When CONFIG_LTO_CLANG=y, additional intermediate *.prelink.o is created
for each module. Also, objtool is postponed until LLVM IR is converted
to ELF.
CONFIG_X86_KERNEL_IBT works in a similar way to postpone objtool until
objects are merged together.
This commit stops generating *.prelink.o, so the build flow will look
similar with/without LTO.
The following figures show how the LTO build currently works, and
how this commit is changing it.
Current build flow
==================
[1] single-object module
$(LD)
$(CC) +objtool $(LD)
foo.c --------------------> foo.o -----> foo.prelink.o -----> foo.ko
(LLVM IR) (ELF) | (ELF)
|
foo.mod.o --/
(LLVM IR)
[2] multi-object module
$(LD)
$(CC) $(AR) +objtool $(LD)
foo1.c -----> foo1.o -----> foo.o -----> foo.prelink.o -----> foo.ko
| (archive) (ELF) | (ELF)
foo2.c -----> foo2.o --/ |
(LLVM IR) foo.mod.o --/
(LLVM IR)
One confusion is that foo.o in multi-object module is an archive
despite of its suffix.
New build flow
==============
[1] single-object module
Since there is only one object, there is no need to keep the LLVM IR.
Use $(CC)+$(LD) to generate an ELF object in one build rule. When LTO
is disabled, $(LD) is unneeded because $(CC) produces an ELF object.
$(CC)+$(LD)+objtool $(LD)
foo.c ----------------------------> foo.o ---------> foo.ko
(ELF) | (ELF)
|
foo.mod.o --/
(LLVM IR)
[2] multi-object module
Previously, $(AR) was used to combine LLVM IR files into an archive,
but there was no technical reason to do so. Use $(LD) to merge them
into a single ELF object.
$(LD)
$(CC) +objtool $(LD)
foo1.c ---------> foo1.o ---------> foo.o ---------> foo.ko
| (ELF) | (ELF)
foo2.c ---------> foo2.o ----/ |
(LLVM IR) foo.mod.o --/
(LLVM IR)
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Similar cleanup to commit 5c8166419a ("kbuild: replace $(if A,A,B)
with $(or A,B)").
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
When building an external module, if users don't need to separate the
compilation output and source code, they run the following command:
"make -C $(LINUX_SRC_DIR) M=$(PWD)". At this point, "$(KBUILD_EXTMOD)"
and "$(src)" are the same.
If they need to separate them, they run "make -C $(KERNEL_SRC_DIR)
O=$(KERNEL_OUT_DIR) M=$(OUT_DIR) src=$(PWD)". Before running the
command, they need to copy "Kbuild" or "Makefile" to "$(OUT_DIR)" to
prevent compilation failure.
So the kernel should change the included path to avoid the copy operation.
Signed-off-by: Jing Leng <jleng@ambarella.com>
[masahiro: I do not think "M=$(OUT_DIR) src=$(PWD)" is the official way,
but this patch is a nice clean up anyway.]
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Change comment "create one <module>.mod.c file pr. module"
to "create one <module>.mod.c file per module"
Signed-off-by: Ramji Jiyani <ramjiyani@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
With CONFIG_LTO_CLANG, we currently link modules into native
code just before modpost, which means with TRIM_UNUSED_KSYMS
enabled, we still look at the LLVM bitcode in the .o files when
generating the list of used symbols. As the bitcode doesn't
yet have calls to compiler intrinsics and llvm-nm doesn't see
function references that only exist in function-level inline
assembly, we currently need a whitelist for TRIM_UNUSED_KSYMS to
work with LTO.
This change moves module LTO linking to happen earlier, and
thus avoids the issue with LLVM bitcode and TRIM_UNUSED_KSYMS
entirely, allowing us to also drop the whitelist from
gen_autoksyms.sh.
Link: https://github.com/ClangBuiltLinux/linux/issues/1369
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Alexander Lobakin <alobakin@pm.me>
Tested-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Nathan reports that the mips defconfig emits the following warning:
WARNING: modpost: Symbol info of vmlinux is missing. Unresolved symbol check will be entirely skipped.
This false-positive happens when CONFIG_TRIM_UNUSED_KSYMS is enabled,
but no CONFIG option is set to 'm'.
Commit a0590473c5 ("nfs: fix PNFS_FLEXFILE_LAYOUT Kconfig default")
turned the last 'm' into 'y' for the mips defconfig, and uncovered
this issue.
In this case, the module feature itself is enabled, but we have no
module to build. As a result, CONFIG_TRIM_UNUSED_KSYMS drops all the
instances of EXPORT_SYMBOL. Then, modpost wrongly assumes vmlinux is
missing because vmlinux.symvers is empty. (As another false-positive
case, you can create a module that does not use any symbol of vmlinux).
The current behavior is to entirely suppress the unresolved symbol
warnings when vmlinux is missing just because there are too many.
I found the origin of this code in the historical git tree. [1]
If this is a matter of noisiness, I think modpost can display the
first 10 warnings, and the number of suppressed warnings at the end.
You will get a bit noisier logs when you run 'make modules' without
vmlinux, but such warnings are better to show because you never know
the resulting modules are actually loadable or not.
This commit changes the following:
- If any of input *.symver files is missing, pass -w option to let
the module build keep going with warnings instead of errors.
- If there are too many (10+) unresolved symbol warnings, show only
the first 10, and also the number of suppressed warnings.
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=1cc0e0529569bf6a94f6d49770aa6d4b599d2c46
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The -w option is meaningless for the first pass of modpost (vmlinux.o).
We know there are unresolved symbols in vmlinux.o, hence we skip
check_exports() and other checks when mod->is_vmlinux is set.
See the following part in the for-loop.
if (mod->is_vmlinux || mod->from_dump)
continue;
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The external module build shows the following warning if Module.symvers
is missing in the kernel tree.
WARNING: Symbol version dump "Module.symvers" is missing.
Modules may not have dependencies or modversions.
I think this is an important heads-up because the resulting modules may
not work as expected. This happens when you did not build the entire
kernel tree, for example, you might have prepared the minimal setups
for external modules by 'make defconfig && make modules_preapre'.
A problem is that 'make modules' creates Module.symvers even without
vmlinux. In this case, that warning is suppressed since Module.symvers
already exists in spite of its incomplete content.
The incomplete (i.e. invalid) Module.symvers should not be created.
This commit changes the second pass of modpost to dump symbols into
modules-only.symvers. The final Module.symvers is created by
concatenating vmlinux.symvers and modules-only.symvers if both exist.
Module.symvers is supposed to collect symbols from both vmlinux and
modules. It might be a bit confusing, and I am not quite sure if it
is an official interface, but presumably it is difficult to rename it
because some tools (e.g. kmod) parse it.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
VPATH is used in Kbuild to make pattern rules search for prerequisites
in both $(objtree) and $(srctree). Some of *.c, *.S files are not real
sources, but generated by tools such as flex, bison, perl.
In contrast, I doubt the benefit of --include-dir=$(abs_srctree) because
it is always clear which Makefiles are real sources, and which are not.
So, my hope is to add $(srctree)/ prefix to all check-in Makefiles,
then remove --include-dir=$(abs_srctree) flag in the future.
I am touching only some Kbuild core parts for now. Treewide fixes will
be needed to achieve this goal.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
With CONFIG_MODVERSIONS, version information is linked into each
compilation unit that exports symbols. With LTO, we cannot use this
method as all C code is compiled into LLVM bitcode instead. This
change collects symbol versions into .symversions files and merges
them in link-vmlinux.sh where they are all linked into vmlinux.o at
the same time.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20201211184633.3213045-4-samitolvanen@google.com
This change adds build system support for Clang's Link Time
Optimization (LTO). With -flto, instead of ELF object files, Clang
produces LLVM bitcode, which is compiled into native code at link
time, allowing the final binary to be optimized globally. For more
details, see:
https://llvm.org/docs/LinkTimeOptimization.html
The Kconfig option CONFIG_LTO_CLANG is implemented as a choice,
which defaults to LTO being disabled. To use LTO, the architecture
must select ARCH_SUPPORTS_LTO_CLANG and support:
- compiling with Clang,
- compiling all assembly code with Clang's integrated assembler,
- and linking with LLD.
While using CONFIG_LTO_CLANG_FULL results in the best runtime
performance, the compilation is not scalable in time or
memory. CONFIG_LTO_CLANG_THIN enables ThinLTO, which allows
parallel optimization and faster incremental builds. ThinLTO is
used by default if the architecture also selects
ARCH_SUPPORTS_LTO_CLANG_THIN:
https://clang.llvm.org/docs/ThinLTO.html
To enable LTO, LLVM tools must be used to handle bitcode files, by
passing LLVM=1 and LLVM_IAS=1 options to make:
$ make LLVM=1 LLVM_IAS=1 defconfig
$ scripts/config -e LTO_CLANG_THIN
$ make LLVM=1 LLVM_IAS=1
To prepare for LTO support with other compilers, common parts are
gated behind the CONFIG_LTO option, and LTO can be disabled for
specific files by filtering out CC_FLAGS_LTO.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20201211184633.3213045-3-samitolvanen@google.com
The same code exists a few lines above.
Fixes: 436b2ac603 ("modpost: invoke modpost only when input files are updated")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The -s option was added by commit 8d8d8289df ("kbuild: do not do
section mismatch checks on vmlinux in 2nd pass").
Now that the second pass does not parse vmlinux, this option is
unneeded.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
If modpost fails to load a symbol dump file, it cannot check unresolved
symbols, hence module dependency will not be added. Nor CRCs can be added.
Currently, external module builds check only $(objtree)/Module.symvers,
but it should check files specified by KBUILD_EXTRA_SYMBOLS as well.
Move the warning message from the top Makefile to scripts/Makefile.modpost
and print the warning if any dump file is missing.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Currently, the second pass of modpost is always invoked when you run
'make' or 'make modules' even if none of modules is changed.
Use if_changed to invoke it only when it is necessary.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The full build runs modpost twice, first for vmlinux.o and second for
modules.
The first pass dumps all the vmlinux symbols into Module.symvers, but
the second pass parses vmlinux again instead of reusing the dump file,
presumably because it needs to avoid accumulating stale symbols.
Loading symbol info from a dump file is faster than parsing an ELF object.
Besides, modpost deals with various issues to parse vmlinux in the second
pass.
A solution is to make the first pass dumps symbols into a separate file,
vmlinux.symvers. The second pass reads it, and parses module .o files.
The merged symbol information is dumped into Module.symvers in the same
way as before.
This makes further modpost cleanups possible.
Also, it fixes the problem of 'make vmlinux', which previously overwrote
Module.symvers, throwing away module symbols.
I slightly touched scripts/link-vmlinux.sh so that vmlinux is re-linked
when you cross this commit. Otherwise, vmlinux.symvers would not be
generated.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Previously, the -i option had two functions; load a symbol dump file,
and set the external_module flag.
I want to assign a dedicate option for each of them.
Going forward, the -i is used to load a symbol dump file, and the -e
to set the external_module flag.
With this, we will be able to use -i for loading in-kernel symbols.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Now that there is no difference between -i and -e, they can be unified.
Make modpost accept the -i option multiple times, then remove -e.
I will reuse -e for a different purpose.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The built-in only code is not required to have MODULE_IMPORT_NS() to
use symbols. So, the namespace is not checked for vmlinux(.o).
Do not pass the meaningless -N option to the first pass of modpost.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The '-T -' option reads the file list from stdin.
It is clearer to put it close to the piped command.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
$(filter -i,$(MAKEFLAGS)) works only in limited use-cases.
The representation of $(MAKEFLAGS) depends on various factors:
- GNU Make version (version 3.8x or version 4.x)
- The presence of other flags like -j
In my experiments, $(MAKEFLAGS) is expanded as follows:
* GNU Make 3.8x:
* without -j option:
--no-print-directory -Rri
* with -j option:
--no-print-directory -Rr --jobserver-fds=3,4 -j -i
* GNU Make 4.x:
* without -j option:
irR --no-print-directory
* with -j option:
irR -j --jobserver-fds=3,4 --no-print-directory
For GNU Make 4.x, the flags are grouped as 'irR', which does not work.
For the single thread build with GNU Make 3.8x, the flags are grouped
as '-Rri', which does not work either.
To make it work for all cases, do likewise as commit 6f0fa58e45
("kbuild: simplify silent build (-s) detection").
BTW, since commit ff9b45c55b ("kbuild: modpost: read modules.order
instead of $(MODVERDIR)/*.mod"), you also need to pass -k option to
build final *.ko files. 'make -i -k' ignores compile errors in modules,
and build as many remaining *.ko as possible.
Please note this feature is kind of dangerous if other modules depend
on the broken module because the generated modules will lack the correct
module dependency or CRC. Honestly, I am not a big fan of it, but I am
keeping this feature.
Fixes: eed380f3f5 ("modpost: Optionally ignore secondary errors seen if a single module build fails")
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
$(firstword ...) in scripts/Makefile.modpost was added by commit
3f3fd3c055 ("[PATCH] kbuild: allow multi-word $M in Makefile.modpost")
to build multiple external module directories.
It was a solution to resolve symbol dependencies when an external
module depends on another external module.
Commit 0d96fb20b7 ("kbuild: Add new Kbuild variable
KBUILD_EXTRA_SYMBOLS") introduced another solution by passing symbol
info via KBUILD_EXTRA_SYMBOLS, then broke the multi-word M= support.
include $(if $(wildcard $(KBUILD_EXTMOD)/Kbuild), \
$(KBUILD_EXTMOD)/Kbuild, $(KBUILD_EXTMOD)/Makefile)
... does not work if KBUILD_EXTMOD contains multiple words.
This feature has been broken for more than a decade. Remove the
bitrotten code, and stop parsing if M or KBUILD_EXTMOD contains
multiple words.
As Documentation/kbuild/modules.rst explains, if your module depends
on another one, there are two solutions:
- add a common top-level Kbuild file
- use KBUILD_EXTRA_SYMBOLS
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Currently when CONFIG_MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS=n, modpost
only warns when a module is missing namespace imports. Under this
configuration, such a module cannot be loaded into the kernel anyway, as
the module loader would reject it. We might as well return a build
error when a module is missing namespace imports under
CONFIG_MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS=n, so that the build
warning does not go ignored/unnoticed.
Signed-off-by: Jessica Yu <jeyu@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
scripts/nsdeps is written to take care of only in-tree modules.
Perhaps, this is not a bug, but just a design. At least,
Documentation/core-api/symbol-namespaces.rst focuses on in-tree modules.
Having said that, some people already tried nsdeps for external modules.
So, it would be nice to support it.
Reported-by: Steve French <smfrench@gmail.com>
Reported-by: Jessica Yu <jeyu@kernel.org>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Tested-by: Jessica Yu <jeyu@kernel.org>
Acked-by: Jessica Yu <jeyu@kernel.org>
Reviewed-by: Matthias Maennich <maennich@google.com>
Tested-by: Matthias Maennich <maennich@google.com>
The modpost, with the -d option given, generates per-module .ns_deps
files.
Kbuild generates per-module .mod files to carry module information.
This is convenient because Make handles multiple jobs in parallel
when the -j option is given.
On the other hand, the modpost always runs as a single thread.
I do not see a strong reason to produce separate .ns_deps files.
This commit changes the modpost to generate just one file,
modules.nsdeps, each line of which has the following format:
<module_name>: <list of missing namespaces>
Please note it contains *missing* namespaces instead of required ones.
So, modules.nsdeps is empty if the namespace dependency is all good.
This will work more efficiently because spatch will no longer process
already imported namespaces. I removed the '(if needed)' from the
nsdeps log since spatch is invoked only when needed.
This also solves the stale .ns_deps problem reported by Jessica Yu:
https://lkml.org/lkml/2019/10/28/467
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Tested-by: Jessica Yu <jeyu@kernel.org>
Acked-by: Jessica Yu <jeyu@kernel.org>
Reviewed-by: Matthias Maennich <maennich@google.com>
Tested-by: Matthias Maennich <maennich@google.com>
'make nsdeps' invokes the modpost three times at most; before linking
vmlinux, before building modules, and finally for generating .ns_deps
files. Running the modpost again and again is not efficient.
The last two can be unified. When the -d option is given, the modpost
still does the usual job, and in addition, generates .ns_deps files.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Tested-by: Matthias Maennich <maennich@google.com>
Reviewed-by: Matthias Maennich <maennich@google.com>
Since commit 040fcc819a ("kbuild: improved modversioning support for
external modules"), the external module build reads Module.symvers in
the directory of the module itself, then dumps symbols back into it.
It accumulates stale symbols in the file when you build an external
module incrementally.
The idea behind it was, as the commit log explained, you can copy
Modules.symvers from one module to another when you need to pass symbol
information between two modules. However, the manual copy of the file
sounds questionable to me, and containing stale symbols is a downside.
Some time later, commit 0d96fb20b7 ("kbuild: Add new Kbuild variable
KBUILD_EXTRA_SYMBOLS") introduced a saner approach.
So, this commit removes the former one. Going forward, the external
module build dumps symbols into Module.symvers to be carried via
KBUILD_EXTRA_SYMBOLS, but never reads it automatically.
With the -I option removed, there is no one to set the external_module
flag unless KBUILD_EXTRA_SYMBOLS is passed. Now the -i option does it
instead.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
When building external modules, $(objtree)/Module.symvers is scanned
for symbol information of vmlinux and in-tree modules.
Additionally, vmlinux is parsed if it exists in $(objtree)/.
This is totally redundant since all the necessary information is
contained in $(objtree)/Module.symvers.
Do not parse vmlinux at all for external module builds. This makes
sense because vmlinux is deleted by 'make clean'.
'make clean' leaves all the build artifacts for building external
modules. vmlinux is unneeded for that.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
The comment line "When building external modules ..." explains
the same thing as "Include the module's Makefile ..." a few lines
below.
The comment "they may be used when building the .mod.c file" is no
longer true; .mod.c file is compiled in scripts/Makefile.modfinal
since commit 9b9a3f20cb ("kbuild: split final module linking out
into Makefile.modfinal"). I still keep the code in case $(obj) or
$(src) is used in the external module Makefile.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Summary of modules changes for the 5.4 merge window:
- Introduce exported symbol namespaces.
This new feature allows subsystem maintainers to partition and
categorize their exported symbols into explicit namespaces. Module
authors are now required to import the namespaces they need.
Some of the main motivations of this feature include: allowing kernel
developers to better manage the export surface, allow subsystem
maintainers to explicitly state that usage of some exported symbols
should only be limited to certain users (think: inter-module or
inter-driver symbols, debugging symbols, etc), as well as more easily
limiting the availability of namespaced symbols to other parts of the
kernel. With the module import requirement, it is also easier to spot
the misuse of exported symbols during patch review. Two new macros are
introduced: EXPORT_SYMBOL_NS() and EXPORT_SYMBOL_NS_GPL(). The API is
thoroughly documented in Documentation/kbuild/namespaces.rst.
- Some small code and kbuild cleanups here and there.
-----BEGIN PGP SIGNATURE-----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=72Uy
-----END PGP SIGNATURE-----
Merge tag 'modules-for-v5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux
Pull modules updates from Jessica Yu:
"The main bulk of this pull request introduces a new exported symbol
namespaces feature. The number of exported symbols is increasingly
growing with each release (we're at about 31k exports as of 5.3-rc7)
and we currently have no way of visualizing how these symbols are
"clustered" or making sense of this huge export surface.
Namespacing exported symbols allows kernel developers to more
explicitly partition and categorize exported symbols, as well as more
easily limiting the availability of namespaced symbols to other parts
of the kernel. For starters, we have introduced the USB_STORAGE
namespace to demonstrate the API's usage. I have briefly summarized
the feature and its main motivations in the tag below.
Summary:
- Introduce exported symbol namespaces.
This new feature allows subsystem maintainers to partition and
categorize their exported symbols into explicit namespaces. Module
authors are now required to import the namespaces they need.
Some of the main motivations of this feature include: allowing
kernel developers to better manage the export surface, allow
subsystem maintainers to explicitly state that usage of some
exported symbols should only be limited to certain users (think:
inter-module or inter-driver symbols, debugging symbols, etc), as
well as more easily limiting the availability of namespaced symbols
to other parts of the kernel.
With the module import requirement, it is also easier to spot the
misuse of exported symbols during patch review.
Two new macros are introduced: EXPORT_SYMBOL_NS() and
EXPORT_SYMBOL_NS_GPL(). The API is thoroughly documented in
Documentation/kbuild/namespaces.rst.
- Some small code and kbuild cleanups here and there"
* tag 'modules-for-v5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/jeyu/linux:
module: Remove leftover '#undef' from export header
module: remove unneeded casts in cmp_name()
module: move CONFIG_UNUSED_SYMBOLS to the sub-menu of MODULES
module: remove redundant 'depends on MODULES'
module: Fix link failure due to invalid relocation on namespace offset
usb-storage: export symbols in USB_STORAGE namespace
usb-storage: remove single-use define for debugging
docs: Add documentation for Symbol Namespaces
scripts: Coccinelle script for namespace dependencies.
modpost: add support for generating namespace dependencies
export: allow definition default namespaces in Makefiles or sources
module: add config option MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
modpost: add support for symbol namespaces
module: add support for symbol namespaces.
export: explicitly align struct kernel_symbol
module: support reading multiple values per modinfo tag
A script that uses the '<module>.ns_deps' files generated by modpost to
automatically add the required symbol namespace dependencies to each
module.
Usage:
1) Move some symbols to a namespace with EXPORT_SYMBOL_NS() or define
DEFAULT_SYMBOL_NAMESPACE
2) Run 'make' (or 'make modules') and get warnings about modules not
importing that namespace.
3) Run 'make nsdeps' to automatically add required import statements
to said modules.
This makes it easer for subsystem maintainers to introduce and maintain
symbol namespaces into their codebase.
Co-developed-by: Martijn Coenen <maco@android.com>
Signed-off-by: Martijn Coenen <maco@android.com>
Acked-by: Julia Lawall <julia.lawall@lip6.fr>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Matthias Maennich <maennich@google.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
I think splitting the modpost and linking modules into separate
Makefiles will be useful especially when more complex build steps
come in. The main motivation of this commit is to integrate the
proposed klp-convert feature cleanly.
I moved the logging 'Building modules, stage 2.' to Makefile.modpost
to avoid the code duplication although I do not know whether or not
this message is needed in the first place.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Currently, the timestamp of module linker scripts are not checked.
Add them to the dependency of modules so they are correctly rebuilt.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Update the build scripts and the version magic to reflect when
CONFIG_PREEMPT_RT is enabled in the same way as CONFIG_PREEMPT is treated.
The resulting version strings:
Linux m 5.3.0-rc1+ #100 SMP Fri Jul 26 ...
Linux m 5.3.0-rc1+ #101 SMP PREEMPT Fri Jul 26 ...
Linux m 5.3.0-rc1+ #102 SMP PREEMPT_RT Fri Jul 26 ...
The module vermagic:
5.3.0-rc1+ SMP mod_unload modversions
5.3.0-rc1+ SMP preempt mod_unload modversions
5.3.0-rc1+ SMP preempt_rt mod_unload modversions
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
I removed the single target %.ko in commit ff9b45c55b ("kbuild:
modpost: read modules.order instead of $(MODVERDIR)/*.mod") because
the modpost stage does not work reliably. For instance, the module
dependency, modversion, etc. do not work if we lack symbol information
from the other modules.
Yet, some people still want to build only one module in their interest,
and it may be still useful if it is used within those limitations.
Fixes: ff9b45c55b ("kbuild: modpost: read modules.order instead of $(MODVERDIR)/*.mod")
Reported-by: Don Brace <don.brace@microsemi.com>
Reported-by: Arend Van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Since commit ff9b45c55b ("kbuild: modpost: read modules.order instead
of $(MODVERDIR)/*.mod"), 'make vmlinux' emits a warning, like this:
$ make defconfig vmlinux
[ snip ]
LD vmlinux.o
cat: modules.order: No such file or directory
MODPOST vmlinux.o
MODINFO modules.builtin.modinfo
KSYM .tmp_kallsyms1.o
KSYM .tmp_kallsyms2.o
LD vmlinux
SORTEX vmlinux
SYSMAP System.map
When building only vmlinux, KBUILD_MODULES is not set. Hence, the
modules.order is not generated. For the vmlinux modpost, it is not
necessary at all.
Separate scripts/Makefile.modpost for the vmlinux/modules stages.
This works more efficiently because the vmlinux modpost does not
need to include .*.cmd files.
Fixes: ff9b45c55b ("kbuild: modpost: read modules.order instead of $(MODVERDIR)/*.mod")
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
__modpost is a phony target. The dependency on FORCE is pointless.
All the objects have been built in the previous stage, so the
dependency on the objects are not necessary either.
Count the number of modules in a more straightforward way.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
KBUILD_EXTRA_SYMBOLS makes sense only when building external modules.
Moreover, the modpost sets 'external_module' if the -e option is given.
I replaced $(patsubst %, -e %,...) with simpler $(addprefix -e,...)
while I was here.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
If a build rule fails, the .DELETE_ON_ERROR special target removes the
target, but does nothing for the .*.cmd file, which might be corrupted.
So, .*.cmd files should be included only when the corresponding targets
exist.
Commit 392885ee82 ("kbuild: let fixdep directly write to .*.cmd
files") missed to fix up this file.
Fixes: 392885ee82 ("kbuild: let fixdep directly write to .*.cmd")
Cc: <stable@vger.kernel.org> # v5.0+
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
While descending directories, Kbuild produces objects for modules,
but do not link final *.ko files; it is done in the modpost.
To keep track of modules, Kbuild creates a *.mod file in $(MODVERDIR)
for every module it is building. Some post-processing steps read the
necessary information from *.mod files. This avoids descending into
directories again. This mechanism was introduced in 2003 or so.
Later, commit 551559e13a ("kbuild: implement modules.order") added
modules.order. So, we can simply read it out to know all the modules
with directory paths. This is easier than parsing the first line of
*.mod files.
$(MODVERDIR) has a flat directory structure, that is, *.mod files
are named only with base names. This is based on the assumption that
the module name is unique across the tree. This assumption is really
fragile.
Stephen Rothwell reported a race condition caused by a module name
conflict:
https://lkml.org/lkml/2019/5/13/991
In parallel building, two different threads could write to the same
$(MODVERDIR)/*.mod simultaneously.
Non-unique module names are the source of all kind of troubles, hence
commit 3a48a91901 ("kbuild: check uniqueness of module names")
introduced a new checker script.
However, it is still fragile in the build system point of view because
this race happens before scripts/modules-check.sh is invoked. If it
happens again, the modpost will emit unclear error messages.
To fix this issue completely, create *.mod with full directory path
so that two threads never attempt to write to the same file.
$(MODVERDIR) is no longer needed.
Since modules with directory paths are listed in modules.order, Kbuild
is still able to find *.mod files without additional descending.
I also killed cmd_secanalysis; scripts/mod/sumversion.c computes MD4 hash
for modules with MODULE_VERSION(). When CONFIG_DEBUG_SECTION_MISMATCH=y,
it occurs not only in the modpost stage, but also during directory
descending, where sumversion.c may parse stale *.mod files. It would emit
'No such file or directory' warning when an object consisting a module is
renamed, or when a single-obj module is turned into a multi-obj module or
vice versa.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Towards the goal of removing MODVERDIR, read out modules.order to get
the list of modules to be processed. This is simpler than parsing *.mod
files in $(MODVERDIR).
For external modules, $(KBUILD_EXTMOD)/modules.order should be read.
I removed the single target %.ko from the top Makefile. To make sure
modpost works correctly, vmlinux and the other modules must be built.
You cannot build a particular .ko file alone.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Commit ea837f1c05 ("kbuild: make modpost processing configurable")
was intended to give KBUILD_MODPOST_WARN flexibility to be configurable.
Right now KBUILD_MODPOST_WARN gets just ignored when KBUILD_EXTMOD is
set which happens per default when building modules out of the tree.
This change gives the opportunity to define module build behaving also
in case of out of tree builds and default will become exit on error.
Errors which can be detected by the build should be trapped out of the box
there, unless somebody wants to notice broken stuff later at runtime.
As this patch changes the default behaving from warning to error,
users can consider to fix it for external module builds by:
- providing module symbol table via KBUILD_EXTRA_SYMBOLS for
modules which are dependent
- OR getting old behaving back by passing KBUILD_MODPOST_WARN to the build
Signed-off-by: Wladislav Wiebe <wladislav.wiebe@nokia.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Unless CONFIG_DEBUG_SECTION_MISMATCH is enabled, modpost only shows
the number of section mismatches.
If you want to know the symbols causing the issue, you need to rebuild
with CONFIG_DEBUG_SECTION_MISMATCH. It is tedious.
I think it is fine to show annoying warning when a new section mismatch
comes in.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
In Kbuild, if_changed and friends must have FORCE as a prerequisite.
Hence, $(filter-out FORCE,$^) or $(filter-out $(PHONY),$^) is a common
idiom to get the names of all the prerequisites except phony targets.
Add real-prereqs as a shorthand.
Note:
We cannot replace $(filter %.o,$^) in cmd_link_multi-m because $^ may
include auto-generated dependencies from the .*.cmd file when a single
object module is changed into a multi object module. Refer to commit
69ea912fda ("kbuild: remove unneeded link_multi_deps"). I added some
comment to avoid accidental breakage.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Rob Herring <robh@kernel.org>
Commit a0f97e06a4 ("kbuild: enable 'make CFLAGS=...' to add
additional options to CC") renamed CFLAGS to KBUILD_CFLAGS.
Commit 222d394d30 ("kbuild: enable 'make AFLAGS=...' to add
additional options to AS") renamed AFLAGS to KBUILD_AFLAGS.
Commit 06c5040cdb ("kbuild: enable 'make CPPFLAGS=...' to add
additional options to CPP") renamed CPPFLAGS to KBUILD_CPPFLAGS.
For some reason, LDFLAGS was not renamed.
Using a well-known variable like LDFLAGS may result in accidental
override of the variable.
Kbuild generally uses KBUILD_ prefixed variables for the internally
appended options, so here is one more conversion to sanitize the
naming convention.
I did not touch Makefiles under tools/ since the tools build system
is a different world.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Masahiro Yamada