We got the following error when running systemd on a device with many ports:
"rtnl: kernel receive buffer overrun
Event source 'rtnl-receive-message' returned error, disabling: No buffer space
available"
I think the kernel socket receive buffer queue should be increased. The default
value is taken from:
"/proc/sys/net/core/rmem_default", but we can overwrite it using SO_RCVBUF
socket option.
This is already done in networkd for other sockets.
For example, the bus socket (sd-bus/bus-socket.c) has a receive queue of 8MB.
In our case, the default is 208KB.
Increasing the buffer receive queue for manager socket to 512KB should be enough
to get rid of the above error.
[tomegun: bump the limit even higher to 8M]
TXT records should have at least one character, so enforce this.
Before 0f84a72 parser SIGSEGV'd on ->txt.strings being NULL, but
even if this is fixed we should reject invalid TXT records.
Loops in RR compression were only detected for the first entry.
Multiple redirections should be allowed, each one checking for an
infinite loop on its own starting point.
Also update the pointer on each redirection to avoid longer loops of
labels and redirections, in names like:
(start) [len=1] "A", [ptr to start]
(David: rename variable to "jump_barrier" and add reference to RFC)
The previous fix e0312f4db "core: fix check for transaction
destructiveness" broke test-engine (noticed by Zbyszek).
Apparently I had a wrong idea of the intended semantics of --fail.
The manpage says the operation should fail if it "conflicts with a
pending job (more specifically: causes an already pending start job to
be reversed into a stop job or vice versa)".
So let's check job_type_is_conflicting, instead of !is_superset.
This makes both test-engine and TEST-03-JOBS pass again.
A strv might be NULL if it is empty. The txt.strings comparison doesn't
take that into account. Introduce strv_equal() to provide a proper helper
for this and fix resolve to use it.
Thanks to Stanisław Pitucha <viraptor@gmail.com> for reporting this!
The kdbus module will later get a policy that endpoint-names are
restricted to "<uid>-<name>" just like bus-names. Make sure that systemd
is already compatible to that.
Pointer acceleration for relative input devices (mice, trackballs, etc.)
applies to the deltas of the device. Alas, those deltas have no physical
reference point - a delta of 10 may be caused by a large movement of a
low-dpi mouse or by a minute movement of a high-dpi mouse.
Which makes pointer acceleration a bit useless and high-dpi devices
essentially unusable.
In an ideal world, we could read the DPI from the device directly and work
with that. In the world we actually live in, we need to compile this list
manually. This patch introduces the database, with the usual match formats
and a single property to be set on a device: MOUSE_DPI
That is either a single value for most mice, or a list of values for mice
that can change resolution at runtime. The exact format is detailed in the
hwdb file.
Note that we're explicitly overshooting the requirements we have for
libinput atm. Frequency could be detected in software and we don't
actually use the list of multiple resolutions (because we can't detect
when they change anyway). However, we might as well collect those values
from the get-go, adding/modifying what will eventually amount to hundreds
of entries is a bit cumbersome.
Note: we rely on the input_id builtin to tag us as mouse first, ordering
of the rules is important.
(David: fixed up typos and moved hwdb file into ./hwdb/)
Make screened character set consistent with unit_name_mangle() by splitting off
the escaping loop into a separate function.
Before this fix, unit names such as `foo@bar.target` would get transformed
into `foo\x40bar.target` when unit_name_mangle_with_suffix() is used.
https://bugs.freedesktop.org/show_bug.cgi?id=86711
When I tryed to run journalctl with --follow and --since arguments it
behaved very strangely.
First It prints logs from what I specified in --since argument, then
printed 10 lines (as is default in --follow) and when app put something
new in to log journalctl printed everithing from the last printed line.
How to reproduce:
1. run: journalctl -m --since 14:00 --follow
Then you'll see 10 lines of logs since 14:00. After that wait until some
app add something in the journal or just run `systemd-cat echo test`
2. After that journalctl will print every single line since 14:00 and will
follow as expected.
As long as --since and --follow will eventually print all relevant
lines, I seen no reason why not to print them right away and not after
first new message in journal.
Relevant bugzillas:
https://bugs.freedesktop.org/show_bug.cgi?id=71546https://bugs.freedesktop.org/show_bug.cgi?id=64291
For IPv6, the kernel returns EINVAL if a route is added with the
RTA_GATEWAY attribute set to in6addr_any (::). A route without a
gateway is useful in some situations, such as layer 3 tunneling
(sit, gre, etc.).
This patch prevents the RTA_GATEWAY attribute from being added
when route.in_addr is ip6addr_any (::).
This makes it possible to drop in logind configuration snippets from a
package or other configuration management mechanism.
Add documentation to the header of /etc/logind.conf pointing the user at
/etc/logind.conf.d/*.conf.
Introduce a new helper, conf_parse_many, to parse configuration files in
a search path.
Several different systemd tools define a nulstr containing a standard
series of configuration file directories, in /etc, /run, /usr/local/lib,
/usr/lib, and (#ifdef HAVE_SPLIT_USR) /lib. Factor that logic out into
a new helper macro, CONF_DIRS_NULSTR.
systemd-journald would refuse to start if it received an unknown
socket from systemd. This is annoying, because the failure more for
systemd-journald is unpleasant: systemd will keep restarting journald,
but most likely the same error will occur every time. It is better
to continue. journald will try to open missing sockets on its own,
so things should mostly work.
One question is whether to close the sockets which cannot be parsed or
to keep them open. Either way we might lose some messages. This
failure is most likely for the audit socket (selinux issues), which
can be opened multiple times so this not a problem, so I decided to
keep them open because it makes it easier to debug the issue after the
system is fully started.
systemd stops adding automatic dependencies on swap.target to swap
units. If a dependency is required, it has to be added by unit
configuration. fstab-generator did that already, except that now it is
modified to create a Requires or Wants type dependency, depending on
whether nofail is specified in /etc/fstab. This makes .swap units
obey the nofail/noauto options more or less the same as .mount units.
Documentation is extended to clarify that, and to make
systemd.mount(5) and system.swap(5) more similar. The gist is not
changed, because current behaviour actually matches existing
documentation.
https://bugs.freedesktop.org/show_bug.cgi?id=86488
This adds a new log_emergency() function, which is equivalent to
log_error() for non-PID-1, and logs at the highest priority for PID 1.
Some messages which occur before freezing are converted to use it.
When checking if the transaction is destructive, we need to check if the
previously installed job is a superset of the new job (and hence the new
job will fold into the installed one without changing it), not the other
way around.
Several functions called from transaction_activate() need to correctly
handle the case where a JOB_NOP job is being checked against a unit's
pending job. The assumption that JOB_NOP never merges with other job
types was correct, but since the job_type_is_*() functions are
implemented using the merge lookup, they need to special-case JOB_NOP
to avoid hitting assertion failures.
This makes udevadm trigger mirror udevadm info, except that multiple
device names can be specified. Instructions in 60-keyboard.hwdb should
now actually work.
udevadm(8) is updated, but it could use a bit more polishing.
https://bugs.freedesktop.org/show_bug.cgi?id=82311
On the system and user busses we create it's the receiver that chooses
which metadata is attched, not the sender, hence set the requirement
mask to ANY, to allow any current of future credential bit to be
attached.
Lennart Poettering