When a reconfiguration of an interface is triggered, previously we
call link_foreignize_config(), which sets all static configurations as
foreign, then later call link_drop_foreign_config(), which drops
unnecessary foreign configurations.
This commit merges these two steps into one, link_drop_unmanaged_config(),
which drops unnecessary static and foreign configurations.
Also, this renames link_drop_managed_configs() to
link_drop_static_config(), as it only drops static configurations.
Note that dynamically aquired configurations are dropped by
link_stop_engines().
Effectively no functional changes, just refactoring and preparation for
later changes.
- convert boolean flag 'force' to LinkReconfigurationFlag enum,
- merge link_reconfigure() and reconfigure_handler_on_bus_method_reload() as
link_reconfigure_full(),
- Rename ReconfigureData -> LinkReconfigurationData,
- make Reconfigure() DBus message wait for reconfiguration being
started before sending reply.
We enumerate interfaces at first, then enumerate other configurations
like addresses and so on. If we are running on a container, previously
we started to configure the enumerated interfaces before enumerating other
configurations.
Let's configure interfaces after all configurations are enumerated.
No functional change, at least now. Preparation for later commits.
But we are planning to extend KeepConfiguration= and also keep
addresses and so on assigned by other dynamic configuration protocol
like DHCPv6 or NDisc.
However, when link_free_engines() is called here, acquired addresses so
on by NDisc will be removed, even if link_stop_engines() handles
restarting networkd or KeepConfiguration= gracefully.
So, let's not free engines here, but free them later in link_free().
It is not necessary to be called here anyway.
This makes networkd process all queued remove requests when a
terminating or restarting signal is received. Otherwise, e.g. DHCPv4
address will not be removed on stop, especially when
KeepConfiguration=no.
Fixes a bug introduced by 85a6f300c1 and
its subsequent commits.
Fixes#34837.
Co-authored-by: Will Fancher <elvishjerricco@gmail.com>
The kernel resets the IPv6 MTU of an interface when its link MTU is changed.
But it seems the operation is asynchronous, and even when we detect that
the link MTU is changed, the IPv6 MTU may not be reset yet.
====
[ 2257.067613] systemd-networkd[447122]: veth99: MTU is changed: 1500 →1600 (min: 68, max: 65535)
[ 2257.067641] systemd-networkd[447122]: Setting '/proc/sys/net/ipv6/conf/veth99/mtu' to '1410'
[ 2257.067711] systemd-networkd[447122]: No change in value '1410', suppressing write
====
As you can see, even if the link MTU is changed to 1600, the IPv6 MTU is
unchanged (in this case, still 1410).
When an interface enters unmanaged state, there are two possibilities:
- no matching .network file found,
- found a matching .network with Unmanaged=yes.
When a matching .network file is found, networkd logs the filename.
Let's also log when no matching .network file is found.
This also slightly adjust the log message when a matching .network file
found.
Closes#34436.
Monitor the sysctl set by networkd for writes, if a sysctl is
overwritten with a different value than the one we set, emit a warning.
Writes are detected with an eBPF program attached as BPF_CGROUP_SYSCTL
which reports the sysctl writes only in net/.
The eBPF program only reports sysctl writes from a different cgroup than networkd.
To do this, it uses the `bpf_current_task_under_cgroup_proto()` helper,
which will be available allowed in BPF_CGROUP_SYSCTL from kernel 6.12[1].
Loading a BPF_CGROUP_SYSCTL program requires the CAP_SYS_ADMIN capability,
so drop it just after the program load, whether it loads successfully or not.
Writes are logged but permitted, in future the functionality can be
extended to also deny writes to managed sysctls.
[1] https://lore.kernel.org/bpf/20240819162805.78235-3-technoboy85@gmail.com/
When the interface is in the failed state, link_getlink_handler_internal()
will do nothing and return zero, thus the interface will not be
reconfigured, especially when the reconfiguration is triggered in
link_enter_failed().
Follow-up for c2eb7753dd.
Otherwise, when an interface gained its carrier, the interface may not
have matching .network file yet, then link_reconfigure_impl() returns
zero, and link_handle_bound_by_list() is skipped.
Fixes#33837.
Even if a timespan specified to IgnoreCarrierLoss= for an interface,
when the carrier of the interface lost, bound interfaces might be bring
down immediately.
Let's also postpone bringing down bound interfaces with the specified
timespan.
Otherwise, writing IPv6 MTU may fail.
This also makes link_set_ipv6_mtu() take log level about rounding IPv6 MTU,
and downgrade the log level from LOG_WARNING -> LOG_INFO, as we usually
use LOG_WARNING for per-interface critical failure.
The linux/ headers include linux/libc-compat.h that makes sure the
linux/ headers won't redeclare symbols already declared by net/if.h, but
glibc's net/if.h doesn't do that, so if the include order is reversed
we'll end up with a bunch of errors about redeclared stuff:
[3/519] Compiling C object test-network-tables.p/src_network_test-network-tables.c.o
FAILED: test-network-tables.p/src_network_test-network-tables.c.o
cc -Itest-network-tables.p -I. -I.. -Isrc/basic -I../src/basic -Isrc/fundamental -I../src/fundamental -Isrc/systemd -I../src/systemd -I../src/libsystemd/sd-bus -I../src/libsystemd/sd-device -I../src/libsystemd/sd-event -I../src/libsystemd/sd-hwdb -I../src/libsystemd/sd-id128 -I../src/libsystemd/sd-journal -I../src/libsystemd/sd-netlink -I../src/libsystemd/sd-network -I../src/libsystemd/sd-resolve -Isrc/shared -I../src/shared -Isrc/libsystemd-network -I../src/libsystemd-network -Isrc/network -I../src/network -I../src/network/netdev -I../src/network/tc -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -std=gnu11 -O0 -g -Wno-missing-field-initializers -Wno-unused-parameter -Wno-nonnull-compare -Warray-bounds -Warray-bounds=2 -Wdate-time -Wendif-labels -Werror=format=2 -Werror=format-signedness -Werror=implicit-function-declaration -Werror=implicit-int -Werror=incompatible-pointer-types -Werror=int-conversion -Werror=missing-declarations -Werror=missing-prototypes -Werror=overflow -Werror=override-init -Werror=return-type -Werror=shift-count-overflow -Werror=shift-overflow=2 -Werror=strict-flex-arrays -Werror=undef -Wfloat-equal -Wimplicit-fallthrough=5 -Winit-self -Wlogical-op -Wmissing-include-dirs -Wmissing-noreturn -Wnested-externs -Wold-style-definition -Wpointer-arith -Wredundant-decls -Wshadow -Wstrict-aliasing=2 -Wstrict-prototypes -Wsuggest-attribute=noreturn -Wunused-function -Wwrite-strings -Wzero-length-bounds -fdiagnostics-show-option -fno-common -fstack-protector -fstack-protector-strong -fstrict-flex-arrays --param=ssp-buffer-size=4 -Wno-unused-result -Werror=shadow -fno-strict-aliasing -fstrict-flex-arrays=1 -fvisibility=hidden -fno-omit-frame-pointer -include config.h -pthread -DTEST_CODE=1 -MD -MQ test-network-tables.p/src_network_test-network-tables.c.o -MF test-network-tables.p/src_network_test-network-tables.c.o.d -o test-network-tables.p/src_network_test-network-tables.c.o -c ../src/network/test-network-tables.c
In file included from ../src/basic/linux/if_bonding.h:47,
from ../src/network/netdev/bond.h:5,
from ../src/network/test-network-tables.c:3:
../src/basic/linux/if.h:111:41: error: redeclaration of enumerator ‘IFF_UP’
111 | #define IFF_UP IFF_UP
| ^~~~~~
../src/basic/linux/if.h:84:9: note: previous definition of ‘IFF_UP’ with type ‘enum net_device_flags’
84 | IFF_UP = 1<<0, /* sysfs */
| ^~~~~~
../src/basic/linux/if.h:112:41: error: redeclaration of enumerator ‘IFF_BROADCAST’
112 | #define IFF_BROADCAST IFF_BROADCAST
| ^~~~~~~~~~~~~
...
This also drops remaining workarounds from the last time this issue was
brought up (6f270e6bd8) since they shouldn't be needed anymore if the
order of the includes is the "correct" one. I also added a comment to
each affected include when this is inevitably encountered again in the
future.
Resolves: #32160
Otherwise, even if the persistent storage is not ready, the DHCP server
may be started e.g. by unplugging and plugging cable.
Follow-up for 5582b36c38.
This does not change anything for DHCPv4, as a DHCPv4 address is always
requested anyway. However for DHCPv6, the client may not request IA_NA
addresses by UseAddress=no, or even if it is requested, the server may
not provide any IA_NA addresses. Even in such cases, here the check is
for delegated prefixes, hence it is not necessary to check if DHCPv6
IA_NA addresses are configured.
Fixes a bug introduced by 195b83edf8.
Fixes#31349.
As networkd does not configure nexthops, neighbors, or routing policy rules
dynamically, this only changes link_drop_managed_addresses(). Previously
it also removed dynamic addresses, but with this commit, now it only
removes static addresses. Note, link_drop_managed_routes() has removed
only static routes, so this commit only changes its name.
Note. as commented in the code, the dynamic addresses will be removed
on lease expiration or stopping client, and link_drop_managed_xyz() are
called on reconfiguration or so, and at that time, all dynamic
addressing clients are stopped or restarted, so this should not
effective behavior change.
Prompted by https://github.com/systemd/systemd/pull/30085#discussion_r1401534107.
Note, like Reconfigure bus method, even reconfiguration for an interface is
triggered by Reload method, the method only wait for the link enters
configuring state (or unmanaged state if no matching .network file exists).
Users still need to invoke systemd-networkd-wait-online if it is
necessary to wait for the interface enters configured state after Reload
medhod.
Previously, a Route object is owned by a Link object corresponding to the
outgoing interface of the route, and a Route object that does not have
outgoing interface is owned by the Manager object.
However, there were several issues:
- if a route has a nexthop ID, then the corresponding nexthop may be
changed to use another interface, hence the outgoing interface of the
route may be changed.
- if a route requested with MultiPathRoute=, then the link who requests
the route is different from the outgoing interface of the configured
route. So, we need to find routes on other interfaces on reconfiguring
or so.
By this change, the limit of the number of routes per-interface is
tentatively dropped. Let's re-introduce the limit later in a nicer way.
- introduce link_required_operstate_for_online() helper function,
- use recently introduced macros and helper functions,
- unconditionally serialize the minimum and maximum of required
operational state.
Otherwise, e.g. when we enumerate a bridge port first, then the bridge
main interface, then the port cannot be managed by the main interface.
Fixes#30682.
Follow-up for b732606950 and
6706ce2fd2.
If Network.ignore_carrier_loss_set flag is set, then the timeout value
is always used, hence the logic implemented by
b732606950 never worked.
Yu Watanabe
Lennart Poettering
Matteo Croce
Sebastian Pucilowski
Frantisek Sumsal
Luca Boccassi