This fixes the following error:
In function ‘get_composition’,
inlined from ‘cfg_srv_pkt’ at mesh/cfgmod-server.c:801:8:
mesh/cfgmod-server.c:758:9: error: ‘comp’ may be used
uninitialized [-Werror=maybe-uninitialized]
758 | memcpy(buf, comp, len);
| ^~~~~~~~~~~~~~~~~~~~~~
mesh/cfgmod-server.c: In function ‘cfg_srv_pkt’:
mesh/cfgmod-server.c:739:24: note: ‘comp’ was declared here
739 | const uint8_t *comp;
| ^~~~
This aligns the behavior of command line option --io to
add new "auto" value and modify the behavior of "generic"
value:
*auto* - Use first available controller: via MGMT interface
if kernel supports it, otherwise, via raw HCI socket (i.e.,
default to *generic*).
*generic:[hci]<index>* - Use generic HCI io on interface hci<index>
The default value is now *auto*, whereas *generic* is used
only if the specific HCI controller is explicitly specified.
This fixes the out-of-bounds array access in mesh-io-mgmt.c caught
by address sanitizer. Similar fixes were applied earlier to
generic and unit IOs. With this patch, the common code is factored
into a centralized location.
If remote device does not respond to a Prov Link Open request, then the
callbacks do not get established, and attempting to send Failure
messages on the non-existent link rersult in seg fault.
The mesh daemon can process incoming mesh packets on more than one
controller, but if a Provisioning data packet that originated from the
local daemon is received by a different controller, it must be filtered
and disregarded, or it will break the provisioning protocol.
Because the daemon explicitly supports multiple nodes, we need
local Config Clients to be able to see local unprovisioned devices.
This loops the unprovisioned beacon, so that local Provisioning servers
can see it.
Fixes Issue: https://github.com/bluez/bluez/issues/341
This fixes adding mandatory models (config server, remote provisioner)
to a node whose configuration is being loaded from storage:
mesh_model_add() was called with a wrong argument.
Was: mesh_model_add(..., PRIMARY_ELE_IDX, ...);
Correct: mesh_model_add(..., ele->models, ...);
When transitioning from Phase 1 to Phase 2 of a network key and we are
beaconing, we need to halt the beaconing on the old key version, and
begin beaconing on the new key version.
With this change, we start evaluating received Mesh Private Beacons in
addition to the legacy Secure Network Beacons. We also add the ability
to request Tx of Mesh Private Beacons, which are regenerated with new
Random Nonce a minimum of every 0 - 2550 seconds.
This initial server supports only the Mesh Private Beacon and returns
"Not Suppoerted" for Get/Set of Private GATT Proxy and Private Node
Identity beacons.
When a SAR transmission has been completed or cancelled, the recipent
should store the block authentication values for at least 10 seconds
and ignore new segments with the same values during this period. See
MshPRFv1.0.1 section 3.5.3.4.
If we receive a message from one of our Low Power nodes, which is
encrypted using frendship credentials and where the destination is not
us, we must relay that message using master credentials.
See MshPRFv1.0.1 section 3.6.6.2.
If we are in Key Refresh Phase 1, and receive a Secure Network beacon
using the new NetKey and with KR flag set to 0, Phase 2 should be
skipped. See MshPRFv1.0.1 section 3.10.4.1.
If we are in IV update in progress state, and receive a Secure Network
beacon with an IV index equal to last known IV index + 1, and IV update
flag set to 1, it should be ignored. See MshPRFv1.0.1 section 3.10.5.
If this node is a member of a primary subnet and receives a Secure Network
beacon on a secondary subnet with an IV Index greater than the last known
IV Index of the primary subnet, the Secure Network beacon shall be ignored.
See MshPRFv1.0.1 section 3.10.5.
When replying to a HB subscription get message, and the current state of
source or destination fields is zero (which means that HB subscription
is disabled), all fields in the status reply should be zero.
If heartbeat subscription is disabled, all fields should be set to zero
but collected data should be preserved. If HB subscription is enabled,
the collected data should be reset (which includes Min Hops = 0x7f).
HB subscription is disabled by setting any of the following fields to
zero: Source, destination or period log.
HB subscription is enabled by setting all the same fields to valid values.
If a Config Heartbeat Publication Set message is unsuccessfully
processed, the fields in the status reply should be the same as in the
original message. See MshPRFv1.0.1 section 4.4.1.2.15.
1. Re-structures MGMT handling such that it is used to detect kernel
support of the mesh MGMT opcodes and events before selecting between
using MGMT or the legacy raw HCI socket method.
2. Re-structures main() to allow command line to prefer MGMT over HCI or
visa versa, plus optionally pass an explicte controller.
3. Adds mesh-io-mgmt as a transport.
Because provisioning is not speed dependent, Timing on outbound PB-ADV
packets have been modified to be less likely missed by remote controlers
with looser timing capabilities.
If processing failed for a Config Model Subscription Virtual Address
Add/Delete/Overwrite message, the address field in the status reply
should be set to zero. See MshPRFv1.0.1 section 4.4.1.2.8.
It is not stated in the mesh profile that the RFU bits in a Model
Publication Set message have to be zero. In fact, PTS test
MESH/NODE/CFG/MP/BV-01-C is sending that command with non-zero RFU and
expects a reply.
For received CTL mesh messages, the wrong network ID variable was
passed to ctl_received. This patch changes to the correct variable.
Also, changed type of net_key_id argument in ctl_received function
to be consistent wich the whole call chain.
When there is a backlog of mesh packets to be sent, the packet sender
Fix bug where bluetooth-meshd stops sending
When there is a backlog of mesh packets to be sent, the packet sender
incorrectly infers that the tx worker thread is already running
and therefore needn't be invoked. As a result, the mesh daemon will
sometimes stop broadcasting while there are still packets in the queue.
It will not resume broadcasting.
This patch will invoke the tx worker thread correctly in that case.
The logic to send packets at least twice when the transmitter is idle
was slightly modified accordingly, and should behave the same way as
before.
snprintf performs it's own range checking and returns a negative value
if string construction fails. Not checking the return value throws a
warning at compile time on GCC 12 and later. This patch removes
redundent range chacking and checks all snprintf return values.
