Version 1.14.8 was not buildable with kirkstone.
The patches have been ported to a new branch on our fork of
ModemManager.
Part of yocto upgrade from dunfell to kirkstone
This commit holds also the following patch:
Same as the upstream fix for building dtc with gcc 10.
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
Whereas an additional file needed to be patched too to get rid of
this multiple definition compile error.
Part of yocto upgrade from dunfell to kirkstone
Signed-off-by: Marc Mattmueller <marc.mattmueller@netmodule.com>
Deploying directly the DEPLOY_DIR_IMAGE bypasses the shared state
mechanism implemented in the deploy class, leading to the image not
being properly deployed when building entirely from sstate.
BugzID: 80331
As explained in the comment, the start of this service may include some
power cycles of the modem and also the startup of ModemManager. All of
this may take longer than the default 90 seconds in some special cases.
This should anyway not be a problem since this service is self resilient
and can handle errors during startup.
BugzID: 80178
As for the kernel, we don't need to build them for every MACHINE and
will save some build time when building them only once per architecture.
BugzID: 80015
As explained in the comments, in yocto the kernel is expected to be
built for each MACHINE and not for each architecture as we do.
Therefor all tools related to kernel must also be adapted.
BugzID: 80015
These dependencies lead to a partial rebuild of the kernel when
switching machine, even where they are actually using the exact same
kernel sources and config.
BugzID: 80015
Instead of having one kernel recipe with two hashes, it makes more sense
and is easier to maintain two recipes with the same name but different
versions. The shared part is in an include file.
Each machine then chooses which version it wants to use with
PREFERRED_VERSION variable.
BugzID: 79987
It happens that a soft reset is failing and blocking the chip in a state
where it always times out. The only wait to get out of this is to hard
reset the bluetooth chip.
BugzID: 77171
Detecting if the mac address must be inverted means doing two resets of
the chip in a short period when the mac must be inverted.
These two resets is sometimes letting the BT chip in a state where it
does not answer for more than a minute.
Also fixes getopts string: options -w and -b don't have arguments.
BugzID: 77171
settin mac address of bt and wlan is now triggered in two separated
services.
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
added several enhancements to the mac-address-set script:
- splittet wlan and bt address setting (use -w / -b options)
- fixed upper and lower case problem with vcu1 addresses
- fixed hexadecimal support with vcu2 addresses
- added testify support (incl. log_test function)
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
adding a unit test script which uses testify framework to test the
mac-address-set script for vuc1 and vcu2.
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
if a device is never getting ready on startup sequence, the wait_on()
function runs forever. with this patch it aborts after 10 seconds.
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
If NetworkManager starts before, it may change the mac address for
scanning and store the chip mac address. When finaly using a static mac
address (for an established connection or AP) it will restore the chip
address instead of our address.
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
After an overall refactoring this script now does:
* Wait on devices being ready before changing the address
* Handle errors when setting wlan or bluetooth mac address
* Handle different MAC address schemes
* Handle different behaviors when setting wlan mac:
On HW21, the address must be set while the interface is down, but on
HW23 it must be set while the interface is up and if it is down, no
error is thrown, so the modified address is checked.
* Handle different behaviors while setting BT mac:
On HW21, the address must be given with the bytes in reverse order.
We therefore first try with the right order and check the address and
retry with the reversed address if it failed.
BugzID: 77171
Signed-off-by: Patrick Zysset <patrick.zysset@netmodule.com>
hw23: Force wpa-supplicant and hostapd versions
meta-netmodule-wlan provides modified versions of these tools, but they
seem to not work properly on HW23. Since this HW is anyway not supported
in meta-netmodule-wlan, we force the usage of latest stable version.
BugzID: 77173
Co-authored-by: Patrick Zysset <patrick.zysset@netmodule.com>
Reviewed-on: https://git.netmodule.intranet/yoctoproject/meta-netmodule-bsp/pulls/111
Co-Authored-By: Alexandre Bard <alexandre.bard@netmodule.com>
Co-Committed-By: Alexandre Bard <alexandre.bard@netmodule.com>
those files reside now in the repository itself, thus the recipe
needed to be adapted
BugzID: 75350
Signed-off-by: Marc Mattmueller <marc.mattmueller@netmodule.com>
Alexandre Bard
Marc Mattmueller
Mr Jenkins
Patrick Zysset
Lucien Mueller