Let sysupgrade run a --test upgrade to verify that the image is
compatible before attempting an upgrade.
This fixes an issue where a router can get stuck without network
connectivity when a a remote autoupgrade was triggered.
Fixes#193
X-FIRMWARE-VERSION gets set to the content of the file referenced by UCI setting "autoupdater.settings.version_file"
If "autoupdater.settings.version_file" is unset X-FIRMWARE-VERSION is
omitted
In case a radio exists, but is disabled, the survey request will return
successfully, but without data.
The code could not handle this case, as the newest_element_index would
be negative in this case, leading to json_object_array_get_idx returning
a null-pointer.
Check for a null pointer prior adding the PHY index, avoiding a
null-pointer dereference.
Signed-off-by: David Bauer <mail@david-bauer.net>
If an autoupdater long option (prefixed with '--') are unknown then a segmentation fault occurs.
Example:
root@node:~# autoupdater --xyz
Segmentation fault
[Matthias Schiffer: changed sentinel syntax]
* include bugfix to allow sending unicast packets to neighbours
* FIX crash due to wrong format string
* FIX packet forwarding on mesh devices added via socket
* allow to build statically
* avoid adding mesh interfaces multiple times
This renames batman-adv.ko to batman-adv-legacy.ko. And also moves
scripts to a "-legacy" variant to avoid build errors when both
batman-adv and batman-adv-legacy are selected. Finally, moves the
batman-adv-legacy config to /etc/config/batman-adv-legacy.
This allows to select and install the batman-adv mainline and
batman-adv-legacy simultaneously.
Instead of unconditionally loading this module on boot the gluon_bat0
netifd protocol script will later take care of loading either the batman-adv
or batman-adv-legacy module, depending on the configured routing
algorithm in UCI.
* memory alignment fixes
* fixes for uninitialized memory
* fixes for neighbour detection (removal of neighbours)
* scheduling fixes in taskqueue
* schedule removal of nodes in NUD_PROBE
simple-uci was modifying the metatable of UCI cursor objects, thus changing
the behaviour of *all* UCI cursors, and not only UCI cursors that are
created through simple-uci.
Fortunately, the destination buffer is always big enough, so no buffer
overflow can happen. Use memcpy instead of strncpy to make this explicit
(and slightly more efficient).
Previously, delays were specific for interfaces, but not for the group
address. This behaviour was confusing when the same interface was passed
for multiple group addresses. Fix by keeping track of both groups and
interfaces.