ld -r is an incremental link used to create built-in.o files in build
subdirectories. It produces relocatable object files containing all
its input files, and these are are then pulled together and relocated
in the final link. Aside from the bloat, this constrains the final
link relocations, which has bitten large powerpc builds with
unresolvable relocations in the final link.
Alan Modra has recommended the kernel use thin archives for linking.
This is an alternative and means that the linker has more information
available to it when it links the kernel.
This patch enables a config option architectures can select, which
causes all built-in.o files to be built as thin archives. built-in.o
files in subdirectories do not get symbol table or index attached,
which improves speed and size. The final link pass creates a
built-in.o archive in the root output directory which includes the
symbol table and index. The linker then uses takes this file to link.
The --whole-archive linker option is required, because the linker now
has visibility to every individual object file, and it will otherwise
just completely avoid including those without external references
(consider a file with EXPORT_SYMBOL or initcall or hardware exceptions
as its only entry points). The traditional built works "by luck" as
built-in.o files are large enough that they're going to get external
references. However this optimisation is unpredictable for the kernel
(due to above external references), ineffective at culling unused, and
costly because the .o files have to be searched for references.
Superior alternatives for link-time culling should be used instead.
Build characteristics for inclink vs thinarc, on a small powerpc64le
pseries VM with a modest .config:
inclink thinarc
sizes
vmlinux 15 618 680 15 625 028
sum of all built-in.o 56 091 808 1 054 334
sum excluding root built-in.o 151 430
find -name built-in.o | xargs rm ; time make vmlinux
real 22.772s 21.143s
user 13.280s 13.430s
sys 4.310s 2.750s
- Final kernel pulled in only about 6K more, which shows how
ineffective the object file culling is.
- Build performance looks improved due to less pagecache activity.
On IO constrained systems it could be a bigger win.
- Build size saving is significant.
Side note, the toochain understands archives, so there's some tricks,
$ ar t built-in.o # list all files you linked with
$ size built-in.o # and their sizes
$ objdump -d built-in.o # disassembly (unrelocated) with filenames
Implementation by sfr, minor tweaks by npiggin.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
(cherry-picked from a5967db9af51a84f5e181600954714a9e4c69f1f)
Change-Id: I2569b083fc15ed8c423fc5c66d179055182e09c1
Signed-off-by: Dan Aloni <daloni@magicleap.com>
Signed-off-by: Davide Garberi <dade.garberi@gmail.com>
This change postpones recordmcount processing until after all LLVM IR
has been compiled into object files. In order to exclude the critical
functions in kernel/trace/ftrace.c from inspection, we add a new
code section .text..ftrace, where recordmcount doesn't look, and a
__norecordmcount attribute that allows functions to be moved to this
section.
Bug: 62093296
Change-Id: I4f79112d7c0919315ee61968890eb255d56282cf
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
This change adds CONFIG_CLANG_LTO for clang's Link Time Optimization.
Bug: 62093296
Change-Id: Id9784f453189c02cea199437f5ff6736f2a75409
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
commit a7df4716d19594b7b3f106f0bc0ca1c548e508e6 upstream.
Similarly to commit fb1770aa78, with gcc 5
on Ubuntu and CONFIG_STATIC_LINK=y I was seeing these linker errors:
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/librt.a(timer_create.o): In function `__timer_create_new':
(.text+0xcd): undefined reference to `pthread_once'
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/librt.a(timer_create.o): In function `__timer_create_new':
(.text+0x126): undefined reference to `pthread_attr_init'
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/librt.a(timer_create.o): In function `__timer_create_new':
(.text+0x168): undefined reference to `pthread_attr_setdetachstate'
[...]
Obviously we also need -lpthread for librt.a.
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On gcc Ubuntu 4.8.4-2ubuntu1~14.04, linking vmlinux fails with:
arch/um/os-Linux/built-in.o: In function `os_timer_create':
/android/kernel/android/arch/um/os-Linux/time.c:51: undefined reference to `timer_create'
arch/um/os-Linux/built-in.o: In function `os_timer_set_interval':
/android/kernel/android/arch/um/os-Linux/time.c:84: undefined reference to `timer_settime'
arch/um/os-Linux/built-in.o: In function `os_timer_remain':
/android/kernel/android/arch/um/os-Linux/time.c:109: undefined reference to `timer_gettime'
arch/um/os-Linux/built-in.o: In function `os_timer_one_shot':
/android/kernel/android/arch/um/os-Linux/time.c:132: undefined reference to `timer_settime'
arch/um/os-Linux/built-in.o: In function `os_timer_disable':
/android/kernel/android/arch/um/os-Linux/time.c:145: undefined reference to `timer_settime'
This is because -lrt appears in the generated link commandline
after arch/um/os-Linux/built-in.o. Fix this by removing -lrt from
arch/um/Makefile and adding it to the UM-specific section of
scripts/link-vmlinux.sh.
Signed-off-by: Lorenzo Colitti <lorenzo@google.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
The Makefiles call the respective interpreter explicitly, but this makes
it easier to use the scripts manually.
Signed-off-by: Michal Marek <mmarek@suse.cz>
x86-64 has a problem: per-cpu variables are actually represented by
their absolute offsets within the per-cpu area, but the symbols are
not emitted as absolute. Thus kallsyms naively creates them as offsets
from _text, meaning their values change if the kernel is relocated
(especially noticeable with CONFIG_RANDOMIZE_BASE):
$ egrep ' (gdt_|_(stext|_per_cpu_))' /root/kallsyms.nokaslr
0000000000000000 D __per_cpu_start
0000000000004000 D gdt_page
0000000000014280 D __per_cpu_end
ffffffff810001c8 T _stext
ffffffff81ee53c0 D __per_cpu_offset
$ egrep ' (gdt_|_(stext|_per_cpu_))' /root/kallsyms.kaslr1
000000001f200000 D __per_cpu_start
000000001f204000 D gdt_page
000000001f214280 D __per_cpu_end
ffffffffa02001c8 T _stext
ffffffffa10e53c0 D __per_cpu_offset
Making them absolute symbols is the Right Thing, but requires fixes to
the relocs tool. So for the moment, we add a --absolute-percpu option
which makes them absolute from a kallsyms perspective:
$ egrep ' (gdt_|_(stext|_per_cpu_))' /proc/kallsyms # no KASLR
0000000000000000 A __per_cpu_start
000000000000a000 A gdt_page
0000000000013040 A __per_cpu_end
ffffffff802001c8 T _stext
ffffffff8099b180 D __per_cpu_offset
ffffffff809a3000 D __per_cpu_load
$ egrep ' (gdt_|_(stext|_per_cpu_))' /proc/kallsyms # With KASLR
0000000000000000 A __per_cpu_start
000000000000a000 A gdt_page
0000000000013040 A __per_cpu_end
ffffffff89c001c8 T _stext
ffffffff8a39d180 D __per_cpu_offset
ffffffff8a3a5000 D __per_cpu_load
Based-on-the-original-screenplay-by: Andy Honig <ahonig@google.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Kees Cook <keescook@chromium.org>
Actually CONFIG_PAGE_OFFSET isn't same with PAGE_OFFSET, so
it isn't easy to figue out PAGE_OFFSET defined in header
file from scripts.
Because CONFIG_PAGE_OFFSET may not be defined in some ARCHs(
64bit ARCH), or defined as bogus value in !MMU case, so
this patch only applys the filter on ARM when CONFIG_PAGE_OFFSET
is defined as the original problem is only on ARM.
Cc: <stable@vger.kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Fixes: f6537f2f0e
Singed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
This patch uses CONFIG_PAGE_OFFSET to filter symbols which
are not in kernel address space because these symbols are
generally for generating code purpose and can't be run at
kernel mode, so we needn't keep them in /proc/kallsyms.
For example, on ARM there are some symbols which may be
linked in relocatable code section, then perf can't parse
symbols any more from /proc/kallsyms, this patch fixes the
problem (introduced b9b32bf70f)
Cc: Russell King <linux@arm.linux.org.uk>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Michal Marek <mmarek@suse.cz>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: stable@vger.kernel.org
We have CONFIG_SYMBOL_PREFIX, which three archs define to the string
"_". But Al Viro broke this in "consolidate cond_syscall and
SYSCALL_ALIAS declarations" (in linux-next), and he's not the first to
do so.
Using CONFIG_SYMBOL_PREFIX is awkward, since we usually just want to
prefix it so something. So various places define helpers which are
defined to nothing if CONFIG_SYMBOL_PREFIX isn't set:
1) include/asm-generic/unistd.h defines __SYMBOL_PREFIX.
2) include/asm-generic/vmlinux.lds.h defines VMLINUX_SYMBOL(sym)
3) include/linux/export.h defines MODULE_SYMBOL_PREFIX.
4) include/linux/kernel.h defines SYMBOL_PREFIX (which differs from #7)
5) kernel/modsign_certificate.S defines ASM_SYMBOL(sym)
6) scripts/modpost.c defines MODULE_SYMBOL_PREFIX
7) scripts/Makefile.lib defines SYMBOL_PREFIX on the commandline if
CONFIG_SYMBOL_PREFIX is set, so that we have a non-string version
for pasting.
(arch/h8300/include/asm/linkage.h defines SYMBOL_NAME(), too).
Let's solve this properly:
1) No more generic prefix, just CONFIG_HAVE_UNDERSCORE_SYMBOL_PREFIX.
2) Make linux/export.h usable from asm.
3) Define VMLINUX_SYMBOL() and VMLINUX_SYMBOL_STR().
4) Make everyone use them.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Tested-by: James Hogan <james.hogan@imgtec.com> (metag)
The shell '.' command is not required to search the current directory as
a fallback and in fact newer versions of bash in sh-mode do not do this.
Force reading the file from the current directory if $KCONFIG_CONFIG
contains no '/'.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Its possible to superseed the config file with KCONFIG_CONFIG and have
completely no .config in the tree. The current script is sourcing
.config in every case, so the kernel will never build succesfully. This
patch fixes that issue by sourcing KCONFIG_CONFIG instead.
Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Pull blackfin updates from Bob Liu:
"One kbuild and a smp build fix."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/lliubbo/blackfin:
kbuild: add symbol prefix arg to kallsyms
blackfin: smp: adapt to generic smp helpers
Commit 1f2bfbd00e ("kbuild: link of
vmlinux moved to a script") introduced in v3.5-rc1 broke kallsyms on
architectures which have symbol prefixes.
The --symbol-prefix argument used to be added to the KALLSYMS command
line from the architecture Makefile, however this isn't picked up by the
new scripts/link-vmlinux.sh. This resulted in symbols like
kallsyms_addresses being added which weren't correctly overriding the
weak symbols such as _kallsyms_addresses. These could then trigger
BUG_ONs in kallsyms code.
This is fixed by removing the KALLSYMS addition from the architecture
Makefile, and using CONFIG_SYMBOL_PREFIX in the link-vmlinux.sh script
to determine whether to add the --symbol-prefix argument.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Bob Liu <lliubbo@gmail.com>
... at least in the top-level Makefile and scripts/link-vmlinux.sh.
There are some more instances of the 'echo <error>; exit 1' pattern in
some arch Makefiles and kconfig.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Pull kbuild updates from Michal Marek.
Fixed up nontrivial merge conflict in Makefile as per Stephen Rothwell
and linux-next (and trivial arch/sparc/Makefile changes due to removed
sparc32 logic).
* 'kbuild' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
mips: Fix KBUILD_CPPFLAGS definition
kbuild: fix ia64 link
kbuild: document KBUILD_LDS, KBUILD_VMLINUX_{INIT,MAIN} and LDFLAGS_vmlinux
kbuild: link of vmlinux moved to a script
kbuild: refactor final link of sparc32
kbuild: drop unused KBUILD_VMLINUX_OBJS from top-level Makefile
kbuild: Makefile: remove unnecessary check for m68knommu ARCH
ia64 build failed like this:
CC init/version.o
LD init/built-in.o
KSYM .tmp_kallsyms1.o
ld: .tmp_kallsyms1.o: linking constant-gp files with non-constant-gp files
ld: failed to merge target specific data of file .tmp_kallsyms1.o
make: *** [vmlinux] Error 1
This was introduced when link of vmlinux was migrated to a script.
Add missing option to as to fix this.
Reported-by: Tony Luck <tony.luck@gmail.com>
Tested-by: Tony Luck <tony.luck@gmail.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Michal Marek <mmarek@suse.cz>
Move the final link of vmlinux to a script to improve
readability and maintainability of the code.
The Makefile fragments used to link vmlinux has over the
years seen far too many changes and the logic had become
hard to follow.
As the process by nature is serialized there was
nothing gained including this in the Makefile.
"um" has special link requirments - and the
only way to handle this was to hard-code the linking
of "um" in the script.
This was better than trying to modularize it only for the
benefit of "um" anyway.
The shell script has been improved after input from:
Arnaud Lacombe <lacombar@gmail.com>
Nick Bowler <nbowler@elliptictech.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Arnaud Lacombe <lacombar@gmail.com>
Cc: Nick Bowler <nbowler@elliptictech.com>
Cc: Richard Weinberger <richard@nod.at>
Signed-off-by: Michal Marek <mmarek@suse.cz>
