Frantisek Sumsal [Thu, 21 Oct 2021 16:06:36 +0000 (18:06 +0200)]
test: loosen sandbox restrictions for integration tests as well
Otherwise we miss quite a lot of coverage (mainly from logind,
hostnamed, networkd, and possibly others), since they can't write their
reports with `ProtectSystem=strict`.
Frantisek Sumsal [Thu, 21 Oct 2021 15:34:43 +0000 (17:34 +0200)]
test: loosen certain sandbox restrictions when collecting coverage
With `ProtectSystem=strict` gcov is unable to write the *.gcda files
with collected coverage. Let's add a yet another switch to make such
restriction less strict to make gcov happy.
This addresses following errors:
```
...
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/binfmt-util.c.gcda:Cannot open
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/base-filesystem.c.gcda:Cannot open
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/barrier.c.gcda:Cannot open
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/ask-password-api.c.gcda:Cannot open
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/apparmor-util.c.gcda:Cannot open
systemd-networkd[272469]: profiling:/systemd-meson-build/src/shared/libsystemd-shared-249.a.p/acpi-fpdt.c.gcda:Cannot open
...
```
Frantisek Sumsal [Thu, 21 Oct 2021 13:59:57 +0000 (15:59 +0200)]
test: wait a bit for the given PID to die if it's still alive
When playing around with the coverage-enabled build I kept hitting
an issue where dnsmasq failed to start because the previous instance was
still shutting down. This should, hopefully, help to mitigate that.
Luca Boccassi [Wed, 20 Oct 2021 21:10:56 +0000 (22:10 +0100)]
CI: disable code coverage in GH Action
It is now ran on the nightly CentOS build, so that it can cover
integration tests too, and not just unit tests. It's nightly as
it considerably increases the integration test runtime, so it's
not appropriate for all PRs.
loop-util: call loop_device_make_internal() at the right place
The whole reason loop_device_make_internal() exists (as opposed to just
loop_device_make()) is to avoid mangling the loop flags value/call
getenv twice. Hence let's actually call it when we already mangled the
flags value.
Yu Watanabe [Thu, 14 Oct 2021 20:57:30 +0000 (05:57 +0900)]
network: dhcp6pd: set default metric 256 for delegated prefix
When Assign= in [DHCPv6PrefixDelegation] is enabled, then the kernel
will create the prefix route for the assigned address with metric 256.
When Assign= is disabled, then the kernel will create the route with
metric 1024.
For the default value, we should choose a smaller value (higher priority)
than 1024, as the unreachable routes for delegated prefix will be
configured with 1024.
Yu Watanabe [Thu, 14 Oct 2021 21:00:22 +0000 (06:00 +0900)]
network: dhcp6: explicitly specify metric for unreachable route
For static IPv6 routes without metric is specified, then we use 1024.
But such an adjustment is not performed to dynamic routes. So, let's
specify the metric explicitly.
Otherwise, configured routes will be handled as foreign.
Yu Watanabe [Thu, 14 Oct 2021 19:21:21 +0000 (04:21 +0900)]
network: dhcp6pd: set lifetime to routes for assigned prefixes
Note that the kernel (at least 5.14.11) seems not to support lifetime
for IPv6 unreachable routes. The lifetime for routes of the type will be
handled by sd-event's timer event source.
So, we cannot confirm the lifetime with 'ip route' command.
Yu Watanabe [Fri, 15 Oct 2021 01:07:57 +0000 (10:07 +0900)]
network: route: drop kernel version check for route expiration
Even in newer kernel version, it seems that some route type does not
support expiration, e.g. IPv4 route or unreachable route. Let's use
timer event source for such routes.
Vincent Bernat [Mon, 18 Oct 2021 18:58:43 +0000 (20:58 +0200)]
utmp: remove /dev from line
utmp(5) says `ut_line` is the device name minus the leading "/dev/". Therefore,
remove it. Without that, when using UtmpMode=user, we get `/dev/tty` in the
output of `last`/`w`.
meson: use partial_dependency() to get include directory
Getting the variable directly from pkg-config (without
adding the sysroot prefix) is prone to host contamination
when building in sysroots as the compiler starts looking for the
headers on the host in addition to the sysroot.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
homework: move allocation/destruction into outer/generic scope
Previously in most cases we'd allocate the HomeSetup context object
in generic code in homework.c. But for some cases we allocated them
instead inside the specific code in homework-{cifs,directory,luks}.c
Let's clean that up, and systematically allocate it in the outer
"entrypoint" calls in homework.c instead of the inner ones.
This doesn't change much in behaviour (i.e. it just means when something
fails we'll now clean it up one stack frame further up). But it will
allow is to more easily work with the context objects, since we'll have
them around in all stack frames.
Yu Watanabe [Wed, 13 Oct 2021 18:10:31 +0000 (03:10 +0900)]
network: dhcp6: manage assigned downstream prefixes by using Hashmap
When a system has thousands of downstream interfaces, previously the
total cost of finding free subnet ID was O(n^2), where n is the number
of downstream interfaces.
This makes assigned prefixes are managed by Manager with Hashmap. So,
the cost becomes O(n log n).
Daan De Meyer [Tue, 12 Oct 2021 10:20:59 +0000 (11:20 +0100)]
journal: Expand rotate log messages in journald
Make sure we always log when we rotate journals and always do so at
least at INFO log level. Doing so we make sure there's always a clear
reason available explaining why we rotated a journal.
Daan De Meyer [Tue, 12 Oct 2021 10:00:00 +0000 (11:00 +0100)]
journal: Add log level argument to journal_file_rotate_suggested()
When journald is rotating a file, we'd like to log the reason at
LOG_INFO or higher instead of LOG_DEBUG. For journalctl --header,
logging the reason at a level higher than LOG_DEBUG doesn't really
make sense. To accomodate both use cases, make the log level used
by journal_file_rotate_suggested() configurable.
docs: mark spdx headers with the default license to website files
Those were added in b41a3f66c97e3d861faed04b727daf929383b827 without
an explicit license, so they are under the default license. Some files
already got a header previously, so this only touches the remaining.
The same should be done for docs/_data/extra_pages.json, but it's json, and
json doesn't allow comments.
It wasn't picked up automatically because it's not in
test/fuzz/fuzz-fido-id-desc/. But looking at the contents, it doesn't seem to
be in the expected input format either.
gitattributes: introduce and use "generated" attribute
I want to mark some files to be ignored for licensing purposes,
e.g. output from fuzzers and other samples. By using the gitattribute
machinery for this we don't need to design a custom protocol:
$ git check-attr generated test/test-sysusers/unhappy-*
test/test-sysusers/unhappy-1.expected-err: generated: set
test/test-sysusers/unhappy-1.input: generated: unspecified
test/test-sysusers/unhappy-2.expected-err: generated: set
test/test-sysusers/unhappy-2.input: generated: unspecified
test/test-sysusers/unhappy-3.expected-err: generated: set
test/test-sysusers/unhappy-3.input: generated: unspecified
tests: add spdx license header to test unit/link/network/conf files
Those are all consumed by our parser, so they all support comments.
I was considering whether they should have a license header at all,
but in the end I decided to add it because those files are often created
by copying parts of real unit files. And if the real ones have a license,
then those might as well. It's easier to add it than to make an exception.
gitattributes: mark fuzz inputs that are pure binary as such
We also have a bunch of files that have some bytes and a lot
of text, like the journal export format. For those, it is still quite
useful when the tools try to diff them, so let's not mark those.
Jan Janssen [Thu, 30 Sep 2021 10:11:56 +0000 (12:11 +0200)]
sd-boot: Get rid of uefi_call_wrapper
The uefi_call_wrapper exists to convert to the right calling convention
and presumably predates compilers that can do so natively. The only
architecture where this is even needed is x86_64.
But because we are building with GNU_EFI_USE_MS_ABI defined, the
EFIAPI macro tells the compiler to use the right calling convention
for EFI functions. Our shim callback (which is called by EFI itself)
already relies on this.
This also adds a safety check to make se we are compiling with
GNU_EFI_USE_MS_ABI defined and also adds it to the compiler args
unconditionally. It is only used with x86_64 anyways, so it should
be fine to do so. EFI_FUNCTION_WRAPPER is unused in gnu-efi, so
it is dropped.
Jan Janssen [Thu, 30 Sep 2021 09:17:22 +0000 (11:17 +0200)]
sd-boot: Convert VOID -> void
We are already using void in several places and having a screaming
typedef for void feels pointless. There are also CONST, IN, OUT
and OPTIONAL which we aren't using either.
This leaves missing_efi.h to keep it in line with how they are
defined in gnu-efi and/or the specs.
Jan Janssen [Thu, 30 Sep 2021 07:46:55 +0000 (09:46 +0200)]
sd-boot: Rework print_status()
A little helper function and some unusual formatting makes this
whole thing a lot easier on the eyes. Also, right-aligning the
properties for better readability at runtime.
This function is a destructor, hence it should be named like one.
(We usually use xyz_free() for a destructor that frees the object passed
itself. xyz_unref() we typically use for destructors that are similar,
but ref counted. xyz_done() usually is used for destructors which free
the members of an object, but not the object itself – to allow stack
allocation of objects. We don't strictly follow this, but it's good to
stick to rules wherever we can.)
These set of functions are constructors for an object called HomeSetup,
which has a destructor home_setup_undo(), hence to be reasonably
symmetric, let's call it home_setup*() too, instead of using a new verb
"prepare" for its name.
homework: reuse home_validate_update_luks() at one more place
We have the same code at two places, let's reuse it. Given the more
generic scope let's rename the function home_get_state() since it
retrieve the current setup state of the LUKS logic.
homed,shutdown: call valgrind magic after LOOP_GET_STATUS64
valgrind doesn't understand LOOP_GET_STATUS64. We already work around
this in various placed, via VALGRIND_MAKE_MEM_DEFINE(), but we forgot
three places. Let's fix that.