The old open-source driver for AMD/ATI GPUs is commonly known as "radeon"
despite the historical package name xf86-video-ati. For example it presents
itself as RADEON in the Xorg log. So adding "radeon" to videoDrivers should
work.
Also changed the docs for the videoDrivers option to use "radeon" in the
default value instead of "ati".
Fixes#37917
A new internal option `hardware.opengl.setLdLibraryPath` is added which controls if `LD_LIBRARY_PATH` should be set to `/run/opengl-driver(-32)/lib`. It is false by default and is meant to be set to true by any driver which requires it. If this option is false, then `opengl.nix` and `xserver.nix` will not set `LD_LIBRARY_PATH`.
Currently Mesa and NVidia drivers don't set `setLdLibraryPath` because they work with libglvnd and do not override libraries, while `amdgpu-pro`, `ati` and `parallels-guest` set it to true (the former two really need it, the last one doesn't build so is presumed to).
Additionally, the `libPath` attribute within entries of `services.xserver.drivers` is removed. This made `xserver.nix` add the driver path directly to the `LD_LIBRARY_PATH` for the display manager (including X server). Not only is it redundant when the driver is added to `hardware.opengl.package` (assuming that `hardware.opengl.enable` is true), in fact all current drivers except `ati` set it incorrectly to the package path instead of package/lib.
This removal of `LD_LIBRARY_PATH` could break certain packages using CUDA, but only those that themselves load `libcuda` or other NVidia driver libraries using `dlopen` (not if they just use `cudatoolkit`). A few have already been fixed but it is practically impossible to test all because most packages using CUDA are libraries/frameworks without a simple way to test.
Fixes#11434 if only Mesa or NVidia graphics drivers are used.
I'm not 100% sure about the incompatibility lines,
but I believe it's better to discourage these anyway.
If you find better information, feel free to amend...
The 32-bit thing is completely GPU-agnostic, so I can't see why we had
it separately for proprietary drivers and missing for the rest.
Instead of searching `/usr` it should search for the `xkb`,
$XDG_DATA_DIRS will be searched. With this approach we allow compliance
on NixOS and non-NixOS systems to find `symbols` in the `xkb` directory.
The patch has been accepted by upstream, but isn't released yet, so this
is mainly a temporary fix until we can bump ZSH to the next stable version.
The `xserver` module links `/share/X11/xkb` to `/run/current-system` to
make this possible.
The fix can be tested inside the following VM:
```
{
zshtest = {
programs.zsh.enable = true;
users.extraUsers.vm = {
password = "vm";
isNormalUser = true;
};
services.xserver.enable = true;
};
}
```
Fixes#46025
Switch from slim to lightdm as the display-manager.
If plasma5 is used as desktop-manager use sdddm.
If gnome3 is used as desktop-manager use gdm.
Based on #12516
This adds configuration options which automate the configuration of NVIDIA Optimus using PRIME. This allows using the NVIDIA proprietary driver on Optimus laptops, in order to render using the NVIDIA GPU while outputting to displays connected only to the integrated Intel GPU. It also adds an option for enabling kernel modesetting for the NVIDIA driver (via a kernel command line flag); this is particularly useful together with Optimus/PRIME because it fixes tearing on PRIME-connected screens.
The user still needs to enable the Optimus/PRIME feature and specify the bus IDs of the Intel and NVIDIA GPUs, but this is still much easier for users and more reliable. The implementation handles both the X configuration file as well as getting display managers to run certain necessary `xrandr` commands just after X has started.
Configuration of commands run after X startup is done using a new configuration option `services.xserver.displayManager.setupCommands`. Support for this option is implemented for LightDM, GDM and SDDM; all of these have been tested with this feature including logging into a Plasma session.
Note: support of `setupCommands` for GDM is implemented by making GDM run the session executable via a wrapper; the wrapper will run the `setupCommands` before execing. This seemed like the simplest and most reliable approach, and solves running these commands both for GDM's X server and user X servers (GDM starts separate X servers for itself and user sessions). An alternative approach would be with autostart files but that seems harder to set up and less reliable.
Note that some simple features for X configuration file generation (in `xserver.nix`) are added which are used in the implementation:
- `services.xserver.extraConfig`: Allows adding arbitrary new sections. This is used to add the Device section for the Intel GPU.
- `deviceSection` and `screenSection` within `services.xserver.drivers`. This allows the nvidia configuration module to add additional contents into the `Device` and `Screen` sections of the "nvidia" driver, and not into such sections for other drivers that may be enabled.
The original idea behind this change (described in ticket #11064) was to
improve the assertions to avoid that users of the X server accidentally
forget to configure a DM or WM.
However this caused several issues with setups that require X, but no DM
or WM. The keymap testcases became instable as well as now disabling DMs
needs to be done explicitly.
(see https://github.com/NixOS/nixpkgs/pull/31268#issuecomment-347080036)
In the end the idea behind the change and #11064 was obviously a
mistake, so reverting it completely for now should be fine.
X libraries in LD_LIBRARY_PATH seem to not be needed anymore.
I've tracked this addition as far as I could
(02cef04c81) and they seem to be added for unfree
NVIDIA and ATI drivers but at least for NVIDIA they are not needed anymore. We
can add them with patchelf instead if it turns out to be the case with ATI.
This makes memoization of Nixpkgs evaluation less effective, since
some Nixpkgs invocations may have 'config = {}' while others may have
'config = { xorg = {}; }'.
Instead set 'config = {}'.
With libinput used for keyboard, base rules produce incorrect keyboard
layouts. We are removing the option as recommended in the XKB configuration
guide [1] to let X server choose the ruleset. It looks like it chooses
evdev rules which seem to work for some reason
[1]: https://www.x.org/releases/X11R7.6/doc/xorg-docs/input/XKB-Config.html#id2521360
This reverts commit 93c54acf97.
This reopens#30517 @nbp @Ma27
Breaking people's config for this is hardly reasonable as is. If it
absolutely cannot be avoided, at the very least, we need to provide
clear instructions on what people need to upgrade in their config. I
actually had to bisect to the commit, to even find out what property I
should change or define, as the error message was useless. It didn't
even mention a property name.
Discussion on the PR seems to be ongoing, so I'm reverting this, so we
don't break people's systems on unstable.
Checking the keyboard layout has been a long set of hurdles so far, with
several attempts. Originally, the checking was introduced by @lheckemann
in #23709.
The initial implementation just was trying to check whether the symbols/
directory contained the layout name.
Unfortunately, that wasn't enough and keyboard variants weren't
recognized, so if you set layout to eg. "dvorak" it will fail with an
error (#25526).
So my improvement on that was to use sed to filter rules/base.lst and
match the layout against that. I fucked up twice with this, first
because layout can be a comma-separated list which I didn't account for
and second because I ran into a Nix issue (NixOS/nix#1426).
After fixing this, it still wasn't enough (and this is btw. what
localectl also does), because we were *only* matching rules but not
symbols, so using "eu" as a layout won't work either.
I decided now it's the time to actually use libxkbcommon to try
compiling the keyboard options and see whether it succeeds. This comes
in the form of a helper tool called xkbvalidate.
IMHO this approach is a lot less error-prone and we can be sure that we
don't forget about anything because that's what the X server itself uses
to compile the keymap.
Another advantage of this is that we now validate the full set of XKB
options rather than just the layout.
Tested this against a variety of wrong and correct keyboard
configurations and against the "keymap" NixOS VM tests.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @lheckemann, @peti, @7c6f434c, @tohl, @vcunat, @lluchs
Fixes: #27597
* lib: introduce imap0, imap1
For historical reasons, imap starts counting at 1 and it's not
consistent with the rest of the lib.
So for now we split imap into imap0 that starts counting at zero and
imap1 that starts counting at 1. And imap is marked as deprecated.
See c71e2d4235 (commitcomment-21873221)
* replace uses of lib.imap
* lib: move imap to deprecated.nix
Regression introduced by 44c64fef16.
The services.xserver.layout option allows to specify more than one
layout separated by comma, which the commit above didn't take into
account.
This is very similar to @lheckemann's pull request (#26984) but differs
in the following ways:
* Print out the full list available layouts (as suggested by @0xABAB
in [1]).
* Loop over $layout using the default IFS (and thus no need for
escaping ${cfg.layout}), because the layouts won't contain white
spaces.
* Re-do the error message, which now uses multiple echos instead of a
heredoc, so the line is wrapped according to the viewers terminal
width.
I've tested this with several good and bad layouts and also against the
keymap NixOS VM subtests.
[1]: https://github.com/NixOS/nixpkgs/pull/26984#discussion_r125146700
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Fixes: #26961Closes: #26984
First of all, thanks to @pbogdan for getting this problem reproduced:
https://github.com/NixOS/nixpkgs/commit/2014db3efcd2a#commitcomment-22815396
Also thanks to @vcunat for bringing this to my attention:
https://github.com/NixOS/nixpkgs/commit/44c64fef16ed5#commitcomment-22813503
Although it is not entirely clear why Nix has killed the build prior to
finishing, it seems to be related to the process substition I was using.
So instead of using "exec touch", let's wrap this inside an if so we
don't exit too early.
Tested this against all sub-tests in nixos/tests/keymap.nix and also a
few configurations with wrong keyboard layout definitions.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Enumerating the symbols directory doesn't include variants, so we're now
basically doing what "localectl list-x11-keymap-layouts" does but we use
sed instead.
The reason I'm not using localectl directly is because the path to
rules/base.lst is hardcoded in the systemd source.
Of course, the XKB specification allows for much more complicated rules,
but at least this should cover the most basic ones including variants.
So the sed expression itself is just for listing the available layouts
and variants and we use a grep with -xF to match only full lines without
interpreting regular expressions.
This should again allow to set "dvorak" as the layout option.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @lheckemann
Fixes: #25526
When you have a setup consisting of multiple monitors, the default is
that the first monitor detected by xrandr is set to the primary monitor.
However this may not be the monitor you need to be set as primary. In
fact this monitor set to primary may in fact be disconnected.
This has happened for the original submitter of the pull request and it
affected these programs:
* XMonad: Gets confused with Super + {w,e,r}
* SDDM: Puts the login screen on the wrong monitor, and does not
currently duplicate the login screen on all monitors
* XMobar: Puts the XMobar on the wrong monitor, as it only puts the
taskbar on the primary monitor
These changes should fix that not only by setting a primary monitor in
xrandrHeads but also make it possible to make a different monitor the
primary one.
The changes are also backwards-compatible.
It was asked by @CMCDragonkai to elaborate on that, so let's just do
this by actually providing a code comment.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Using invalid module options in the submodule isn't very nice, because
it doesn't give very useful errors in case of type mismatch, also we
don't get descriptions of these options as they're effecively
nonexistent to the module system. Another downside of this is that
merging of these options isn't done correctly as well (eg. for
types.lines).
So we now have proper submodules for each xrandrHead and we also use
corcedTo in the type of xrandrHeads so that we can populate the
submodule's "output" option in case a plain string is defined for a list
item.
Instead of silently skipping multiple primary heads, we now have an
assertion, which displays a message and aborts configuration evaluation
appropriately.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
It was already ordered after systemd-udev-settle.service, but that
doesn't do anything if no other units require
systemd-udev-settle.service. This was causing random failures during X
server startup, e.g.
machine# [ 12.691372] display-manager[607]: (EE) open /dev/dri/card0: No such file or directory
http://hydra.nixos.org/build/41062823
gnome-x-session provides good defaults which we really should not
override.
We have to add assertions to gdm.nix if the user specified one of those.
enableTCP must be configured through a gnome setting
dunno why we have terminate but it probably breaks stuff
We should expose configFile so we can use it from gdm module.
* x11 module: don't restart the display manager indefinitely
If the display managers crashes continuously in loops it prevents the
user from switching to the console and try to fix things. Especially
when using the "auto" display manager it can happen quite easily.
* x11 module: fix display manager restart timeouts
It takes more than 1 second to boot the X server.
---
Using the configure option relieves us of the patch and passing the path
via the env var in many places. Also the env var may not be inherited
when components like gdm spawn new sessions.
We were pulling in 44 MiB of fonts in the default configuration, which
is a bit excessive for headless configurations like EC2
instances. Note that dejavu_minimal ensures that remote X11-forwarded
applications still have a basic font regardless.
This ensures that "journalctl -u display-manager" does what you would
expect in 2016. However, the main reason is to ensure that our VM
tests show the output of the X server.
A slight problem is that with KDE user switching, messages from the
various X servers end up in the same place. However, that's an
improvement over the previous situation, where the second X server
would overwrite the /var/log/X.0.log of the first. (This was caused by
the fact that we were passing a hard-coded value for -logfile.)
Commit 98d9bba introduced this option as a nullOr type and it actually
checks whether null has been set and only appends -dpi if that's the
case. So let's actually set the default to null instead of 0.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
GnuPG 2.1.x changed the way the gpg-agent works, and that new approach no
longer requires (or even supports) the "start everything as a child of the
agent" scheme we've implemented in NixOS for older versions.
To configure the gpg-agent for your X session, add the following code to
~/.xsession or some other appropriate place that's sourced at start-up:
gpg-connect-agent /bye
GPG_TTY=$(tty)
export GPG_TTY
If you want to use gpg-agent for SSH, too, also add the settings
unset SSH_AGENT_PID
export SSH_AUTH_SOCK="${HOME}/.gnupg/S.gpg-agent.ssh"
and make sure that
enable-ssh-support
is included in your ~/.gnupg/gpg-agent.conf.
The gpg-agent(1) man page has more details about this subject, i.e. in the
"EXAMPLES" section.
- if xserver.tty and/or display are set to null, then don't specify
them, or the -logfile argument in the xserverArgs
- For lightdm, we set default tty and display to null and we determine
those at runtime based on arguments passed. This is necessary because
we run multiple X servers so they can't all be on the same display
