Jan Engelhardt [Wed, 20 Mar 2019 09:58:41 +0000 (10:58 +0100)]
rpm: avoid hiding errors from systemd commands
Humans are susceptible to making orthographic errors sometimes. A
misspelled "%systemd_post caek.service" would go unnoticed due to all
output from systemctl being discarded if and when %post runs.
To alleviate this, cease hiding outputs. Then, to account for the
potential absence of systemd from the system, add file checks so as
to not generate a "command not found" error.
Jan Engelhardt [Wed, 20 Mar 2019 13:48:47 +0000 (14:48 +0100)]
rpm: check argument counts for systemd macros
Invoking %systemd_tmpfiles (in %post) without any arguments, while
possible, will cause systemd-tmpfiles to process the entire system
configuration, rather than just the newly installed configuration
files. In https://github.com/systemd/systemd/pull/12048, it was
established that processing everything constitutes unusual practice,
and should be flagged as a mistake at build time.
Furthermore, invoking %systemd_post without any arguments will cause
the underlying `systemctl preset` to outright return an error ("Too
few arguments") when run. This can be flagged during build time in
the same manner.
As I have found no ways to successfully nest %if clauses inside a
macro[1], I am helping myself by reusing the recursive variable
expansion technique pioneered in [2].
Now, when %systemd_post or %systemd_tmpfiles is incorrectly used,
rpm gives accurate line number reporting, too:
error: This macro requires some arguments
error: line 11: %{systemd_post}
error: This macro requires two arguments
error: line 13: %{tmpfiles_create_package meh more more}
[1] what has been tried: %{expand:%%if "%#" == 0 \\\
%%{error:you have given me %# args} \\\
%%endif}
Chris Morin [Wed, 20 Mar 2019 08:34:23 +0000 (01:34 -0700)]
journal-file: handle SIGBUS on offlining thread
The thread launched in journal_file_set_offline() accesses a memory
mapped file, so it needs to handle SIGBUS. Leave SIGBUS unblocked on the
offlining thread so that it uses the same handler as the main thread.
The result of triggering SIGBUS in a thread where it's blocked is
undefined in Linux. The tested implementations were observed to cause
the default handler to run, taking down the whole journald process.
We can leave SIGBUS unblocked in multiple threads since it's handler is
thread-safe. If SIGBUS is sent to the journald process asynchronously
(i.e. with kill, sigqueue, or raise), either thread handling it will
result in the same behavior: it will install the default handler and
reraise the signal, killing the process.
Franck Bui [Tue, 19 Mar 2019 09:59:26 +0000 (10:59 +0100)]
core: only watch processes when it's really necessary
If we know that main pid is our child then it's unnecessary to watch all
other processes of a unit since in this case we will get SIGCHLD when the main
process will exit and will act upon accordingly.
So let's watch all processes only if the main process is not our child since in
this case we need to detect when the cgroup will become empty in order to
figure out when the service becomes dead. This is only needed by cgroupv1.
Franck Bui [Mon, 18 Mar 2019 19:59:36 +0000 (20:59 +0100)]
core: reduce the number of stalled PIDs from the watched processes list when possible
Some PIDs can remain in the watched list even though their processes have
exited since a long time. It can easily happen if the main process of a forking
service manages to spawn a child before the control process exits for example.
However when a pid is about to be mapped to a unit by calling unit_watch_pid(),
the caller usually knows if the pid should belong to this unit exclusively: if
we just forked() off a child, then we can be sure that its PID is otherwise
unused. In this case we take this opportunity to remove any stalled PIDs from
the watched process list.
If we learnt about a PID in any other form (for example via PID file, via
searching, MAINPID= and so on), then we can't assume anything.
alloc-util: use malloc_usable_size() to determine allocated size
It's a glibc-specific API, but supported on FreeBSD and musl too at
least, hence fairly common. This way we can reduce our calls to
realloc() as much as possible.
units: turn off keyring handling for user@.service
This service uses PAM anyway, hence let pam_keyring set things up for
us. Moreover, this way we ensure that the invocation ID is not set for
this service as key, and thus can't confuse the user service's
invocation ID.
Claudius Ellsel [Tue, 19 Mar 2019 00:30:22 +0000 (01:30 +0100)]
Change Razer Abyssus DPI in 70-mouse.hwdb (#12029)
As discussed in https://gitlab.freedesktop.org/libinput/libinput/issues/198#note_100642 the DPI for the Razer Abyssus mouse is not 3500 by default, but around 1600-1700 when measured with the mouse-dpi-tool.
So I have done some measurements now and always got a value of about 21000 device units on a distance of 12.5 inch. This would result in a calculated resolution of about 1680 DPI. Since such an odd number does not occur in the hwdb file I decided to round to 1600 DPI.
The function so far always returned -ECANCELLED, which is ignored in all
cases the function is invoked, except one: in unit_test_start_limit()
where -ECANCELLED is returned when the start limit is hit, which is part
of unit_start()'s protocol of return values.
Since the emergency_action() logic should be relatively generic and is
used in many places, let's drop the return value from it, since it's
constant anyway, and in alll cases useless. Instead, let's return it in
unit_test_start_limit(), where it's part of the protocol.
Reproducer:
for i in `seq 1 100`; do gdbus call --session -d org.freedesktop.systemd1 -m org.freedesktop.systemd1.Manager.StartUnit -o "/$(for x in `seq 0 28000`; do echo -n $x; done)" & done
socket-util: add wrappers for binding socket to ifindex/ifname
socket_bind_to_ifindex() uses the the SO_BINDTOIFINDEX sockopt of kernel
5.0, with a fallback to SO_BINDTODEVICE on older kernels.
socket_bind_to_ifname() is a trivial wrapper around SO_BINDTODEVICE, the
only benefit of using it instead of SO_BINDTODEVICE directly is that it
determines the size of the interface name properly so that it also works
for unbinding. Moreover, it's an attempt to unify our invocations of the
sockopt with a size of strlen(ifname) rather than strlen(ifname)+1...
basic/fd-util: refuse "infinite" loop in close_all_fds()
I had a test machine with ulimit -n set to 1073741816 through pam
("session required pam_limits.so set_all", which copies the limits from PID 1,
left over from testing of #10921).
test-execute would "hang" and then fail with a timeout when running
exec-inaccessiblepaths-proc.service. It turns out that the problem was in
close_all_fds(), which would go to the fallback path of doing close() 1073741813 times. Let's just fail if we hit this case. This only matters
for cases where both /proc is inaccessible, and the *soft* limit has been
raised.
(gdb) bt
#0 0x00007f7e2e73fdc8 in close () from target:/lib64/libc.so.6
#1 0x00007f7e2e42cdfd in close_nointr ()
from target:/home/zbyszek/src/systemd-work3/build-rawhide/src/shared/libsystemd-shared-241.so
#2 0x00007f7e2e42d525 in close_all_fds ()
from target:/home/zbyszek/src/systemd-work3/build-rawhide/src/shared/libsystemd-shared-241.so
#3 0x0000000000426e53 in exec_child ()
#4 0x0000000000429578 in exec_spawn ()
#5 0x00000000004ce1ab in service_spawn ()
#6 0x00000000004cff77 in service_enter_start ()
#7 0x00000000004d028f in service_enter_start_pre ()
#8 0x00000000004d16f2 in service_start ()
#9 0x00000000004568f4 in unit_start ()
#10 0x0000000000416987 in test ()
#11 0x0000000000417632 in test_exec_inaccessiblepaths ()
#12 0x0000000000419362 in run_tests ()
#13 0x0000000000419632 in main ()
test-execute: allow filtering test cases by pattern
When debugging failure in one of the cases, it's annoying to have to wade
through the output from all the other cases. Let's allow picking select
cases.
seccomp: allow shmat to be a separate syscall on architectures which use a multiplexer
After
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=0d6040d46817,
those syscalls have their separate numbers and we can block them.
But glibc might still use the old ones. So let's just do a best-effort
block and not assume anything about how effective it is.
../src/shared/bootspec.c: In function ‘find_sections’:
../src/shared/bootspec.c:425:23: warning: comparison of integer expressions of different signedness: ‘ssize_t’ {aka ‘int’} and ‘uint32_t’ {aka ‘unsigned int’} [-Wsign-compare]
425 | if (n != size)
| ^~
man: reorder and add examples to systemd-analyze(1)
The number of verbs supported by systemd-analyze has grown quite a bit, and the
man page has become an unreadable wall of text. Let's put each verb in a
separate subsection, grouping similar verbs together, and add a lot of examples
to guide the user.
Frantisek Sumsal [Fri, 15 Mar 2019 09:05:33 +0000 (10:05 +0100)]
test: use PBKDF2 instead of Argon2 in cryptsetup...
to reduce memory requirements for volume manipulation. Also,
to further improve the test performance, reduce number of PBKDF
iterations to 1000 (allowed minimum).
tests: install /usr/bin/dbus-broker when using dbus-broker
We'd install the service file, and then dbus-broker-launcher because it is
mentioned in ExecStart=, but not the main executable, so nothing would work.
Let's just install dbus-broker executables if found. They are small, so this
doesn't matter much, and is much easier than figuring the exact conditions
under which dbus-broker will be used instead of dbus-daemon.
The "include" files had type "book" for some raeason. I don't think this
is meaningful. Let's just use the same everywhere.
$ perl -i -0pe 's^..DOCTYPE (book|refentry) PUBLIC "-//OASIS//DTD DocBook XML V4.[25]//EN"\s+"http^<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN"\n "http^gms' man/*.xml
projects / thirdparty / systemd.git / log
