…ith fd passing enabled
Let's add a simple flag that enables fd passing for all connections of a
server. It's much easier to use this than to install a connect handler
which manually enables this for each connection.
Devices under /sys/firmware/ do not have subsystems. Hence, the
validation in sd_device_new_from_subsystem_sysname() ->
device_new_from_path_join() always failed.
Fixes a bug introduced by cd7c71154c
(v257).
Fixes#35861.
Supersedes #35308 (cherry-picked one commit and replaced the rest)
(I left a few comments that's folded by GitHub. Please make sure to
check them too.)
Let's add a simple flag that enables fd passing for all connections of a
server. It's much easier to use this than to install a connect handler
which manually enables this for each connection.
Devices under /sys/firmware/ do not have subsystems. Hence, the
validation in sd_device_new_from_subsystem_sysname() ->
device_new_from_path_join() always failed.
Fixes a bug introduced by cd7c71154c (v257).
Fixes#35861.
If bus_add_match_full() is called without install callback and we failed
to install the signal match e.g. by timeout, then add_match_callback()
will disconnect from the bus. Let's use a custom install handler and
handle failures gracefully.
This does not *solve* the root cause of issue
https://github.com/systemd/systemd/issues/30573, but should improve the
situation when the issue is triggered.
The new json_dispatch_const_path() is to json_dispatch_path() what
sd_json_dispatch_const_string() is to sd_json_dispatch_ string(), i.e.
doesn't implicitly strdup() the string, but gives you the pointer into
the JSON structure, and thus requires you to keep it pinned.
This is a minor compat break, but given the slow adoption of the
sd-path.h APIs I think it's one we should take. Basically, the idea is
that if the user provides a suffix path with a trailing slash (thus
encoding in the path that the last element must be a dir), we should
keep it in place, and not suppress it, in order to not willy nilly
reduce the amount of information contained in the path.
Simplifications that do not alter meaning, and do not suppress
information should be fine to apply to a path, but otherwise we really
should be conservative on this.
Now that we only support serialization into a memfd (rather than a file
in /run/) there's no point to check the free space in /run/. Let's drop it.
One error scenario gone. Yay.
This is just like memfd_new(), but allows fine grained control of the
sealing flags.
This switches over all uses of memfd_new() where we actually want
sealing to use memfd_new_full().
This then allows use to use memfd_new() for two further calls, where we
previously used the more lowlevel memfd_create_wrapper().
This commit add the `-i` option to `udevadm trigger` that force it to
match parent devices even if they're excluded from filters.
The rationale is that some embedded devices have a huge number of
platform devices ( ~ 4k for MX8 ) they are there because they're defined
in the device tree but there isn't any action or udev rules associated
with them.
So at boot a significant time is spend triggering and processing rules
for devices that don't produce any effect and we would like to filter
them by calling:
```
udevadm trigger --type=device --action=add -s block -s tty
```
instead of the normal
```
udevadm trigger --type=device --action=add
```
so we can use filter to filter out only subsystems for we we know that
we have rules in place that do something useful.
On the other side action / rules are not triggered until the parent is
triggered ( which is part of another subsystem), so the additional option
will allows udev to complete the coldplug with only the devices we care.
Example on iMX8:
.Without the new option
```
root@dev:~# udevadm trigger --dry-run -s block --action=add -v
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0boot0
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0boot1
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p1
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p2
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p3
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p4
```
.With the new option
```
root@dev:~# udevadm trigger --dry-run -i -s block --action=add -v
/sys/devices/platform
/sys/devices/platform/bus@5b000000
/sys/devices/platform/bus@5b000000/5b010000.mmc
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0boot0
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0boot1
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p1
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p2
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p3
/sys/devices/platform/bus@5b000000/5b010000.mmc/mmc_host/mmc0/mmc0:0001/block/mmcblk0/mmcblk0p4
```
Boot time reduction with this is place is ~ 1 second.
This function was listed in the public sd-varlink.h header, but not
actually made public. Fix that. It's quite useful, the comment in it
describes the usecase nicely.
Fixes: #35554
The values that were previously hardcoded in sd-varlink.c are now defined
in new varlink_set_info_systemd() and that function is called everywhere
where we create a server.
Fixes https://github.com/systemd/systemd/issues/35508.
As reported in the bug, the values were hardcoded for the systemd project.
https://varlink.org/Service lists vendor, product, version, url, and interfaces
as the mandatory parameters, so add an interface to set the first four. The
last field is set automatically based on the registered interfaces as before.
If the values are not filled in, we return empty strings. With NULL,
'varlinkctl info' would say:
(string):1:25: Object field 'vendor' has wrong type null, expected string.
When determining the poll events to wait for we need to take the queue
of pending messages that carry fds into account. Otherwise we might end
up not waking up if such an fd-carrying message is enqueued
asynchronously (i.e. not from a dispatch callback).
The auditing subsystem is still not virtualized for containers, hence the two
values don't really make sense inside them, they will just leak
information from outside into the container. Hence don't make use of the
data if we detect we are run inside of a container.
This has visible effects: logind will no longer try to reuse the
auditing session ids as its own session ids when run inside a container.
While are at it, modernize the calls in more ways:
1. switch to pidref behaviour, all but one of our uses are using pidref
anyway already.
2. use read_virtual_file() + proc_mounted()
3. reasonable distinguish ENOENT errors when reading the process proc
files: distinguish the case where /proc is not mounted, from the case
where the process is already gone, from where auditing is not enabled
in the kernel build.
The previous commit removed the UINT_MAX check for the fd array. Let's
now re-add one, but at a better place, and with a more useful limit. As
it turns out the kernel does not allow passing more than 253 fds at the
same time, hence use that as limit. And do so immediately before
calculating the control buffer size, so that we catch multiplication
overflows.
Let's move the helper from nss-resolve.c to generic code, as it's going
to be useful in #34640.
Also, let's tighten the rules, and refuse negative ifindexes, because
they are invalid.
We fucked that up in the original sd_listen() calls, and then we fixed
that on the newer flavours. But pour internal common implementation
should of course use the full range size_t, as it should be.
This then allows us to drop a redundant range check.
