systemd: fix uninitialized memory access in selinux
==6129== Syscall param socketcall.getsockopt(optlen) points to uninitialised byte(s)
==6129== at 0x312D0F325A: getsockopt (syscall-template.S:81)
==6129== by 0x4615E6: get_audit_data (selinux-access.c:268)
==6129== by 0x461A81: selinux_access_check (selinux-access.c:394)
==6129== by 0x453B23: bus_manager_message_handler (dbus-manager.c:1395)
==6129== by 0x313581D9C4: _dbus_object_tree_dispatch_and_unlock (dbus-object-tree.c:862)
==6129== by 0x313580FC1F: dbus_connection_dispatch (dbus-connection.c:4672)
==6129== by 0x44CD4C: bus_dispatch (dbus.c:528)
==6129== by 0x41292F: manager_loop (manager.c:1573)
==6129== by 0x40CD8E: main (main.c:1735)
==6129== Address 0x7fefff2c0 is on thread 1's stack
David Strauss [Thu, 7 Feb 2013 06:23:59 +0000 (01:23 -0500)]
build-sys: add libsystemd-id128 as libudev dependency
After d848b9cbfa0 'Move generic specifier functions to shared' libudev
depends (through) libsystemd-shared.la on libsystemd-id128.so. The
problem only appears when the linker does not support --gc-sections
and manifests itself as the inability to resolve sd_id128_get_machine
and other libsystemd-id128 functions, which aren't really used.
Mirco Tischler [Mon, 4 Feb 2013 14:13:24 +0000 (15:13 +0100)]
logs-show: show messages for all unit types in systemctl status
I can't find a reason why we shouldn't try to output messages for other
unit types than .service, .socket, .mount and .swap as well. It's probably
a leftover from before we started logging UNIT= from inside PID 1.
Eelco Dolstra [Mon, 4 Feb 2013 14:56:26 +0000 (15:56 +0100)]
nss-myhostname: copy first result to preallocated buffer
Fixes a segfault in nscd when using nss-myhostname.
Nscd expects that an NSS module's gethostbyname4_r function returns
its first result in the pre-allocated gaih_addrtuple denoted by **pat.
(See nscd/aicache.c in the Glibc sources.) However, nss-myhostname
doesn't fill in **pat but allocates the first result in ‘buffer’, then
sets *pat. So nscd crashes (e.g. when running ‘getent ahosts
my-machine’).
Hard to tell if this is a bug in nscd, since there doesn't seem to be
a proper API spec for gethostbyname4_r. But in any case, this patch
fixes the crash by copying the first result to **pat.
In other cases where multiple directories are searched for unit files,
the list of directories is described in the man page describing the
format. I think this makes sense too in case of systemd directories,
since the systemd(1) manpage already has an overview of many different
topics.
systemd: do not remove empty paths from unit lookup path
The ability to start a new unit with 'systemctl start ...' should not
depend on whether there are other units in the directory. Previously,
an additional 'systemctl daemon-reload' would be necessary to tell
systemd to update the list of unit lookup paths.
man rules were repeating the same information in too many places,
which was error prone. Those rules can be easily generated from .xml
files. For efficiency and because python is not a required dependency,
Makefile-man.am is only regenerated when requested with
make update-man-list
If no metadata in man/*.xml changed, this file should not change. So
only when a new man page or a new alias is added, this file should
show up in 'git diff'. The change should then be committed.
If the support for building from git without python was dropped, we
could drop Makefile-man.am from version control. This would also
increase the partial build time (since more stuff would be rebuild
whenever sources in man/*.xml would be modified), so it would probably
wouldn't be worth it.
Failed to open /dev/tty0: Permission denied
Failed to create root cgroup hierarchy: Permission denied
Assertion 'manager_new(SYSTEMD_SYSTEM, &m) == 0' failed at src/test/test-unit-name.c:125, function test_unit_printf(). Aborting.
Michal Schmidt [Thu, 31 Jan 2013 10:03:09 +0000 (11:03 +0100)]
cryptsetup: accept both "read-only" and "readonly" spellings
Mukund Sivaraman pointed out that cryptsetup(5) mentions the "read-only"
option, while the code understands "readonly".
We could just fix the manpage, but for consistency in naming of
multi-word options it would be prettier to have "read-only". So let's
accept both spellings.
Sometimes it is useful to look at them, and they don't take
up any significant amount of space. Keeping them also avoids
the message about files being removed at the end of make
run.
- fix typo
- use compiled systemd-nspawn
- drop --capability=... from systemd-nspawn invocation, is is the default now
- simplify sudo make invocations
William Giokas [Sat, 26 Jan 2013 17:15:32 +0000 (11:15 -0600)]
man: add Arch Linux entry to systemd-nspawn(5)
Archlinux has a similar tool to debbotstrap in the arch-install-scripts
package that will install to a specified directory. This is generally
used for installation, so the -d flag must be passed to tell it to
install to a non-mountpoint directory.
man: extend systemd.directives(7) to all manual pages
New sections are added: PAM options, crypttab options, commandline
options, miscellaneous. The last category will be used for all
untagged <varname> elements.
Commandline options sections is meant to be a developer tool: when
adding an option it is sometimes useful to be able to check if
similarly named options exist elsewhere.
Michal Schmidt [Fri, 25 Jan 2013 23:16:13 +0000 (00:16 +0100)]
nspawn: assume stdout is always writable if it does not support epoll
stdout can be redirected to a regular file. Regular files don't support epoll.
nspawn failed with: "Failed to register fds in epoll: Operation not permitted".
If stdout does not support epoll, assume it's always writable.
Michal Schmidt [Fri, 25 Jan 2013 21:19:19 +0000 (22:19 +0100)]
test: factor out testsuite.target, end.service
Tests can use the same testsuite.target.
Add end.service to call poweroff instead of doing it from ExecStopPost
where it may be skipped on failure of ExecStart.
Michal Sekletar [Fri, 18 Jan 2013 15:13:08 +0000 (16:13 +0100)]
tmpfiles: introduce type X
Type X will exclude path itself from clean-up. However, if the path is a
directory systemd-tmpfiles will clean-up its content.
In contrast to type x, where path is ignored completely, type X needs some
Age parameter. In order to determine Age parameter, we will look for config
entries of type d or D and pick the best match. Best match is either
exact match or longest prefix match.
Add _cleanup_pclose_ and fix mismatching pipe close opened by popen()
Based-on-patch-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
cppcheck reported:
[src/bootchart/svg.c:791]: (error) Mismatching allocation and deallocation: f
- Don't allow any locks to be taken while we are in the process of
executing the specific operation, so that apps are not surprised if a
suspend/shutdown happens while they rely on their inhibitor.
- Get rid of the Resumed signal, it was a bad idea, and redundant due to
PrepareForSleep(false), see below.
- Always send out PrepareFor{Shutdown,Sleep} signals, instead of only if
a delay lock is taken.
- Move PrepareForSleep(false) after we come back from the suspend, so
that apps can use this as "Resumed" notification. This also has the
benefit that apps know when to take a new lock.
loginctl: don't show [UACCESS] info in device tree
As the tree doesn't really necessarily show all device node devices and
only those are marked for uaccess it's kinda pointless showing this at
all, since it would give a pretty incomplete impression of the uaccess
information.
while working on another bug, I discovered the "strange" way systemd is
parsing Environment= in .service and thought it was worth documenting
(because I don't expect people to find this syntax by themselves unless
they read the parsing code ;)
Be more verbose about using space in Environment field and not
using value of other variables