chiark / gitweb /
elogind.git
6 years agofileio: minor tweak to executable_is_script()
Lennart Poettering [Sat, 30 Dec 2017 14:44:29 +0000 (15:44 +0100)]
fileio: minor tweak to executable_is_script()

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".

6 years agofileio: tweak write_string_stream_ts() to write out trailing \n in one go even if...
Lennart Poettering [Sat, 30 Dec 2017 14:42:03 +0000 (15:42 +0100)]
fileio: tweak write_string_stream_ts() to write out trailing \n in one go even if buffering is off

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.

6 years agosd-bus: port one use of SO_PEERCRED by getpeercred()
Lennart Poettering [Sat, 30 Dec 2017 14:20:38 +0000 (15:20 +0100)]
sd-bus: port one use of SO_PEERCRED by getpeercred()

6 years agosd-bus: let's use mfree() where we can
Lennart Poettering [Sat, 30 Dec 2017 14:19:53 +0000 (15:19 +0100)]
sd-bus: let's use mfree() where we can

6 years agosd-bus: use SO_PEERGROUPS when available to identify groups of peer
Lennart Poettering [Sat, 30 Dec 2017 14:19:15 +0000 (15:19 +0100)]
sd-bus: use SO_PEERGROUPS when available to identify groups of peer

6 years agosocket-util: slight rework of getpeersec()
Lennart Poettering [Sat, 30 Dec 2017 14:15:44 +0000 (15:15 +0100)]
socket-util: slight rework of getpeersec()

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)

6 years agotree-wide: use {pid,uid,gid}_is_valid() where appropriate
Lennart Poettering [Sat, 30 Dec 2017 14:15:03 +0000 (15:15 +0100)]
tree-wide: use {pid,uid,gid}_is_valid() where appropriate

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.

6 years agouser-util: minor updates of in_gid()
Lennart Poettering [Sat, 30 Dec 2017 13:05:41 +0000 (14:05 +0100)]
user-util: minor updates of in_gid()

Let's use newa() and use the right type for storing sysconf() return
values.

6 years agosocket-util: add new getpeergroups() call
Lennart Poettering [Sat, 30 Dec 2017 13:02:36 +0000 (14:02 +0100)]
socket-util: add new getpeergroups() call

It's a wrapper around the new SO_PEERGROUPS sockopt, similar in style as
getpeersec() and getpeercred().

6 years agoprocess-util: add new FORK_NEW_MOUNTNS flag to safe_fork()
Lennart Poettering [Fri, 29 Dec 2017 17:52:20 +0000 (18:52 +0100)]
process-util: add new FORK_NEW_MOUNTNS flag to safe_fork()

That way we can move one more code location to use safe_fork()

6 years agopager: let's move static variables up, to the rest of them
Lennart Poettering [Fri, 29 Dec 2017 17:14:37 +0000 (18:14 +0100)]
pager: let's move static variables up, to the rest of them

let's keep static variables together, and before the function
definitions.

6 years agopager,agent: insist that we are called from the main thread
Lennart Poettering [Fri, 29 Dec 2017 17:13:38 +0000 (18:13 +0100)]
pager,agent: insist that we are called from the main thread

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.

6 years agotree-wide: make use of wait_for_terminate_and_check() at various places
Lennart Poettering [Fri, 29 Dec 2017 17:07:00 +0000 (18:07 +0100)]
tree-wide: make use of wait_for_terminate_and_check() at various places

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.

6 years agoprocess-spec: add another flag FORK_WAIT to safe_fork()
Lennart Poettering [Fri, 29 Dec 2017 17:01:37 +0000 (18:01 +0100)]
process-spec: add another flag FORK_WAIT to safe_fork()

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.

6 years agoexec-util: drop redundant log message in do_spawn()
Lennart Poettering [Fri, 29 Dec 2017 16:11:49 +0000 (17:11 +0100)]
exec-util: drop redundant log message in do_spawn()

safe_fork() logs that anyway, hence no need to do this twice.

6 years agoprocess-util: rework wait_for_terminate_and_warn() to take a flags parameter
Lennart Poettering [Wed, 27 Dec 2017 23:51:19 +0000 (00:51 +0100)]
process-util: rework wait_for_terminate_and_warn() to take a flags parameter

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.

6 years agoprocess-util: add another fork_safe() flag for enabling LOG_ERR/LOG_WARN logging
Lennart Poettering [Wed, 27 Dec 2017 20:49:19 +0000 (21:49 +0100)]
process-util: add another fork_safe() flag for enabling LOG_ERR/LOG_WARN logging

6 years agoraw-clone: beef up raw_clone() wrapper a bit
Lennart Poettering [Fri, 29 Dec 2017 15:45:04 +0000 (16:45 +0100)]
raw-clone: beef up raw_clone() wrapper a bit

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.

6 years agocoverity: don't use (void)0 under coverity
Zbigniew Jędrzejewski-Szmek [Thu, 4 Jan 2018 08:45:23 +0000 (09:45 +0100)]
coverity: don't use (void)0 under coverity

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>.

6 years agoNEWS: add in a missing change back from 229 (#7804)
Lennart Poettering [Thu, 4 Jan 2018 12:52:50 +0000 (13:52 +0100)]
NEWS: add in a missing change back from 229 (#7804)

See:

https://lists.freedesktop.org/archives/systemd-devel/2017-July/039255.html
http://fromthecodefront.blogspot.de/2017/07/systemd-pitfalls.html

6 years agomeson: don't search PATH for telinit (#7770)
Mike Gilbert [Thu, 4 Jan 2018 12:14:20 +0000 (07:14 -0500)]
meson: don't search PATH for telinit (#7770)

On a typical system running systemd, the telinit in PATH is very likely to be a symlink
to systemctl. Setting TELINIT to this may result in an infinite recursion if telinit is called
and sd_booted() == 0. This may commonly occur in a chroot environment.

Bug: https://bugs.gentoo.org/642724

[zj:
The path was originally hardcoded as "/lib/upstart/telinit", but was made configurable without
changing the default in 4ad61fd1806dde23d2c99043b4bed91a196d2c82. Then the default was
changed to `/lib/sysvinit/telinit` in abaaabf40a9891014ed4c402d7beb5a67ac256b1. Then it
started being  autodetected when meson support was added in
5c23128daba7236a6080383b2a5649033cfef85c. This patch restores the behaviour that was
implemented in configure.ac at the time of its removal.]

6 years agotest-sizeof: also add uid_t
Zbigniew Jędrzejewski-Szmek [Thu, 4 Jan 2018 10:11:16 +0000 (11:11 +0100)]
test-sizeof: also add uid_t

We had gid_t, and pid_t, but not uid_t. Add for completeness.

6 years agofix machinectl shell (in machined) (#7785)
Shawn Landden [Thu, 4 Jan 2018 11:37:15 +0000 (03:37 -0800)]
fix machinectl shell (in machined) (#7785)

4c253ed broke machined
$machinectl shell arch
Failed to get shell PTY: Input/output error

Closes: #7779
v2: do not drop DEATHSIG flag

6 years agonsflags: add namespace_flag_to_string_many_with_check()
Yu Watanabe [Fri, 29 Dec 2017 08:11:42 +0000 (17:11 +0900)]
nsflags: add namespace_flag_to_string_many_with_check()

The function will be used in later commits.

6 years agobasic: introduce *_to_string_with_check() functions
Yu Watanabe [Fri, 29 Dec 2017 08:03:54 +0000 (17:03 +0900)]
basic: introduce *_to_string_with_check() functions

They are used in later commits.

6 years agosocket-util: introduce parse_socket_address_bind_ipv6_only_or_bool()
Yu Watanabe [Mon, 1 Jan 2018 15:15:03 +0000 (00:15 +0900)]
socket-util: introduce parse_socket_address_bind_ipv6_only_or_bool()

6 years agocore: add dbus-util.[ch] to simplify creating transient units
Yu Watanabe [Mon, 1 Jan 2018 17:22:36 +0000 (02:22 +0900)]
core: add dbus-util.[ch] to simplify creating transient units

The functions and macros introduced by them will be used in later commits.

6 years agomeson: add a single .h file to shared libs with no sources
Zbigniew Jędrzejewski-Szmek [Wed, 20 Dec 2017 08:12:08 +0000 (09:12 +0100)]
meson: add a single .h file to shared libs with no sources

Otherwise stuff doesn't build on old Ubuntu with meson-0.42.1-1~xenial.

6 years agomeson: use a convenience lib for shared resolve files
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 21:46:01 +0000 (22:46 +0100)]
meson: use a convenience lib for shared resolve files

This reduces the man=false meson target count from 1281 to 1253.

--

A fully scientific test:
  git grep _sources, :/*.build|cut -d: -f2|tr -d ' '|sort|uniq -c
reveals that libudev_sources is the only source list now reused twice.  There's
some ugly circular dependency between libudev and libshared, and anyway I'm not
sure if we don't want to use different compilation options (LOG_REALM_…) in
those two cases, so I'm leaving that alone for now.

6 years agomeson: rename libudev_internal to libudev_static and link into libudev
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 19:54:46 +0000 (20:54 +0100)]
meson: rename libudev_internal to libudev_static and link into libudev

This reduces the meson man=false target count to 1281.

v2:
- link test-engine with libshared instead of libsystemd_static
Previous version built fine on F27, but fails on F26 with the following error:
/usr/bin/ld: /tmp/ccr8HRGw.ltrans6.ltrans.o: undefined reference to symbol '__start_BUS_ERROR_MAP@@SD_SHARED'
/home/zbyszek/fedora/systemd/systemd-9d5aae75c64f5583a110f03b94816aacc03bbf4d/x86_64-redhat-linux-gnu/src/shared/libsystemd-shared-236.so: error adding symbols: DSO missing from command line

v3:
- add libudev_basic

6 years agomeson: use a convenience lib for journal user sources
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 18:38:43 +0000 (19:38 +0100)]
meson: use a convenience lib for journal user sources

Instead of compiling those files twice, once for libsystemd and once for
libshared, compile once as a static archive and then link into both.
This reduce the meson target for man=no compile to 1291.

6 years agomeson: link libsystemd_static in libshared instead of recompiling
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 18:14:03 +0000 (19:14 +0100)]
meson: link libsystemd_static in libshared instead of recompiling

This is similar to the great-grandpa commit. This time the number
of meson targets compilation without man is reduced from 1347 to 1302.

6 years agoMove gcrypt-util to basic/
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 18:06:56 +0000 (19:06 +0100)]
Move gcrypt-util to basic/

We were including gcrypt-util.[ch] by hand in the few places where it
was used. Create a convenience library to avoid compiling the same
files multiple times.

v2:
- use a separate static library instead of mergin into libbasic

6 years agomeson: link libbasic and libshared_static into libshared
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 13:19:46 +0000 (14:19 +0100)]
meson: link libbasic and libshared_static into libshared

gcrypt_util_sources had to be moved because otherwise they appeared twice
in libshared.so halfproducts, causing an error.

-fvisibility=default is added to libbasic, libshared_static so that the symbols
appear properly in the exported symbol list in libshared.

The advantage is that files are not compiled twice. When configured with -Dman=false,
the ninja target list is reduced from 1588 to 1347 targets. The difference in compilation
time is small (<10%). I think this is because of -O0 and ccache and multiple cores, and
in different settings the compilation time could be reduced. The main advantage is that
errors and warnings are not reported twice.

6 years agomeson: reuse libsystemd_static for libsystemd
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 10:53:53 +0000 (11:53 +0100)]
meson: reuse libsystemd_static for libsystemd

This gets rid of recompilation, making things faster and avoids duplicated warnings.
The result seems to be the same:
$ ls -l build/libsystemd.so.0.20.0 build2/libsystemd.so.0.20.0
-rwxrwxr-x. 1 zbyszek zbyszek 3071312 Dec 19 11:45 build2/libsystemd.so.0.20.0
-rwxrwxr-x. 1 zbyszek zbyszek 3071760 Dec 19 11:11 build/libsystemd.so.0.20.0
$ diff -U1 <(objdump -T build/libsystemd.so.0.20.0|sed -r 's/[0-9a-f]{16}/________________/g') <(objdump -T build2/libsystemd.so.0.20.0|sed -r 's/[0-9a-f]{16}/________________/g')
-build/libsystemd.so.0.20.0:     file format elf64-x86-64
+build2/libsystemd.so.0.20.0:     file format elf64-x86-64

6 years agomeson: rename libsystemd_internal to libsystem_static
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 10:35:01 +0000 (11:35 +0100)]
meson: rename libsystemd_internal to libsystem_static

We already use the "_static" suffix for libshared_static ("shared" is the name
of the library, "static" is the format) and other libs, so let's rename for
consistency.

Also change libsystemd_static_sources to libsystemd_sources, since the same
list is used for both and shorter is better.

6 years agobasic: detect_vm_cpuid: fix hypervisor detection
Mike Gilbert [Sat, 30 Dec 2017 17:48:20 +0000 (12:48 -0500)]
basic: detect_vm_cpuid: fix hypervisor detection

The __get_cpuid() function only calls __cpuid() if __get_cpuid_max()
returns a value that is less than or equal to the leaf value.

In QEMU/KVM, I found that the special hypervisor leaf value (0x40000000U)
is always larger than the value retured by __get_cpuid_max().

Avoid this problem by calling the __cpuid() macro directly once we have
checked the hypervisor bit from leaf 1.

Fixes: d31b0033b7743393562a2e9d3c1e74afea981c13
6 years agofileio: write_string_stream_ts: check for file errors immediately
Mike Gilbert [Thu, 28 Dec 2017 15:03:47 +0000 (10:03 -0500)]
fileio: write_string_stream_ts: check for file errors immediately

6 years agofileio: write_string_stream_ts: return errors from fputs and fputc
Mike Gilbert [Thu, 28 Dec 2017 02:46:52 +0000 (21:46 -0500)]
fileio: write_string_stream_ts: return errors from fputs and fputc

Ignoring errors from these functions may mask errors returned by the
kernel.

Fixes: https://github.com/systemd/systemd/issues/7744
6 years agobasic: detect_vm_cpuid: use gcc's __get_cpuid() function (#7758)
Mike Gilbert [Fri, 29 Dec 2017 18:30:38 +0000 (13:30 -0500)]
basic: detect_vm_cpuid: use gcc's __get_cpuid() function (#7758)

The __get_cpuid() function includes a safety check to ensure that
executing the cpuid instruction is valid/safe.

This method also works with clang.

https://lists.freedesktop.org/archives/systemd-devel/2017-December/040054.html

6 years agoupdate TODO
Lennart Poettering [Tue, 26 Dec 2017 15:02:10 +0000 (16:02 +0100)]
update TODO

6 years agosd-daemon: use sockaddr_port() helper
Lennart Poettering [Tue, 26 Dec 2017 12:36:07 +0000 (13:36 +0100)]
sd-daemon: use sockaddr_port() helper

6 years agosocket-util: clarify why sockaddr_port returns unsigned rather than uint16_t
Lennart Poettering [Tue, 26 Dec 2017 12:35:52 +0000 (13:35 +0100)]
socket-util: clarify why sockaddr_port returns unsigned rather than uint16_t

6 years agocondition: extend ConditionKernelVersion= with relative version checks
Lennart Poettering [Sat, 23 Dec 2017 14:02:58 +0000 (15:02 +0100)]
condition: extend ConditionKernelVersion= with relative version checks

Now that we have str_verscmp() in our source tree anyway, let's make it
generic and reuse it for ConditionKernelVersion=.

6 years agoman: systemd: fix typo (#7746)
nulsoh [Tue, 26 Dec 2017 19:45:06 +0000 (20:45 +0100)]
man: systemd: fix typo (#7746)

Add missing _r_ for ctl-alt-del.target

6 years agoupdate TODO
Lennart Poettering [Fri, 22 Dec 2017 14:28:17 +0000 (15:28 +0100)]
update TODO

6 years agoprocess-util: debug log if PR_SET_NAME fails.
Lennart Poettering [Sun, 24 Dec 2017 17:50:17 +0000 (18:50 +0100)]
process-util: debug log if PR_SET_NAME fails.

6 years agoprocess-util: allow rename_process() only in the main thread
Lennart Poettering [Sun, 24 Dec 2017 17:48:49 +0000 (18:48 +0100)]
process-util: allow rename_process() only in the main thread

We make assumptions about the comm name we set via PR_SET_NAME: that it
would reflect the process name, but that's only the case for the main
thread. Moreover, we cache the mmap() region without locking.

Let's hence be safe rather than sorry and support all this only in the
main thread.

6 years agobasic: split out blockdev-util.[ch] from util.h
Lennart Poettering [Fri, 22 Dec 2017 14:22:59 +0000 (15:22 +0100)]
basic: split out blockdev-util.[ch] from util.h

With three functions it makes sense to split this out now.

6 years agoagents: use kill_and_sigcont() where appropriate
Lennart Poettering [Fri, 22 Dec 2017 14:10:56 +0000 (15:10 +0100)]
agents: use kill_and_sigcont() where appropriate

6 years agoprocess-util: move fork_agent() to process-util.[ch]
Lennart Poettering [Fri, 22 Dec 2017 14:07:22 +0000 (15:07 +0100)]
process-util: move fork_agent() to process-util.[ch]

It's a relatively small wrapper around safe_fork() now, hence let's move
it over, and make its signature even more alike. Also, set a different
process name for the polkit and askpw agents.

6 years agotree-wide: use EXIT_SUCCESS/EXIT_FAILURE in exit() where we can
Lennart Poettering [Fri, 22 Dec 2017 12:24:40 +0000 (13:24 +0100)]
tree-wide: use EXIT_SUCCESS/EXIT_FAILURE in exit() where we can

6 years agotree-wide: introduce new safe_fork() helper and port everything over
Lennart Poettering [Fri, 22 Dec 2017 12:08:14 +0000 (13:08 +0100)]
tree-wide: introduce new safe_fork() helper and port everything over

This adds a new safe_fork() wrapper around fork() and makes use of it
everywhere. The new wrapper does a couple of things we previously did
manually and separately in a safer, more correct and automatic way:

1. Optionally resets signal handlers/mask in the child

2. Sets a name on all processes we fork off right after forking off (and
   the patch assigns useful names for all processes we fork off now,
   following a systematic naming scheme: always enclosed in () – in order
   to indicate that these are not proper, exec()ed processes, but only
   forked off children, and if the process is long-running with only our
   own code, without execve()'ing something else, it gets am "sd-" prefix.)

3. Optionally closes all file descriptors in the child

4. Optionally sets a PR_SET_DEATHSIG to SIGTERM in the child, in a safe
   way so that the parent dying before this happens being handled
   safely.

5. Optionally reopens the logs

6. Optionally connects stdin/stdout/stderr to /dev/null

7. Debug logs about the forked off processes.

6 years agoterminal-util: open /dev/null with O_CLOEXEC in make_stdio_null()
Lennart Poettering [Sat, 23 Dec 2017 23:54:40 +0000 (00:54 +0100)]
terminal-util: open /dev/null with O_CLOEXEC in make_stdio_null()

Ultimately, O_CLOEXEC should be off in fd 0, 1, 2, but when we open
/dev/null here it's unlikely to be < 0, and after dupping the fd to 0,
1, 2 we turn off O_CLOEXEC explicitly anyway.

Unless we know that what we are about to open will return 0, 1 or 2 we
should always set O_CLOEXEC in order to be safe to other threads forking
of subprocesses at the wrong moment.

6 years agoterminal-util: return first error, not last in make_stdio()
Lennart Poettering [Fri, 22 Dec 2017 12:05:33 +0000 (13:05 +0100)]
terminal-util: return first error, not last in make_stdio()

Just a minor tweak, making sure we execute as much as we can of the
funciton, but return the first error instead of the last we encounter.

This is usuelly how we do things when we have functions that continue on
the first error, so let's do it like that here too.

6 years agofd-util: use close_nointr() return value instead of errno
Lennart Poettering [Fri, 22 Dec 2017 12:04:24 +0000 (13:04 +0100)]
fd-util: use close_nointr() return value instead of errno

Our own calls return errors in their return values, hence use that
rather than errno when checking errors.

6 years agofd-util: add some (void) casts
Lennart Poettering [Fri, 22 Dec 2017 12:03:35 +0000 (13:03 +0100)]
fd-util: add some (void) casts

6 years agoman: add a systemd-rc-local-generator(8) man page
Lennart Poettering [Sun, 24 Dec 2017 15:50:19 +0000 (16:50 +0100)]
man: add a systemd-rc-local-generator(8) man page

Most importantly, let's highlight the differences to the rc-local
behaviour in SysV.

Fixes: #7703
6 years agoseparate flags from shebang
bleep_blop [Sun, 24 Dec 2017 06:53:20 +0000 (06:53 +0000)]
separate flags from shebang

6 years agomeson: hopefully renameat2() will show up where renameat() is defined
Lennart Poettering [Mon, 25 Dec 2017 11:35:43 +0000 (12:35 +0100)]
meson: hopefully renameat2() will show up where renameat() is defined

Should glibc add this eventually, let's try to be smart where to look
for it.

6 years agomeson: look for gettid() definition where getpid() is defined
Lennart Poettering [Mon, 25 Dec 2017 11:35:28 +0000 (12:35 +0100)]
meson: look for gettid() definition where getpid() is defined

Hopefully, should gettid() show up one day in glibc it'll show up where
getpid() is defined too.

6 years agomeson: when pivot_root() is added one day, look for it in <unistd.h>
Lennart Poettering [Mon, 25 Dec 2017 11:07:40 +0000 (12:07 +0100)]
meson: when pivot_root() is added one day, look for it in <unistd.h>

We of course don't know in which header glibc will export pivot_root()
and if it ever will. But there's a good chance they'll place it where
chroot() is located, given the similarity in the operations, hence let's
try our luck and look for it at the same place.

If we are lucky this means we don't have to patch our code if glibc
decides to expose the call one day.

6 years agomeson: use "args" for setting _GNU_SOURCE when checking for functions
Lennart Poettering [Mon, 25 Dec 2017 11:01:14 +0000 (12:01 +0100)]
meson: use "args" for setting _GNU_SOURCE when checking for functions

This reworks how we set _GNU_SOURCE when checking for the availability
of functions:

1. We set it for most of the functions we look for. After all we set it
for our entire built anyway, and it's usually how Linux-specific
definitions in glibc are protected these days. Given that we usually
have checks for such modern stuff only anyway, let's just blanket enable
it.

2. Use "args" instead of "prefix" to set the macro. This is what is
suggested in the meson docs, hence let's do it.

6 years agomeson: define _GNU_SOURCE to detect copy_file_range() (#7734)
Yu Watanabe [Mon, 25 Dec 2017 10:42:42 +0000 (19:42 +0900)]
meson: define _GNU_SOURCE to detect copy_file_range() (#7734)

Follow-up for bad7a0c81f501fbbcc79af9eaa4b8254441c4a1f of git
repository for glibc.

Recently glibc added `copy_file_range()`, but to use it,
`_GNU_SOURCE` needs to be defined. This adds the flag in
meson.build to detect the function by meson correctly.

6 years agoMove mkdir_label() to mkdir-label.c
Zbigniew Jędrzejewski-Szmek [Sat, 16 Dec 2017 12:32:42 +0000 (13:32 +0100)]
Move mkdir_label() to mkdir-label.c

It just seems strange to have it in a different file if mkdir-label.c exists.

6 years agosmack-util: remove unneeded initalization
Zbigniew Jędrzejewski-Szmek [Sat, 16 Dec 2017 12:22:57 +0000 (13:22 +0100)]
smack-util: remove unneeded initalization

6 years agoAdd mkdir_errno_wrapper() and use instead of mkdir() in various places
Zbigniew Jędrzejewski-Szmek [Fri, 15 Dec 2017 16:08:13 +0000 (17:08 +0100)]
Add mkdir_errno_wrapper() and use instead of mkdir() in various places

We'd pass pointers to mkdir and mkdir_label to call in various places. mkdir
returns the error in errno while mkdir_label returns the error directly.

6 years agobasic: introduce socket_protocol_{from,to}_name()
Yu Watanabe [Sat, 23 Dec 2017 10:32:04 +0000 (19:32 +0900)]
basic: introduce socket_protocol_{from,to}_name()

And use them where they can be applicable.

6 years agosocket-util: add socket_address_type_{from,to}_string()
Yu Watanabe [Sat, 23 Dec 2017 07:55:36 +0000 (16:55 +0900)]
socket-util: add socket_address_type_{from,to}_string()

6 years agonetworkd: Add support for ipvlan L3s and flags (#7726)
Susant Sahani [Sat, 23 Dec 2017 17:55:03 +0000 (23:25 +0530)]
networkd: Add support for ipvlan L3s and flags (#7726)

This works supports to configure L3S mode and flags
such as bridge, private and vepa

6 years agosd-bus: drop check for selinux before calling getsockopt(SO_PEERSEC)
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 14:12:50 +0000 (15:12 +0100)]
sd-bus: drop check for selinux before calling getsockopt(SO_PEERSEC)

Quoting Lennart Poettering in
https://github.com/systemd/systemd/pull/6464#issuecomment-319029293:
> If the kernel allows us to query that data we should also be Ok with passing
> it on to our own caller, regardless if selinux is technically on or off...

The advantage is that this allows gcc to be smarter and reduce linkage:
(before)$ ldd build/libnss_systemd.so.2
linux-vdso.so.1 (0x00007ffeb46ff000)
librt.so.1 => /lib64/librt.so.1 (0x00007f2f60da6000)
libcap.so.2 => /lib64/libcap.so.2 (0x00007f2f60ba1000)
libselinux.so.1 => /lib64/libselinux.so.1 (0x00007f2f60978000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f2f60759000)
libc.so.6 => /lib64/libc.so.6 (0x00007f2f60374000)
/lib64/ld-linux-x86-64.so.2 (0x00007f2f61294000)
libpcre2-8.so.0 => /lib64/libpcre2-8.so.0 (0x00007f2f600f0000)
libdl.so.2 => /lib64/libdl.so.2 (0x00007f2f5feec000)
(after )$ ldd build/libnss_systemd.so.2
linux-vdso.so.1 (0x00007ffe5f543000)
librt.so.1 => /lib64/librt.so.1 (0x00007f427dcaa000)
libcap.so.2 => /lib64/libcap.so.2 (0x00007f427daa5000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f427d886000)
libc.so.6 => /lib64/libc.so.6 (0x00007f427d4a1000)
/lib64/ld-linux-x86-64.so.2 (0x00007f427e196000)

Note that this only works in conjuction with the previous commit: either
of the two commits alone does not have the desired effect on linkage.

Replaces #6464.

6 years agoMove selinux-related stuff from btrfs-util.c to label.c
Zbigniew Jędrzejewski-Szmek [Tue, 19 Dec 2017 13:33:11 +0000 (14:33 +0100)]
Move selinux-related stuff from btrfs-util.c to label.c

In preparation for future changes.

6 years agomore portable perl shebangs (#7701)
Jörg Thalheim [Tue, 19 Dec 2017 10:13:34 +0000 (10:13 +0000)]
more portable perl shebangs (#7701)

same motivation as in #5816:

- distributions have scripts to rewrite shebangs on installation and
  they know what locations to rely on.
- For tests/compilation we should rather rely on the user to have setup
  there PATH correctly.

6 years agomissing: Add DM_DEFERRED_REMOVE
Henrik Grindal Bakken [Thu, 23 Mar 2017 15:19:15 +0000 (16:19 +0100)]
missing: Add DM_DEFERRED_REMOVE

Also include missing.h in dissect-image.c to pick it up.

6 years agomissing: Define SMACK_MAGIC if it's missing
Henrik Grindal Bakken [Thu, 23 Mar 2017 15:09:10 +0000 (16:09 +0100)]
missing: Define SMACK_MAGIC if it's missing

6 years agomissing: Define EFIVARFS_MAGIC if missing
Henrik Grindal Bakken [Thu, 23 Mar 2017 15:02:20 +0000 (16:02 +0100)]
missing: Define EFIVARFS_MAGIC if missing

6 years agomissing: Add MAX_HANDLE_SZ
Henrik Grindal Bakken [Thu, 23 Mar 2017 15:07:33 +0000 (16:07 +0100)]
missing: Add MAX_HANDLE_SZ

6 years agomissing: Add PR_SET_MM_{ARG,ENV}_{START,END}
Henrik Grindal Bakken [Thu, 23 Mar 2017 14:42:15 +0000 (15:42 +0100)]
missing: Add PR_SET_MM_{ARG,ENV}_{START,END}

6 years agomissing: Add some more btrfs structs and constants
Henrik Grindal Bakken [Thu, 23 Mar 2017 14:33:06 +0000 (15:33 +0100)]
missing: Add some more btrfs structs and constants

6 years agobuild-sys: install TRANSIENT-SETTINGS.md and UIDS-GIDS.md (#7690)
Felipe Sateler [Mon, 18 Dec 2017 13:58:13 +0000 (10:58 -0300)]
build-sys: install TRANSIENT-SETTINGS.md and UIDS-GIDS.md (#7690)

6 years agomount-setup: fix MNT_CHECK_WRITABLE error handling, and log about the issue
Lennart Poettering [Fri, 15 Dec 2017 16:37:16 +0000 (17:37 +0100)]
mount-setup: fix MNT_CHECK_WRITABLE error handling, and log about the issue

Let's correct the error handling (the error is in errno, not r), and
let's add logging like the rest of the function has it.

6 years agomeson: libudev_core and udevadm should have LOG_REALM=LOG_REALM_UDEV (#7666)
Franck Bui [Sat, 16 Dec 2017 08:36:36 +0000 (09:36 +0100)]
meson: libudev_core and udevadm should have LOG_REALM=LOG_REALM_UDEV (#7666)

Otherwise, setting udev_log=debug in /etc/udev/udev.conf has no effects since
systemd-udevd is built with LOG_REALM=LOG_REALM_UDEV.

However using LOG_REALM_UDEV (for libudev_core) reveals another similar bug for
udevadm which should also define LOG_REALM_UDEV.

6 years agologind: use free_and_replace in one spot
Zbigniew Jędrzejewski-Szmek [Thu, 7 Dec 2017 21:16:16 +0000 (22:16 +0100)]
logind: use free_and_replace in one spot

No functional change.

6 years agotree-wide: use SPECIAL_ROOT_SLICE
Zbigniew Jędrzejewski-Szmek [Thu, 7 Dec 2017 13:28:00 +0000 (14:28 +0100)]
tree-wide: use SPECIAL_ROOT_SLICE

6 years agologind: fix misleading message
Zbigniew Jędrzejewski-Szmek [Thu, 7 Dec 2017 12:12:13 +0000 (13:12 +0100)]
logind: fix misleading message

This message would also be emitted at boot for any user with linger
enabled, so "logged in" is the wrong term to use.

6 years agologind: simplify one conditional
Zbigniew Jędrzejewski-Szmek [Thu, 7 Dec 2017 11:57:31 +0000 (12:57 +0100)]
logind: simplify one conditional

Don't bother with removing the directory if we didn't create it.

6 years agoFix logical error in meson.build (#7658)
Max Harmathy [Fri, 15 Dec 2017 15:05:25 +0000 (16:05 +0100)]
Fix logical error in meson.build (#7658)

sysvinit_path and sysvrcnd_path have to be set both to activate Sysv compatibility.

6 years agoman: Clarify when OnFailure= activates after restarts (#7646)
Ken (Bitsko) MacLeod [Fri, 15 Dec 2017 10:10:41 +0000 (04:10 -0600)]
man: Clarify when OnFailure= activates after restarts (#7646)

6 years agotree-wide: add DEBUG_LOGGING macro that checks whether debug logging is on (#7645)
Lennart Poettering [Fri, 15 Dec 2017 10:09:00 +0000 (11:09 +0100)]
tree-wide: add DEBUG_LOGGING macro that checks whether debug logging is on (#7645)

This makes things a bit easier to read I think, and also makes sure we
always use the _unlikely_ wrapper around it, which so far we used
sometimes and other times we didn't. Let's clean that up.

6 years agoTODO: update
Yu Watanabe [Thu, 14 Dec 2017 13:17:17 +0000 (22:17 +0900)]
TODO: update

6 years agoverbs: Rename VERB_OFFLINE to VERB_ONLINE_ONLY, expand MUSTBEROOT
Colin Walters [Thu, 14 Dec 2017 20:23:02 +0000 (15:23 -0500)]
verbs: Rename VERB_OFFLINE to VERB_ONLINE_ONLY, expand MUSTBEROOT

Followup to previous commit. Suggested by @poettering.
Reindented the `verbs[]` tables to match the apparent previous
whitespace rules (indent to one flag, allow multiple flags to overflow?).

6 years agosystemctl,verbs: Introduce SYSTEMD_OFFLINE environment variable
Colin Walters [Wed, 13 Dec 2017 21:04:41 +0000 (16:04 -0500)]
systemctl,verbs: Introduce SYSTEMD_OFFLINE environment variable

A lot of code references the `running_in_chroot()` function; while
I didn't dig I'm pretty certain this arose to deal with situations
like RPM package builds in `mock` - there we don't want the `%post`s
to `systemctl start` for example.

And actually this exact same use case arises for
[rpm-ostree](https://github.com/projectatomic/rpm-ostree/)
where we implement offline upgrades by default; the `%post`s are
always run in a new chroot using [bwrap](https://github.com/projectatomic/bubblewrap).

And here's the problem: bwrap creates proper mount roots, so it
passes `running_in_chroot()`, and then if a script tries to do
`systemctl start` we get:
`System has not been booted with systemd as init system (PID 1)`
but that's an *error*, unlike the `running_in_chroot()` case where we ignore.

Further complicating things is there are real world RPM packages
like `glusterfs` which end up invoking `systemctl start`.

A while ago, the `SYSTEMD_IGNORE_CHROOT` environment variable was
added for the inverse case of running in a chroot, but still wanting
to use systemd as PID 1 (presumably some broken initramfs setups?).

Let's introduce a `SYSTEMD_OFFLINE` environment variable for cases like
mock/rpm-ostree so we can force on the "ignore everything except preset" logic.
This way we'll still not start services even if mock switches to use nspawn or
bwrap or something else that isn't a chroot.

We also cleanly supercede the `SYSTEMD_IGNORE_CHROOT=1` which is now spelled
`SYSTEMD_OFFLINE=0`.  (Suggested by @poettering)

Also I made things slightly nicer here and we now print the ignored operation.

6 years agoCODING_STYLE: provide better explanation why /* */ over // (#7647)
Lennart Poettering [Fri, 15 Dec 2017 09:26:07 +0000 (10:26 +0100)]
CODING_STYLE: provide better explanation why /* */ over // (#7647)

Let's provide a real reason why /* */ should be used for commenting,
rather than //, beyond mere taste.

(This ultimately simply codifies how I use // vs. /* */ comments, and I
think this is useful as an explanation and reason hence.)

6 years agofinal v236 update (#7649)
Lennart Poettering [Thu, 14 Dec 2017 22:09:57 +0000 (23:09 +0100)]
final v236 update (#7649)

6 years agoNEWS: update NEWS again, and prepare for a release tomorrow
Lennart Poettering [Wed, 13 Dec 2017 17:27:59 +0000 (18:27 +0100)]
NEWS: update NEWS again, and prepare for a release tomorrow

6 years agomeson: increase version numbers
Lennart Poettering [Mon, 11 Dec 2017 15:10:25 +0000 (16:10 +0100)]
meson: increase version numbers

6 years agocore: Implement timeout based umount/remount limit
Kyle Walker [Wed, 13 Dec 2017 17:49:26 +0000 (12:49 -0500)]
core: Implement timeout based umount/remount limit

Remount, and subsequent umount, attempts can hang for inaccessible network
based mount points. This can leave a system in a hard hang state that
requires a hard reset in order to recover. This change moves the remount,
and umount attempts into separate child processes. The remount and umount
operations will block for up to 90 seconds (DEFAULT_TIMEOUT_USEC). Should
those waits fail, the parent will issue a SIGKILL to the child and continue
with the shutdown efforts.

In addition, instead of only reporting some additional errors on the final
attempt, failures are reported as they occur.

6 years agotree-wide: make use of new STRLEN() macro everywhere (#7639)
Lennart Poettering [Thu, 14 Dec 2017 18:02:29 +0000 (19:02 +0100)]
tree-wide: make use of new STRLEN() macro everywhere (#7639)

Let's employ coccinelle to do this for us.

Follow-up for #7625.

6 years agobasic: turn off stdio locking for a couple of helper calls
Lennart Poettering [Mon, 11 Dec 2017 19:01:55 +0000 (20:01 +0100)]
basic: turn off stdio locking for a couple of helper calls

These helper calls are potentially called often, and allocate FILE*
objects internally for a very short period of time, let's turn off
locking for them too.

6 years agocore: add EXTEND_TIMEOUT_USEC={usec} - prevent timeouts in startup/runtime/shutdown...
Daniel Black [Thu, 14 Dec 2017 11:17:43 +0000 (22:17 +1100)]
core: add EXTEND_TIMEOUT_USEC={usec} - prevent timeouts in startup/runtime/shutdown (#7214)

With Type=notify services, EXTEND_TIMEOUT_USEC= messages will delay any startup/
runtime/shutdown timeouts.

A service that hasn't timed out, i.e, start time < TimeStartSec,
runtime < RuntimeMaxSec and stop time < TimeoutStopSec, may by sending
EXTEND_TIMEOUT_USEC=, allow the service to continue beyond the limit for
the execution phase (i.e TimeStartSec, RunTimeMaxSec and TimeoutStopSec).

EXTEND_TIMEOUT_USEC= must continue to be sent (in the same way as
WATCHDOG=1) within the time interval specified to continue to reprevent
the timeout from occuring.

Watchdog timeouts are also extended if a EXTEND_TIMEOUT_USEC is greater
than the remaining time on the watchdog counter.

Fixes #5868.