Given that we are no longer inspecting the target of the /proc/self/exe
symlink, stop asserting that it has any properties. Remove the plumbing
for wrappersDir, which is no longer used.
Asserting that the binary is located in the specific place is no longer
necessary, because we don't rely on that location being writable only by
privileged entities (we used to rely on that when assuming that
readlink(/proc/self/exe) will continue to point at us and when assuming
that the `.real` file can be trusted).
Assertions about lack of write bits on the file were
IMO meaningless since inception: ignoring the Linux's refusal to honor
S[UG]ID bits on files-writeable-by-others, if someone could have
modified the wrapper in a way that preserved the capability or S?ID
bits, they could just remove this check.
Assertions about effective UID were IMO just harmful: if we were
executed without elevation, the caller would expect the result that
would cause in a wrapperless distro: the targets gets executed without
elevation. Due to lack of elevation, that cannot be used to abuse
privileges that the elevation would give.
This change partially fixes#98863 for S[UG]ID wrappers. The issue for
capability wrappers remains.
/proc/self/exe is a "fake" symlink. When it's opened, it always opens
the actual file that was execve()d in this process, even if the file was
deleted or renamed; if the file is no longer accessible from the current
chroot/mount namespace it will at the very worst fail and never open the
wrong file. Thus, we can make a much simpler argument that we're reading
capabilities off the correct file after this change (and that argument
doesn't rely on things such as protected_hardlinks being enabled, or no
users being able to write to /run/wrappers, or the verification that the
path readlink returns starts with /run/wrappers/).
Before this change it was crucial that nonprivileged users are unable to
create hardlinks to SUID wrappers, lest they be able to provide a
different `.real` file alongside. That was ensured by not providing a
location writable to them in the /run/wrappers tmpfs, (unless
disabled) by the fs.protected_hardlinks=1 sysctl, and by the explicit
own-path check in the wrapper. After this change, ensuring
that property is no longer important, and the check is most likely
redundant.
The simplification of expectations of the wrapper will make it
easier to remove some of the assertions in the wrapper (which currently
cause the wrapper to fail in no_new_privs environments, instead of
executing the target with non-elevated privileges).
Note that wrappers had to be copied (not symlinked) into /run/wrappers
due to the SUID/capability bits, and they couldn't be hard/softlinks of
each other due to those bits potentially differing. Thus, this change
doesn't increase the amount of memory used by /run/wrappers.
We set[1] ASPELL_CONF to the last nix profile containing lib/aspell in
2013. In 2017, aspell is patched[2] to search NIX_PROFILES, which
makes [1] not needed any more.
Deleting it is also agreed in this discussion[3].
[1]: 0192c02720
[2]: ba4cefe4ae
[3]: https://github.com/NixOS/nixpkgs/pull/30234
exiftool is written in Perl which appears to call `chown` as part of startup. This is blocked by the `@privileged` system call group. This causes a failure when changing image orientation.
Fixes: https://github.com/NixOS/nixpkgs/issues/249120
This change also generates the invidious config by putting JSON
snippets into a bash array and then using jq to merge them all into
a single configuration where later elements override previous
elements.
No content was changed, new titles are wrapped with () to signal that
they will need to be decided on in a future commit.
Section in the manual have been preserved with a simple redirect to
GitHub, the proper anchors should be filled out in a future commit once
the new section names are decided.
- Move contents of README.md from
nixos/modules/installer/tools/manpages/ to
nixos/doc/manual/contributing-to-this-manual.chapter.md.
- Don't mention DocBook as its obsolete and too specific.
- Rename derivation attribute name of configuration.nix(5) manual page,
both on the `contributing-to-this-manual.chapter.md`, and in other
places.
Since each such `nixos-*` tool has it's own derivation, exposed in pkgs,
There is no point in separating the manuals from the packages. If
someone wishes to have the tools without the manuals, they can use
meta.outputsToInstall to disable the installation of the manpages of
these packages. This Fixes#244450.
adwaita-qt tries valiantly to ensure a visual consistency
but unfortunately, it often falls into an uncanny valley instead.
Let’s make it opt-in again for more vanilla default experience.
Related: https://pagure.io/fedora-workstation/issue/351
Since Linux 5.7 it's possible to set `SO_BINDTODEVICE` via `setsockopt(2)`
as unprivileged user if this operation doesn't imply escaping a VRF
interface[1].
Dropping the wrapper is actually desirable because `captive-browser`
itself doesn't drop capabilities and as a result, the capabilities are
passed on to `chromium` itself[2].
For older kernels, this is still necessary, hence the wrapper will only
be added nowadays if the kernel is older than 5.7.
[1] c427bfec18
[2] 08450562e5/bind_device_linux.go (L11-L14)
and because our setcap wrapper makes all capabilities
inheritable.
In user namespaces where an unprivileged user is mapped as root and root
is unmapped, setuid bits have no effect. However setuid root
executables like mount are still usable *in the namespace* as the user
already has the required privileges. This commit detects the situation
where the wrapper gained no privileges that the parent process did not
already have and in this case does less sanity checking. In short there
is no need to be picky since the parent already can execute the foo.real
executable themselves.
Details:
man 7 user_namespaces:
Set-user-ID and set-group-ID programs
When a process inside a user namespace executes a set-user-ID
(set-group-ID) program, the process's effective user (group) ID
inside the namespace is changed to whatever value is mapped for
the user (group) ID of the file. However, if either the user or
the group ID of the file has no mapping inside the namespace, the
set-user-ID (set-group-ID) bit is silently ignored: the new
program is executed, but the process's effective user (group) ID
is left unchanged. (This mirrors the semantics of executing a
set-user-ID or set-group-ID program that resides on a filesystem
that was mounted with the MS_NOSUID flag, as described in
mount(2).)
The effect of the setuid bit is that the real user id is preserved and
the effective and set user ids are changed to the owner of the wrapper.
We detect that no privilege was gained by checking that euid == suid
== ruid. In this case we stop checking that euid == owner of the
wrapper file.
As a reminder here are the values of euid, ruid, suid, stat.st_uid and
stat.st_mode & S_ISUID in various cases when running a setuid 42 executable as user 1000:
Normal case:
ruid=1000 euid=42 suid=42
setuid=2048, st_uid=42
nosuid mount:
ruid=1000 euid=1000 suid=1000
setuid=2048, st_uid=42
inside unshare -rm:
ruid=0 euid=0 suid=0
setuid=2048, st_uid=65534
inside unshare -rm, on a suid mount:
ruid=0 euid=0 suid=0
setuid=2048, st_uid=65534
Removing a network from the joinNetworks list does not make the machine leave the network which is confusing.
This behavior is now clarified via the options description
Caddy 2.7.x does no longer return etags for files with unix modtimes of
0 and 1.
Files in /nix/store have a modtime of 1.
This is something that has been specifically implemented for nix.
For now, we decided to remove the test.
But I might reimplement a similar etag subtest some time in the future.
For non-interactive installation it's quite handy to be able to nix copy additional dependencies to the system.
While this is possible for the root user, we cannot easily ssh into it, as we don't allow root login with a password.
By making nixos a trusted user, we can do "passwd && sudo systemctl start sshd" and than run nixos-anywhere
This option only has an effect on the initial setup of Nextcloud and
changes later won't have any effect.
Same issue as with `adminpassFile` - it's only passed to the installer
command - but even worse because the username is frequently used as
unique ID in the database, so there's no trivial way to change it, even
imperatively.
https://github.com/NixOS/nixpkgs/pull/191540 indirectly broke kernel
passthru.tests; calling the testsForLinuxPackages and testsForKernel functions
with some args intended for some other exposed test-internal function.
Organise the passed-through functions under `passthru` to prevent this from
happening.
discoverTests tries to discover some sort of internal function and tries to call
it with the arguments for that internal function. This poses an issue when you
want to expose some other functions (i.e. a parameterisation for a test) in
nixosTests.
This commit allows a test to pass through arbitrary values via `.passthru`
without them having discovery applied to them; including functions.
As suggested by @nickcao this commit moves the defaults back to the
options. Only `homeserver.domain` stays in the config section since the
documentation module does not support referencing attributes of other
modules.
The upstream default for the thumbnail font is set to "Liberation Serif
Regular" located at /usr/share/fonts which is inaccessible under nix.
(2a2bf3bf55/src/paperless/settings.py (L894))
Paperless throws an error when parsing plaintext files without a valid
font. This change sets a nix default using the liberation_ttf package.
Netdata creates its control socket at /tmp/netdata-ipc by default, which
is insecure and actually inaccessible with systemd's PrivateTmp enabled.
Originally we patched its source code to move the socket to
/run/netdata/ipc. However, it was removed due to incompatibility when
upgrading to v1.41.0: 1d2a2dc7d0
Fortunately, this new version of netdata adds support for setting the
location of the control socket via the environment variable
NETDATA_PIPENAME. So let's set it for the netdata service and the
command line utility so that they can communicate properly.
Right now, the settings aren't additive which means that when I do
services.nextcloud.phpOptions."opcache.interned_strings_buffer = "23";
all other options are discarded because of how the module system works.
This isn't very nice in this case, though because wanting to override
a single option doesn't mean I want to discard the rest of the -
reasonable - defaults. Hence, the settings are showed as default in the
option's manual section, but are added with normal priority.
That means, to override _all_ options at once, an expression like
services.nextcloud.phpOptions = mkForce {
/* ... */
};
is needed. This is also way more intuitive IMHO because the `mkForce`
explicitly tells that everything will be modified.
Also, APCu enable and the memory & file-size limits are also written
into `services.nextcloud.phpOptions` rather than adding them
silently before passing all options to the PHP package. This has the
benefit that users will realize on evaluation time that they configured
options that would otherwise be set by the module on its own.
Allow giving a custom package containing the `systemd-repart` binary.
Defaults to `pkgs.systemd`. This option opens up the possibility to use
a different package for the image builder and the system configuration.
For example, someone could use this option to build an image with a
patched systemd while still using the upstream nixpkgs systemd package
(i.e., `pkgs.systemd`) for the system configuration installed to the
created image.
This avoids creating a build-time reference on `boot.kernelParams` if
the configuration does not use a kernel, i.e., `boot.kernel.enable` is
set to `false`.
