kernel 3.4+ needs cifs-utils to mount CIFS filesystems.
the kernel itself (and busybox's cifs mount code) are no longer able
to do this in some/most cases and will error out saying:
"CIFS VFS: connecting to DFS root not implemented yet"
Nixos' qemu-vm target is hurt by this, as it wants to mount /nix/store
via cifs very early in the boot process.
This commit makes sure the initrd for affected kernels is built with
cifs-utils if needed.
proxy_arp (and proxy_ndp for ipv6) can be turned on on a few
interfaces (at least 2).
This is mainly useful for creating pseudo-bridges between a real
interface and a virtual network such as VPN or a virtual machine for
interfaces that don't support real bridging (most wlan interfaces).
As ARP proxying acts slightly above the link-layer, below-ip traffic
isn't bridged, so things like DHCP won't work. The advantage above
using NAT lies in the fact that no IP addresses are shared, so all
hosts are reachable/routeable.
This especially annoyed me whenver I was doing nixos-rebuild switch and getting
logged out on all consoles. With this there now is services.mingetty.dontRestart
for heavy VT users to deactivate this behaviour.
As non-QWERTY keyboards don't feel so warm and cozy if they hug QWERTY LUKS
password prompts, it was on honor for me to serve King Dvorak XV to fight the
glorious keyboard war against... what?! Yes, I'm awake!
We're fighting with loadkeys to spit out busybox binary keymaps against loadkmap
(loadkeys does have a special target -b for that).
And yep, I'm somewhat abusing preLVMCommands, if someone got issues with that,
feel free to introduce a new substitute in stage-i-init.sh.
Sent from my iPhone
We had a "mount -o remount,rw none /" that was setting back 'relatime',
although we had set 'noatime' at initrd mount. Removing the word 'none' fixed
it.
Specifying a device (in this case 'none'), makes mount to forget previous
device options. According to manpage, it says not to read fstab or mtab. But the
effect is that of setting 'relatime', if it was mounted 'noatime.
Unless we search the entire filesystem to do a chown *and* restart
existing processes owned by that user, there is no sensible way that
we can change uids/gids. So don't try.
The option is disabled by default so that previously existing installations
aren't affected.
If you'd like to migrate to the fixed numeric id for Apache, set "fixUidAndGid
= true", edit the file "/etc/groups" and replace the old GID value with 54.
(NixOS can't do that for you because it refuses to change a GID that identifies
the primary group of a user.) Then run
find / -xdev -uid $oldUID -exec chown 54 {} +
find / -xdev -gid $oldGID -exec chgrp 54 {} +
to update ownership of all files that are supposed to be owned by Apache.
- Renamed system.build.menuBuilder to system.build.installBootLoader.
- ‘install-grub.pl’ (formerly grub-menu-builder.pl) now generates the
GRUB menu *and* installs GRUB (if necessary).
- ‘switch-to-configuration.sh’ has no boot loader specific knowledge
anymore. It just calls installBootLoader.
The old GRUB menu builder script is quite slow, typically taking
several seconds. This is a real annoyance since it's run every time
you switch to a new configuration. Therefore this patch replaces the
Bash script with a much faster Perl script. In a VirtualBox test, the
execution time went from 2.7s to 0.1s. The Perl version is also more
correct because it uses XML to get the GRUB configuration (through
builtins.toXML), so there are no shell escaping issues.
The new script currently lacks support for subconfigurations defined
through "nesting.children".
- The new option 'apacheHttpd' determines the version of the Apache
HTTP Server that's being used by this module. The default version
is Apache 2.2.x, as before.
- The new option 'configFile' allows users specify their own custom
config file for the web server instead of being limited to the one
that this module generates.
The environment variable "NIXOS_EXTRA_MODULES" is now checked to
contain a path to a file similar to modules/module-list.nix.
This gives the ability to include nixos modules that are not in the
nixos source tree.
This can be useful for modules that are still experimental, or which
aren't useful for other nixos users. Of course, this was already
possible to do this using a forked nixos tree, but with this
functionality, you can just rely on the nixos channel, easing things a
lot.
Ugly hack to get around the error "a string that refers to a store
path cannot be appended to a path". The underlying problem is that
you cannot do
"${./file1} ${./file2}"
but you can do
" ${./file1} ${./file2}"
Obviously we should allow the first case as well.
This allows hiding the implementation details for how to represent logstash
config types that don't directly map to nix expressions, particularly floats,
hashes, and name-value pair sets with repeated names. Instead of setting
__type and value directly, the user now uses these convenience functions to
generate their logstash config.
Since the logstash config file seemed very similar to a nixexpr, I decided
to map directly from nixexprs to logstash configs. I didn't realize until
too far in that this solution was probably way over-engineered, but it
works.
The new option ‘skins’ allows specifying a list of directories
providing skins to be added to the MediaWiki installation. The
‘defaultSkin’ option just sets the default.
resolvconf prefers a locally running BIND resolver over the forwarders; we just
have to tell it whether we have one or not. We use 'config.services.bind.enable'
to make that decision, assuming that people are not going to configure a local
BIND that won't respond to queries on 127.0.0.1. If we run into such a (weird)
case, then we'll need to introduce another variable for that purpose which can
be set independently from 'config.services.bind.enable'.
BusyBox doesn't handle the "auto" filesystem type very well: fsck will
just ignore such filesystems, and mount will only work properly if the
required kernel module is already loaded. Therefore, use blkid to
determine the filesystem type.
Also generate an /etc/fstab in the initrd rootfs on the fly. This is
useful if you're dropped into an emergency shell since it allows you
to say "fsck /dev/sda1" or "mount /dev/sda" and have the right thing
happen.
So it turns out that BusyBox doesn't auto-load modules for filesystems
that have type "auto", e.g. it doesn't figure out that it should load
the "iso9660" module should be loaded when mounting the CD-ROM. We
don't want to give the root FS on the CD type "iso9660", because that
breaks booting from a USB stick created by unetbootin. So make sure
the "iso9660" module is loaded.
Using BusyBox instead of Bash plus a bunch of other tools gives us a
much more feature-full, yet smaller initrd. In particular, BusyBox
contains networking commands such as ip and a DHCP client, useful for
NFS boots. It's also much more convenient for rescue situations
because the shell has builtin readline support and there are many more
tools (including vi).