Changes ERR to WARN message when relay buffers
run out, as this is not really an error but a
consequence of high traffic.
Signed-off-by: Omkar Kulkarni <omkar.kulkarni@nordicsemi.no>
Don't replace destination of segmented message when acknowledgement is
received from a friend node, otherwise, when segments are retransmitted,
the destination will be used and the friend node will treat it as a new
segmented message.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
Heartbeat publication sent callback can be used
as a watch mechanism to determine whether
the mesh stack is still capable of sending messages or not.
Signed-off-by: Aleksandr Khromykh <aleksandr.khromykh@nordicsemi.no>
According to Kconfig guidelines, boolean prompts must not start with
"Enable...". The following command has been used to automate the changes
in this patch:
sed -i "s/bool \"[Ee]nables\? \(\w\)/bool \"\U\1/g" **/Kconfig*
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
The extended advertiser would fail to build due to a missing kconfig
option dependency when relay was disabled.
Fixes#43172.
Signed-off-by: Trond Einar Snekvik <Trond.Einar.Snekvik@nordicsemi.no>
When local node poweroff and power on again, will
receive iv update but within 96hours.
When 96h after, due to cache, we can't process this
beacon any more.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
According Mesh Profile 3.10.6 IV Index Recovery procedure
Upon receiving and successfully authenticating a Secure Network
beacon for a primary subnet whose IV Index is 1 or more higher
than the current known IV Index, the node shall set its current
IV Index and its current IV Update procedure state from the
values in this Secure Network beacon.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
Since `CONFIG_BT_MESH_PB_GATT` represent to pb-gatt-srv.
We use `CONFIG_BT_MESH_PB_GATT_COMMON` to represent common.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
When Proxy advertising or PB-GATT Advertising Enabled and use a same
advertising sets.
As `adv_start` will call multi HCI Command will cause syswork_q yield.
At same time, if another thread(BT RX) all `schedule_send` will cause
unable send mesh message, because `ADV_FLAG_ACTIVE` was be set, but
`ADV_FLAG_PROXY` not set currentlly.
Add `ADV_FLAG_SCHEDULE_PENDING` indicate mesh buf has been pendings
but not scheduled, so when proxy advertising enable, let's take again,
as we can't break or terminated `adv_start`, so we must waiting proxy
advertising enabled.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
As use for simple message with no-segment send or receive.
This will be useful for ram-resource-constrained device.
such as bbc-microbit-v1.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
Before this change, enabling CONFIG_BT_SETTINGS and calling
settings_load(), but delaying / not calling bt_enable would trigger an
assertion error due to a timeout. The fault is that the settings load
handler for the Bluetooth host assumes bt_enable has already been called
and sends HCI commands to the controller. This times out if HCI is not
running.
The fix is to skip loading Bluetooth settings before bt_enable. The doc
is updated to guide the user on how to enable Bluetooth after settings
have been loaded before.
Signed-off-by: Aleksander Wasaznik <aleksander.wasaznik@nordicsemi.no>
After PR: https://github.com/zephyrproject-rtos/zephyr/pull/38296
If we receive the second recovery after 192 hours after the first
recovery, we may still not be able to perform any recovery, even
if the IV Index differs by 42, because the `ivi_was_recovered` flag
will only be cleared during the iv update.
But still May lose iv update, as we did the first iv recovery.
This patch by checking the iv duration, if 192h later, auto clear
`ivi_was_recovered` flag.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
A common pattern here was to take the work item as the subfield of a
containing object. But the contained field is not a k_work, it's a
k_work_delayable.
Things were working only because the work field was first, so the
pointers had the same value. Do things right and fix things to
produce correct code if/when that field ever moves within delayable.
Signed-off-by: Yong Cong Sin <yongcong.sin@gmail.com>
Adds clause so that the LPN must have a established friend
connection for the connection terminate callback to trigger
upon clearing a friendship.
Signed-off-by: Anders Storrø <anders.storro@nordicsemi.no>
Adds state check to the friend_req_sent CB to ensure
that the CB does not alter the LPN state if it is in a
disabled state. This resolves behavioural issue for lpn_set(false).
Signed-off-by: Anders Storrø <anders.storro@nordicsemi.no>
bt_mesh_subnet_find calls a callback for every subnet, and returns the
subnet that got a non-zero return code from the callback. As pointed out
in #41693, the callback should return a boolean, not an int.
Signed-off-by: Trond Einar Snekvik <Trond.Einar.Snekvik@nordicsemi.no>
Adds fix so that a node running with extended adveriser is able to
enable GATT proxy correctly. This fixes a corner case issue where a
device with no other ongoing message sending is unable to advertise
the GATT proxy through
bt_mesh_gatt_proxy_set(BT_MESH_FEATURE_ENABLED) .
Signed-off-by: Anders Storrø <anders.storro@nordicsemi.no>
The signature of bt_mesh_app_key_resolve expresses pointer to a 16 byte
array as a const uint8_t *app_key[16], which is actually an array of 16
byte pointers. The intended type is equivalent to a double pointer to
const uint8_t, but trips up GCC 11, which actually checks this.
Signed-off-by: Trond Einar Snekvik <Trond.Einar.Snekvik@nordicsemi.no>
The crypto.h declaration of the bt_mesh_beacon_auth function declares
the net_id parameter to be a 16 byte array, but the function definition
says 8 bytes. This breaks compilation in GCC 11, which feeds an 8 byte
array into this API, triggering a warning. Change the header declaration
to 8 bytes, which is the right size.
Signed-off-by: Trond Einar Snekvik <Trond.Einar.Snekvik@nordicsemi.no>
Period measurement should be started right after a message is updated.
That is because `struct bt_mesh_send_cb.start` callback is called from
the advertiser thread which is not necessary to be scheduled immeditely
and a user may see some delays between update handler calls.
The publication timer still needs to be scheduled after a message being
actually sent out to the air so that access layer doesn't overflow the
advertiser's buffer. To achieve some specific use cases, where a
published message needs to be retransmitted with 50ms interval, a user
has to switch off network retransmissions to make the legacy advertiser
sleep for the shortest possible time.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
Period measurement should be started right after a message is updated.
That is because `struct bt_mesh_send_cb.start` callback is called from
the advertiser thread which is not necessary to be scheduled immeditely
and a user may see some delays between update handler calls.
The publication timer still needs to be scheduled after a message being
actually sent out to the air so that access layer doesn't overflow the
advertiser's buffer. To achieve some specific use cases, where a
published message needs to be retransmitted with 50ms interval, a user
has to switch off network retransmissions to make the legacy advertiser
sleep for the shortest possible time.
Signed-off-by: Pavel Vasilyev <pavel.vasilyev@nordicsemi.no>
The previous PR was not fully baked, so this PR fixes issues by:
- Adding checks when dereferncing pointers stored in params.
- Clearing ack context if a return argument is NULL to not block and
wait for response.
Signed-off-by: Michał Narajowski <michal.narajowski@codecoup.pl>
- Do not ignore unsolicited status messages to allow for the API to be
- used asynchronously
Signed-off-by: Michał Narajowski <michal.narajowski@codecoup.pl>
- Add optional callback to receive status messages even when using async
API
- Split acked and unacked API
Signed-off-by: Michał Narajowski <michal.narajowski@codecoup.pl>
