This is a 277K (as of right now) addition that can greatly help in some
last recourse scenarios. The specific rEFInd setup will not be able to
boot the installer image, but this is not why it has been added. It has
been added to make use of its volumes scanning capabilities to boot
existing EFI images on the target computer, which is sometimes necessary
with buggy EFI. While is isn't NixOS's job to fix buggy EFI, shipping
this small bit with the installer will help the unlucky few.
Example scenario: two wildly different EFI implementation I have
encountered have fatal flaws in which they sometimes will lose all the
settings, this includes boot configuration. This is compounded by the
fact that the two specific and distinct implementation do not allow
manually adding ESP paths from their interface. The only recourse is to
let the EFI boot the default paths, EFI/boot/boot{platform}.efi, which
is not a default location used by the NixOS bootloaders. rEFInd is able
to scan the volumes and detect the existing efi bootloaders, and boot
them successfully.
Following up https://github.com/NixOS/nixpkgs/pull/23665
Bootable USB-drives are not limited to ISO-images, there can be "normal" MBR/GPT-partitioned disk connected via USB-rack.
Also, "uas" implies "usb-storage", so there is no need to mention both.
This reverts commit 095fe5b43d.
Pointless renames considered harmful. All they do is force people to
spend extra work updating their configs for no benefit, and hindering
the ability to switch between unstable and stable versions of NixOS.
Like, what was the value of having the "nixos." there? I mean, by
definition anything in a NixOS module has something to do with NixOS...
When running e.g. `nixos-option boot.kernelPackages` I get an output
like this on the current unstable channel (18.09pre144959.be1461fc0ab):
```
$ nixos-option boot.kernelPackages
Value:
*exit 1*
```
This is fairly counter-intuitive as I have no clue what might went
wrong. `strace` delivers an output like this:
```
read(3, "error: Package \342\200\230cryptodev-linu"..., 128) = 128
read(3, "ux/cryptodev/default.nix:22 is m"..., 128) = 128
read(3, "lowBroken = true; }\nin configura"..., 128) = 128
read(3, "you can add\n { allowBroken = tr"..., 128) = 128
read(3, "n)\n", 128) = 3
read(3, "", 128) = 0
```
`nixos-option` evaluates the system config using `nix-instantiate` which
might break when the evaluation fails (e.g. due to broken or unfree
packages that are prohibited to evaluate by default). The script aborts
due to the shebang `@shell@ -e`.
In order to ensure that no unexpected
behavior occurs due to removing `-e` from the interpreter the easiest
way to work around this was to wrap `nix-instantiate` in `evalNix()`
with a `set +e`. The function checks the success of the evaluation with
`$?` in the end. Additionally `evalNix` shouldn't break, if one
evaluation (e.g. the values that contain a package set by default) to
return additional information like a description.
With the change `nixos-option boot.kernelPackages` delivers the
following output for me:
```
Value:
error: Package ‘cryptodev-linux-1.9-4.14.52’ in /nix/store/47z2s8cwppymmgzw6n7pbcashikyk5jk-nixos/nixos/pkgs/os-specific/linux/cryptodev/default.nix:22 is marked as broken, refusing to evaluate.
Default:
{ __unfix__ = <LAMBDA>; acpi_call = <CODE>; amdgpu-pro = <CODE>; ati_drivers_x11 = <CODE>; batman_adv = <CODE>; bbswitch = <CODE>; bcc = <CODE>; beegfs-module = <CODE>; blcr = <CODE>; broadcom_sta = <CODE>; callPackage = <CODE>; cpupower = <CODE>; cryptodev = <CODE>; dpdk = <CODE>; e1000e = <CODE>; ena = <CODE>; evdi = <CODE>; exfat-nofuse = <CODE>; extend = <CODE>; facetimehd = <CODE>; fusionio-vsl = <CODE>; hyperv-daemons = <CODE>; ixgbevf = <CODE>; jool = <CODE>; kernel = <CODE>; lttng-modules = <CODE>; mba6x_bl = <CODE>; mwprocapture = <CODE>; mxu11x0 = <CODE>; ndiswrapper = <CODE>; netatop = <CODE>; nvidiaPackages = <CODE>; nvidia_x11 = <CODE>; nvidia_x11_beta = <CODE>; nvidia_x11_legacy304 = <CODE>; nvidia_x11_legacy340 = <CODE>; nvidiabl = <CODE>; odp-dpdk = <CODE>; openafs = <CODE>; openafs_1_8 = <CODE>; perf = <CODE>; phc-intel = <CODE>; pktgen = <CODE>; ply = <CODE>; prl-tools = <CODE>; recurseForDerivations = true; rtl8192eu = <CODE>; rtl8723bs = <CODE>; rtl8812au = <CODE>; rtl8814au = <CODE>; rtlwifi_new = <CODE>; sch_cake = <CODE>; spl = <CODE>; splLegacyCrypto = <CODE>; splStable = <CODE>; splUnstable = <CODE>; stdenv = <CODE>; sysdig = <CODE>; systemtap = <CODE>; tbs = <CODE>; tmon = <CODE>; tp_smapi = <CODE>; usbip = <CODE>; v4l2loopback = <CODE>; v86d = <CODE>; vhba = <CODE>; virtualbox = <CODE>; virtualboxGuestAdditions = <CODE>; wireguard = <CODE>; x86_energy_perf_policy = <CODE>; zfs = <CODE>; zfsLegacyCrypto = <CODE>; zfsStable = <CODE>; zfsUnstable = <CODE>; }
Example:
{ _type = "literalExample"; text = "pkgs.linuxPackages_2_6_25"; }
Description:
"This option allows you to override the Linux kernel used by\nNixOS. Since things like external kernel module packages are\ntied to the kernel you're using, it also overrides those.\nThis option is a function that takes Nixpkgs as an argument\n(as a convenience), and returns an attribute set containing at\nthe very least an attribute <varname>kernel</varname>.\nAdditional attributes may be needed depending on your\nconfiguration. For instance, if you use the NVIDIA X driver,\nthen it also needs to contain an attribute\n<varname>nvidia_x11</varname>.\n"
Declared by:
"/nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/nixos/modules/system/boot/kernel.nix"
Defined by:
"/home/ma27/Projects/nixos-config/system/boot.nix"
```
This partially reverts a change from e88f28965a
which removed the `mount --rbind /sys`.
While true that the activation scripts will mount `sysfs` at `/sys`,
none of the mountpoints lower in the `/sys` tree are handled by the
activation script, which includes `efivarfs`.
This fixes #38477 since it ensures the presence of `efivarfs` in the
`/sys` tree, which is why the systemd-boot installation failed.
Nothing probably uses this, but let's be pedantic and have the
pre-included channel on the install media be as close as possible to
what 'nix-channel --update' will give them.
The only remaining difference is that the channel adds programs.sqlite,
which is fundamentally unfixable.
Because it improves out-of-the-box user experience a lot (IMHO).
(zsh completion is already on by default.)
Remove "programs.bash.enableCompletion = true" from
nixos-generate-config.pl, which feels superflous now.
- Add `imageName` and `imageBaseName` options similar to the `isoName`
and `isoBaseName` options
- Make the filename of the iso match what iso-image.nix does
- Generate a nix-support/hydra-build-products like iso-image.nix does
The function arguments for make-squashfs.nix have changed in
df117acab7, so we need to change them here
as well.
The boot.netboot NixOS VM test now succeeds again.
Signed-off-by: aszlig <aszlig@nix.build>
Cc: @edolstra
Of course, you'll get a bunch of warnings from the activation script:
$ nixos-enter --root /tmp/mnt/
setting up /etc...
mount: /dev: permission denied.
mount: /dev/pts: permission denied.
mount: /dev/shm: permission denied.
mount: /sys: permission denied.
/nix/var/nix/profiles/system/activate: line 74: /proc/sys/kernel/modprobe: Permission denied
chown: changing ownership of '/run/wrappers/wrappers.0pKlU8JsvV/dbus-daemon-launch-helper': Invalid argument
NOTE: Under Linux, effective file capabilities must either be empty, or
exactly match the union of selected permitted and inheritable bits.
Failed to set capabilities on file `/run/wrappers/wrappers.0pKlU8JsvV/ping' (Operation not permitted)
chown: changing ownership of '/run/wrappers/wrappers.0pKlU8JsvV/unix_chkpwd': Invalid argument
[root@nixos:/]#
The use of Nix 2.0 significantly simplifies the installer, since we
can just pass a different store URI (--store /mnt) - it's no longer
needed to set up a chroot environment for the build, and to bootstrap
Nix into the chroot.
Also, commands that need to run in the installation (namely boot
loader installation and setting a root password) are now executed
using nixos-enter.
This also removes the need for nixos-prepare-root since any required
initialisation is done by Nix or by the activation script.
In particular, it conflicts with `services.tlp.enable` option.
There exists workaround:
```
powerManagement.cpuFreqGovernor = lib.mkForce null;
services.tlp.enable = true;
```
But should it?
Regression introduced by 801c920e95.
Since then, the btrfsSimple subtest of the installer VM test fails with:
Btrfs did not return a path for the subvolume at /
The reason for this is that the output for "btrfs subvol show" has
changed between version 4.8.2 and 4.13.1.
For example the output of "btrfs subvol show /" in version 4.8.2 was:
/ is toplevel subvolume
In version 4.13.1, the output now is the following and thus the regular
expressions used in nixos-generate-config.pl and install-grub.pl now
match (which results in the error mentioned above):
/
Name: <FS_TREE>
UUID: -
Parent UUID: -
Received UUID: -
Creation time: -
Subvolume ID: 5
Generation: 287270
Gen at creation: 0
Parent ID: 0
Top level ID: 0
Flags: -
Snapshot(s):
In order to fix this I've changed nixos-generate-config.pl and
install-grub.pl, because both use "btrfs subvol show" in a similar vein,
so the regex for parsing the output now doesn't match anymore whenever
the volume path is "/", which should result in the same behaviour as we
had with btrfs-progs version 4.8.2.
Tested against the btrfsSimple, btrfsSubvols and btrfsSubvolDefault
subtests of the installer VM test and they all succeed now.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
All available options were just enabling
hardware.enableRedistributableFirmware. There were nix files without
modules which weren't referenced anywhere.
Add another option for debugging instead. Lots of users have been
complaining about this default behaviour.
This patch also cleans up the EFI bootloader entries in the ISO.
This version should have more conventional regexes that work across many
platforms and regex engines. This is an issue because up until Nix 1.11,
Nix called out to the libc regex matcher, which behaved differently on
Darwin and Linux. And in Nix 1.12, we're moving to std::regex which will
also behave differently here.
And yes, I do actually evaluate make-disk-image.nix on Darwin ;)
Due the recent inclusion of broadcom-bt-firmware in enableAllFirmware,
it was required to set `nixpkgs.config.allowUnfree` to obtain the full
list. To make this dependency more explicit an assertion is added and an
alternative option `enableRedistributalFirmware` is provided to only
obtain firmware with an license allowing redistribution.
This changes much of the make-disk-image.nix logic (and thus most NixOS
image building) to use LKL to set up the target directory structure rather
than a Linux VM. The only work we still do in a VM is less IO-heavy stuff
that while still time-consuming, is less of the overall load. The goal is
to kill more of that stuff, but that will require deeper changes to NixOS
activation scripts and switch-to-configuration.pl, and I don't want to
bite off too much at once.
The key distinction I'm drawing is that there's a component that deals
with the store of the machine being built, and another component for
the store building it. The inner part of it assumes nothing from the
builder (doesn't need chroot or root powers) so it can run comfortably
inside a Nix build, as well as nixos-rebuild. I have some upcoming work
that will use that to significantly speed up and streamline image builds
for NixOS, especially on virtualized hosts like EC2, but it's also a
reasonable speedup on native hosts.
Currently, if you try to build a NixOS config including channel.nix, e.g.:
nix-build -I nixpkgs=. -I nixos-config=nixos/modules/installer/cd-dvd/installation-cd-minimal.nix nixos -A config.system.build.isoImage
twice in a row, you end up with two different build results. This is
caused by the 'result' symlink of the first build affecting the channel
contents of the second build.
If we use filterSource with a predicate that ignores the 'result'
symlinks, the problem is gone. Do the same thing for VIM/Emacs
swap/backup files to avoid even more 'spurious' rebuilds.
Additionally, filter out the '.git' directory at the same time, as we
'rm -rf' it from the result anyway. This avoids a considerable amount of
unnecessary file I/O copying and deleting the .git directory.
Turns out all variants of start.elf and fixup.dat are needed (depending
on what's in config.txt). I was under the mistaken impression that you
were supposed to rename one of the variants to switch using them, but
nope.
