NM core uses nm-logging which is entirely configurable at runtime.
Other components use glib-logging, which can also be partly configured
via G_MESSAGES_DEBUG.
It makes sense to have a compile time option to enable some
logging statements that are only useful for heavy debugging.
For glib-logging, this is a way to enable/disable extra logging.
For nm-logging, we could alternatively configure a least log-level
that is enabled at compile time (that way, we could configure to prune all
LOGL_TRACE logging). While that might be useful (too), this gives
an alternative way to disable/enable logging.
Add a configure option --enable-more-logging and a NM_MORE_LOGGING define
for that.
If we don't find this useful after a while, we can simply remove it,
because our logging statements are not part of a "stable" behavior.
(cherry picked from commit 63593a19d8)
Even if asserts are not enabled, still let the compiler see what we would
assert. Otherwise, we get warnings about unused variables or we migth miss
compile errors inside nm_assert().
(cherry picked from commit fe3e1849b7)
NM already has two kinds of assertions:
- g_assert*(), conditionally compiled via #ifndef G_DISABLE_ASSERT
- g_return*(), conditionally compiled via #ifndef G_DISABLE_CHECKS
In theory, one should be able to disable both asserts and NM should
still work correctly (and possibly more efficient). In practice,
hardly anybody is testing such a configuration and it might be broken.
Especially, we don't disable asserts for production builds, both because
of less test coverage and because it might reduce our ability to debug.
Add a new configure option --enable-more-asserts, which defines
NM_MORE_ASSERTS and nm_assert(). This is for expensive asserts,
that -- contrary to the asserts above -- are disabled by default.
This is useful for extended debugging.
(cherry picked from commit 08ecafd2bf)
The Interface held a reference to the manager to listen for the 'available'
signal. Instead of that, let's make the child unaware of the master to
keep the inheritance cleaner.
(cherry picked from commit 7ed2d7a809)
The user-requested reason is not really appropriate and blocks autoreconnect.
We added a new reason precisely for that.
Fixes: 4cb97cf66f
(cherry picked from commit 2bfbf33720)
Fixes build on Ubuntu 12.04.
systemd/src/libsystemd-network/dhcp-network.c: In function '_bind_raw_socket':
systemd/src/libsystemd-network/dhcp-network.c:75:17: error: 'BPF_XOR' undeclared (first use in this function)
systemd/src/libsystemd-network/dhcp-network.c:75:17: note: each undeclared identifier is reported only once for each function it appears in
make[4]: *** [libsystemd_nm_la-dhcp-network.lo] Error 1
(cherry picked from commit 3811a68389)
It seems like a poor default for various downstream toolchains. We can't
anticipate the compiler warnings for future compiler versions and older
ones are prone to false positives. Also, older gdbus-codegen is known
to generate code that triggers compiler warnings.
Let's keep it enabled for maintainer builds and distcheck so that we're
sure a tool chain that builds releases without warnings exists.
(cherry picked from commit 84021454eb)
The Ubuntu 12.04 introspection data don't contain it. However, the default
constructor works just well and even looks a bit more Python-y.
(cherry picked from commit b9b7bb1958)
It yields completely unpredictable results on Ubuntu 12.04 (the global variable
successfully comparing to NULL despite demonstrably not NULL). Possibly a
toolchain bug.
(cherry picked from commit ccb0ca4493)
Add a netlink implementation for reading InfiniBand properties, but fall back to
sysfs when that isn't supported by the kernel.
(cherry picked from commit 5cf226463a)
We need to know whether we can create interfaces of any given
NMDevice subclass or not. So don't rely on just the NMPlatformLink
for that information, because we won't have a platform link for
software devices before we create them.
(cherry picked from commit d4e0a1e8cc)
We can't have devices with duplicate interface names so we might
as well use that for dupe checking instead of the (mostly useless)
UDI.
(cherry picked from commit 81db512997)
==21573== Syscall param mount(type) points to unaddressable byte(s)
==21573== at 0x854B9BA: mount (syscall-template.S:81)
==21573== by 0x158922: main (test-common.c:295)
==21573== Address 0x0 is not stack'd, malloc'd or (recently) free'd
==21573==
{
<insert_a_suppression_name_here>
Memcheck:Param
mount(type)
fun:mount
fun:main
}
Fixes: d6aef9c188
(cherry picked from commit 0d7012faab)
We'd be able to do so for already existing devices, but not for devices that
are added afterwards, since gudev is hardwired not to listen for events from
kernel.
(cherry picked from commit 26aeb12749)
Mount a private sysfs instance. Otherwise gudev sees the devices from the
parent netns as opposed to our netns.
We do, however need a writable /sys/devices subtree for testing the bridge
code. There doesn't seem to be any other way to get a writable subtree of a
read-only filesystem than remounting it with no parameters after the initial
mount. We use this to get a writable sysfs instance and then bindmount it so
that it fits properly in the sysfs hierarchy.
Co-Authored-By: Thomas Haller <thaller@redhat.com>
(cherry picked from commit d6aef9c188)
When compiling with -fexceptions (as we build our RPM), we must
initialize all variables with a cleanup attribute.
Fixes: 29ccb8851c
Fixes: f8a9574f7e
(cherry picked from commit f55a272ade)
Instead of hacky stuff in the Manager, let plugins themselves indicate
which links should be ignored (because they are really child links that
are controlled by a different device that the plugin handles).
(cherry picked from commit 8fa0f4690f)
Ethernet, WiFi, and VLAN used the same implementation for initial address.
Ethernet and WiFi used the same implementation (and duplicated code) for
permanent MAC address, plus they both used ethtool in what should be
generic code, which is better done in the platform.
(cherry picked from commit aba250a7d4)
If the caller requested that a bond be created, don't return success
if there was an existing bridge with the same name.
(cherry picked from commit 29ccb8851c)
Instead of having a bunch of logic in the Manager for determining the
VLAN and Infiniband virtual interface names, move the type-specific
logic into the plugins themselves.
(cherry picked from commit 179d56c73c)
Instead of looping over all plugins and asking each plugin whether it
can handle a link or a connection, have them advertise the link and
connection types they support, and use that when creating new devices.
(cherry picked from commit 71bde20c30)
For existing devices, depending on the order that netlink sends interfaces to
us, the parent may be found after the VLAN interface and not be available when
the VLAN interface is constructed. Instead of failing construction, when a
NMDeviceVlan has no parent keep it unavailable for activation. Then have
the Manager notify existing devices when a new device is found, and let
NMDeviceVlan find the parent later and become available via that mechanism.
This doesn't apply to VLANs created by NM itself, because the kernel requires
a parent ifindex when creating a VLAN device. Thus this fix only applies to
VLANs created outside NetworkManager, or existing when NM starts up.
(cherry picked from commit cd3df12c8f)
beb18050 made this code run for all devices instead of just ethernet
devices, which means any kind of connection gets compared to any
device. But only compatible connections should be considered.
(cherry picked from commit 77d01c9094)
If the queued activation request failed before the device is
finished deactiving the old request, don't start activating
the failed queued request.
(cherry picked from commit fb1d04099f)
The link test was disabled in commit 67ad3fcb5b.
The previous issues are not fixed, but apparently disabling the test doesn't
help to get it fixed.
Re-enable it and if it fails we have a better reason to fix it.
Or maybe it works now (?). Didn't fail for me...
(cherry picked from commit f614ebe6f5)
