fork() is not async-signal-safe and calling it from the signal handler
could result in a deadlock when at_fork() handlers are called. Using
the raw clone() syscall sidesteps that problem.
The tricky part is that raise() does not work, since getpid() does not
work. Add raw_getpid() to get the real pid, and use kill() instead of
raise().
https://bugs.freedesktop.org/show_bug.cgi?id=86604
If child supports, but the parent does not, or when the child does
not support, but the parent does, assume the child is a mount point.
Only if neither supports use the fallback.
dup3() allows setting O_CLOEXEC which we are not interested in. However,
it also fails if called with the same fd as input and output, which is
something we don't want. Hence use dup2().
Also, we need to explicitly turn off O_CLOEXEC for the fds, in case the
input fd was O_CLOEXEC and < 3.
[zj: When we lstat the target path, symlinks above the last component
will be followed by both stat and lstat. So when we look at the
parent, we should follow symlinks.]
'systemctl cat' now works for templates too.
'systemctl edit' does not refuse to edit units that have changed on
disk. That restriction didn't seem useful, actually editing units that
have changed on disk before they are started is very reasonable.
'edit' with instances and templates works again:
Now:
$ build/systemctl edit getty@
Failed to copy /etc/systemd/system/getty@.service.d/override.conf to /etc/systemd/system/getty@.service.d/.override.confdff6290408c86369: Permission denied
$ build/systemctl edit getty@tty3
Failed to create directories for /etc/systemd/system/getty@tty3.service.d/override.conf: Permission denied
$ build/systemctl edit --full getty@tty3
Failed to copy /usr/lib/systemd/system/getty@.service to /etc/systemd/system/.getty@tty3.serviced3d175087e7e439b: Permission denied
Failed to create temporary file for /etc/systemd/system/getty@tty3.service: Permission denied
$ build/systemctl edit --full getty@
Failed to copy /usr/lib/systemd/system/getty@.service to /etc/systemd/system/.getty@.servicea3caad491c0f2f3d: Permission denied
Failed to create temporary file for /etc/systemd/system/getty@.service: Permission denied
Containers do not really support .device, .automount or .swap units;
Systems compiled without support for swap do not support .swap units;
Systems without kdbus do not support .busname units.
With this change attempts to start a unsupported unit types will result
in an immediate "unsupported" job result, which is a lot more
descriptive then before. Also, attempts to start device units in
containers will now immediately fail instead of causing jobs to be
enqueued that never go away.
Previously, if we provided getty@.service to systemctl edit it would
have failed when using the bus because it is an invalid unit name.
But it would have succeeded when searching in the filesystem.
Now, we check if we have a template, if we do we search in the
filesystem, if we don't have a templae and we can use the bus, we do.
Furthermore, if we provided getty@tty1.service it would not have worked
when searching the filesystem, but it would have worked with the bus.
So now, when using the filesystem we use the template name and not the
unit name, and the same when logging errors.
(Also did a refactoring to avoid a long function)
try_context() is such a hot path that the hashmap lookup is expensive.
The number of contexts is small - it is the number of object types.
Using a hashmap is overkill. A plain array will do.
Before:
$ time ./journalctl --since=2014-06-01 --until=2014-07-01 > /dev/null
real 0m9.445s
user 0m9.228s
sys 0m0.213s
After:
$ time ./journalctl --since=2014-06-01 --until=2014-07-01 > /dev/null
real 0m5.438s
user 0m5.266s
sys 0m0.170s
Note that numbers 0 and -1 are both replaced with OBJECT_UNUSED,
because they are treated the same everywhere (e.g. type_to_context()
translates them both to 0).
If type==0 and a non-NULL object were given as arguments to
journal_file_hmac_put_object(), its object type check would fail and it
would return -EBADMSG.
All existing callers use either a positive type or -1. Still, for
behavior consistency with journal_file_move_to_object() let's allow
type 0 to pass.
The only user is sd_journal_enumerate_unique() and, as explained in
the previous commit (fed67c38e3 "journal: map objects to context set by
caller, not by actual object type"), the use of them there is now
superfluous. Let's remove them.
This reverts major parts of commits:
ae97089d49 journal: fix access to munmapped memory in
sd_journal_enumerate_unique
06cc69d44c sd-journal: fix sd_journal_enumerate_unique skipping values
Tested with an "--enable-debug" build and "journalctl --list-boots".
It gives the expected number of results. Additionally, if I then revert
the previous commit ("journal: map objects to context set by caller, not
to actual object type"), it crashes with SIGSEGV, as expected.
When the caller of journal_file_move_to_object() specifies type==0,
the object header is at first mapped in context 0. Then after the header
is checked, the whole object is mapped in a context determined by
the actual object type (which is not even range-checked using
type_to_context()). This looks wrong. It should map in the
caller-specified context.
An old comment in sd_journal_enumerate_unique() supports this view:
/* We do not use the type context here, but 0 instead,
* so that we can look at this data object at the same
* time as one on another file */
Clearly the expectation was that the data object will remain mapped
in context 0 without being pushed away by mapping other objects in
context OBJECT_DATA.
I suspect that this was the real bug that got fixed by ae97089d49
"journal: fix access to munmapped memory in sd_journal_enumerate_unique".
In other words, journal_file_object_keep/release are superfluous after
applying this patch.
This is useful for exposing unsafe access to mmapped objects after
the context that they were mapped in was already moved.
For example:
journal_file_move_to_object(f1, OBJECT_DATA, p1, &o1);
journal_file_move_to_object(f2, OBJECT_DATA, p2, &o2);
t = o1->object.type; /* this usually works, but is unsafe */
There will be more debugging options later.
--enable-debug will enable them all.
--enable-debug=hashmap will enable only hashmap debugging.
Also rename the C #define to ENABLE_DEBUG_* pattern.
Michal Schmidt
Zbigniew Jędrzejewski-Szmek
tomsod-m ya ru
Umut Tezduyar Lindskog
Lennart Poettering
Ken Werner
Tom Gundersen
Thomas Hindoe Paaboel Andersen
Susant Sahani
Jan Synacek
Harald Hoyer
Martin Pitt
Gabriel de Perthuis
Ivan Shapovalov
Ronny Chevalier