applied
in order for math operations to be valid.
Fixes#105101
Signed-off-by: Jordan Yates <jordan@embeint.com>
Add an early check whether FAT fs is already mounted similar to
littlefs mount command. This fixes a null pointer access if the mount
command is used twice.
Fixes#104777
Signed-off-by: Jonas Spinner <jonas.spinner@burkert.com>
When USE_SWITCH=y, the thread struct is now mostly degenerate. Only
the two words for ICI/IT state tracking are required. Eliminate all
the extra fields when not needed and save a bunch of SRAM.
Note a handful of spots in coredump/debug that need a location for the
new stack pointer (stored as the switch handle now) are also updated.
Signed-off-by: Andy Ross <andyross@google.com>
The SWO backend's panic() handler was a no-op. When LOG_PANIC() is
called (e.g., before a bootloader jumps to an app, or during a fatal
error), the logging subsystem flushes software buffers to the ITM
FIFO, but the TPIU hardware pipeline still has bytes in-flight on
the SWO pin. The debug probe loses these final messages.
Spin on ITM_TCR.BUSY until the formatter is idle, then busy-wait
for the TPIU output stage to clock remaining bytes out at the
configured SWO baud rate.
Signed-off-by: Jay Beavers <jay@tolttechnologies.com>
When routing packets, check if the destination address matches an
on-link prefix on any interface before using the default router.
This ensures packets destined for on-link prefixes are not
incorrectly sent to the default router.
Signed-off-by: Jani Hirsimäki <jani.hirsimaki@nordicsemi.no>
Logging a message in PRE_KERNEL with CONFIG_LOG_THREAD_ID_PREFIX=y
triggers a assert in k_current_get() because the current thread is not
yet available.
Signed-off-by: Tim Pambor <tim.pambor@codewrights.de>
Resume from cli->block.number instead of block 0,
so BLOCK_START matches the server's current block.
Blob client does not drop server due to this anymore.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
Accept WAITING_FOR_CHUNK phase on resume
when server responds SUCCESS
(server's XFER_START idempotency path).
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
Add a dedicated Kconfig option for this previously hidden and
automatically enabled Mbed TLS configuration.
Make the relevant key exchange Kconfig options depend on it.
Enable it explicitly where needed.
Fix the condition for defining MBEDTLS_X509_USE_C (based on
mbedtls_config.h).
Check for CONFIG_MBEDTLS_X509_CRT_PARSE_C instead of
MBEDTLS_X509_CRT_PARSE_C now that we have a Kconfig option for it.
Signed-off-by: Tomi Fontanilles <tomi.fontanilles@nordicsemi.no>
Add function to get an instance pointer by index.
This works the same as bt_tbs_client_get_by_ccid except that
it uses the index instead of the CCID.
Signed-off-by: Emil Gydesen <emil.gydesen@nordicsemi.no>
Support for using v3 of the FUSE library instead of v2 was added in
b86e52bd6f
As time has been passing more distros are removing v2 support, and at
this point probably all include the v3 in their package repos.
So, let's switch the default to v3.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
When calling bt_avrcp_ct_register_notification,
it sets ct->ct_notify[event_id].cb = cb.
In process_register_notification_rsp, it clears the callback
by setting ct->ct_notify[event_id].cb = NULL. However, when
the connection is disconnected, the callback
ct->ct_notify[event_id].cb is not set to NULL , causing
subsequent notification registrations to fail immediately with
an error because the callback pointer is still active.
Similarly, this issue also exists in the AVRCP TG.
Signed-off-by: Lu Jia <jialu@xiaomi.com>
Refactored the Stream End Point (SEP) callback mechanism to use a
unified bt_avdtp_sep_ops structure instead of individual function
pointers. This change consolidates the connected, disconnected, and
media_data_cb callbacks into a single operations structure. The stream
established callback should be called when the stream l2cap is
connected not the open cmd is finished.
Signed-off-by: Mark Wang <yichang.wang@nxp.com>
Do not add a newline when flushing the buffer to the host terminal.
The caller may not want these newlines to be added, and the host
terminal can handle line wrapping on its own.
Signed-off-by: Joakim Andersson <joerchan@gmail.com>
Add buffer size configuration for posix log backend.
This allows to increase the buffer size for the posix console, which is
needed to avoid message flushed in the middle of ansi escape sequences.
Signed-off-by: Joakim Andersson <joerchan@gmail.com>
Adds compile-time option for overriding UAC2 Feedback endpoint to allow
usage with non-class-compliant Windows UAC2 driver.
Signed-off-by: James Bennion-Pedley <James.Bennion-Pedley@thinksmartbox.com>
Remove the k_is_user_context call from the Z_LOG_LEVEL_ALL_CHECK macro.
Add the Z_LOG_LEVEL_ALL_CHECK_BREAK macro, which safely checks whether
logging is performed in user context before checking the dynamic log level
that requires access to kernel structure.
Signed-off-by: Adrian Warecki <adrian.warecki@intel.com>
Add missing memory barriers after branching on k_is_user_context() to
prevent reordering possible of privileged memory access.
Signed-off-by: Adrian Warecki <adrian.warecki@intel.com>
Remove the unnecessary CONFIG_USERSPACE checks around k_is_user_context
calls. This function is inline and already performs the config check
internally.
Signed-off-by: Adrian Warecki <adrian.warecki@intel.com>
Add WEP security support in Wi-Fi mgmt ops.
Need to enable Kconfig CONFIG_WIFI_NM_WPA_SUPPLICANT_WEP.
Signed-off-by: Kapil Bhatt <kapil.bhatt@nordicsemi.no>
Only set state to ACTIVE after RESUME succeeds, if it fails,
keep SUSPENDED STATE and attach an error code.
Signed-off-by: Zhaoxiang Jin <Zhaoxiang.Jin_1@nxp.com>
