* Disable ext2/ext3 (as ext4 can be used to mount them).
* Enable AUFS/Squshfs support
* Misc improvements (e.g. ARM optimized AES code, disable
drivers/frameworks unlikely to be used)
Jan prefers to store the fully extended .config file so we will
do it. One advantage is to more easily see what has changed between
two kernel versions.
cp ../build.sysmobts/tmp/work/sysmobts_v2-poky-linux-gnueabi/linux-sysmocom/3.10.80+gitAUTOINC+3122273386-r43/git/.config ./recipes-bsp/linux/files/sysmobts-v2/defconfig
cp ../build.sysmobts/tmp/work/sysmocom_odu-poky-linux-gnueabi/linux-sysmocom/3.2.63+gitAUTOINC+3ed283bea6-r41/git/.config ./recipes-bsp/linux/files/sysmocom-odu/defconfig
cp ../build.sysmobts/tmp/work/sysmocom_bsc-poky-linux/linux-sysmocom/3.10.80+gitAUTOINC+3122273386-r43/git/.config ./recipes-bsp/linux/files/sysmocom-bsc/defconfig
I think the last device we sold was in 2011 and I am not aware
if anyone has used it for the last couple of years. Remove it
to lower the complexity of the builds.
* add interface for connection to Inmarsat
* remove 'br0' and work directly with 'eth2' instead, to avoid bridges
in vlans of bridges
* bridge the video LAN also into the backbone (for later break-out)
linux-sysmocom_git.bb is used for the BTS, BSC and IDU. Let's
really use 3.10.76 (from 3.10.50 just rebased and not tested
on the device yet). This should fix the Yocto-Master build
tonight.
9610e792ffde7fb2599d19ca6d2a03230f2043c6 can not be found
in the linux-3.10.76 branch. Just avoid this branch game.
ERROR: Fetcher failure: Unable to find revision 9610e792ffde7fb2599d19ca6d2a03230f2043c6 in branch linux-3.10.76 even from upstream
ERROR: Function failed: Fetcher failure for URL: 'git://git.sysmocom.de/sysmo-bts/linux.git;protocol=git;branch=linux-3.10.76'. Unable to fetch URL from any source.
ERROR: Logfile of failure stored in: /home/builder/jenkins/workspace/Yocto-Master/label/OE/build/tmp/work/sysmobts_v2-poky-linux-gnueabi/linux-sysmocom/3.10.76+gitAUTOINC+9610e792ff-r43/temp/log.do_fetch.16350
NOTE: recipe linux-sysmocom-3.10.76+gitAUTOINC+9610e792ff-r43: task do_fetch: Failed
ERROR: Task 126 (/home/builder/jenkins/workspace/Yocto-Master/label/OE/meta-sysmocom-bsp/recipes-bsp/linux/linux-sysmocom_git.bb, do_fetch) failed with exit code '1
It seems like the kernel/package/upgrade/... proceess doesn't deal well
with the fact that some options used to be modules, which are now
built-in. Let's keep it as modules for the time being.
this enables conntrack events, ipset, ebtables, tc
scheduler/classifiers/ematches and much more.
it also disables the ATM code for which I don't think we have any use
case.
Create the /run/network directory for all of our systems. The
proper way with tmpfiles.d ends with a depedency mess that has
no point in being debugged.
My local build hasn't completed yet but we need to rush. Let the
CI test this change and we will see.
Related: SYS#1514
It looks like on my tests with the sysmobts I was saved by
busybox-ifplogd executing ifup again. I couldn't figure out
the dependencies between the LSB networking service and the
tmpfiles.d interaction. So instead of winning a beauty contest
we will need to create the directory by hand.
Cause: SYS#1514
Revert "ifupdown: Package the new file as well"
This reverts commit 15dc869cda.
Revert "ifupdown: Try the workdir for the files we copied"
This reverts commit 1ef790672e.
Revert "ifupdown: Create /run/network through a tmpfiles.d"
This reverts commit e0c7f3c621.
This way the ifupdown application will not only work
for the IDU/ODU but the BTS, BSC and any other machine
we want to build.
(untested OE modification)
To be able to talk to the equipment we need to set a route
after the link is up. Assume that no other (e.g. 3G) equipment
is using the same IPv4 address range.
Message-ID: 55279904.9020802@sysmocom.de
As pointed out by Harald the PR is lower than the package
that is already on the device. Bump it to 6 then.
Current:
opkg list_installed | grep ntp
ntp - 4.2.8p1-r5.0
ntp-tickadj - 4.2.8p1-r5.0
ntpdate - 4.2.8p1-r5.0
Import ntp 4.2.8 from meta-networking. Our version has some weird
dependency issues that create build breakages and more.
Remove the pps-tools dependency compared to meta-networking and
leave the ntp.conf as doneby Harald (e.g. drift on /etc and not
in /var...). Now I have not verified the upgrade path but I sadly
merged testing to stable when the ntp changes were comitted on
monday. This means I can't easily test the upgrade.
This new version of ntp has quite a number of changes, e.g.
certain utilities requiring perl, and some programs are now in
/usr/sbin rather than /usr/bin.
I edited the config through make menuconfig and then used
savedefconfig but the diff from our config to a minimal
defconfig is too big. So I enabled the temp sensors and i2c
by hand.
Fixes: SYS#1148
For the V-ODU we want to have vlan support as well. This requires
to use the debian ifupdown which means we need a script that creates
/run/network early enough.
On eth1 we should also use a network from the 172.16 block rather than
the 10.23 that we used on the sysmoBSC
Also, we now use 172.16.4.0/23 natively on the bridged onboard network,
this is a temporary measure so existing JBs will still connect and get
an IP address before their software is updated to use VLAN6. After they
(and all APs, BTSs, etc.) have migrated, the 172.16.4.0/23 can be
removed.
For systemd and the Debian ifupdown application we need to
have the /run/network directory. The debian init file is
creating it on the fly and we should do the same for now.
* needed to copy over the oe-core kernel-module classes
and modify them because there was no other way to use them and
name the packages kernel-module-backports
Related: SYS#850
This means that currently only Yocto Dora and Yocto Master are
being targeted. The amount of maintenance to get it to work on
Edison again is too high and it is better to focus on a systemd
only world. Remove the symlinks for yocto-edison.
A quick look and some notes from henning pointed me to wget
which is in the default layers, the sysmobts-2050 which is now
handle dby sysmobts and some older notes that are unlikely
to be addressed.
Okay "make modules" will lead to the kernel.release file
being re-created. So kernel.bbclass of Yocto master needs
to be fixed.
This reverts commit 3cdbc49cbb.
Somehow the do_shared_workdir tries to access a file that
should be created by do_configure but isn't present then.
Add these items to see what happens during the build.
We want to use backported modules this means we need to make
some thing a module for it to work. Take Daniel's .config file,
run make savedefconfig and update our defconfig.
Related: SYS#850
Go back to have our own do_configure. The kernel.bbclass would
execute "oldnoconfig" where we simply executed "oldconfig". Keep
on using our old way (for the legacy of it). E.g. with the question
how we ended up having SMP enabled in the BSC build this might
be the answer.
OE-Core seems to unpack the kernel sources into another
directory now (e.g. to share the sources with builds for
modules). Use the kernel.bbclass kernel_do_configure and
copy the .config to the build directory.
The kernel is now built in a dedicated build directory.
We actually only use the PMIC (TPS65910) RTC in the hardware,
and not the AM335x RTC. Having a driver for a hardware that doesn't
receive actual backup voltage doesn't make sense.
The odu-gpiotool is a small script created by roh to selectively reset
individual peripherals of the system by the administrator. It originates
from ticket https://projects.sysmocom.de/ortelius/issues/655
We use the new sysfs file to obtain the board version number and
configure the port swapping accordingly. If the board_version sysfs
attribute is not present, we default to v2.
We also integrate resetting the hub chip from within this tool right
befeore writing the new register values via I2C.
* split out rtl-nic firmware from the big
linux-firmware package
[hfreyther: For whatever reasons the NIC works without
additional firmware as well. I couldn't see in the driver
source when that is the case and which features are missing]
Fixes: SYS#732
[hfreyther: I removed the MACHINE_ESSENTIAL_EXTRA_RDEPENDS as applications
that use the symlinks should rdepend on the package]
[hfreyther: Changed mismatch of tabs/spaces]
We want to enable the watchdog in the bootloader already and
for this to be failsafe we need to have a kernel that can boot
and get ethernet without any additional modules. Now systemd
would not be able to trigger the watchdog and the system would
reboot within 60 seconds. I am not sure how the kernel upgrade
will go and we need to try this.
Related: SYS#374
We have changed the strategy to support multiple BTS with a single
build but don't remove support from files that are used with Edison.
Linux 3.10 and systemd were not used with edison though. So we can
remove it. As the sysmobts-2050 extended the sysmobts-v2 build most
of the -sysmobts-2050 changes are not needed anyway.
This variable doesn't exist on dora and will not break things.
Fixes:
ERROR: Please ensure module rtfifo is listed in KERNEL_MODULE_PROBECONF since module_conf_rtfifo is set
The sysmobts-2050 machine does inherit from the generic bts so
we do not need an exact copy of the configuration file.
The watchdog is not used in dora or later images as this functionality
is integrated into the systemd.
Related: SYS#559
If the year is smaller than < 2014 an initial date will be set. This
has been manually verified by copying ntpdate to ntpdate.off and making
sure that the VPN is coming up. Due the short SSL lifetime of CACert
we will need to update the script once in a while. But our VPN should
now always come up.
systemd doesn't look in /etc/rcS.d for init scripts so we never
start networking unless we are using ifplugd. We need to check
how this behves on the sysmobts and sysmobsc in case the network
is being started twice.
* to be able to mount vfat formatted
usbsticks we need the nls subsystem
* providing nls as modules is a bad idea because
you need to add them at image creation
* better include them into the kernel,
like the kernel for bts and bts-2050 does
We will change the MTD partitioning for our devices soon. The
kernel needs to honor the cmdline given by the kernel for that.
The rootfs will move a few pages to the back with newer bootloaders.
The name/structure of the NAT configuration was changed. Catch up
and put the new names into the old defconfig too. It is an ugly hack
to have it work with v3.2 and v3.10 kernels. The sysmobsc has been
fixed independently.
* use seperate defconfig, because of option changed from 2.6.39 to 3.10
* use python magic to switch between linux and linux-sysmocom
* add sysmocom-bsc to compatible machines for linux-sysmocom
* add new confiugration parameters for syslinux
This currently requires the meta-ti layer but we will make sure
that nothing from it will be required in the end. The kernel is
based on linux-3.2 as this is the "default" TI kernel that supports
the SGX (not that we care)
This is the latest stable kernel. I had hoped some crashes to
go away with it but it turns out to be a compiler issue that
we backported a patch for separately.
* The Linux 3.2.X kernel has some funny issues with GCC 4.8 in the
fs/namei.c look up code. It is easier to move to 3.10 than to fully
debug the issue.
* disable the mISDN loop patch for v3.10
* use the hack to make v3.2 compilable for edison and v3.10 for dora
this might help us in some installations where there are multiple
back-hauls and we want to base routing decisions not only based
on the destination IP.
The build differs in that it doesn't use a getty on ttyS0. In the fuutre,
a modified kernel serial driver as well as a 2050-specific utility
for managing the supervisor processor will follow.
The previous change modified the defconfig to not include the legacy
pseudo terminal support. Increase the version for the sysmoBTS to have
a kernel without old terminal files.
There is no need to have the various pty files clobber the /dev
directory when they are never used. Do not force a PR upgrade
right now as we don't have a good upgreade strategy for the v1
and the BSC config.
systemd is using the depedencies of the LSB section and according to
the debian documentation all applications accessing /var should depend
on the localfs. This makes the watchdog start properly on systemd systems.
Some of our target applications require ext3/ext4 support for mounting
an MMC card. While the sysmobts-v1 hardware does not have an external
MMC card slot it is nice to have the same image available for this
hardware.
Install the sysmobts firmware headers on top of the Osmocom toolchain.
This way the SDK can be used to re-compile both the OpenBSC and the
osmo-bts software.
Make sure the -dev package of the femtobts-api has not runtime
dependencies to a library that does not exist.
The kernel has been rebased to target 3.2.40 with various bug
fixes in the upstream work. The sysmobts v2 config has been
changed to allocate space for two more UARTs. These are required
for the multi-trx configuration and they make a warning go away.
Holger Hans Peter Freyther
Jan Luebbe
Henning Heinold
Harald Welte
Daniel Willmann
Holger Hans Peter Freyther