nordic-dev.net/nixpkgs
nordic-dev.net/nixpkgs
2
0
Fork 0
mirror of https://github.com/NixOS/nixpkgs.git synced 2024-12-14 17:53:37 +00:00
Code Issues Packages Projects Releases Wiki Activity
3d02a041ba
nixpkgs/pkgs/os-specific/linux/kernel/common-config.nix

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1033 lines
40 KiB
Nix
Raw Normal View History

linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# WARNING/NOTE: whenever you want to add an option here you need to either
# * mark it as an optional one with `option`,
# * or make sure it works for all the versions in nixpkgs,
# * or check for which kernel versions it will work (using kernel
# changelog, google or whatever) and mark it with `whenOlder` or
# `whenAtLeast`.
# Then do test your change by building all the kernels (or at least
# their configs) in Nixpkgs or else you will guarantee lots and lots
# of pain to users trying to switch to an older kernel because of some
# hardware problems with a new one.
kernel: add a warning/note at the top of common-config so that people would hopefully stop breaking the older kernels
2015-03-24 21:34:54 +00:00
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Configuration
pkgs/os-specific: stdenv.lib -> lib
2021-01-15 14:45:37 +00:00
{ lib, stdenv, version
kernel: add a warning/note at the top of common-config so that people would hopefully stop breaking the older kernels
2015-03-24 21:34:54 +00:00
linux: remove remnants of grsecurity Nixpkgs hasn't supported grsecurity kernels since 2017, so unless anybody is manually enabling the grsecurity feature to make these small kernel tweaks this is dead code. This means we don't actually support any "features" in the kernel common-config any more, but I've left the argument there because it's conceivable we could have some again in future.
2021-04-05 10:01:58 +00:00
, features ? {}
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
}:
pkgs/os-specific: stdenv.lib -> lib
2021-01-15 14:45:37 +00:00
with lib;
with lib.kernel;
with (lib.kernel.whenHelpers version);
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
let
linux: fix features system
2018-06-25 00:12:17 +00:00
# configuration items have to be part of a subattrs
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
flattenKConf = nested: mapAttrs (_: head) (zipAttrs (attrValues nested));
linux config: add support for xdp sockets and ebpf jit xdp socket support (AF_XDP) is the new way of implementing high performance networking on linux. on arch linux and debian this is already enabled (checked via the links from the nixos manual). moreover, these flags are suggested by the bpf documentation at cilium: https://cilium.readthedocs.io/en/latest/bpf/#compiling-the-kernel additionally the flag `BPF_JIT_ALWAYS_ON` on was suggested to help spectre attack mitigations: https://github.com/torvalds/linux/commit/290af86629b25ffd1ed6232c4e9107da031705cb
2019-11-11 06:57:30 +00:00
whenPlatformHasEBPFJit =
mkIf (stdenv.hostPlatform.isAarch32 ||
stdenv.hostPlatform.isAarch64 ||
stdenv.hostPlatform.isx86_64 ||
lib/systems/inspect.nix: remove isPowerPC Very confusingly, the `isPowerPC` predicate in `lib/systems/inspect.nix` does *not* match `powerpc64le`! This is because `isPowerPC` is defined as isPowerPC = { cpu = cpuTypes.powerpc; }; Where `cpuTypes.powerpc` is: { bits = 32; significantByte = bigEndian; family = "power"; }; This means that the `isPowerPC` predicate actually only matches the subset of machines marketed under this name which happen to be 32-bit and running in big-endian mode which is equivalent to: with stdenv.hostPlatform; isPower && isBigEndian && is32bit This seems like a sharp edge that people could easily cut themselves on. In fact, that has already happened: in `linux/kernel/common-config.nix` there is a test which will always fail: (stdenv.hostPlatform.isPowerPC && stdenv.hostPlatform.is64bit) A more subtle case of the strict isPowerPC being used instead of the moreg general isPower accidentally are the GHC expressions: Update pkgs/development/compilers/ghc/8.10.7.nix Update pkgs/development/compilers/ghc/8.8.4.nix Update pkgs/development/compilers/ghc/9.2.2.nix Update pkgs/development/compilers/ghc/9.0.2.nix Update pkgs/development/compilers/ghc/head.nix Since the remaining legitimate use sites of isPowerPC are so few, remove the isPowerPC predicate completely. The alternative expression above is noted in the release notes as an alternative. Co-authored-by: sternenseemann <sternenseemann@systemli.org>
2022-04-10 08:59:20 +00:00
(stdenv.hostPlatform.isPower && stdenv.hostPlatform.is64bit) ||
linux config: add support for xdp sockets and ebpf jit xdp socket support (AF_XDP) is the new way of implementing high performance networking on linux. on arch linux and debian this is already enabled (checked via the links from the nixos manual). moreover, these flags are suggested by the bpf documentation at cilium: https://cilium.readthedocs.io/en/latest/bpf/#compiling-the-kernel additionally the flag `BPF_JIT_ALWAYS_ON` on was suggested to help spectre attack mitigations: https://github.com/torvalds/linux/commit/290af86629b25ffd1ed6232c4e9107da031705cb
2019-11-11 06:57:30 +00:00
(stdenv.hostPlatform.isMips && stdenv.hostPlatform.is64bit));
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
options = {
debug = {
linux-kernel: Enable BTF BTF is a new, lightweight debug information format tailored specifically for the needs of eBPF, allowing eBPF programs to be portable across various kernel versions, configurations and distributions. This is used by bpftrace and lots of new eBPF-based tooling to avoid a dependency on LLVM on the host. BTF debug information is enabled on all major distributions: Fedora 31+, RHEL 8.2+, Ubuntu 20.10, Debian 11 and ArchLinux all have enabled it. Enabling BTF debug information requires adding two new dependencies to the kernel build: Python3 and pahole. Those will be used to generate the BTF debugging information.
2021-06-22 15:55:14 +00:00
# Necessary for BTF
linuxManualConfig: install GDB scripts These are required to debug kernel modules. Since we're now able to do that, there's another reason besides BTF to enable DEBUG_INFO, so I've done that for pre-BTF kernel modules as well here. For GDB to get configured correctly, vmlinux-gdb.py has to be two directories up from scripts/gdb, and vmlinux has to be next to vmlinux-gdb.py. The least invasive way to satisfy these constraints is to make vmlinux a symlink, which GDB will resolve before looking for vmlinux-gdb.py. Tested both ways of getting the scripts into GDB that I know of: gdb /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/vmlinux \ -iex 'add-auto-load-safe-path /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/lib/modules/6.1.19/build/vmlinux-gdb.py' \ -ex 'lx-version' \ -ex 'q' gdb /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/vmlinux \ -ex 'source /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/lib/modules/6.1.19/build/vmlinux-gdb.py' \ -ex 'lx-version' \ -ex 'q' Also tested that the strip changes don't result in meaningful output size changes (there's some small variation due to BTF data not always coming out the same size, which is unrelated), and built every kernel I can on x86_64 to make sure I'm not relying on build system behaviour specific to newer kernels.
2023-03-17 18:44:06 +00:00
DEBUG_INFO = yes;
linux_latest: 5.17.9 -> 5.18 NSFD_V3 is now always enabled, and enabling debug info now requires selecting a DWARF version instead of just setting DEBUG_INFO=y.
2022-05-23 09:19:42 +00:00
DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT = whenAtLeast "5.18" yes;
linux-kernel config: disable DEBUG_INFO_REDUCED Linux's aarch64 defconfig has been updated in 5.13 to enable "reduced" debug infos (upstream commit ed938a4bfc58 ("arm64: defconfig: Use DEBUG_INFO_REDUCED"), but that commits locks DEBUG_INFO_BTF as noticed in #175467 This disables it back which should fix bpftrace usage of BTF not working on newer kernels.
2022-07-09 01:29:40 +00:00
# Reduced debug info conflict with BTF and have been enabled in
# aarch64 defconfig since 5.13
DEBUG_INFO_REDUCED = whenAtLeast "5.13" (option no);
Revert "linux-kernel: disable BTF on 32-bit platforms on kernels 5.15+" This reverts commit 79e05fb16b1af292e50cc0c479809cc66b47b087. broken 32bit BTF builds got fixed in #175467 by switching libbpf from libelf to elfutils, as a side-product of the upgrade, so we don't need this anymore.
2022-07-11 01:14:06 +00:00
DEBUG_INFO_BTF = whenAtLeast "5.2" (option yes);
linux: enable MODULE_ALLOW_BTF_MISMATCH Right now it looks like the BTFs are not reproducible between different builds of the same kernel, and the kernel will refuse to load modules if the BTF doesn't match. This can cause some interesting side effects when Nix uses different substituters for different parts of the kernel. This is far from ideal, and we _really_ should figure out how to actually make the BTF building consistently reproducible, but that seems more complicated, so maybe we should do this to get affected systems booting. See also: https://lore.kernel.org/bpf/YfK18x%2FXrYL4Vw8o@syu-laptop/ , where the openSUSE people ran into similar issues.
2022-07-14 09:11:55 +00:00
# Allow loading modules with mismatched BTFs
# FIXME: figure out how to actually make BTFs reproducible instead
# See https://github.com/NixOS/nixpkgs/pull/181456 for details.
MODULE_ALLOW_BTF_MISMATCH = whenAtLeast "5.18" (option yes);
linux-kernel: enable BPF_LSM Enables instrumentation of the security hooks with BPF programs for implementing dynamic MAC and Audit Policies. The BPF LSM was merged into the Linux kernel 5.7 This has already been enabled in Fedora (version 33 onwards), Ubuntu (Hirsute Hippo), Flatcar Linux, Arch Linux. Distros like Ubuntu don't enable bpf in CONFIG_LSM by default to avoid any performance penalty so similar to that this commit enables CONFIG_BPF_LSM but doesn't add bpf to the default list in CONFIG_LSM, users willing to use this feature could boot with the lsm=...bpf parameter on the kernel https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=641cd7b06c911c5935c34f24850ea18690649917 https://outflux.net/blog/archives/2020/09/21/security-things-in-linux-v5-7/ https://lwn.net/Articles/813057/ https://github.com/flatcar-linux/Flatcar/issues/343 https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1905975 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=983329
2021-11-02 11:13:55 +00:00
BPF_LSM = whenAtLeast "5.7" (option yes);
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
DEBUG_KERNEL = yes;
DEBUG_DEVRES = no;
DYNAMIC_DEBUG = yes;
DEBUG_STACK_USAGE = no;
RCU_TORTURE_TEST = no;
SCHEDSTATS = no;
DETECT_HUNG_TASK = yes;
CRASH_DUMP = option no;
# Easier debugging of NFS issues.
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
SUNRPC_DEBUG = yes;
kernel/common-config: enable SCHED_DEBUG
2019-09-24 10:55:46 +00:00
# Provide access to tunables like sched_migration_cost_ns
SCHED_DEBUG = yes;
linuxManualConfig: install GDB scripts These are required to debug kernel modules. Since we're now able to do that, there's another reason besides BTF to enable DEBUG_INFO, so I've done that for pre-BTF kernel modules as well here. For GDB to get configured correctly, vmlinux-gdb.py has to be two directories up from scripts/gdb, and vmlinux has to be next to vmlinux-gdb.py. The least invasive way to satisfy these constraints is to make vmlinux a symlink, which GDB will resolve before looking for vmlinux-gdb.py. Tested both ways of getting the scripts into GDB that I know of: gdb /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/vmlinux \ -iex 'add-auto-load-safe-path /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/lib/modules/6.1.19/build/vmlinux-gdb.py' \ -ex 'lx-version' \ -ex 'q' gdb /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/vmlinux \ -ex 'source /nix/store/7n77ijlxkxr6d613h02lr707kvjx6j1k-linux-6.1.19-dev/lib/modules/6.1.19/build/vmlinux-gdb.py' \ -ex 'lx-version' \ -ex 'q' Also tested that the strip changes don't result in meaningful output size changes (there's some small variation due to BTF data not always coming out the same size, which is unrelated), and built every kernel I can on x86_64 to make sure I'm not relying on build system behaviour specific to newer kernels.
2023-03-17 18:44:06 +00:00
GDB_SCRIPTS = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
power-management = {
CPU_FREQ_DEFAULT_GOV_PERFORMANCE = yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
CPU_FREQ_GOV_SCHEDUTIL = yes;
linux: common-config condition power-management to required platform
2022-03-23 21:42:41 +00:00
PM_ADVANCED_DEBUG = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
PM_WAKELOCKS = yes;
kernel: set POWERCAP and INTEL_RAPL These enable some power management settings. They are needed for [s-tui to show power readings](https://github.com/amanusk/s-tui/issues/105). The values chosen here match what Arch Linux has. In particular the Intel specific code is loadable as a module rather than compiled in.
2019-08-22 10:23:37 +00:00
POWERCAP = yes;
linux: enable ACPI_FPDT, ACPI_HMAT, ACPI_APEI, ACPI_APEI_GHES, ACPI_DPTF
2023-01-08 10:18:48 +00:00
# ACPI Firmware Performance Data Table Support
linux: fix-build on i686
2023-02-19 20:04:50 +00:00
ACPI_FPDT = whenAtLeast "5.12" (option yes);
linux: enable ACPI_FPDT, ACPI_HMAT, ACPI_APEI, ACPI_APEI_GHES, ACPI_DPTF
2023-01-08 10:18:48 +00:00
# ACPI Heterogeneous Memory Attribute Table Support
linux: fix-build on i686
2023-02-19 20:04:50 +00:00
ACPI_HMAT = whenAtLeast "5.2" (option yes);
linux: enable ACPI_FPDT, ACPI_HMAT, ACPI_APEI, ACPI_APEI_GHES, ACPI_DPTF
2023-01-08 10:18:48 +00:00
# ACPI Platform Error Interface
linux: fix-build on i686
2023-02-19 20:04:50 +00:00
ACPI_APEI = (option yes);
linux: enable ACPI_FPDT, ACPI_HMAT, ACPI_APEI, ACPI_APEI_GHES, ACPI_DPTF
2023-01-08 10:18:48 +00:00
# APEI Generic Hardware Error Source
linux: fix-build on i686
2023-02-19 20:04:50 +00:00
ACPI_APEI_GHES = (option yes);
linux: common-config condition power-management to required platform
2022-03-23 21:42:41 +00:00
} // optionalAttrs (stdenv.hostPlatform.isx86) {
INTEL_IDLE = yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
INTEL_RAPL = whenAtLeast "5.3" module;
linux: common-config condition power-management to required platform
2022-03-23 21:42:41 +00:00
X86_INTEL_LPSS = yes;
X86_INTEL_PSTATE = yes;
linux: set X86_AMD_PSTATE=y instead of =m
2022-12-02 20:37:00 +00:00
X86_AMD_PSTATE = whenAtLeast "5.17" yes;
linux: enable ACPI_FPDT, ACPI_HMAT, ACPI_APEI, ACPI_APEI_GHES, ACPI_DPTF
2023-01-08 10:18:48 +00:00
# Intel DPTF (Dynamic Platform and Thermal Framework) Support
ACPI_DPTF = whenAtLeast "5.10" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
external-firmware = {
# Support drivers that need external firmware.
STANDALONE = no;
};
proc-config-gz = {
# Make /proc/config.gz available
IKCONFIG = yes;
IKCONFIG_PROC = yes;
};
optimization = {
# Optimize with -O2, not -Os
CC_OPTIMIZE_FOR_SIZE = no;
};
memtest = {
MEMTEST = yes;
};
# Include the CFQ I/O scheduler in the kernel, rather than as a
# module, so that the initrd gets a good I/O scheduler.
scheduler = {
linux: 4.20-rc7 -> 5.0-rc1
2019-01-07 12:57:32 +00:00
IOSCHED_CFQ = whenOlder "5.0" yes; # Removed in 5.0-RC1
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BLK_CGROUP = yes; # required by CFQ"
linux: enable io cost and io latency block group controller They are enabled by default on Arch.
2020-07-18 12:00:00 +00:00
BLK_CGROUP_IOLATENCY = whenAtLeast "4.19" yes;
BLK_CGROUP_IOCOST = whenAtLeast "5.4" yes;
linux: 4.20-rc7 -> 5.0-rc1
2019-01-07 12:57:32 +00:00
IOSCHED_DEADLINE = whenOlder "5.0" yes; # Removed in 5.0-RC1
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
MQ_IOSCHED_DEADLINE = yes;
BFQ_GROUP_IOSCHED = yes;
MQ_IOSCHED_KYBER = yes;
IOSCHED_BFQ = module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
linux: Set CONFIG_NO_HZ_FULL=y. CONFIG_NO_HZ_FULL=y should be set to enable the `nohz_full=` and `rcu_nocbs=` options. These carry no additional performance penalty compared to CONFIG_NO_HZ_IDLE and behaves like it by default, but allows disabling the tick interrupts on cores for power or performance reasons. [Debian][1] also applied the change to all their kernels. Like the Kernel says: "If you're a distro say Y." [1]: https://salsa.debian.org/kernel-team/linux/-/commit/f6aad27f05c007d6f30b34ff77bc7ea47844f117
2022-10-30 15:47:12 +00:00
timer = {
# Enable Full Dynticks System.
linux: avoid NO_HZ_FULL on i686-linux This is just a stop-gap; seemed better than a real revert. The issue is from commit 8d3fe232e (PR #198666).
2022-11-02 22:02:43 +00:00
NO_HZ_FULL = mkIf stdenv.is64bit yes; # TODO: more precise condition?
linux: Set CONFIG_NO_HZ_FULL=y. CONFIG_NO_HZ_FULL=y should be set to enable the `nohz_full=` and `rcu_nocbs=` options. These carry no additional performance penalty compared to CONFIG_NO_HZ_IDLE and behaves like it by default, but allows disabling the tick interrupts on cores for power or performance reasons. [Debian][1] also applied the change to all their kernels. Like the Kernel says: "If you're a distro say Y." [1]: https://salsa.debian.org/kernel-team/linux/-/commit/f6aad27f05c007d6f30b34ff77bc7ea47844f117
2022-10-30 15:47:12 +00:00
};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Enable NUMA.
numa = {
NUMA = option yes;
};
networking = {
NET = yes;
linux: add policy routing config flag for aarch64 CONFIG_IP_MULTIPLE_TABLES is part of the default x86 kernel config but absent from the Aarch64 one. Adding explicitely this flag together with its dependency IP_ADVANCED_ROUTER. Both of these config flags are needed to use the routing policy facilities.
2020-02-29 22:57:43 +00:00
IP_ADVANCED_ROUTER = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
IP_PNP = no;
IP_VS_PROTO_TCP = yes;
IP_VS_PROTO_UDP = yes;
IP_VS_PROTO_ESP = yes;
IP_VS_PROTO_AH = yes;
linux: Enable IPv6 support for IPVS (CONFIG_IP_VS_IPV6)
2019-06-08 08:23:48 +00:00
IP_VS_IPV6 = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
IP_DCCP_CCID3 = no; # experimental
CLS_U32_PERF = yes;
CLS_U32_MARK = yes;
linux config: add support for xdp sockets and ebpf jit xdp socket support (AF_XDP) is the new way of implementing high performance networking on linux. on arch linux and debian this is already enabled (checked via the links from the nixos manual). moreover, these flags are suggested by the bpf documentation at cilium: https://cilium.readthedocs.io/en/latest/bpf/#compiling-the-kernel additionally the flag `BPF_JIT_ALWAYS_ON` on was suggested to help spectre attack mitigations: https://github.com/torvalds/linux/commit/290af86629b25ffd1ed6232c4e9107da031705cb
2019-11-11 06:57:30 +00:00
BPF_JIT = whenPlatformHasEBPFJit yes;
linux: fix kernel config options for i686 Even the default pkgsi686Linux.linux was broken.
2020-06-12 06:52:09 +00:00
BPF_JIT_ALWAYS_ON = whenPlatformHasEBPFJit no; # whenPlatformHasEBPFJit yes; # see https://github.com/NixOS/nixpkgs/issues/79304
linux config: add support for xdp sockets and ebpf jit xdp socket support (AF_XDP) is the new way of implementing high performance networking on linux. on arch linux and debian this is already enabled (checked via the links from the nixos manual). moreover, these flags are suggested by the bpf documentation at cilium: https://cilium.readthedocs.io/en/latest/bpf/#compiling-the-kernel additionally the flag `BPF_JIT_ALWAYS_ON` on was suggested to help spectre attack mitigations: https://github.com/torvalds/linux/commit/290af86629b25ffd1ed6232c4e9107da031705cb
2019-11-11 06:57:30 +00:00
HAVE_EBPF_JIT = whenPlatformHasEBPFJit yes;
BPF_STREAM_PARSER = whenAtLeast "4.19" yes;
XDP_SOCKETS = whenAtLeast "4.19" yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
XDP_SOCKETS_DIAG = whenAtLeast "5.1" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
WAN = yes;
linux: enable TCP_CONG_ADVANCED TCP_CONG_ADVANCED is enabled by default on x86_64[1] in the upstream. Although it is not the case for aarch64[2], many distributions, such as Debian[3], Fedora[4] and Gentoo[5], choose to enable it in their distribution kernel. With this patch, aarch64 users can choose many other TCP congestion algorithms, which may improve their network performance. [1]: https://github.com/torvalds/linux/blob/7ddb58cb0ecae8e8b6181d736a87667cc9ab8389/arch/x86/configs/x86_64_defconfig#L71 [2]: https://github.com/torvalds/linux/blob/7ddb58cb0ecae8e8b6181d736a87667cc9ab8389/arch/arm64/configs/defconfig [3]: https://salsa.debian.org/kernel-team/linux/-/blob/e2d14375d74cdcb2cfed6518d7a5ad784c30df1a/debian/config/config#L7063 [4]: https://src.fedoraproject.org/rpms/kernel/raw/836165dd2dff34e4f2c47ca8f9c803002c1e6530/f/kernel-aarch64-fedora.config [5]: https://github.com/gentoo/gentoo/blob/5808eb2f066ea2ece19c35a1978b7f79077a467b/sys-kernel/gentoo-kernel/gentoo-kernel-5.10.77.ebuild#L27
2021-11-04 10:54:34 +00:00
TCP_CONG_ADVANCED = yes;
linux config: Set TCP_CONG_CUBIC=yes to restore the default This will switch the default TCP congestion control algorithm from new Reno to CUBIC. CUBIC is the default since Linux kernel 2.6.19 (see 597811ec167fa) and most (all?) distributions keep this default (e.g. Debian and Ubuntu). On NixOS the default was still new Reno because generate-config.pl changes TCP_CONG_CUBIC from y to m (since we try to build everything as a module by default). To check the active and available algorithms: $ sysctl net.ipv4.tcp_congestion_control net.ipv4.tcp_congestion_control = cubic $ sysctl net.ipv4.tcp_available_congestion_control net.ipv4.tcp_available_congestion_control = cubic reno Note: E.g. x86_64_defconfig sets TCP_CONG_CUBIC=y indirectly via CONFIG_TCP_CONG_ADVANCED=y (but CUBIC is also the default if set to no, see net/ipv4/Kconfig).
2020-02-23 17:45:29 +00:00
TCP_CONG_CUBIC = yes; # This is the default congestion control algorithm since 2.6.19
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Required by systemd per-cgroup firewalling
CGROUP_BPF = option yes;
CGROUP_NET_PRIO = yes; # Required by systemd
IP_ROUTE_VERBOSE = yes;
IP_MROUTE_MULTIPLE_TABLES = yes;
IP_MULTICAST = yes;
linux: add policy routing config flag for aarch64 CONFIG_IP_MULTIPLE_TABLES is part of the default x86 kernel config but absent from the Aarch64 one. Adding explicitely this flag together with its dependency IP_ADVANCED_ROUTER. Both of these config flags are needed to use the routing policy facilities.
2020-02-29 22:57:43 +00:00
IP_MULTIPLE_TABLES = yes;
kernel config: explicitly enable CONFIG_IPV6 We currently build CONFIG_IPV6=m. This seems to be not really well-supported in mainline kernels - see https://lore.kernel.org/netdev/20201115224509.2020651-1-flokli@flokli.de/T/#u Compiling it as a module doesn't give too much benefit - even for people who did explicitly set `enableIPv6` to false, the `ipv6` module was still loaded, as soon as another module was loaded that requires it (bridge,br_netfilter,wireguard,ip6table_mangle,sctp,…). By compiling it in, we only loose the possibility to not add it to `boot.kernelModules` anymore (as it's part of the kernel directly). The space savings are negligible. People wanting to disable IPv6 still get the appropriate sysctls and options set (while having the kernel code loaded), nothing is really changing here.
2020-11-16 12:02:48 +00:00
IPV6 = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
IPV6_ROUTER_PREF = yes;
IPV6_ROUTE_INFO = yes;
IPV6_OPTIMISTIC_DAD = yes;
IPV6_MULTIPLE_TABLES = yes;
IPV6_SUBTREES = yes;
IPV6_MROUTE = yes;
IPV6_MROUTE_MULTIPLE_TABLES = yes;
IPV6_PIMSM_V2 = yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
IPV6_FOU_TUNNEL = module;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
IPV6_SEG6_LWTUNNEL = yes;
IPV6_SEG6_HMAC = yes;
linux: Enable SRv6 options
2021-03-25 04:54:30 +00:00
IPV6_SEG6_BPF = whenAtLeast "4.18" yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
NET_CLS_BPF = module;
NET_ACT_BPF = module;
linux/kernel: enable QoS and/or fair queueing This allows to set queueing disciplines different than a simple fifo, like fq_codel, which is the default in systemd since quite some time. NET_SCHED is already set in the kernels x86_64_defconfig, but not on arm/aarch64, so let's set it here.
2019-09-01 00:24:38 +00:00
NET_SCHED = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
L2TP_V3 = yes;
L2TP_IP = module;
L2TP_ETH = module;
BRIDGE_VLAN_FILTERING = yes;
BONDING = module;
NET_L3_MASTER_DEV = option yes;
NET_FOU_IP_TUNNELS = option yes;
IP_NF_TARGET_REDIRECT = module;
PPP_MULTILINK = yes; # PPP multilink support
PPP_FILTER = yes;
# needed for iwd WPS support (wpa_supplicant replacement)
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
KEY_DH_OPERATIONS = yes;
nftables: enable works in linux kernel 4.17 and 4.18
2018-08-24 13:54:50 +00:00
# needed for nftables
nftables: enable all features
2019-10-14 13:41:39 +00:00
# Networking Options
NETFILTER = yes;
NETFILTER_ADVANCED = yes;
# Core Netfilter Configuration
NF_CONNTRACK_ZONES = yes;
NF_CONNTRACK_EVENTS = yes;
NF_CONNTRACK_TIMEOUT = yes;
NF_CONNTRACK_TIMESTAMP = yes;
NETFILTER_NETLINK_GLUE_CT = yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
NF_TABLES_INET = mkMerge [ (whenOlder "4.17" module)
(whenAtLeast "4.17" yes) ];
NF_TABLES_NETDEV = mkMerge [ (whenOlder "4.17" module)
(whenAtLeast "4.17" yes) ];
nftables: add netdev REJECT support
2021-01-17 16:21:56 +00:00
NFT_REJECT_NETDEV = whenAtLeast "5.11" module;
nftables: enable all features
2019-10-14 13:41:39 +00:00
# IP: Netfilter Configuration
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
NF_TABLES_IPV4 = mkMerge [ (whenOlder "4.17" module)
(whenAtLeast "4.17" yes) ];
NF_TABLES_ARP = mkMerge [ (whenOlder "4.17" module)
(whenAtLeast "4.17" yes) ];
nftables: enable all features
2019-10-14 13:41:39 +00:00
# IPv6: Netfilter Configuration
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
NF_TABLES_IPV6 = mkMerge [ (whenOlder "4.17" module)
(whenAtLeast "4.17" yes) ];
nftables: enable all features
2019-10-14 13:41:39 +00:00
# Bridge Netfilter Configuration
NF_TABLES_BRIDGE = mkMerge [ (whenBetween "4.19" "5.3" yes)
(whenAtLeast "5.3" module) ];
kernel/common-config: enable INET_{TCP,UDP,RAW}_DIAG and INET_DIAG_DESTROY
2019-09-24 19:52:42 +00:00
linux: Enable `CONFIG_NET_DROP_MONITOR` by default. Needed for subscribing to dropped packets (e.g. via `dropwatch`).
2020-04-13 01:36:15 +00:00
# needed for `dropwatch`
# Builtin-only since https://github.com/torvalds/linux/commit/f4b6bcc7002f0e3a3428bac33cf1945abff95450
NET_DROP_MONITOR = yes;
kernel/common-config: enable INET_{TCP,UDP,RAW}_DIAG and INET_DIAG_DESTROY
2019-09-24 19:52:42 +00:00
# needed for ss
linux: fix fallout from conflicting kernel configs The parent commit forbids conflicting kernel config options. Fix the hardened kernels by allowing options in common-config.nix to be overridden by conflicting ones in hardened/config.nix. I'm explicitly avoiding using a higher priority (e.g. using mkForce) in hardened/config.nix so that the user can easily override the options in that file.
2020-06-10 14:29:14 +00:00
# Use a lower priority to allow these options to be overridden in hardened/config.nix
INET_DIAG = mkDefault module;
INET_TCP_DIAG = mkDefault module;
INET_UDP_DIAG = mkDefault module;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
INET_RAW_DIAG = mkDefault module;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
INET_DIAG_DESTROY = mkDefault yes;
linux: build with multipath tcp support Starting from Linx 5.6, there is partial upstream support for the Multipath TCP protocol. There are no downsides to enabling it afaict, since applications need to opt-in when creating a socket. From https://github.com/multipath-tcp/mptcp_net-next/wiki: "[...] users of regular TCP continue to get the same type of connection and performance unless MPTCP is requested."
2021-01-07 00:49:23 +00:00
# enable multipath-tcp
MPTCP = whenAtLeast "5.6" yes;
MPTCP_IPV6 = whenAtLeast "5.6" yes;
linux: fix fallout from conflicting kernel configs The parent commit forbids conflicting kernel config options. Fix the hardened kernels by allowing options in common-config.nix to be overridden by conflicting ones in hardened/config.nix. I'm explicitly avoiding using a higher priority (e.g. using mkForce) in hardened/config.nix so that the user can easily override the options in that file.
2020-06-10 14:29:14 +00:00
INET_MPTCP_DIAG = whenAtLeast "5.9" (mkDefault module);
linux: enable kTLS
2021-11-22 09:34:00 +00:00
# Kernel TLS
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
TLS = module;
linux: enable kTLS
2021-11-22 09:34:00 +00:00
TLS_DEVICE = whenAtLeast "4.18" yes;
kernel: Enable IPoIB Connected Mode `INFINIBAND` and `INFINIBAND_IPOIB` are here for clarity - They along with other required flags are enabled already in the default config.
2021-06-05 20:40:13 +00:00
# infiniband
INFINIBAND = module;
INFINIBAND_IPOIB = module;
INFINIBAND_IPOIB_CM = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
wireless = {
CFG80211_WEXT = option yes; # Without it, ipw2200 drivers don't build
IPW2100_MONITOR = option yes; # support promiscuous mode
IPW2200_MONITOR = option yes; # support promiscuous mode
HOSTAP_FIRMWARE = option yes; # Support downloading firmware images with Host AP driver
HOSTAP_FIRMWARE_NVRAM = option yes;
ATH9K_PCI = option yes; # Detect Atheros AR9xxx cards on PCI(e) bus
ATH9K_AHB = option yes; # Ditto, AHB bus
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
B43_PHY_HT = option yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BCMA_HOST_PCI = option yes;
linux: build rtw88 module Adds support for Realtek wireless/bluetooth cards found in some Lenovo laptops. The old `r8822be` module was removed in favour of this one.
2019-09-09 23:08:28 +00:00
RTW88 = whenAtLeast "5.2" module;
linux: 5.7-rc6 -> 5.8-rc1
2020-06-15 15:10:00 +00:00
RTW88_8822BE = mkMerge [ (whenBetween "5.2" "5.8" yes) (whenAtLeast "5.8" module) ];
RTW88_8822CE = mkMerge [ (whenBetween "5.2" "5.8" yes) (whenAtLeast "5.8" module) ];
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
fb = {
FB = yes;
FB_EFI = yes;
FB_NVIDIA_I2C = yes; # Enable DDC Support
FB_RIVA_I2C = yes;
FB_ATY_CT = yes; # Mach64 CT/VT/GT/LT (incl. 3D RAGE) support
FB_ATY_GX = yes; # Mach64 GX support
FB_SAVAGE_I2C = yes;
FB_SAVAGE_ACCEL = yes;
FB_SIS_300 = yes;
FB_SIS_315 = yes;
FB_3DFX_ACCEL = yes;
FB_VESA = yes;
FRAMEBUFFER_CONSOLE = yes;
linux: Enable fbcon deferred takeover when possible This config value ensures that when booting through e.g. UEFI, the existing framebuffer contents stay put until the first character is printed. As the default NixOS stage-1 immediately outputs a welcome message on init, this does not impact it, but it will allow for a cleaner boot when configured as such.
2020-05-17 17:30:11 +00:00
FRAMEBUFFER_CONSOLE_DEFERRED_TAKEOVER = whenAtLeast "4.19" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
FRAMEBUFFER_CONSOLE_ROTATION = yes;
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
FB_GEODE = mkIf (stdenv.hostPlatform.system == "i686-linux") yes;
linux: do not build in DRM_SIMPLEDRM on newer kernels After linux 5.14.11 FB_SIMPLE conflicts with DRM_SIMPLEDRM, which will fail configuration, when DRM_SIMPLEDRM is configured as a module and FB_SIMPLE gets requested as builtin. Do not enable DRM_SIMPLEDRM as a temporary workaround, until good enough migration path is found.
2021-10-17 17:06:35 +00:00
# On 5.14 this conflicts with FB_SIMPLE.
DRM_SIMPLEDRM = whenAtLeast "5.14" no;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
kernel: enable Terminus 16x32 font This font was added into the kernel for high-resolution displays. It has been available since 5.0.
2023-01-11 14:16:36 +00:00
fonts = {
FONTS = yes;
# Default fonts enabled if FONTS is not set
FONT_8x8 = yes;
FONT_8x16 = yes;
# High DPI font
FONT_TER16x32 = whenAtLeast "5.0" yes;
};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
video = {
linux: Disable DRM_LEGACY, NOUVEAU_LEGACY_CTX_SUPPORT This currently gets enabled as generate-config.pl will enable all the drivers below it as modules. Is “not set” in [Arch][1], [Debian][2], [Fedora][3]. See also [summary of setting from various distros in April 2020][4]. Recommended disabled by [CLIP OS][5] and per current [Kernel config description][6]: > bool "Enable legacy drivers (DANGEROUS)" > Enable legacy DRI1 drivers. Those drivers expose unsafe and dangerous > APIs to user-space, which can be used to circumvent access > restrictions and other security measures. For backwards compatibility > those drivers are still available, but their use is highly > inadvisable and might harm your system. > > You are recommended to use the safe modeset-only drivers instead, and > perform 3D emulation in user-space. > > Unless you have strong reasons to go rogue, say "N". Also disable NOUVEAU_LEGACY_CTX_SUPPORT, as this does `select DRM_LEGACY`. Per Kernel config docs: >There was a version of the nouveau DDX that relied on legacy > ctx ioctls not erroring out. But that was back in time a long > ways, so offer a way to disable it now. For uapi compat with > old nouveau ddx this should be on by default, but modern distros > should consider turning it off. and the [commit][7]: > These driver functions contain several bugs and security holes. This > change makes these functions optional can be turned on by a setting, > they are turned off by default for modeset driver with the exception of > the nouvea driver that may require them with an old version of libdrm. Referenced earlier commit elaborates that > libdrm_nouveau before 2.4.33 used contexts Since nixpkgs here has a much newer version (2.4.33 is from March 2012), should not be a concern. NOUVEAU_LEGACY_CTX_SUPPORT is also “not set” in the linked Arch, Debian, & Fedora configs. [1]: https://github.com/archlinux/svntogit-packages/blob/66d72ee54afc604391b618fc3eecc43f29e479e8/trunk/config#L6637 [2]: https://salsa.debian.org/kernel-team/linux/-/blob/07731f5956cf29876a7abc13f4ecbdf4d9459592/debian/config/config#L713 [3]: https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-x86_64-fedora.config#_1528 [4]: https://github.com/a13xp0p0v/kconfig-hardened-check/issues/38#issuecomment-608639217 [5]: https://docs.clip-os.org/clipos/kernel.html#configuration [6]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpu/drm/Kconfig#n421 [7]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b30a43ac7132cdda833ac4b13dd1ebd35ace14b7
2022-08-27 21:10:06 +00:00
DRM_LEGACY = no;
linux_testing: 6.2-rc6 -> 6.3-rc1
2023-03-10 20:15:17 +00:00
NOUVEAU_LEGACY_CTX_SUPPORT = whenBetween "5.2" "6.3" no;
linux: Disable DRM_LEGACY, NOUVEAU_LEGACY_CTX_SUPPORT This currently gets enabled as generate-config.pl will enable all the drivers below it as modules. Is “not set” in [Arch][1], [Debian][2], [Fedora][3]. See also [summary of setting from various distros in April 2020][4]. Recommended disabled by [CLIP OS][5] and per current [Kernel config description][6]: > bool "Enable legacy drivers (DANGEROUS)" > Enable legacy DRI1 drivers. Those drivers expose unsafe and dangerous > APIs to user-space, which can be used to circumvent access > restrictions and other security measures. For backwards compatibility > those drivers are still available, but their use is highly > inadvisable and might harm your system. > > You are recommended to use the safe modeset-only drivers instead, and > perform 3D emulation in user-space. > > Unless you have strong reasons to go rogue, say "N". Also disable NOUVEAU_LEGACY_CTX_SUPPORT, as this does `select DRM_LEGACY`. Per Kernel config docs: >There was a version of the nouveau DDX that relied on legacy > ctx ioctls not erroring out. But that was back in time a long > ways, so offer a way to disable it now. For uapi compat with > old nouveau ddx this should be on by default, but modern distros > should consider turning it off. and the [commit][7]: > These driver functions contain several bugs and security holes. This > change makes these functions optional can be turned on by a setting, > they are turned off by default for modeset driver with the exception of > the nouvea driver that may require them with an old version of libdrm. Referenced earlier commit elaborates that > libdrm_nouveau before 2.4.33 used contexts Since nixpkgs here has a much newer version (2.4.33 is from March 2012), should not be a concern. NOUVEAU_LEGACY_CTX_SUPPORT is also “not set” in the linked Arch, Debian, & Fedora configs. [1]: https://github.com/archlinux/svntogit-packages/blob/66d72ee54afc604391b618fc3eecc43f29e479e8/trunk/config#L6637 [2]: https://salsa.debian.org/kernel-team/linux/-/blob/07731f5956cf29876a7abc13f4ecbdf4d9459592/debian/config/config#L713 [3]: https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-x86_64-fedora.config#_1528 [4]: https://github.com/a13xp0p0v/kconfig-hardened-check/issues/38#issuecomment-608639217 [5]: https://docs.clip-os.org/clipos/kernel.html#configuration [6]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpu/drm/Kconfig#n421 [7]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=b30a43ac7132cdda833ac4b13dd1ebd35ace14b7
2022-08-27 21:10:06 +00:00
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Allow specifying custom EDID on the kernel command line
DRM_LOAD_EDID_FIRMWARE = yes;
VGA_SWITCHEROO = yes; # Hybrid graphics support
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
DRM_GMA500 = whenAtLeast "5.12" module;
linux-kernel: update config for 5.13
2021-06-14 12:25:02 +00:00
DRM_GMA600 = whenOlder "5.13" yes;
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
DRM_GMA3600 = whenOlder "5.12" yes;
linuxPackages_testing: remove unused options for 6.2
2023-02-02 20:54:46 +00:00
DRM_VMWGFX_FBCON = whenOlder "6.2" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# (experimental) amdgpu support for verde and newer chipsets
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
DRM_AMDGPU_SI = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# (stable) amdgpu support for bonaire and newer chipsets
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
DRM_AMDGPU_CIK = yes;
kernel config: CONFIG_DRM_DP_AUX_CHARDEV
2019-08-06 22:29:54 +00:00
# Allow device firmware updates
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
DRM_DP_AUX_CHARDEV = yes;
linux: enable display core support for all amd gpus newer amd gpus require display core to function normally
2021-01-12 11:29:07 +00:00
# amdgpu display core (DC) support
DRM_AMD_DC_DCN1_0 = whenBetween "4.15" "5.6" yes;
DRM_AMD_DC_PRE_VEGA = whenBetween "4.15" "4.18" yes;
DRM_AMD_DC_DCN2_0 = whenBetween "5.3" "5.6" yes;
DRM_AMD_DC_DCN2_1 = whenBetween "5.4" "5.6" yes;
DRM_AMD_DC_DCN3_0 = whenBetween "5.9" "5.11" yes;
Merge branch 'staging-next' into staging Resolved the following conflicts: - kernel flags between 09176d28a0ebbd8bae35f92406a48cded78ac9f7 and 2b28822d8dec38fd62ef17643ec71da3bb488619 - clojure-lsp between 3fa00685ce21fb1a4218d444d0862ebcb2b99706 and e03c068af5c9ca950214cd6c3b5cfe67f595da88
2021-02-19 16:15:31 +00:00
DRM_AMD_DC_DCN = whenAtLeast "5.11" yes;
linux: enable display core support for all amd gpus newer amd gpus require display core to function normally
2021-01-12 11:29:07 +00:00
DRM_AMD_DC_HDCP = whenAtLeast "5.5" yes;
DRM_AMD_DC_SI = whenAtLeast "5.10" yes;
kernel config: Fix recent i686 builds
2018-06-30 11:01:22 +00:00
} // optionalAttrs (stdenv.hostPlatform.system == "x86_64-linux") {
# Intel GVT-g graphics virtualization supports 64-bit only
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
DRM_I915_GVT = whenAtLeast "4.16" yes;
DRM_I915_GVT_KVMGT = whenAtLeast "4.16" module;
linux: kernel: enable DRM_HYPERV
2022-04-15 22:27:40 +00:00
# Enable Hyper-V Synthetic DRM Driver
DRM_HYPERV = whenAtLeast "5.14" module;
linux: enable vc4 HDMI-CEC by default (#176762)
2022-06-10 09:20:50 +00:00
} // optionalAttrs (stdenv.hostPlatform.system == "aarch64-linux") {
# enable HDMI-CEC on RPi boards
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
DRM_VC4_HDMI_CEC = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
sound = {
SND_DYNAMIC_MINORS = yes;
SND_AC97_POWER_SAVE = yes; # AC97 Power-Saving Mode
SND_HDA_INPUT_BEEP = yes; # Support digital beep via input layer
SND_HDA_RECONFIG = yes; # Support reconfiguration of jack functions
# Support configuring jack functions via fw mechanism at boot
SND_HDA_PATCH_LOADER = yes;
linux config: SND_HDA_CODEC_CA0132_DSP for <5.7 The default enable for SND_HDA_CODEC_CA0132_DSP was already merged into 5.7-rc1 [1], which means we can adjust the whenOlder to 5.7. [1]: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=652bb5d8df4b3a79ed350db35cda12637e63efa7
2020-04-17 06:43:49 +00:00
SND_HDA_CODEC_CA0132_DSP = whenOlder "5.7" yes; # Enable DSP firmware loading on Creative Soundblaster Z/Zx/ZxR/Recon
linux: enable SND_OSSEMUL Fixes #54074
2019-02-18 16:11:37 +00:00
SND_OSSEMUL = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
SND_USB_CAIAQ_INPUT = yes;
linux config: enable Sound Open Firmware support
2020-04-29 03:34:41 +00:00
# Enable Sound Open Firmware support
} // optionalAttrs (stdenv.hostPlatform.system == "x86_64-linux" &&
versionAtLeast version "5.5") {
linux config: SND_SOC_INTEL_SOUNDWIRE_SOF_MACH >= 5.10 Enable for SND_SOC_INTEL_SOUNDWIRE_SOF_MACH kernel module. This is used on some 10/11th gen Intel laptops such as the XPS 17 97[00|10] Enable SND_SOC_INTEL_USER_FRIENDLY_LONG_NAMES as well - this is required dep
2022-01-10 02:33:41 +00:00
SND_SOC_INTEL_SOUNDWIRE_SOF_MACH = whenAtLeast "5.10" module;
SND_SOC_INTEL_USER_FRIENDLY_LONG_NAMES = whenAtLeast "5.10" yes; # dep of SOF_MACH
linuxKernel.kernels: also enable SND_SOC_SOF_INTEL_SOUNDWIRE_LINK between 5.10-5.11
2022-01-10 14:11:49 +00:00
SND_SOC_SOF_INTEL_SOUNDWIRE_LINK = whenBetween "5.10" "5.11" yes; # dep of SOF_MACH
linux config: enable Sound Open Firmware support
2020-04-29 03:34:41 +00:00
SND_SOC_SOF_TOPLEVEL = yes;
SND_SOC_SOF_ACPI = module;
SND_SOC_SOF_PCI = module;
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
SND_SOC_SOF_APOLLOLAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_APOLLOLAKE_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_CANNONLAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_CANNONLAKE_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_COFFEELAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_COFFEELAKE_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_COMETLAKE = whenAtLeast "5.12" module;
linux: 5.8-rc2 -> 5.8-rc4
2020-07-10 14:23:46 +00:00
SND_SOC_SOF_COMETLAKE_H_SUPPORT = whenOlder "5.8" yes;
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
SND_SOC_SOF_COMETLAKE_LP_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_ELKHARTLAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_ELKHARTLAKE_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_GEMINILAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_GEMINILAKE_SUPPORT = whenOlder "5.12" yes;
linux config: enable Sound Open Firmware support
2020-04-29 03:34:41 +00:00
SND_SOC_SOF_HDA_AUDIO_CODEC = yes;
linux: fix kernel config options for 5.7
2020-06-10 12:07:15 +00:00
SND_SOC_SOF_HDA_COMMON_HDMI_CODEC = whenOlder "5.7" yes;
linux config: enable Sound Open Firmware support
2020-04-29 03:34:41 +00:00
SND_SOC_SOF_HDA_LINK = yes;
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
SND_SOC_SOF_ICELAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_ICELAKE_SUPPORT = whenOlder "5.12" yes;
linux config: enable Sound Open Firmware support
2020-04-29 03:34:41 +00:00
SND_SOC_SOF_INTEL_TOPLEVEL = yes;
linux-testing: 5.11-rc5 -> 5.12-rc4 - Sound SOC support was refactored in torvalds/linux@43df5242af4ed67e8811257ab1bfe6a07e4a5858 - GMA3600 is part of GMA500 as of torvalds/linux@26499e0518a77de29e7db2c53fb0d0e9e15be8fb
2021-03-24 14:03:46 +00:00
SND_SOC_SOF_JASPERLAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_JASPERLAKE_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_MERRIFIELD = whenAtLeast "5.12" module;
SND_SOC_SOF_MERRIFIELD_SUPPORT = whenOlder "5.12" yes;
SND_SOC_SOF_TIGERLAKE = whenAtLeast "5.12" module;
SND_SOC_SOF_TIGERLAKE_SUPPORT = whenOlder "5.12" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
usb-serial = {
USB_SERIAL_GENERIC = yes; # USB Generic Serial Driver
} // optionalAttrs (versionOlder version "4.16") {
kernel config: Minor text fixes
2018-06-30 11:05:29 +00:00
# Include firmware for various USB serial devices.
# Only applicable for kernels below 4.16, after that no firmware is shipped in the kernel tree.
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
USB_SERIAL_KEYSPAN_MPR = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
USB_SERIAL_KEYSPAN_USA28 = yes;
USB_SERIAL_KEYSPAN_USA28X = yes;
USB_SERIAL_KEYSPAN_USA28XA = yes;
USB_SERIAL_KEYSPAN_USA28XB = yes;
USB_SERIAL_KEYSPAN_USA19 = yes;
USB_SERIAL_KEYSPAN_USA18X = yes;
USB_SERIAL_KEYSPAN_USA19W = yes;
USB_SERIAL_KEYSPAN_USA19QW = yes;
USB_SERIAL_KEYSPAN_USA19QI = yes;
USB_SERIAL_KEYSPAN_USA49W = yes;
USB_SERIAL_KEYSPAN_USA49WLC = yes;
};
usb = {
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
USB_DEBUG = { optional = true; tristate = whenOlder "4.18" "n";};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
USB_EHCI_ROOT_HUB_TT = yes; # Root Hub Transaction Translators
USB_EHCI_TT_NEWSCHED = yes; # Improved transaction translator scheduling
linux: enable config option for raw HID devices This flag allows for things like Uninterruptable Power Supplies to work properly. See https://cateee.net/lkddb/web-lkddb/USB_HIDDEV.html for more information.
2019-06-14 22:15:12 +00:00
USB_HIDDEV = yes; # USB Raw HID Devices (like monitor controls and Uninterruptable Power Supplies)
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
# Filesystem options - in particular, enable extended attributes and
# ACLs for all filesystems that support them.
filesystem = {
kernel: common-config.nix: enable FANOTIFY_ACCESS_PERMISSIONS Required for, eg, ClamAV's OnAccessPrevention feature.
2022-03-15 17:43:54 +00:00
FANOTIFY = yes;
FANOTIFY_ACCESS_PERMISSIONS = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
TMPFS = yes;
TMPFS_POSIX_ACL = yes;
linux: further cleanup config after drop of 4.9
2022-10-23 16:28:32 +00:00
FS_ENCRYPTION = if (versionAtLeast version "5.1") then yes else option module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
EXT2_FS_XATTR = yes;
EXT2_FS_POSIX_ACL = yes;
EXT2_FS_SECURITY = yes;
EXT3_FS_POSIX_ACL = yes;
EXT3_FS_SECURITY = yes;
EXT4_FS_POSIX_ACL = yes;
EXT4_FS_SECURITY = yes;
linux: fix unused option warnings on 5.x kernels
2022-03-19 23:04:51 +00:00
EXT4_ENCRYPTION = whenOlder "5.1" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: disable NTFS_FS, enable NTFS3_LZX_XPRESS and NTFS3_FS_POSIX_ACL
2022-08-09 20:43:56 +00:00
NTFS_FS = whenAtLeast "5.15" no;
NTFS3_LZX_XPRESS = whenAtLeast "5.15" yes;
NTFS3_FS_POSIX_ACL = whenAtLeast "5.15" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
REISERFS_FS_XATTR = option yes;
REISERFS_FS_POSIX_ACL = option yes;
REISERFS_FS_SECURITY = option yes;
JFS_POSIX_ACL = option yes;
JFS_SECURITY = option yes;
XFS_QUOTA = option yes;
XFS_POSIX_ACL = option yes;
XFS_RT = option yes; # XFS Realtime subvolume support
linux: XFS_ONLINE_SCRUB=y (#195266)
2022-10-12 18:10:07 +00:00
XFS_ONLINE_SCRUB = option yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
OCFS2_DEBUG_MASKLOG = option no;
BTRFS_FS_POSIX_ACL = yes;
UBIFS_FS_ADVANCED_COMPR = option yes;
F2FS_FS = module;
F2FS_FS_SECURITY = option yes;
linux: fix unused option warnings on 5.x kernels
2022-03-19 23:04:51 +00:00
F2FS_FS_ENCRYPTION = whenOlder "5.1" yes;
linux: enable F2FS compression Fixes https://github.com/NixOS/nixpkgs/issues/110534
2021-01-22 21:35:34 +00:00
F2FS_FS_COMPRESSION = whenAtLeast "5.6" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
UDF_FS = module;
linuxPackages_testing: remove unused options for 6.2
2023-02-02 20:54:46 +00:00
NFSD_V2_ACL = whenOlder "6.2" yes;
linux_latest: 5.17.9 -> 5.18 NSFD_V3 is now always enabled, and enabling debug info now requires selecting a DWARF version instead of just setting DEBUG_INFO=y.
2022-05-23 09:19:42 +00:00
NFSD_V3 = whenOlder "5.18" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
NFSD_V3_ACL = yes;
NFSD_V4 = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
NFSD_V4_SECURITY_LABEL = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
NFS_FSCACHE = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
NFS_SWAP = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
NFS_V3_ACL = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
NFS_V4_1 = yes; # NFSv4.1 client support
NFS_V4_2 = yes;
NFS_V4_SECURITY_LABEL = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
CIFS_XATTR = yes;
kernel-config: CIFS_POSIX no longer exists
2018-12-21 14:10:17 +00:00
CIFS_POSIX = option yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
CIFS_FSCACHE = yes;
linux: 4.18-rc8 -> 4.19-rc1
2018-08-27 01:50:19 +00:00
CIFS_STATS = whenOlder "4.19" yes;
linux_testing: 5.13-rc6 -> 5.15-rc6
2021-10-21 19:27:44 +00:00
CIFS_WEAK_PW_HASH = whenOlder "5.15" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
CIFS_UPCALL = yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
CIFS_ACL = whenOlder "5.3" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
CIFS_DFS_UPCALL = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
CEPH_FSCACHE = yes;
CEPH_FS_POSIX_ACL = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
SQUASHFS_FILE_DIRECT = yes;
linuxPackages_testing: remove unused options for 6.2
2023-02-02 20:54:46 +00:00
SQUASHFS_DECOMP_MULTI_PERCPU = whenOlder "6.2" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
SQUASHFS_XATTR = yes;
SQUASHFS_ZLIB = yes;
SQUASHFS_LZO = yes;
SQUASHFS_XZ = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
SQUASHFS_LZ4 = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
SQUASHFS_ZSTD = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Native Language Support modules, needed by some filesystems
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
NLS = yes;
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
NLS_DEFAULT = freeform "utf8";
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
NLS_UTF8 = module;
kernel config: Minor text fixes
2018-06-30 11:05:29 +00:00
NLS_CODEPAGE_437 = module; # VFAT default for the codepage= mount option
NLS_ISO8859_1 = module; # VFAT default for the iocharset= mount option
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: enable ISO9660_FS module
2022-02-18 01:26:55 +00:00
# Needed to use the installation iso image. Not included in all defconfigs (e.g. arm64)
ISO9660_FS = module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
DEVTMPFS = yes;
linux: Enable CONFIG_UNICODE to support casefold Required to mount filesystems with casefold enabled. Fixes #109432
2021-01-29 12:38:41 +00:00
UNICODE = whenAtLeast "5.2" yes; # Casefolding support for filesystems
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
security = {
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
FORTIFY_SOURCE = option yes;
linux: enable FORTIFY_SOURCE
2022-01-10 20:44:59 +00:00
linux: enable DEBUG_LIST
2022-01-09 19:41:33 +00:00
# https://googleprojectzero.blogspot.com/2019/11/bad-binder-android-in-wild-exploit.html
DEBUG_LIST = yes;
linux: Enable HARDENED_USERCOPY Enabled in [Arch][1], [Debian][2], [Fedora][3]. Recommended by [Kernel Self Protection Project][4]. Originally [reported to have no noticeable performance impact][5]. [1]: https://github.com/archlinux/svntogit-packages/blob/66d72ee54afc604391b618fc3eecc43f29e479e8/trunk/config#L10252 [2]: https://salsa.debian.org/kernel-team/linux/-/blob/07731f5956cf29876a7abc13f4ecbdf4d9459592/debian/config/config#L7710 [3]: https://src.fedoraproject.org/rpms/kernel/blob/6d6ad72f0ccfe72146f2876f90fe609548caa349/f/kernel-x86_64-fedora.config#_2202 [4]: https://kernsec.org/wiki/index.php/Kernel_Self_Protection_Project/Recommended_Settings [5]: https://lwn.net/Articles/695991/
2022-08-27 15:40:23 +00:00
HARDENED_USERCOPY = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
RANDOMIZE_BASE = option yes;
linuxKernel.kernels: mark {IO_,}STRICT_DEVMEM optional to unbreak hardened kernels
2022-01-10 20:45:21 +00:00
STRICT_DEVMEM = mkDefault yes; # Filter access to /dev/mem
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
IO_STRICT_DEVMEM = mkDefault yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
SECURITY_SELINUX_BOOTPARAM_VALUE = whenOlder "5.1" (freeform "0"); # Disable SELinux by default
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Prevent processes from ptracing non-children processes
SECURITY_YAMA = option yes;
linux config: enable Landlock LSM
2021-08-12 12:07:12 +00:00
# The goal of Landlock is to enable to restrict ambient rights (e.g. global filesystem access) for a set of processes.
# This does not have any effect if a program does not support it
SECURITY_LANDLOCK = whenAtLeast "5.13" yes;
linux-kernel: update config for 5.13
2021-06-14 12:25:02 +00:00
DEVKMEM = whenOlder "5.13" no; # Disable /dev/kmem
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
USER_NS = yes; # Support for user namespaces
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
SECURITY_APPARMOR = yes;
DEFAULT_SECURITY_APPARMOR = yes;
linuxPackages_testing: remove unused options for 6.2
2023-02-02 20:54:46 +00:00
RANDOM_TRUST_CPU = whenOlder "6.2" (whenAtLeast "4.19" yes); # allow RDRAND to seed the RNG
RANDOM_TRUST_BOOTLOADER = whenOlder "6.2" (whenAtLeast "5.4" yes); # allow the bootloader to seed the RNG
linux: enable RANDOM_TRUST_CPU Allows the RDRAND instruction to seed the kernel RNG. This should lead to faster CRNG initialization, since the kernel can use the processors capabilities directly. This is default enabled in Debian [1] and Fedora [2] as well. [1]: https://salsa.debian.org/kernel-team/linux/-/blob/debian/5.10.13-1/debian/config/config#L459 [2]: https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-x86_64-fedora.config#_4907
2021-02-24 04:35:03 +00:00
linux: disable module signing Without this, the kernel would generate a random one for us which obviously isn't reproducible. `nix-build -A linux --check` succeeds now! (Tested at different times with different kernel)
2020-12-23 06:33:25 +00:00
MODULE_SIG = no; # r13y, generates a random key during build and bakes it in
# Depends on MODULE_SIG and only really helps when you sign your modules
# and enforce signatures which we don't do by default.
linux: fix unused option warnings on 5.x kernels
2022-03-19 23:04:51 +00:00
SECURITY_LOCKDOWN_LSM = whenAtLeast "5.4" no;
linux: enable PERSISTENT_KEYRINGS and KEYS_REQUEST_CACHE PERSISTENT_KEYRINGS provides a register of persistent per-UID keyrings, useful for encrypting storage pools in stratis. KEYS_REQUEST_CACHE enable temporary caching of the last request_key() result.
2022-08-02 07:03:56 +00:00
# provides a register of persistent per-UID keyrings, useful for encrypting storage pools in stratis
PERSISTENT_KEYRINGS = yes;
# enable temporary caching of the last request_key() result
KEYS_REQUEST_CACHE = whenAtLeast "5.3" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
} // optionalAttrs (!stdenv.hostPlatform.isAarch32) {
# Detect buffer overflows on the stack
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
CC_STACKPROTECTOR_REGULAR = {optional = true; tristate = whenOlder "4.18" "y";};
linux: enable X86_SGX{_KVM} on x86_64 only The config option X86_SGX is available on x86_64-linux only; i686-linux is not supported. https://github.com/torvalds/linux/blob/55a677b/arch/x86/Kconfig#L1914
2022-01-04 23:19:09 +00:00
} // optionalAttrs stdenv.hostPlatform.isx86_64 {
linux: enable X86_SGX and X86_SGX_KVM on x86 Enable Intel Software Guard eXtensions (SGX) on x86 when using Linux 5.11.0 or later. Also enable KVM guests to create SGX enclaves if running Linux 5.13.0 or later.
2021-11-29 07:03:26 +00:00
# Enable Intel SGX
X86_SGX = whenAtLeast "5.11" yes;
# Allow KVM guests to load SGX enclaves
X86_SGX_KVM = whenAtLeast "5.13" yes;
linux: enable AMD SME, SEV, SEV-SE, SEV-SNP on x86_64 Enables the following kernel config options for AMD CPUs on x86_64: - `CRYPTO_DEV_CCP`: Enables offloading of crypto operations to AMD's Cryptographic Coprocessor (CCP). Also required by `KVM_AMD_SEV`. - `AMD_MEM_ENCRYPT`: Enables support for Secure Memory Encryption (SME). Please note that `AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT` is not enabled; yet, you you can enable memory encryption by passing `mem_encrypt=on` as a kernal command line option. - `KVM_AMD_SEV`: Enables launching Encrypted VMs (SEV) and Secure VMs with Encrypted State (SEV-ES). - `SEV_GUEST`: Enables support for AMD Secure Encrypted Virtualization with Secure Nested Paging (SEV-SNP). Built as module. Enabling these options is in line with other distros, e.g., Debian, Fedora or Arch Linux.
2022-12-01 10:32:45 +00:00
# AMD Cryptographic Coprocessor (CCP)
CRYPTO_DEV_CCP = yes;
# AMD SME
AMD_MEM_ENCRYPT = yes;
# AMD SEV and AMD SEV-SE
KVM_AMD_SEV = whenAtLeast "4.16" yes;
# AMD SEV-SNP
SEV_GUEST = whenAtLeast "5.19" module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
microcode = {
MICROCODE = yes;
MICROCODE_INTEL = yes;
MICROCODE_AMD = yes;
# Write Back Throttling
# https://lwn.net/Articles/682582/
# https://bugzilla.kernel.org/show_bug.cgi?id=12309#c655
BLK_WBT = yes;
linux: 4.20-rc7 -> 5.0-rc1
2019-01-07 12:57:32 +00:00
BLK_WBT_SQ = whenOlder "5.0" yes; # Removed in 5.0-RC1
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BLK_WBT_MQ = yes;
};
container = {
linux: enable CGROUP_HUGETLB, CGROUP_PERF, CGROUP_RDMA (#50225) These options were added in: NAMESPACES 2.6.25 CGROUP_DEVICE 2.6.26 CGROUP_HUGETLB 3.6 CGROUP_PERF 2.6.39 CGROUP_RDMA 4.11
2018-11-13 01:13:00 +00:00
NAMESPACES = yes; # Required by 'unshare' used by 'nixos-install'
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
RT_GROUP_SCHED = no;
linux: enable CGROUP_HUGETLB, CGROUP_PERF, CGROUP_RDMA (#50225) These options were added in: NAMESPACES 2.6.25 CGROUP_DEVICE 2.6.26 CGROUP_HUGETLB 3.6 CGROUP_PERF 2.6.39 CGROUP_RDMA 4.11
2018-11-13 01:13:00 +00:00
CGROUP_DEVICE = yes;
CGROUP_HUGETLB = yes;
CGROUP_PERF = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
CGROUP_RDMA = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
MEMCG = yes;
linuxKernel.kernels.linux_testing: 6.0-rc5 -> 6.1-rc3
2022-10-31 13:21:39 +00:00
MEMCG_SWAP = whenOlder "6.1" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BLK_DEV_THROTTLING = yes;
linux: 4.20-rc7 -> 5.0-rc1
2019-01-07 12:57:32 +00:00
CFQ_GROUP_IOSCHED = whenOlder "5.0" yes; # Removed in 5.0-RC1
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
CGROUP_PIDS = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
staging = {
# Enable staging drivers. These are somewhat experimental, but
# they generally don't hurt.
STAGING = yes;
};
proc-events = {
# PROC_EVENTS requires that the netlink connector is not built
# as a module. This is required by libcgroup's cgrulesengd.
CONNECTOR = yes;
PROC_EVENTS = yes;
};
tracing = {
FTRACE = yes;
KPROBES = yes;
FUNCTION_TRACER = yes;
FTRACE_SYSCALLS = yes;
SCHED_TRACER = yes;
STACK_TRACER = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
UPROBE_EVENTS = option yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
BPF_SYSCALL = yes;
linux: Enable BPF_UNPRIV_DEFAULT_OFF in 5.15
2022-03-21 01:19:07 +00:00
BPF_UNPRIV_DEFAULT_OFF = whenBetween "5.10" "5.16" yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
BPF_EVENTS = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
FUNCTION_PROFILER = yes;
RING_BUFFER_BENCHMARK = no;
};
virtualisation = {
PARAVIRT = option yes;
linux: remove remnants of grsecurity Nixpkgs hasn't supported grsecurity kernels since 2017, so unless anybody is manually enabling the grsecurity feature to make these small kernel tweaks this is dead code. This means we don't actually support any "features" in the kernel common-config any more, but I've left the argument there because it's conceivable we could have some again in future.
2021-04-05 10:01:58 +00:00
HYPERVISOR_GUEST = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
PARAVIRT_SPINLOCKS = option yes;
KVM_ASYNC_PF = yes;
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
KVM_GENERIC_DIRTYLOG_READ_PROTECT = yes;
linux: remove remnants of grsecurity Nixpkgs hasn't supported grsecurity kernels since 2017, so unless anybody is manually enabling the grsecurity feature to make these small kernel tweaks this is dead code. This means we don't actually support any "features" in the kernel common-config any more, but I've left the argument there because it's conceivable we could have some again in future.
2021-04-05 10:01:58 +00:00
KVM_GUEST = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
KVM_MMIO = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
KVM_VFIO = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
KSM = yes;
VIRT_DRIVERS = yes;
Fix some typos in comments
2021-04-01 18:32:45 +00:00
# We need 64 GB (PAE) support for Xen guest support
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
HIGHMEM64G = { optional = true; tristate = mkIf (!stdenv.is64bit) "y";};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
VFIO_PCI_VGA = mkIf stdenv.is64bit yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: Disable VirtualBox guest modules. These break the modules from VirtualBox additions. Fixes issue https://github.com/NixOS/nixpkgs/issues/58127.
2019-05-04 19:13:26 +00:00
# VirtualBox guest drivers in the kernel conflict with the ones in the
# official additions package and prevent the vboxsf module from loading,
# so disable them for now.
VBOXGUEST = option no;
DRM_VBOXVIDEO = option no;
linux: enable Xen everywhere it can be I don't think there's any reason to have a seperate kernel variant because of this, with all the maintenance burden that imposes. Debian and Fedora both enable all these options on their normal kernels. Alias the Linux Xen attributes, so this change should be seemless for people who were using the Xen kernels up to now. All the Xen options are marked as optional anyway, so it should be fine to try to enable them on non-x86 platforms as well. Fixes: https://github.com/NixOS/nixpkgs/issues/115182
2021-04-01 17:30:29 +00:00
XEN = option yes;
XEN_DOM0 = option yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
PCI_XEN = option yes;
HVC_XEN = option yes;
HVC_XEN_FRONTEND = option yes;
XEN_SYS_HYPERVISOR = option yes;
SWIOTLB_XEN = option yes;
XEN_BACKEND = option yes;
XEN_BALLOON = option yes;
XEN_BALLOON_MEMORY_HOTPLUG = option yes;
XEN_EFI = option yes;
XEN_HAVE_PVMMU = option yes;
XEN_MCE_LOG = option yes;
XEN_PVH = option yes;
XEN_PVHVM = option yes;
XEN_SAVE_RESTORE = option yes;
linux: fix unused option warnings on 5.x kernels
2022-03-19 23:04:51 +00:00
XEN_SCRUB_PAGES = whenOlder "4.19" yes;
XEN_SELFBALLOONING = whenOlder "5.3" yes;
linux: enable VIRTIO_MMIO_CMDLINE_DEVICES
2023-02-27 21:39:23 +00:00
# Enable device detection on virtio-mmio hypervisors
VIRTIO_MMIO_CMDLINE_DEVICES = yes;
linux: enable Xen everywhere it can be I don't think there's any reason to have a seperate kernel variant because of this, with all the maintenance burden that imposes. Debian and Fedora both enable all these options on their normal kernels. Alias the Linux Xen attributes, so this change should be seemless for people who were using the Xen kernels up to now. All the Xen options are marked as optional anyway, so it should be fine to try to enable them on non-x86 platforms as well. Fixes: https://github.com/NixOS/nixpkgs/issues/115182
2021-04-01 17:30:29 +00:00
};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
media = {
MEDIA_DIGITAL_TV_SUPPORT = yes;
MEDIA_CAMERA_SUPPORT = yes;
linux: 4.20-rc7 -> 5.0-rc1
2019-01-07 12:57:32 +00:00
MEDIA_CONTROLLER = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
MEDIA_PCI_SUPPORT = yes;
MEDIA_USB_SUPPORT = yes;
MEDIA_ANALOG_TV_SUPPORT = yes;
VIDEO_STK1160_COMMON = module;
};
"9p" = {
# Enable the 9P cache to speed up NixOS VM tests.
"9P_FSCACHE" = option yes;
"9P_FS_POSIX_ACL" = option yes;
};
huge-page = {
TRANSPARENT_HUGEPAGE = option yes;
TRANSPARENT_HUGEPAGE_ALWAYS = option no;
TRANSPARENT_HUGEPAGE_MADVISE = option yes;
};
zram = {
ZRAM = module;
ZSWAP = option yes;
ZBUD = option yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
ZSMALLOC = module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
brcmfmac = {
# Enable PCIe and USB for the brcmfmac driver
BRCMFMAC_USB = option yes;
BRCMFMAC_PCIE = option yes;
};
# Support x2APIC (which requires IRQ remapping)
reewide: Purge all uses `stdenv.system` and top-level `system` It is deprecated and will be removed after 18.09.
2018-08-20 19:11:29 +00:00
x2apic = optionalAttrs (stdenv.hostPlatform.system == "x86_64-linux") {
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
X86_X2APIC = yes;
IRQ_REMAP = yes;
};
kernel config: Minor text fixes
2018-06-30 11:05:29 +00:00
# Disable various self-test modules that have no use in a production system
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
tests = {
# This menu disables all/most of them on >= 4.16
RUNTIME_TESTING_MENU = option no;
} // optionalAttrs (versionOlder version "4.16") {
kernel config: Minor text fixes
2018-06-30 11:05:29 +00:00
# For older kernels, painstakingly disable each symbol.
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
ARM_KPROBES_TEST = option no;
ASYNC_RAID6_TEST = option no;
ATOMIC64_SELFTEST = option no;
BACKTRACE_SELF_TEST = option no;
INTERVAL_TREE_TEST = option no;
PERCPU_TEST = option no;
RBTREE_TEST = option no;
TEST_BITMAP = option no;
TEST_BPF = option no;
TEST_FIRMWARE = option no;
TEST_HASH = option no;
TEST_HEXDUMP = option no;
TEST_KMOD = option no;
TEST_KSTRTOX = option no;
TEST_LIST_SORT = option no;
TEST_LKM = option no;
TEST_PARMAN = option no;
TEST_PRINTF = option no;
TEST_RHASHTABLE = option no;
TEST_SORT = option no;
TEST_STATIC_KEYS = option no;
TEST_STRING_HELPERS = option no;
TEST_UDELAY = option no;
TEST_USER_COPY = option no;
TEST_UUID = option no;
} // {
CRC32_SELFTEST = option no;
CRYPTO_TEST = option no;
EFI_TEST = option no;
GLOB_SELFTEST = option no;
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
DRM_DEBUG_MM_SELFTEST = { optional = true; tristate = whenOlder "4.18" "n";};
LNET_SELFTEST = { optional = true; tristate = whenOlder "4.18" "n";};
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
LOCK_TORTURE_TEST = option no;
MTD_TESTS = option no;
NOTIFIER_ERROR_INJECTION = option no;
linux: further cleanup config after drop of 4.9
2022-10-23 16:28:32 +00:00
RCU_PERF_TEST = whenOlder "5.9" no;
linux: fix unused option warnings on 5.x kernels
2022-03-19 23:04:51 +00:00
RCU_SCALE_TEST = whenAtLeast "5.10" no;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
RCU_TORTURE_TEST = option no;
TEST_ASYNC_DRIVER_PROBE = option no;
WW_MUTEX_SELFTEST = option no;
XZ_DEC_TEST = option no;
kernel/common-config: move criu to its own set It was previously in the "test" set, which seemed an inappropriate placement.
2019-05-08 17:40:39 +00:00
};
linux: enable kcmp() system call Since 2020, Mesa requires the kcmp() system call to be available for some of its functionality. At the moment, this system call is enabled when CHECKPOINT_RESTORE is enabled in the Linux kernel. This option is also enabled by default in Fedora and Debian. This patch enables CHECKPOINT_RESTORE by default on all kernel versions where it is not marked as EXPERT, i.e. everything after and including 4.19. Fixes #112082.
2021-02-15 21:48:12 +00:00
criu = if (versionAtLeast version "4.19") then {
# Unconditionally enabled, because it is required for CRIU and
# it provides the kcmp() system call that Mesa depends on.
CHECKPOINT_RESTORE = yes;
} else optionalAttrs (features.criu or false) ({
# For older kernels, CHECKPOINT_RESTORE is hidden behind EXPERT.
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
EXPERT = yes;
CHECKPOINT_RESTORE = yes;
} // optionalAttrs (features.criu_revert_expert or true) {
RFKILL_INPUT = option yes;
HID_PICOLCD_FB = option yes;
HID_PICOLCD_BACKLIGHT = option yes;
HID_PICOLCD_LCD = option yes;
HID_PICOLCD_LEDS = option yes;
HID_PICOLCD_CIR = option yes;
DEBUG_MEMORY_INIT = option yes;
});
linux: don't compress by ZSTD on 32-bit It doesn't build on i686 and I don't consider it worth deep investigation. I tried a VM test, too (nixosTests.knot).
2021-04-03 08:46:30 +00:00
misc = let
# Use zstd for kernel compression if 64-bit and newer than 5.9, otherwise xz.
# i686 issues: https://github.com/NixOS/nixpkgs/pull/117961#issuecomment-812106375
useZstd = stdenv.buildPlatform.is64bit && versionAtLeast version "5.9";
in {
KERNEL_XZ = mkIf (!useZstd) yes;
KERNEL_ZSTD = mkIf useZstd yes;
linux-kernel: HID_BATTERY_STRENGTH=yes
2019-09-05 15:22:13 +00:00
HID_BATTERY_STRENGTH = yes;
linux: CONFIG_HIDRAW=y
2020-05-02 12:06:43 +00:00
# enabled by default in x86_64 but not arm64, so we do that here
HIDRAW = yes;
linux: enable force feedback support in HID drivers
2020-06-11 10:41:59 +00:00
HID_ACRUX_FF = yes;
DRAGONRISE_FF = yes;
linux: enable FF for many gamepads
2022-01-24 10:01:59 +00:00
GREENASIA_FF = yes;
linux: enable force feedback support in HID drivers
2020-06-11 10:41:59 +00:00
HOLTEK_FF = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
JOYSTICK_PSXPAD_SPI_FF = yes;
linux: enable FF for many gamepads
2022-01-24 10:01:59 +00:00
LOGIG940_FF = yes;
NINTENDO_FF = whenAtLeast "5.16" yes;
PLAYSTATION_FF = whenAtLeast "5.12" yes;
linux: enable force feedback support in HID drivers
2020-06-11 10:41:59 +00:00
SONY_FF = yes;
SMARTJOYPLUS_FF = yes;
THRUSTMASTER_FF = yes;
ZEROPLUS_FF = yes;
linux-kernel: update config for 5.13
2021-06-14 12:25:02 +00:00
MODULE_COMPRESS = whenOlder "5.13" yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
MODULE_COMPRESS_XZ = yes;
kernel: use zstd compression on 5.9+.
2020-12-12 04:56:33 +00:00
linux: explicitly enable SYSVIPC The linux-hardened patch set removes this default, probably because of its original focus on Android kernel hardening.
2020-04-05 03:18:36 +00:00
SYSVIPC = yes; # System-V IPC
linux: explicitly enable AIO This is disabled by default in the linux-hardened patchset, but is required by e.g. LVM. Fixes #87260.
2020-05-10 21:34:19 +00:00
AIO = yes; # POSIX asynchronous I/O
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
UNIX = yes; # Unix domain sockets.
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
MD = yes; # Device mapper (RAID, LVM, etc.)
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Enable initrd support.
BLK_DEV_INITRD = yes;
PM_TRACE_RTC = no; # Disable some expensive (?) features.
ACCESSIBILITY = yes; # Accessibility support
AUXDISPLAY = yes; # Auxiliary Display support
DONGLE = whenOlder "4.17" yes; # Serial dongle support
HIPPI = yes;
MTD_COMPLEX_MAPPINGS = yes; # needed for many devices
SCSI_LOWLEVEL = yes; # enable lots of SCSI devices
SCSI_LOWLEVEL_PCMCIA = yes;
SCSI_SAS_ATA = yes; # added to enable detection of hard drive
SPI = yes; # needed for many devices
SPI_MASTER = yes;
"8139TOO_8129" = yes;
"8139TOO_PIO" = no; # PIO is slower
AIC79XX_DEBUG_ENABLE = no;
AIC7XXX_DEBUG_ENABLE = no;
AIC94XX_DEBUG = no;
BLK_DEV_INTEGRITY = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
BLK_SED_OPAL = yes;
linux: Enable support for Self-Encrypting Drives To quote block/Kconfig: > Builds Logic for interfacing with Opal enabled controllers. > Enabling this option enables users to setup/unlock/lock > Locking ranges for SED devices using the Opal protocol. Without `BLK_SED_OPAL`, it is impossible to resume from sleep when using a locked self-encrypting drive. This configuration option appeared in earlier kernels, but only reached maturity in 4.14 according to discussion at: - https://github.com/Drive-Trust-Alliance/sedutil/issues/90 and - https://github.com/Drive-Trust-Alliance/sedutil/pull/190 This kernel option is enabled in the default kernels shipped with Fedora, Debian, and other mainstream Linux distributions.
2019-02-21 12:05:43 +00:00
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BSD_PROCESS_ACCT_V3 = yes;
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
SERIAL_DEV_BUS = yes; # enables support for serial devices
SERIAL_DEV_CTRL_TTYPORT = yes; # enables support for TTY serial devices
linux: Add kernel config required for QCA6390 bluetooth (XPS 9310) This commit aims to upstream some kernel config patches from the [xps 9310 module in the nixos-hardware repo][1]. These patches have been in use by at least a few community members for around 10 months and seem to consistently and successfully enable bluetooth support for XPS 9310 models that come with the QCA6390 connectivity chip. Without these patches, bluetooth will not work on the Dell XPS 9310 laptop models that come with the QCA6390. *Note that this isn't all XPS 9310 devices, but seems to be all devices shipped with 32GB RAM.* The motivation for upstreaming is pretty simple: currently we have to build the entire kernel every time we want to update it, which takes a good half hour at least on this little laptop :) The `whenAtLeast` version params were determined via the entries for each parameter found at https://cateee.net/lkddb (linked for each below). Added config parameters: - `BT_QCA` - provides the `btqca` module. https://cateee.net/lkddb/web-lkddb/BT_QCA.html - `BT_HCIUART_QCA` - required for QCA6390 bluetooth support. Requires `BT_HCIUART`, `BT_HCIUART_SERDEV` https://cateee.net/lkddb/web-lkddb/BT_HCIUART_QCA.html - `BT_HCIUART_SERDEV` Requires `SERIAL_DEV_BUS`, `BT_HCIUART` https://cateee.net/lkddb/web-lkddb/BT_HCIUART_SERDEV.html - `BT_HCIUART` Requires `SERIAL_DEV_BUS`, `TTY` https://cateee.net/lkddb/web-lkddb/BT_HCIUART.html - `SERIAL_DEV_CTRL_TTYPORT` Requires `TTY`, `SERIAL_DEV_BUS` != module https://cateee.net/lkddb/web-lkddb/SERIAL_DEV_CTRL_TTYPORT.html - `SERIAL_DEV_BUS` https://cateee.net/lkddb/web-lkddb/SERIAL_DEV_BUS.html Fwiw, these parameters are also set in [the default Arch config][2]. [1]: https://github.com/NixOS/nixos-hardware/blob/master/dell/xps/13-9310/default.nix [2]: https://github.com/archlinux/svntogit-packages/blob/packages/linux/trunk/config
2021-11-03 05:09:53 +00:00
linux: add BT_HCIBTUSB_MTK to common kernel config > The MediaTek protocol support enables firmware download support and chip initialization for MediaTek Bluetooth USB controllers. Necessary to make Bluetooth work on some MediaTek controllers.
2021-11-14 00:13:34 +00:00
BT_HCIBTUSB_MTK = whenAtLeast "5.3" yes; # MediaTek protocol support
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
BT_HCIUART_QCA = yes; # Qualcomm Atheros protocol support
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
BT_HCIUART_SERDEV = yes; # required by BT_HCIUART_QCA
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
BT_HCIUART = module; # required for BT devices with serial port interface (QCA6390)
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
BT_HCIUART_BCSP = option yes;
BT_HCIUART_H4 = option yes; # UART (H4) protocol support
BT_HCIUART_LL = option yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
BT_RFCOMM_TTY = option yes; # RFCOMM TTY support
kernel/common-config: clean up after 4.4 removal
2022-02-11 12:54:45 +00:00
BT_QCA = module; # enables QCA6390 bluetooth
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: condition CLEANCACHE to before 5.17 when it was removed
2022-03-23 22:45:14 +00:00
# Removed on 5.17 as it was unused
# upstream: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=0a4ee518185e902758191d968600399f3bc2be31
CLEANCACHE = whenOlder "5.17" (option yes);
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
CRASH_DUMP = option no;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
DVB_DYNAMIC_MINORS = option yes; # we use udev
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
EFI_STUB = yes; # EFI bootloader in the bzImage itself
kernel: enable EFI initrd loader Initrd loader is not enabled by default in some aarch64 kernels, which makes systemd-boot booted kernels fail by default, add this everywhere, since this is a sane default even when it's already in some kernel defaults.
2021-10-05 08:39:07 +00:00
EFI_GENERIC_STUB_INITRD_CMDLINE_LOADER =
linuxPackages_testing: remove unused options for 6.2
2023-02-02 20:54:46 +00:00
whenOlder "6.2" (whenAtLeast "5.8" yes); # initrd kernel parameter for EFI
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
CGROUPS = yes; # used by systemd
FHANDLE = yes; # used by systemd
SECCOMP = yes; # used by systemd >= 231
SECCOMP_FILTER = yes; # ditto
POSIX_MQUEUE = yes;
FRONTSWAP = yes;
FUSION = yes; # Fusion MPT device support
linux: only configure IDE to "no" pre-5.14 When trying to build a 5.14 (rc-*), this fails to build otherwise: > error: unused option: IDE
2021-08-02 18:50:35 +00:00
IDE = whenOlder "5.14" no; # deprecated IDE support, removed in 5.14
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
IDLE_PAGE_TRACKING = yes;
IRDA_ULTRA = whenOlder "4.17" yes; # Ultra (connectionless) protocol
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: 5.2-rc7 -> 5.3-rc5
2019-08-20 18:54:37 +00:00
JOYSTICK_IFORCE_232 = { optional = true; tristate = whenOlder "5.3" "y"; }; # I-Force Serial joysticks and wheels
JOYSTICK_IFORCE_USB = { optional = true; tristate = whenOlder "5.3" "y"; }; # I-Force USB joysticks and wheels
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
JOYSTICK_XPAD_FF = option yes; # X-Box gamepad rumble support
JOYSTICK_XPAD_LEDS = option yes; # LED Support for Xbox360 controller 'BigX' LED
linux: Enable KEYBOARD_APPLESPI This is available for 5.3+ and enables support for Apple trackpads and keyboards Closes #71552
2019-10-23 13:03:16 +00:00
KEYBOARD_APPLESPI = whenAtLeast "5.3" module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
KEXEC_FILE = option yes;
KEXEC_JUMP = option yes;
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
PARTITION_ADVANCED = yes; # Needed for LDM_PARTITION
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Windows Logical Disk Manager (Dynamic Disk) support
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
LDM_PARTITION = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
LOGIRUMBLEPAD2_FF = yes; # Logitech Rumblepad 2 force feedback
LOGO = no; # not needed
MEDIA_ATTACH = yes;
MEGARAID_NEWGEN = yes;
kernel: enable MLX5_CORE_EN when possible Hydra's builders from Packet.net use this option a lot, causing lots of slow deploys.
2018-10-30 16:29:37 +00:00
MLX5_CORE_EN = option yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: enable NVMe Multipath Enable NVME_MULTIPATH so that a single /dev/nvmeXnY device will show up for each NVMe namespaces, even if it is accessible through multiple controllers. Can be disabled at boot with `nvme_core.multipath=0`. This is default enabled in Debian [1], Ubuntu 20.04 [2] , Arch [3] and Fedora 33 [4]. [1]: https://salsa.debian.org/kernel-team/linux/-/blob/debian/5.10.19-1/debian/config/config#L4362 [2]: https://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/focal/tree/debian.master/config/config.common.ubuntu?h=Ubuntu-5.4.0-67.75#n6722 [3]: https://git.archlinux.org/svntogit/packages.git/tree/trunk/config?h=packages/linux#n2423 [4]: https://src.fedoraproject.org/rpms/kernel/blob/f33/f/kernel-x86_64-fedora.config#_4338
2021-03-10 17:40:30 +00:00
NVME_MULTIPATH = whenAtLeast "4.15" yes;
kernel: enable pressure stall information on 4.20+ Closes #56207. Signed-off-by: Austin Seipp <aseipp@pobox.com>
2019-02-22 17:19:36 +00:00
PSI = whenAtLeast "4.20" yes;
linux: CONFIG_MOUSE_ELAN_I2C_SMBUS=y Signed-off-by: Anders Kaseorg <andersk@mit.edu>
2020-05-23 02:35:19 +00:00
MOUSE_ELAN_I2C_SMBUS = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
MOUSE_PS2_ELANTECH = yes; # Elantech PS/2 protocol extension
kernel: enable MOUSE_PS2_VMMOUSE Turns VMware guest mouse support on in the kernel. This is needed for running Wayland and non-root X in a VMWare guest. In a pre-Wayland world the `xf86-input-vmmouse` userspace driver would have handled this for us. This allows the mouse to properly work in a vmware guest (for example it can now leave the vmware window). See: https://github.com/vmware/open-vm-tools/issues/528
2021-07-20 19:41:36 +00:00
MOUSE_PS2_VMMOUSE = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
MTRR_SANITIZER = yes;
NET_FC = yes; # Fibre Channel driver support
kernel: fix touchpads on AMD laptops
2022-08-08 14:28:05 +00:00
# Needed for touchpads to work on some AMD laptops
kernel: only enable PINCTRL_AMD on 5.19+
2022-08-09 00:42:45 +00:00
PINCTRL_AMD = whenAtLeast "5.19" yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# GPIO on Intel Bay Trail, for some Chromebook internal eMMC disks
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
PINCTRL_BAYTRAIL = yes;
linux: build in Cherryview pinctrl driver This allows integrated keyboards on Braswell and Cherryview devices to function properly.
2021-10-27 23:36:36 +00:00
# GPIO for Braswell and Cherryview devices
# Needs to be built-in to for integrated keyboards to function properly
PINCTRL_CHERRYVIEW = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# 8 is default. Modern gpt tables on eMMC may go far beyond 8.
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
MMC_BLOCK_MINORS = freeform "32";
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
REGULATOR = yes; # Voltage and Current Regulator Support
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
RC_DEVICES = option yes; # Enable IR devices
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
RT2800USB_RT53XX = yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
RT2800USB_RT55XX = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
SCHED_AUTOGROUP = yes;
CFS_BANDWIDTH = yes;
SCSI_LOGGING = yes; # SCSI logging facility
SERIAL_8250 = yes; # 8250/16550 and compatible serial support
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
SLAB_FREELIST_HARDENED = yes;
SLAB_FREELIST_RANDOM = yes;
linux: Enable SLAB_FREELIST_HARDENED, SLAB_FREELIST_RANDOM Enabled in [Arch][1], [Debian][2], [Fedora][3]; no others checked. Recommended by [Kernel Self Protection Project][4]. This should also implicitly enable SHUFFLE_PAGE_ALLOCATOR. Performance impact per upstream: For _HARDENED: > The difference gets lost in the noise, but if the above is to be taken > literally, using CONFIG_FREELIST_HARDENED is 0.07% slower. For _RANDOM: > Performance results highlighted no major changes [1]: https://github.com/archlinux/svntogit-packages/blob/66d72ee54afc604391b618fc3eecc43f29e479e8/trunk/config#L1037-L1038 [2]: https://salsa.debian.org/kernel-team/linux/-/blob/07731f5956cf29876a7abc13f4ecbdf4d9459592/debian/config/config#L6742-6743 [3]: https://src.fedoraproject.org/rpms/kernel/blob/6d6ad72f0ccfe72146f2876f90fe609548caa349/f/kernel-x86_64-fedora.config#_6079 [4]: https://kernsec.org/wiki/index.php/Kernel_Self_Protection_Project/Recommended_Settings
2022-08-27 17:45:26 +00:00
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
SLIP_COMPRESSED = yes; # CSLIP compressed headers
SLIP_SMART = yes;
HWMON = yes;
THERMAL_HWMON = yes; # Hardware monitoring support
linux: Enable NVME_HWMON This is available for 5.5+ and enables support for NVMe drives temperature reporting
2020-02-03 17:08:45 +00:00
NVME_HWMON = whenAtLeast "5.5" yes; # NVMe drives temperature reporting
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
UEVENT_HELPER = no;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
USERFAULTFD = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
X86_CHECK_BIOS_CORRUPTION = yes;
X86_MCE = yes;
linux: explicitly enable RAS This is needed for EDAC support.
2020-11-24 10:44:22 +00:00
RAS = yes; # Needed for EDAC support
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Our initrd init uses shebang scripts, so can't be modular.
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
BINFMT_SCRIPT = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# For systemd-binfmt
BINFMT_MISC = option yes;
# Disable the firmware helper fallback, udev doesn't implement it any more
FW_LOADER_USER_HELPER_FALLBACK = option no;
linux: support loading compressed firmware Supported since Linux 5.3.
2022-05-12 15:21:04 +00:00
FW_LOADER_COMPRESS = option yes;
kernel: remove conditions based on 3.X kernels The oldest kernel in nixpkgs being 4.4, we get rid of checks for older kernels.
2018-06-22 23:54:44 +00:00
HOTPLUG_PCI_ACPI = yes; # PCI hotplug using ACPI
HOTPLUG_PCI_PCIE = yes; # PCI-Expresscard hotplug support
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
linux: Enable AMD's amdkfd kernel driver in >= 4.20 The amdkfd driver was built (and loaded when support was detected) without explicit configuration in kernels 4.17-4.19, but not in newer kernels.
2019-03-18 02:28:25 +00:00
# Enable AMD's ROCm GPU compute stack
linux: fix kernel config options for i686 Even the default pkgsi686Linux.linux was broken.
2020-06-12 06:52:09 +00:00
HSA_AMD = mkIf stdenv.hostPlatform.is64bit (whenAtLeast "4.20" yes);
ZONE_DEVICE = mkIf stdenv.hostPlatform.is64bit (whenAtLeast "5.3" yes);
linux: fix amdgpu memory mapping with kernel >= 5.3 These configuration options are needed for the ROCm GPU compute stack on kernels >= 5.3.
2019-12-07 16:06:42 +00:00
HMM_MIRROR = whenAtLeast "5.3" yes;
DRM_AMDGPU_USERPTR = whenAtLeast "5.3" yes;
linux: Enable AMD's amdkfd kernel driver in >= 4.20 The amdkfd driver was built (and loaded when support was detected) without explicit configuration in kernels 4.17-4.19, but not in newer kernels.
2019-03-18 02:28:25 +00:00
kernel/common-config: PREEMPT -> PREEMPT_VOLUNTARY
2019-05-08 17:42:24 +00:00
PREEMPT = no;
PREEMPT_VOLUNTARY = yes;
linux: remove deprecated kernel config options Our oldest kernel is 4.4. That's why we can remove configurations that only applies to kernels older than that.
2019-10-14 10:40:35 +00:00
kernel: Enable X86_AMD_PLATFORM_DEVICE This is needed to get the toupad working on my Acer Nitro laptop.
2019-09-08 14:22:27 +00:00
X86_AMD_PLATFORM_DEVICE = yes;
linux-kernel: Add dell drivers on 5.12+
2021-06-23 13:30:09 +00:00
X86_PLATFORM_DRIVERS_DELL = whenAtLeast "5.12" yes;
kernel/common-config: PREEMPT -> PREEMPT_VOLUNTARY
2019-05-08 17:42:24 +00:00
linux: enable LIRC
2021-06-23 14:41:35 +00:00
LIRC = mkMerge [ (whenOlder "4.16" module) (whenAtLeast "4.17" yes) ];
kernel: enable core scheduling on 5.14+ kernels Core scheduling is a recent innovation in newer kernels to help run certain untrusted compute workloads more safely in the face of vulnerabilities like Spectre. In short: it lets processes assign a unique "cookie" to some group of processes to indicate they are allowed to be scheduled together on the same SMT-capable core. This helps mitigate attacks that rely on observing usage of CPU execution units by cohabitated threads. Some extra details are available via Linux Weekly News: "Core scheduling lands in 5.14", https://lwn.net/Articles/861251/ Signed-off-by: Austin Seipp <aseipp@pobox.com>
2021-11-13 06:29:51 +00:00
SCHED_CORE = whenAtLeast "5.14" yes;
linux: build with support for Multi-Gen LRU
2022-12-09 21:35:36 +00:00
LRU_GEN = whenAtLeast "6.1" yes;
linux: enable Multi-Gen LRU by default
2022-12-09 21:35:51 +00:00
LRU_GEN_ENABLED = whenAtLeast "6.1" yes;
linux: build with support for Multi-Gen LRU
2022-12-09 21:35:36 +00:00
linux: enable FSL_MC_UAPI_SUPPORT
2021-12-17 23:05:49 +00:00
FSL_MC_UAPI_SUPPORT = mkIf (stdenv.hostPlatform.system == "aarch64-linux") (whenAtLeast "5.12" yes);
Merge branch 'staging' into staging-next Conflicts: pkgs/os-specific/linux/kernel/common-config.nix
2021-12-25 22:16:26 +00:00
linux: disable ASHMEM on >= 5.18 Dropped by upstream, see https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=721412ed3d819e767cac2b06646bf03aa158aaec It was marked as optional, so it didn't break our builds but resulted in: warning: unused option: ASHMEM Explicitly disable ASHMEM on kernels >=5.18 for clarity and fewer warnings
2022-07-07 12:24:33 +00:00
ASHMEM = { optional = true; tristate = whenBetween "5.0" "5.18" "y";};
android: with new kernel versions this option is no longer necessary Description: - ANDROID = { optional = true; tristate = whenBetween "5.0" "5.19" "y";}; - starting from 5.20 and 6.0, this parameter no longer exists. Change-Id: I34d6638e01eb539de34afe2152ff5927a317b68a Signed-off-by: Ionut Nechita <ionut_n2001@yahoo.com>
2023-03-29 05:21:45 +00:00
ANDROID = { optional = true; tristate = whenBetween "5.0" "5.19" "y";};
linux: CONFIG_ASHMEM=y, CONFIG_ANDROID=y This enables ashmem, binder so waydroid/anbox works with the provided linux kernel Cherry-picked from https://github.com/NixOS/nixpkgs/pull/102341
2020-11-01 06:08:13 +00:00
ANDROID_BINDER_IPC = { optional = true; tristate = whenAtLeast "5.0" "y";};
ANDROID_BINDERFS = { optional = true; tristate = whenAtLeast "5.0" "y";};
ANDROID_BINDER_DEVICES = { optional = true; freeform = whenAtLeast "5.0" "binder,hwbinder,vndbinder";};
linux: enable TASKSTATS, TASK_XACCT, TASK_DELAY_ACCT and TASK_IO_ACCOUNTING iotop needs TASKSTATS, TASK_DELAY_ACCT, TASK_XACCT and TASK_IO_ACCOUNTING to work. For x86_64, all these options are enabled by upstream[1]. For aarch64, however, only TASK_XACCT and TASK_IO_ACCOUNTING are enabled by upstream[2]. This patch enables all these four options for aarch64, which have been enabled by many other distributions, e.g. debian[3], fedora[4], rhel[5] and gentoo[6]. I tried to only enable TASKSTATS and TASK_DELAY_ACCT since the other two options are enabled by upstream, but it turns out that it's necessary to explicitly enable all four options. I do not figure out the reason though. Additionally, given that debian enables these four options for all arch[3], I think it's safe for us to do the same thing. [1]: https://github.com/torvalds/linux/blob/56e337f2cf1326323844927a04e9dbce9a244835/arch/x86/configs/x86_64_defconfig#L8-L11 [2]: https://github.com/torvalds/linux/blob/56e337f2cf1326323844927a04e9dbce9a244835/arch/arm64/configs/defconfig#L10-L11 [3]: https://salsa.debian.org/kernel-team/linux/-/blob/da6ddc7d8f1a95980d9a1c499fe58066cfe1986b/debian/config/config#L6356-6359 [4]: https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-aarch64-fedora.config#_7398 [5]: https://src.fedoraproject.org/rpms/kernel/blob/rawhide/f/kernel-aarch64-rhel.config#_5885 [6]: https://github.com/gentoo/gentoo/blob/b839fccce25d49df2fcfe5ed184b557796b0d6bd/sys-kernel/gentoo-kernel/gentoo-kernel-5.15.29.ebuild#L27
2022-03-17 09:45:40 +00:00
TASKSTATS = yes;
TASK_DELAY_ACCT = yes;
TASK_XACCT = yes;
TASK_IO_ACCOUNTING = yes;
linux: disable WERROR by default gcc update frequently breaks most recent kernel releases due to blanket -Werror flag. Let's avoid -Werror in a default build to ease kernel and gcc maintenance.
2022-05-30 06:15:14 +00:00
# Fresh toolchains frequently break -Werror build for minor issues.
WERROR = whenAtLeast "5.15" no;
linux: avoid memory hotplug support on most platforms It broke i686 build, and it's probably not worth it on others, too. /cc #54095 e63414078a7.
2019-01-19 18:19:28 +00:00
} // optionalAttrs (stdenv.hostPlatform.system == "x86_64-linux" || stdenv.hostPlatform.system == "aarch64-linux") {
linux/kernel: enable automatic cpu hotplug support
2019-07-22 17:22:16 +00:00
# Enable CPU/memory hotplug support
# Allows you to dynamically add & remove CPUs/memory to a VM client running NixOS without requiring a reboot
ACPI_HOTPLUG_CPU = yes;
Enable memory hotplug support
2019-01-16 19:58:08 +00:00
ACPI_HOTPLUG_MEMORY = yes;
MEMORY_HOTPLUG = yes;
MEMORY_HOTREMOVE = yes;
linux/kernel: enable automatic cpu hotplug support
2019-07-22 17:22:16 +00:00
HOTPLUG_CPU = yes;
Enable memory hotplug support
2019-01-16 19:58:08 +00:00
MIGRATION = yes;
SPARSEMEM = yes;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
# Bump the maximum number of CPUs to support systems like EC2 x1.*
# instances and Xeon Phi.
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
NR_CPUS = freeform "384";
linux/common-config.nix: Configure for armv7l-linux too
2021-05-01 20:59:15 +00:00
} // optionalAttrs (stdenv.hostPlatform.system == "armv7l-linux" || stdenv.hostPlatform.system == "aarch64-linux") {
linux: Enables support for the Allwinner Display Engine 2.0
2019-06-11 01:31:20 +00:00
# Enables support for the Allwinner Display Engine 2.0
linux: cleanup common-config after drop of 4.9 linux-4.9 was dropped on 8d9133c67d25c15348ec12720ee2ce90762d4d4c next lowest version in nixpkgs is 4.14 so cleaning up options
2022-11-19 02:48:31 +00:00
SUN8I_DE2_CCU = yes;
linux: Add CRYPTO_AEGIS128_SIMD for aarch64 See comments: https://github.com/NixOS/nixpkgs/commit/9b67ea9106102d882f53d62890468071900b9647
2019-11-28 14:20:15 +00:00
# See comments on https://github.com/NixOS/nixpkgs/commit/9b67ea9106102d882f53d62890468071900b9647
linux: fix kernel config options Some of the options didn't have correct kernel version constraints, others had been removed or made optional unnecessarily in #84032.
2020-05-26 12:09:20 +00:00
CRYPTO_AEGIS128_SIMD = whenAtLeast "5.4" no;
linux: configure aarch64 contiguous memory allocator via kernel config As per the in-line comment, this is where distros should configure it. Not via kernel command line parameters. As found by looking at the implementation, while exploring the cause of a bug on the Raspberry Pi 4, it was found that `cma=` on the command line parameters will overwrite the values a device tree will have configured for a given platform. With this, the more recent 5.4 vendor kernel boots just fine on the Raspberry Pi 4 using our common configuration.
2020-09-14 03:26:21 +00:00
# Distros should configure the default as a kernel option.
# We previously defined it on the kernel command line as cma=
# The kernel command line will override a platform-specific configuration from its device tree.
# https://github.com/torvalds/linux/blob/856deb866d16e29bd65952e0289066f6078af773/kernel/dma/contiguous.c#L35-L44
CMA_SIZE_MBYTES = freeform "32";
linux/common-config.nix: Enable FB_SIMPLE on AArch64
2021-04-19 20:23:11 +00:00
# Many ARM SBCs hand off a pre-configured framebuffer.
# This always can can be replaced by the actual native driver.
# Keeping it a built-in ensures it will be used if possible.
FB_SIMPLE = yes;
linux/common-config.nix: Configure for armv7l-linux too
2021-05-01 20:59:15 +00:00
kernel: Enable Chrome platform modules They were enabled in 5.4 but then removed. Let's enable them explicitly here. To keep the version constraints simple, we match kernel >5.4 even though some of them are available since 4.x.
2021-07-18 17:52:04 +00:00
} // optionalAttrs (versionAtLeast version "5.4" && (stdenv.hostPlatform.system == "x86_64-linux" || stdenv.hostPlatform.system == "aarch64-linux")) {
# Required for various hardware features on Chrome OS devices
CHROME_PLATFORMS = yes;
CHROMEOS_TBMC = module;
CROS_EC = module;
CROS_EC_I2C = module;
CROS_EC_SPI = module;
CROS_EC_LPC = module;
CROS_EC_ISHTP = module;
CROS_KBD_LED_BACKLIGHT = module;
} // optionalAttrs (versionAtLeast version "5.4" && stdenv.hostPlatform.system == "x86_64-linux") {
CHROMEOS_LAPTOP = module;
CHROMEOS_PSTORE = module;
linux: translate config to structured config Instead of using a string to describe kernel config, use a nix attribute set, then converted to a string. - allows to override the config, aka convert 'yes' into 'modules' or vice-versa - while for now merging different configs is still crude (last spec wins), at least there should be only one CONFIG_XYZ value compared to the current string config where the first defined would be used and others ignored. [initial idea by copumpkin in 2016, a major rebase to 2018 by teto]
2016-01-05 14:22:43 +00:00
};
};
linux: ability to merge structured configs This should make the composability of kernel configurations more straigthforward. - now distinguish freeform options from tristate ones - will look for a structured config in kernelPatches too one can now access the structuredConfig from a kernel via linux_test.configfile.structuredConfig in order to reinject it into another kernel, no need to rewrite the config from scratch The following merge strategies are used in case of conflict: -- freeform items must be equal or they conflict (mergeEqualOption) -- for tristate (y/m/n) entries, I use the mergeAnswer strategy which takes the best available value, "best" being defined by the user (by default "y" > "m" > "n", e.g. if one entry is both marked "y" and "n", "y" wins) -- if one item is both marked optional/mandatory, mandatory wins (mergeFalseByDefault)
2018-10-03 09:49:50 +00:00
in
flattenKConf options
Reference in New Issue Copy Permalink