Historically, systemd-tmpfiles was designed to manager temporary
files, but nowadays it has become a generic tool for managing
all kinds of files. To avoid user confusion, let's remove "temporary"
from the tool's description.
As discussed in #33349
currently when dispatching json objects into C structs we either insist
on the field type or we don't. Let's extend this model a bit: depending
on two new fields either allow or refuse null types in addition to the
specified type.
This is useful for example when dispatch enums as this allows us
explicitly refuse null in various scenarios where we allow multiple
types.
Writing the progress bar so far was irritatingly slow, which was caused
by the fact that the various things we output so far resulted in one
write() syscall each because STDERR is unbuffered by default.
Let's fix that, and temporarily turn on full buffering for stderr,
restoring the normal unbuffered output right after.
This makes progress bar print visibly more efficient (and flicker free
too, since terminals no longer will move the cursor around during
output).
Builds with kernels headers < 4.14 fail with:
../src/shared/loop-util.c: In function ‘loop_configure_fallback’:
../src/shared/loop-util.c:237:31: error: ‘LOOP_SET_BLOCK_SIZE’ undeclared (first use in this function); did you mean ‘LOOP_SET_DIRECT_IO’?
if (ioctl(fd, LOOP_SET_BLOCK_SIZE, (unsigned long) c->block_size) < 0)
^~~~~~~~~~~~~~~~~~~
LOOP_SET_DIRECT_IO
Fixes: https://github.com/systemd/systemd/issues/33341
Signed-off-by: Raphaël Mélotte <raphael.melotte@mind.be>
Mention that by default, /home is managed by tmpfiles.d/home.conf, and
recommend that users run systemd-tmpfiles --dry-run --purge first to
see exactly what will be removed.
Let's keep the verb_lock_xyz() and verb_unlock_xyz() calls together, and
move event_log_reduce_to_safe_pcrs() which so far was in betwee them all
further down closer to where the function is actually used.
if we pass NULL boot_entry_token_ensure() will use its own default,
which is the same as what we passed so far explicitly, hence let's make
use of that.
The "service" field that one is supposed to pass to machine is supposed
to indicate the implementation of the client, not the service unit the
client runs in (which is typically even a scope unit, not a system
unit). Hence fix that, and make it closely match what systemd-nspawn
does.
Let's freshly calculate "m" on each iteration and always start with the maximum
size we can. If sendfile() is used we must adhere to its limit of
SSIZE_MAX minus the current offset. Otherwise we can copy more, i.e.
SSIZE_MAX without any restrictions.
Also, if we get too close to having copied SSIZE_MAX, let's turn off
sendfile() for the rest.
Also, make sure the NUL byte iovec becomes an exported constant too.
This is better than the previous situation where this was a macro
resolving to a compount expression, since the lifetime of the expression
is limited to its invoking scope. By turning this into a proper variable
the lifetime becomes unbounded, which makes it easier to use in various
scenarios, such as "if" blocks.