GitLab

Redefine GZIP, BZIP2, LZOP variables as KGZIP, KBZIP2, KLZOP resp.
GZIP, BZIP2, LZOP env variables are reserved by the tools. The original
attempt to redefine them internally doesn't work in makefiles/scripts
intercall scenarios, e.g., "make GZIP=gzip bindeb-pkg" and results in
broken builds. There can be other broken build commands because of this,
so the universal solution is to use non-reserved env variables for the
compression tools.

Fixes: 8dfb61dc

 ("kbuild: add variables for compression tools")
Signed-off-by: Denis Efremov <efremov@linux.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Allow user to use alternative implementations of compression tools,
such as pigz, pbzip2, pxz. For example, multi-threaded tools to
speed up the build:
$ make GZIP=pigz BZIP2=pbzip2

Variables _GZIP, _BZIP2, _LZOP are used internally because original env
vars are reserved by the tools. The use of GZIP in gzip tool is obsolete
since 2015. However, alternative implementations (e.g., pigz) still rely
on it. BZIP2, BZIP, LZOP vars are not obsolescent.

The credit goes to @grsecurity.

As a sidenote, for multi-threaded lzma, xz compression one can use:
$ export XZ_OPT="--threads=0"
Signed-off-by: Denis Efremov <efremov@linux.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Many applications check for available kernel features via:

  - /proc/modules (loaded modules, present if CONFIG_MODULES=y)
  - $(MODLIB)/modules.builtin (builtin modules)

They fail to detect features if the kernel was built with CONFIG_MODULES=n
and modules.builtin isn't installed.

Therefore, add the target "_builtin_inst_" and make "install" and
"modules_install" depend on it.

Tests results:

  - make install: kernel image is copied as before, modules.builtin copied
  - make modules_install: (CONFIG_MODULES=n) nothing is copied, exit 1
Signed-off-by: Jonas Zeiger <jonas.zeiger@talpidae.net>
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>

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. Otherwis...

Make modules.order depend on $(obj-m), and use if_changed to build it.
This will avoid unneeded update of modules.order, which will be useful
to optimize the modpost stage.

Currently, the second pass of modpost is always invoked. By checking the
timestamp of modules.order, we can avoid the unneeded modpost.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Do not overwrite core-y or drivers-y. Remove libs-y1 and libs-y2.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

This will slightly change the link order; drivers-y from arch Makefile
will be linked after virt/built-in.a, but I guess this is not a big
deal.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

No arch Makefile specifies init-y.

Merge init-y into core-y. This does not change the link order.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

This hunk has two 'ifdef CONFIG_TRIM_UNUSED_KSYMS ... endif' blocks
with no other code interleaved. Merge them.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

This code does not work as stated in the comment.

$(CONFIG_MODVERSIONS) is always empty because it is expanded before
include/config/auto.conf is included. Hence, 'make modules' with
CONFIG_MODVERSION=y cannot record the version CRCs.

This has been broken since 2003, commit ("kbuild: Enable modules to be
build using the "make dir/" syntax"). [1]

[1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=15c6240cdc44bbeef3c4797ec860f9765ef4f1a7


Cc: linux-stable <stable@vger.kernel.org> # v2.5.71+
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

As debug information gets larger and larger, it helps significantly save
the size of vmlinux images to compress the information in the debug
information sections. Note: this debug info is typically split off from
the final compressed kernel image, which is why vmlinux is what's used
in conjunction with GDB. Minimizing the debug info size should have no
impact on boot times, or final compressed kernel image size.

All of the debug sections will have a `C` flag set.
$ readelf -S <object file>

$ bloaty vmlinux.gcc75.compressed.dwarf4 -- \
    vmlinux.gcc75.uncompressed.dwarf4

    FILE SIZE        VM SIZE
 --------------  --------------
  +0.0%     +18  [ = ]       0    [Unmapped]
 -73.3%  -114Ki  [ = ]       0    .debug_aranges
 -76.2% -2.01Mi  [ = ]       0    .debug_frame
 -73.6% -2.89Mi  [ = ]       0    .debug_str
 -80.7% -4.66Mi  [ = ]       0    .debug_abbrev
 -82.9% -4.88Mi  [ = ]       0    .debug_ranges
 -70.5% -9.04Mi  [ = ]       0    .debug_line
 -79.3% -10.9Mi  [ = ]       0    .debug_loc
 -39.5% -88.6Mi  [ = ]       0    .debug_info
 -18.2%  -123Mi  [ = ]       0    TOTAL

$ bloaty vmlinux.clang11.compressed.dwarf4 -- \
    vmlinux.clang11.uncompressed.dwarf4

    FILE SIZE        VM SIZE
 --------------  --------------
  +0.0%     +23  [ = ]       0    [Unmapped]
 -65.6%    -871  [ = ]       0    .debug_aranges
 -77.4% -1.84Mi  [ = ]       0    .debug_frame
 -82.9% -2.33Mi  [ = ]       0    .debug_abbrev
 -73.1% -2.43Mi  [ = ]       0    .debug_str
 -84.8% -3.07Mi  [ = ]       0    .debug_ranges
 -65.9% -8.62Mi  [ = ]       0    .debug_line
 -86.2% -40.0Mi  [ = ]       0    .debug_loc
 -42.0% -64.1Mi  [ = ]       0    .debug_info
 -22.1%  -122Mi  [ = ]       0    TOTAL

For x86_64 defconfig + LLVM=1 (before):
Elapsed (wall clock) time (h:mm:ss or m:ss): 3:22.03
Maximum resident set size (kbytes): 43856

For x86_64 defconfig + LLVM=1 (after):
Elapsed (wall clock) time (h:mm:ss or m:ss): 3:32.52
Maximum resident set size (kbytes): 1566776

Thanks to:
Nick Clifton helped us to provide the minimal binutils version.
Sedat Dilek found an increase in size of debug .deb package.

Cc: Nick Clifton <nickc@redhat.com>
Suggested-by: David Blaikie <blaikie@google.com>
Reviewed-by: Fangrui Song <maskray@google.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

$(firstword ...) in scripts/Makefile.modpost was added by commit
3f3fd3c0 ("[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 0d96fb20

 ("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>

Do not try to build any module-related artifacts when CONFIG_MODULES
is disabled.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

I think all the warnings have been fixed by now. Make it a fatal error.

Check it before modpost because we need to stop building *.ko files.
Also, pass modules.order via a script parameter.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Some targets are internal-use only.

It is tedious to care about "what if __build_one_by_one is contained
in $(MAKECMDGOALS)?" etc.

Prefix internal targets with double underscores. Stop parsing Makefile
if they are directly run.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Make it clearer, and self-documenting.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Merge {CLEAN,MRPROPER,DISTCLEAN}_DIRS into {CLEAN,MRPROPER,DISTCLEAN}_FILES
because the difference is just the -r option passed to the 'rm' command.

Do likewise as commit 1634f2bf

 ("kbuild: remove clean-dirs syntax").
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

This comment was added by commit ("kbuild: Restore build nr, improve
vmlinux link") [1].

It was talking about if_changed_rule at that time. Now, it is unclear
what to fix.

[1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=ea52ca1b3e3882b499cc6c043f384958b88b62ff

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

If include/generated/autoconf.h is accidentally lost somehow,
there is no clear way to fix it. Make it self-healing.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

Kbuild supports the infrastructure to build host programs, but there
was no support to build userspace programs for the target architecture
(i.e. the same architecture as the kernel).

Sam Ravnborg worked on this in 2014 (https://lkml.org/lkml/2014/7/13/154),
but it was not merged. One problem at that time was, there was no good way
to know whether $(CC) can link standalone programs. In fact, pre-built
kernel.org toolchains [1] are often used for building the kernel, but they
do not provide libc.

Now, we can handle this cleanly because the compiler capability is
evaluated at the Kconfig time. If $(CC) cannot link standalone programs,
the relevant options are hidden by 'depends on CC_CAN_LINK'.

The implementation just mimics scripts/Makefile.host

The userspace programs are compiled with the same flags as the host
programs. In addition, it uses -m32 or -m64 if it is found in
$(KBUILD_CFLAGS).

This new syntax has two usecases.

- Sample programs

  Several userspace programs under samples/ include UAPI headers
  installed in usr/include. Most of them were previously built for
  the host architecture just to use the 'hostprogs' syntax.

  However, 'make headers' always works for the target architecture.
  This caused the arch mismatch in cross-compiling. To fix this
  distortion, sample code should be built for the target architecture.

- Bpfilter

  net/bpfilter/Makefile compiles bpfilter_umh as the user mode helper,
  and embeds it into the kernel. Currently, it overrides HOSTCC with
  CC to use the 'hostprogs' syntax. This hack should go away.

[1]: https://mirrors.edge.kernel.org/pub/tools/crosstool/

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>

This change adds generic support for Clang's Shadow Call Stack,
which uses a shadow stack to protect return addresses from being
overwritten by an attacker. Details are available here:

  https://clang.llvm.org/docs/ShadowCallStack.html



Note that security guarantees in the kernel differ from the ones
documented for user space. The kernel must store addresses of
shadow stacks in memory, which means an attacker capable reading
and writing arbitrary memory may be able to locate them and hijack
control flow by modifying the stacks.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
[will: Numerous cosmetic changes]
Signed-off-by: Will Deacon <will@kernel.org>

This notice has been here for a while. Remove it entirely now.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>

gcc-10 now warns about passing aliasing pointers to functions that take
restricted pointers.

That's actually a great warning, and if we ever start using 'restrict'
in the kernel, it might be quite useful.  But right now we don't, and it
turns out that the only thing this warns about is an idiom where we have
declared a few functions to be "printf-like" (which seems to make gcc
pick up the restricted pointer thing), and then we print to the same
buffer that we also use as an input.

And people do that as an odd concatenation pattern, with code like this:

    #define sysfs_show_gen_prop(buffer, fmt, ...) \
        snprintf(buffer, PAGE_SIZE, "%s"fmt, buffer, __VA_ARGS__)

where we have 'buffer' as both the destination of the final result, and
as the initial argument.

Yes, it's a bit questionable.  And outside of the kernel, people do have
standard declarations like

    int snprintf( char *restrict buffer, size_t bufsz,
                  const char *restrict format, ... );

where that output buffer is marked as a restrict pointer that cannot
alias with any other arguments.

But in the context of the kernel, that 'use snprintf() to concatenate to
the end result' does work, and the pattern shows up in multiple places.
And we have not marked our own version of snprintf() as taking restrict
pointers, so the warning is incorrect for now, and gcc picks it up on
its own.

If we do start using 'restrict' in the kernel (and it might be a good
idea if people find places where it matters), we'll need to figure out
how to avoid this issue for snprintf and friends.  But in the meantime,
this warning is not useful.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This is the final array bounds warning removal for gcc-10 for now.

Again, the warning is good, and we should re-enable all these warnings
when we have converted all the legacy array declaration cases to
flexible arrays. But in the meantime, it's just noise.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This is another fine warning, related to the 'zero-length-bounds' one,
but hitting the same historical code in the kernel.

Because C didn't historically support flexible array members, we have
code that instead uses a one-sized array, the same way we have cases of
zero-sized arrays.

The one-sized arrays come from either not wanting to use the gcc
zero-sized array extension, or from a slight convenience-feature, where
particularly for strings, the size of the structure now includes the
allocation for the final NUL character.

So with a "char name[1];" at the end of a structure, you can do things
like

       v = my_malloc(sizeof(struct vendor) + strlen(name));

and avoid the "+1" for the terminator.

Yes, the modern way to do that is with a flexible array, and using
'offsetof()' instead of 'sizeof()', and adding the "+1" by hand.  That
also technically gets the size "more correct" in that it avoids any
alignment (and thus padding) issues, but this is another long-term
cleanup thing that will not happen for 5.7.

So disable the warning for now, even though it's potentially quite
useful.  Having a slew of warnings that then hide more urgent new issues
is not an improvement.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This is a fine warning, but we still have a number of zero-length arrays
in the kernel that come from the traditional gcc extension.  Yes, they
are getting converted to flexible arrays, but in the meantime the gcc-10
warning about zero-length bounds is very verbose, and is hiding other
issues.

I missed one actual build failure because it was hidden among hundreds
of lines of warning.  Thankfully I caught it on the second go before
pushing things out, but it convinced me that I really need to disable
the new warnings for now.

We'll hopefully be all done with our conversion to flexible arrays in
the not too distant future, and we can then re-enable this warning.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

We have some rather random rules about when we accept the
"maybe-initialized" warnings, and when we don't.

For example, we consider it unreliable for gcc versions < 4.9, but also
if -O3 is enabled, or if optimizing for size.  And then various kernel
config options disabled it, because they know that they trigger that
warning by confusing gcc sufficiently (ie PROFILE_ALL_BRANCHES).

And now gcc-10 seems to be introducing a lot of those warnings too, so
it falls under the same heading as 4.9 did.

At the same time, we have a very straightforward way to _enable_ that
warning when wanted: use "W=2" to enable more warnings.

So stop playing these ad-hoc games, and just disable that warning by
default, with the known and straight-forward "if you want to work on the
extra compiler warnings, use W=123".

Would it be great to have code that is always so obvious that it never
confuses the compiler whether a variable is used initialized or not?
Yes, it would.  In a perfect world, the compilers would be smarter, and
our source code would be simpler.

That's currently not the world we live in, though.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

As Documentation/kbuild/llvm.rst implies, building the kernel with a
full set of LLVM tools gets very verbose and unwieldy.

Provide a single switch LLVM=1 to use Clang and LLVM tools instead
of GCC and Binutils. You can pass it from the command line or as an
environment variable.

Please note LLVM=1 does not turn on the integrated assembler. You need
to pass LLVM_IAS=1 to use it. When the upstream kernel is ready for the
integrated assembler, I think we can make it default.

We discussed what we need, and we agreed to go with a simple boolean
flag that switches both target and host tools:

  https://lkml.org/lkml/2020/3/28/494
  https://lkml.org/lkml/2020/4/3/43



Some items discussed, but not adopted:

- LLVM_DIR

  When multiple versions of LLVM are installed, I just thought supporting
  LLVM_DIR=/path/to/my/llvm/bin/ might be useful.

  CC      = $(LLVM_DIR)clang
  LD      = $(LLVM_DIR)ld.lld
    ...

  However, we can handle this by modifying PATH. So, we decided to not do
  this.

- LLVM_SUFFIX

  Some distributions (e.g. Debian) package specific versions of LLVM with
  naming conventions that use the version as a suffix.

  CC      = clang$(LLVM_SUFFIX)
  LD      = ld.lld(LLVM_SUFFIX)
    ...

  will allow a user to pass LLVM_SUFFIX=-11 to use clang-11 etc.,
  but the suffixed versions in /usr/bin/ are symlinks to binaries in
  /usr/lib/llvm-#/bin/, so this can also be handled by PATH.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com> # build
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>

The 'AS' variable is unused for building the kernel. Only the remaining
usage is to turn on the integrated assembler. A boolean flag is a better
fit for this purpose.

AS=clang was added for experts. So, I replaced it with LLVM_IAS=1,
breaking the backward compatibility.
Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>

Kbuild supports not only obj-y but also lib-y to list objects linked to
vmlinux.

The difference between them is that all the objects from obj-y are
forcibly linked to vmlinux, whereas the objects from lib-y are linked
as needed; if there is no user of a lib-y object, it is not linked.

lib-y is intended to list utility functions that may be called from all
over the place (and may be unused at all), but it is a problem for
EXPORT_SYMBOL(). Even if there is no call-site in the vmlinux, we need
to keep exported symbols for the use from loadable modules.

Commit 7f2084fa ("[kbuild] handle exports in lib-y objects reliably")
worked around it by linking a dummy object, lib-ksyms.o, which contains
references to all the symbols exported from lib.a in that directory.
It uses the linker script command, EXTERN. Unfortunately, the meaning of
EXTERN of ld.lld is different from that of ld.bfd. Therefore, this does
not work with LD=ld.lld (CBL issue #515).

Anyway, the build rule of lib-ksyms.o is somewhat tricky. So, I want to
get rid of it.

At first, I was thinking of accumulating lib-y objects into obj-y
(or even replacing lib-y with obj-y entirely), but the lib-y syntax
is used beyond the ordinary use in lib/ and arch/*/lib/.

Examples:

 - drivers/firmware/efi/libstub/Makefile builds lib.a, which is linked
   into vmlinux in the own way (arm64), or linked to the decompressor
   (arm, x86).

 - arch/alpha/lib/Makefile builds lib.a which is linked not only to
   vmlinux, but also to bootloaders in arch/alpha/boot/Makefile.

 - arch/xtensa/boot/lib/Makefile builds lib.a for use from
   arch/xtensa/boot/boot-redboot/Makefile.

One more thing, adding everything to obj-y would increase the vmlinux
size of allnoconfig (or tinyconfig).

For less impact, I tweaked the destination of lib.a at the top Makefile;
when CONFIG_MODULES=y, lib.a goes to KBUILD_VMLINUX_OBJS, which is
forcibly linked to vmlinux, otherwise lib.a goes to KBUILD_VMLINUX_LIBS
as before.

The size impact for normal usecases is quite small since at lease one
symbol in every lib-y object is eventually called by someone. In case
you are intrested, here are the figures.

x86_64_defconfig:

   text	   data	    bss	    dec	    hex	filename
19566602 5422072 1589328 26578002 1958c52 vmlinux.before
19566932 5422104 1589328 26578364 1958dbc vmlinux.after

The case with the biggest impact is allnoconfig + CONFIG_MODULES=y.

ARCH=x86 allnoconfig + CONFIG_MODULES=y:

   text	   data	    bss	    dec	    hex	filename
1175162	 254740	1220608	2650510	 28718e	vmlinux.before
1177974	 254836	1220608	2653418	 287cea	vmlinux.after

Hopefully this is still not a big deal. The per-file trimming with the
static library is not so effective after all.

If fine-grained optimization is desired, some architectures support
CONFIG_LD_DEAD_CODE_DATA_ELIMINATION, which trims dead code per-symbol
basis. When LTO is supported in mainline, even better optimization will
be possible.

Link: https://github.com/ClangBuiltLinux/linux/issues/515

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reported-by: kbuild test robot <lkp@intel.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>