First, pass in `self' again so that overriding works properly (thanks
for pointing that out, @edolstra)
Second, instead of having linuxPackages*.kernel mean something different
inside the set and out, add a new attribute linuxPackages*.kernelDev,
which for the generic kernel is simply linuxPackages*.kernel but for the
manual-config kernel is the `dev' output (which has the build tree,
source tree, etc.)
The second change required trivial modifications in a bunch of
expressions, I verified that all of the linuxPackages* sets defined in
all-packages.nix have the same drv paths before and after the change.
Signed-off-by: Shea Levy <shea@shealevy.com>
This makes it possible to still build out-of-tree modules without making a system using this kernel depend on the full source and build tree at runtime.
Note that references to the source tree are removed from kernel modules after build.
Ideally, this would be accomplished by modifying the Makefile that puts the reference there in the first place, but I haven't tracked that down yet.
As timeconst.pl was replaced by a bc script in the merge window for kernel 3.9,
we also need bc during build time, for more information about the reasons,
please have a look at the following commit:
torvalds/linux@70730bca13
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Conflicts:
pkgs/applications/networking/browsers/chromium/default.nix
pkgs/top-level/all-packages.nix
Merge conflicts seemed trivial, but a look from viric and aszlig would be nice.
NFS_FSCACHE and CIFS_FSCACHE are required to enable the local cache support for
NFS and CIFS (the 'fsc' mount option). The reasoning for enabling it from 3.2
and forward is that at least Ubuntu 12.04 (linux 3.2) has it turned on. Could
probably be enabled for earlier kernels too.
Conflicts:
pkgs/development/compilers/gcc/4.6/default.nix
pkgs/development/compilers/gcc/4.7/default.nix
The 4.7 had some weird parameters added in crossAttrs; I've removed
them, but I don't understand where they come from.
This is for consistency with terminology in stdenv (and the terms
"hostDrv" and "buildDrv" are not very intuitive, even if they're
consistent with GNU terminology).
The previos version did a for loop over the output of set, which spits out _all_
defined variables and their contents. This not only is dangerous if there is a
variable starting with CONFIG_ but also can't handle whitespace, as the IFS is
set to any (horizontal _and_ vertical) whitespace by default.
So, imagine (actually don't imagine, something like this is the case in a lot of
kernel configuration files) you have the following variable:
CONFIG_ARCH_HWEIGHT_CFLAGS="-fcall-saved-rdi -fcall-saved-rsi ..."
A loop with for and the default IFS would result in the following variable
pieces:
0: CONFIG_ARCH_HWEIGHT_CFLAGS="-fcall-saved-rdi
1: -fcall-saved-rsi
2: ..."
This obviously leads to the problem that this config variable is being cut off
at the first whitespace.
Another downside of this approach is that set not only returns variables but
functions as well. This could lead to quite a lot of unexpected behaviour and
confusion.
So the new approach doesn't source the kernel configuration anymore but uses
`read` to parse the file line-by line, setting IFS to '=', thus splitting all
configuration lines into key/value pairs.
Using parameter expansion, we ensure that we only read lines starting with
"CONFIG_". This particularily has the advantage of not being bash-specific,
should we choose to change to a different default shell someday.
Now, after we got a correct "CONFIG_" line, we're using a temporary variable to
split off the first quote from the result. Particularily the reason behind this
is shell compatibility again, as ${${foo#"}%"} only works in Bash, Zsh and
whatnot but not in plain SH.
And within the next line we obviously insert the no_firstquote variable without
it's last quote removed.
But, what about escaping?
First of all, if we'd just eval the $val variable, we would correctly unescape
the value, but this has the downside that variables within the content would be
expanded, for example look at this:
CONFIG_DEFCONFIG_LIST="/lib/modules/$UNAME_RELEASE/.config"
Well, obviously this is a bad example at the Nix sense, but just to show that
variables within kernel configuration entries aren't impossible.
And second, which would have been a show stopper if \" would be within $val: It
simply would end up being an invalid Nix expression, because \" would end up as
a ploin " within a double quoted string.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
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 just marks the problematic kernels.
An associated commit in nixos will use this info to fix the problem.
Since we keep the source in the kernel build anyway, no space is wasted by having a separate store path for the unpacked source. The upside is that the same source can be used to build the kernel multiple times, or generate configurations for later kernel builds, without going through the (fairly long) unpack and patch process multiple times.
Instead of manually adding which 'features' a kernel has based on its config here, just generate a nixexpr representation of the whole config and let places that need it (e.g. NixOS modules) decide how to map config options to features.
Note that if your kernel requires firmware whose hex file sits directly in ${src}/firmware rather than in a subdirectory thereof, you may need this patch for now: https://lkml.org/lkml/2012/7/31/724
The goal of this function is to make it possible to build a kernel with a user provided .config.
As a secondary goal, it will extract NixOS-relevant features from the config automatically.
As a tertiary goal, the build will aim to be simpler than the current generic kernel builder.sh.
Unfortunately, that simplicity is offset by the complexity of the feature extraction, especially since nix segfaults when trying to split the file into lines (so an import from a derivation is used)
(less sigill, less sigbus). Related to bad handling of FPU instructions.
I apply them only to linux 3.4, although I think they can apply to many older kernels too.
svn path=/nixpkgs/trunk/; revision=34522
as they are referenced from other kernel headers, this seems like the
best thing to do. Ubuntu seems to do so too.
Fixes issues with nvidia's binary driver and bbswitch on kernels > 3.3
svn path=/nixpkgs/trunk/; revision=34469
This incorporates the btrfs fix, so remove that patch. Also, I will test
that this builds after committing, and fix it if it fails
svn path=/nixpkgs/trunk/; revision=33885
This should solve the problem I had, where I could not boot either 3.3 or 3.3.1
in my system, as I got ENOSPC all the time.
svn path=/nixpkgs/trunk/; revision=33714
The patch is currently being discussed on LKML and hopefully will be included
in mainline in some form in the future. Note that booting from the livecd has
to do a lot of work before anything is output to the console, so if the drive
is still busy don't assume the boot has hanged
svn path=/nixpkgs/trunk/; revision=33235
Not merged r32497 (tree conflict, glibc GNU Hurd update). Ludovic, could you
please look at this?
svn path=/nixpkgs/branches/stdenv-updates/; revision=32520
what the new nix thinks the fuloong is.
Anyone having the old nix should use a nixpkgs previous to this change to build
the new nix. And then, with the new nix, he can use any newer nixpkgs revision.
svn path=/nixpkgs/trunk/; revision=31751
I add newt, checking that it cross-builds too.
I update perf to have newt support, and now it's also finding python, whatever
that means. I've not tested if 'python' as buildInputs is enough.
svn path=/nixpkgs/trunk/; revision=31353
and lower quality than regular drivers. However, there are a lot of drivers
for wireless cards that we really need to have. And it doesn't really hurt
to have these drivers if you don't need them.
* Enable the Radeon KMS option. This shouldn't be a problem since the X driver
supports KMS (I think).
svn path=/nixpkgs/trunk/; revision=29994
The "unset MODULE_DIR" trick was enough to get Linux 3.x kernels compiling, but it was definitely the Wrong Thing
We NEED MODULE_DIR set so that depmod can store the right dependencies during the build. The REAL problem with the
3.x kernels was two-fold: Our module-init-tools was so old that the kernel build needed to introduce a hack when
calling depmod (involving creating a symlink prepending 99.98 to the version number), and the depmod wrapper was
moved out of the Makefile into scripts/depmod.sh, so our substituteInPlace to get rid of '-b $(INSTALL_MOD_PATH)' in
the Makefile was a noop and INSTALL_MOD_PATH was still being passed to depmod. This is now fixed and modprobe can
successfully find dependencies using the modules.dep created during install
svn path=/nixpkgs/trunk/; revision=29559
I wanted to wait for kernel.org to get back up, but there doesn't seem to be any information about when that will be. If you don't trust that this is Torvalds' github, see https://lkml.org/lkml/2011/9/4/92 for how to verify
svn path=/nixpkgs/trunk/; revision=29198
With the new kernel versioning scheme, the first release in a series only has a version number and
a major revision number (e.g. linux 3.0, linux 3.1-rc1, etc.). Unfortunately, the module
directory for these kernels still has a minor revision number (e.g. lib/modules/3.0.0, lib/modules/3.0.1-rc1, etc.).
This causes problems for packages such as broadcom_sta that need to know the module directory, so
this attribute will allow setting the module directory version number separate from the
kernel version number when necessary.
svn path=/nixpkgs/trunk/; revision=28405
Renamed cifs-timeout-2.6.32 patch to cifs-timeout-2.6.29 as this is the
older kernel version this patch applies to.
svn path=/nixpkgs/trunk/; revision=27711
Currently suffixed with 2.6.32.
This pre-patch prepares the landing of several versions of this patch to
support other Linux kernel versions.
svn path=/nixpkgs/trunk/; revision=27709
Moved the hardcoded postBuild hook from the builder to generic.nix:
Some old kernel (such as 2.6.15) did not yet support the unifdef target.
As a result, compiling them with the current Linux builder leads to a
failure.
Fixed by moving this hook as argument of the top-level function of
generic.nix. This allows some kernel nix codes to overrides its default
value.
svn path=/nixpkgs/trunk/; revision=27708
This patch adds a "features.aufs2_1" to the aufs-2.1 patch for Linux
2.6.37 to prevent aufs2_1 and aufs2_1_util from being options for
kernels without an aufs 2.1 patch. There were several Hydra build
failures as a result of attempting to build aufs2.1 against older
kernels.
svn path=/nixpkgs/trunk/; revision=26597
* My motivation for this patch is that kernels < 2.6.36 contain an
e1000e that does not support the ethernet card that is part of the
chipset for the second-generation Core-i Intel CPUs, so in order
to have a more useful livecd I needed to get aufs working with a
newer kernel, and 2.6.37 is the latest kernel with an official
aufs release.
* All sources are downloaded with fetchgit. This is because the aufs
upstream doesn't provide release tarballs, they just add a tag to
their git tree for an official release.
* The make target for the aufs2.1 headers uses a Makefile in the
kernel build directory that requires that unifdef be in the
scripts/ subdirectory of the build directory. The way I've dealt
with this here is by adding "make $makeFlags -C scripts unifdef"
to the postBuild in the kernel builder. Since the builder is used
by all kernel versions, this will require rebuilding every kernel
and kernel-dependent package if the patch is accepted, so one
alternative I thought of would be to create a fake kernel build
directory where everything is symlinked to the real build
directory except scripts/, which is first copied and then make
unifdef is run before building aufs2.1. If that more complicated
solution is preferred, or if anyone has ideas for another one, I
can do that and submit a new patch.
* The patch was tested by building a livecd ISO that uses it, then
running the ISO from within virtualbox and installing aufs2.1-util
from within the livecd environment.
* The livecd was built using installation-cd-minimal.nix, with two
changes to the Nixos tree:
1. boot.kernelPackages = pkgs.linuxPackages_2_6_37 was added to
profiles/minimal.nix
2. config.boot.kernelPackages.aufs2 was changed to
config.boot.kernelPackages.aufs2_1 in iso-image.nix
I would have preferred to keep all changes within
profiles/minimal.nix, but I couldn't figure out how to override
iso-image.nix's definition of boot.extraModulePackages. Livecds
that use an older kernel can't be built with this iso-image.nix,
since we don't have aufs2.1 for them (just aufs2). If someone can
point me to how I can override things set in iso-image.nix, I'd
appreciate it.
make -C scripts unifdef compiles the unifdef application in the
scripts/ directory, and when Nix copies over the build tree to
$out/lib/modules/$version/build for kernel modules to reference, it
copies over all of scripts/ except the .o files. I can't speak for
other kernel versions, but at the least for 2.6.37.1 unifdef is not
built by default. If you look at the Makefile in scripts, unifdef is
listed under a comment saying that the following programs are only
built on-demand.
svn path=/nixpkgs/trunk/; revision=26548
