We currently wait for the RemoveMatch() reply, but then ignore what it
actually says. Let's optimize this a bit, and not even ask for an answer
back: just enqueue the RemoveMatch() operation, and do not request not
wait for any answer.
They do the same thing as their synchronous counterparts, but only
enqueue the operation, thus removing synchronization points during
service initialization.
If the callback function is passed as NULL we'll fallback to generic
implementations of the reply handlers, that terminate the connection if
the requested name cannot be acquired, under the assumption that not
being able to acquire the name is a technical problem.
When kdbus was still around we always had two implementations of the
various control calls: one for dbus1 and one for kdbus. Let'sget rid of
this, simplify things, and just merge the wrappers that used to
multiplex this with the implementations.
No change in behaviour, just some merging of functions
Currently, reply callback timeouts are started the instant the method
calls are enqueued, which can be very early on. For example, the Hello()
method call is enqueued right when sd_bus_start() is called, i.e. before
the socket connection and everything is established.
With this change we instead start the method timeout the moment we
actually leave the authentication phase of the connection. This way, the
timeout the kernel applies on socket connecting, and we apply on the
authentication phase no longer runs in parallel to the Hello() method
call, but all three run serially one after the other, which is
definitely a cleaner approach.
Moreover, this makes the "watch bind" feature a lot more useful, as it
allows enqueuing method calls while we are still waiting for inotify
events, without them timeouting until the connection is actually
established, i.e. when the method call actually has a chance of being
actually run.
This is a change of behaviour of course, but I think the new behaviour
is much better than the old one, since we don't race timeouts against
each other anymore...
This adds a "watch-bind" feature to sd-bus connections. If set and the
AF_UNIX socket we are connecting to doesn't exist yet, we'll establish
an inotify watch instead, and wait for the socket to appear. In other
words, a missing AF_UNIX just makes connecting slower.
This is useful for daemons such as networkd or resolved that shall be
able to run during early-boot, before dbus-daemon is up, and want to
connect to dbus-daemon as soon as it becomes ready.
We'd like to use inotify to get notified when AF_UNIX sockets become
connectable. That happens at the moment of listen(), but this is doesn't
necessarily create in a watchable inotify event. Hence, let's synthesize
one whenever we generically create a socket, or when we know we created
it for a D-Bus server.
Ideally we wouldn't have to do this, and the kernel would generate an
event anyway for this. Doing this explicitly isn't too bad however, as
the event is still nicely associated with the AF_UNIX socket node, and
we generate all D-Bus sockets in our code hence it's safe.
AF_UNIX socket addresses aren't necessarily NUL terminated, however
they are usually used as strings which are assumed to be NUL terminated.
Let's hence add an extra byte to the end of the sockaddr_un structure,
that contains this NUL byte, simply for safety reasons.
Note that actually this patch changes exactly nothing IRL, as the other
sockaddr structures already are large enough to accomodate for an extra
NUL byte. The size of the union hence doesn't change at all by doing
this. The entire value of this patch is hence in the philosophical
feeling of safety, and by making something explicit that before was
implicit.
This changes two things when binding to AF_UNIX file system sockets:
1. When wethe socket already exists in the fs, and unlink() on it fails,
don't bother to bind() a second time: since nothing changed it won't
work either.
2. Also use SELinux-aware bind() for the second attempt.
Let's rework touch_file() so that it works correctly on sockets, fifos,
and device nodes: let's open an O_PATH file descriptor first and operate
based on that, if we can. This is usually the better option as it this
means we can open AF_UNIX nodes in the file system, and update their
timestamps and ownership correctly. It also means we can correctly touch
symlinks and block/character devices without triggering their drivers.
Moreover, by operating on an O_PATH fd we can make sure that we
operate on the same inode the whole time, and it can't be swapped out in
the middle.
While we are at it, rework the call so that we try to adjust as much as
we can before returning on error. This is a good idea as we call the
function quite often without checking its result, and hence it's best to
leave the files around in the most "correct" fashion possible.
If we can't process the bus for some reason we shouldn't just disable
the event source, but log something and give up on the connection. Hence
do that, and disconnect.
Currently, when sd-bus is used to issue a method call, and we get a
reply and the specified reply handler fails, we log this locally at
debug priority and proceed. The idea is that a bad server-side reply
should not be fatal for the program, except when the developer
explicitly terminates the event loop.
The reply to the initial Hello() method call we issue when joining a bus
should not be handled like that however. Instead, propagate the error
immediately, as anything that is wrong with the Hello() reply should be
considered a fatal connection problem.
This is useful so that callers know whether anything at all and how much
was flushed.
This patches through users of this functions to ensure that the return
values > 0 which may be returned now are not propagated in public APIs.
Also, users that ignore the return value are changed to do so explicitly
now.
Including BitsPerSecond or Duplex values in .link files did not work when
set_slinksettings was called because the routine was not copying the base
parameters to the structure given to ioctl. As a result, EINVAL was always
reported, and no change occurred on the Ethernet device.
I'm not sure why this is needed, but apparrently coverity doesn't like
(void)0. With this change, coverity can (almost) build systemd:
CFLAGS='-D_Float128="long double"' meson cov-build -Dman=false && \
CCACHE_DISABLE=1 COVERITY_UNSUPPORTED=1 cov-build --dir cov-int ninja -C cov-build
Patch originially by Marek Cermak <macermak@redhat.com>.
If read_line() returns ENOBFUS this means the line was overly long. When
we use this for checking whether an executable is a script, then this
shouldn't be propagated as-is, but simply as "this is not a script".
This tweaks write_string_stream_ts() in one minor way: when stdio
buffering has been turned off, let's append the newline we shall append
to the buffer we write ourselves so that the kernel only gets one
syscall for the result. When buffering is enabled stdio will take care
of that anyway.
Follow-up for #7750.
We use the same check at two places, let's add a tiny helper function
for it, since it's not entirely trivialy, and we changes this before
multiple times, and it's a good thing if we can change it at one place
only instead of multiple.
Let's call getsockopt() in a loop, so that we can deal correctly with
the label changing while we are trying to read it.
(also, while we are at it, let's make sure that there's always one
trailing NUL byte at the end of the buffer, after all SO_PEERSEC has
zero documentation, and multiple implementing backends, hence let's
better be safe than sorry)
Also, drop UID/GID validity checks from getpeercred() as the kernel will
never pass us invalid UID/GID on userns, but the overflow UID/GID
instead. Add a comment about this.
This ensures that in all threads we fork off in the background in our
code we mask out all signals, so that our thread won't end up getting
signals delivered the main process should be getting.
We always set the signal mask before forking off the thread, so that the
thread has the right mask set from its earliest existance on.
We maintain static process-wide variables in these subsystems without
locking, hence let's refuse operation unless we are called from the main
thread (which we do anyway) just as a safety precaution.
When we check the exit status of a subprocess, let's compare it with
EXIT_SUCCESS rather than 0 when looking for success.
This clarifies in code what kind of variable we are looking at and what
we are doing.
Using wait_for_terminate_and_check() instead of wait_for_terminate()
let's us simplify, shorten and unify the return value checking and
logging of waitid(). Hence, let's use it all over the place.
This new flag will cause safe_fork() to wait for the forked off child
before returning. This allows us to unify a number of cases where we
immediately wait on the forked off child, witout running any code in the
parent after the fork, and without direct interest in the precise exit
status of the process, except recgonizing EXIT_SUCCESS vs everything
else.
This renames wait_for_terminate_and_warn() to
wait_for_terminate_and_check(), and adds a flags parameter, that
controls how much to log: there's one flag that means we log about
abnormal stuff, and another one that controls whether we log about
non-zero exit codes. Finally, there's a shortcut flag value for logging
in both cases, as that's what we usually use.
All callers are accordingly updated. At three occasions duplicate logging
is removed, i.e. where the old function was called but logged in the
caller, too.
First of all, let's return pid_t, which appears to be the correct type
given that we return PIDs, and it#s what fork() uses too.
Most importantly though, flush out our PID cache, so that the call
becomes compatible with our getpid_cached() logic.
