linux/scripts/Makefile.modinst
Linus Torvalds c538944d8e modules-6.3-rc1
Nothing exciting at all for modules for v6.3. The biggest change is
 just the change of INSTALL_MOD_DIR from "extra" to "updates" which
 I found lingered for ages for no good reason while testing the CXL
 mock driver [0]. The CXL mock driver has no kconfig integration and requires
 building an external module... and re-building the *rest* of the production
 drivers. This mock driver when loaded but not the production ones will
 crash. All this crap can obviously be fixed by integrating kconfig
 semantics into such test module, however that's not desirable by
 the maintainer, and so sensible defaults must be used to ensure a
 default "make modules_install" will suffice for most distros which
 do not have a file like /etc/depmod.d/dist.conf with something like
 `search updates extra built-in`. Since most distros rely on kmod and
 since its inception the "updates" directory is always in the search
 path it makes more sense to use that than the "extra" which only
 *some* RH based systems rely on. All this stuff has been on linux-next
 for a while.
 
 For v6.4 I already have queued some initial work by Song Liu which gets
 us slowly going to a place where we *may* see a generic allocator for
 huge pages for module text to avoid direct map fragmentation *and*
 reduce iTLB pressure. That work is in its initial stages, no allocator
 work is done yet. This is all just prep work. Fortunately Thomas Gleixner
 has helped convince Song that modules *need* to be *requirement* if we
 are going to see any special allocator touch x86. So who knows... maybe
 around v6.5 we'll start seeing some *real* performance numbers of the
 effect of using huge pages for something other than eBPF toys.
 
 For v6.4 also, you may start seeing patches from Nick Alcock on different
 trees and modules-next which aims at extending kallsyms *eventually* to provide
 clearer address to symbol lookups. The claim is that this is a *great* *feature*
 tracing tools are dying to have so they can for instance disambiguate symbols as
 coming from modules or from other parts of the kernel. I'm still waiting to see
 proper too usage of such stuff, but *how* we lay this out is still being ironed
 out. Part of the initial work I've been pushing for is to help upkeep our
 modules build optimizations, so being mindful about the work by Masahiro Yamada
 on commit 8b41fc4454 ("kbuild: create modules.builtin without
 Makefile.modbuiltin or tristate.conf") which helps avoid traversing the build
 tree twice. After this commit we now rely on the MODULE_LICENSE() tag to
 determine in a *faster* way if something being built could be a module and
 we dump this into the modules.builtin so that modprobe can simply succeed
 if a module is known to already be built-in. The cleanup work on MODULE_LICENSE()
 simply stems to assist false positives from userspace for things as built-in
 when they *cannot ever* be modules as we don't even tristate the code as
 modular. This work also helps with the SPDX effort as some code is not clearly
 identified with a tag. In the *future*, once all *possible* modules are
 confirmed to have a respective SPDX tag, we *may* just be able to replace the
 MODULE_LICENSE() to instead be generated automatically through inference of
 the respective module SPDX tags.
 
 [0] https://lkml.kernel.org/r/20221209062919.1096779-1-mcgrof@kernel.org
 -----BEGIN PGP SIGNATURE-----
 
 iQJGBAABCgAwFiEENnNq2KuOejlQLZofziMdCjCSiKcFAmP1R4cSHG1jZ3JvZkBr
 ZXJuZWwub3JnAAoJEM4jHQowkoinwKoP/1xzihp1mdV+XKxUkeOiVmnX06EXGjQw
 oT6zUtkxQar9ud2sM+2Qmb2/uTXr+8dWwdcv+oq5WFqMzwFdjlxK4pAPzC+0M8dj
 yS0mlslIQbRY0H3DSHJdZ/gBCUU/XNWNwnWpiMr3jZiAlBznIZMW9ZUMUAsHJd7L
 x2x0aOi6E8Xd9VMQ4+4AraI/fTmoTUUZvie9is38IzjKsgFAMpCN2PoN+8T61iBy
 yn6tGfR7Vn/oh2XY/877yq6OAvYAtOsfCIRiuCb483XdjdE8/SW4/o3BZsEfYsVm
 aMwoc309rfljeTOqz6EnnaJivWPpibVZ3nFq6Rz5a5vrOT6hEvDSazD194UpuHw/
 P5m2AoqHGf35lgDdzQgSpNskK+B32PVh397EJ97mfuHwfHICU8QHoBiWPSGoYftF
 oDS0TJ4fU1lCRQ+p+Fda3n9HhvpnZ6ohYaAi1rMuBqyEr8gxfThsOcWDqT6LCSQo
 DeU8x80XK1VW5ReaEUuVATzUBPzyEsthh3gQcBGoqGElRKmlUDlnjVlhc7NtSvmS
 +ohznxktp7Vsp6lnPtuJGOMpq3B9oaaEz0AqVpdZlsvEHST711OsRQp6b6Lz3bbr
 pDPQw+9uv+YxHrUiOcxcsYEkbMKgz0WNReKPjs5TZW5e6w9fLdETaY3TJRMdVAbz
 G6yl1QdrmGeY
 =aeTV
 -----END PGP SIGNATURE-----

Merge tag 'modules-6.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux

Pull modules updates from Luis Chamberlain:
 "Nothing exciting at all for modules for v6.3.

  The biggest change is just the change of INSTALL_MOD_DIR from "extra"
  to "updates" which I found lingered for ages for no good reason while
  testing the CXL mock driver [0].

  The CXL mock driver has no kconfig integration and requires building
  an external module... and re-building the *rest* of the production
  drivers. This mock driver when loaded but not the production ones will
  crash.

  All this can obviously be fixed by integrating kconfig semantics into
  such test module, however that's not desirable by the maintainer, and
  so sensible defaults must be used to ensure a default "make
  modules_install" will suffice for most distros which do not have a
  file like /etc/depmod.d/dist.conf with something like `search updates
  extra built-in`.

  Since most distros rely on kmod and since its inception the "updates"
  directory is always in the search path it makes more sense to use that
  than the "extra" which only *some* RH based systems rely on.

  All this stuff has been on linux-next for a while"

[0] https://lkml.kernel.org/r/20221209062919.1096779-1-mcgrof@kernel.org

* tag 'modules-6.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux:
  Documentation: livepatch: module-elf-format: Remove local klp_modinfo definition
  module.h: Document klp_modinfo struct using kdoc
  module: Use kstrtobool() instead of strtobool()
  kernel/params.c: Use kstrtobool() instead of strtobool()
  test_kmod: stop kernel-doc warnings
  kbuild: Modify default INSTALL_MOD_DIR from extra to updates
2023-02-23 14:05:08 -08:00

119 lines
2.6 KiB
Makefile

# SPDX-License-Identifier: GPL-2.0
# ==========================================================================
# Installing modules
# ==========================================================================
PHONY := __modinst
__modinst:
include include/config/auto.conf
include $(srctree)/scripts/Kbuild.include
modules := $(call read-file, $(MODORDER))
ifeq ($(KBUILD_EXTMOD),)
dst := $(MODLIB)/kernel
else
INSTALL_MOD_DIR ?= updates
dst := $(MODLIB)/$(INSTALL_MOD_DIR)
endif
$(foreach x, % :, $(if $(findstring $x, $(dst)), \
$(error module installation path cannot contain '$x')))
suffix-y :=
suffix-$(CONFIG_MODULE_COMPRESS_GZIP) := .gz
suffix-$(CONFIG_MODULE_COMPRESS_XZ) := .xz
suffix-$(CONFIG_MODULE_COMPRESS_ZSTD) := .zst
modules := $(patsubst $(extmod_prefix)%.o, $(dst)/%.ko$(suffix-y), $(modules))
__modinst: $(modules)
@:
#
# Installation
#
quiet_cmd_install = INSTALL $@
cmd_install = mkdir -p $(dir $@); cp $< $@
# Strip
#
# INSTALL_MOD_STRIP, if defined, will cause modules to be stripped after they
# are installed. If INSTALL_MOD_STRIP is '1', then the default option
# --strip-debug will be used. Otherwise, INSTALL_MOD_STRIP value will be used
# as the options to the strip command.
ifdef INSTALL_MOD_STRIP
ifeq ($(INSTALL_MOD_STRIP),1)
strip-option := --strip-debug
else
strip-option := $(INSTALL_MOD_STRIP)
endif
quiet_cmd_strip = STRIP $@
cmd_strip = $(STRIP) $(strip-option) $@
else
quiet_cmd_strip =
cmd_strip = :
endif
#
# Signing
# Don't stop modules_install even if we can't sign external modules.
#
ifeq ($(CONFIG_MODULE_SIG_ALL),y)
ifeq ($(filter pkcs11:%, $(CONFIG_MODULE_SIG_KEY)),)
sig-key := $(if $(wildcard $(CONFIG_MODULE_SIG_KEY)),,$(srctree)/)$(CONFIG_MODULE_SIG_KEY)
else
sig-key := $(CONFIG_MODULE_SIG_KEY)
endif
quiet_cmd_sign = SIGN $@
cmd_sign = scripts/sign-file $(CONFIG_MODULE_SIG_HASH) "$(sig-key)" certs/signing_key.x509 $@ \
$(if $(KBUILD_EXTMOD),|| true)
else
quiet_cmd_sign :=
cmd_sign := :
endif
ifeq ($(modules_sign_only),)
$(dst)/%.ko: $(extmod_prefix)%.ko FORCE
$(call cmd,install)
$(call cmd,strip)
$(call cmd,sign)
else
$(dst)/%.ko: FORCE
$(call cmd,sign)
endif
#
# Compression
#
quiet_cmd_gzip = GZIP $@
cmd_gzip = $(KGZIP) -n -f $<
quiet_cmd_xz = XZ $@
cmd_xz = $(XZ) --lzma2=dict=2MiB -f $<
quiet_cmd_zstd = ZSTD $@
cmd_zstd = $(ZSTD) -T0 --rm -f -q $<
$(dst)/%.ko.gz: $(dst)/%.ko FORCE
$(call cmd,gzip)
$(dst)/%.ko.xz: $(dst)/%.ko FORCE
$(call cmd,xz)
$(dst)/%.ko.zst: $(dst)/%.ko FORCE
$(call cmd,zstd)
PHONY += FORCE
FORCE:
.PHONY: $(PHONY)