-Wmissing-field-initializers is too noisy to be useful on W=1. Let's
move it to W=2.
Signed-off-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

Usage of these flags has been deprecated for nearly 4 years by:

    commit f77bf014
    Author: Sam Ravnborg <sam@neptun.(none)>
    Date:   Mon Oct 15 22:25:06 2007 +0200

        kbuild: introduce ccflags-y, asflags-y and ldflags-y

Moreover, these flags (at least EXTRA_CFLAGS) have been documented for
command line use. By default, gmake(1) do not override command line
setting, so this is likely to result in build failure or unexpected
behavior.

Do not advertise for its usage.

Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: linux-kbuild@vger.kernel.org
Signed-off-by: NArnaud Lacombe <lacombar@gmail.com>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

Modifications to recordmcount must be performed on all object
files to stay consistent with what the kernel code may expect.
Add the recordmcount files to the main dependencies to make sure
any change to them causes a full recompile.
Signed-off-by: NMichal Marek <mmarek@suse.cz>
Link: http://lkml.kernel.org/r/20110517133646.GP13293@sepie.suse.czSigned-off-by: NSteven Rostedt <rostedt@goodmis.org>

When mcount is called in a section that ftrace will not modify it into
a nop, we want to warn about this. But not warn about this always. Now
if the user builds the kernel with the option RECORDMCOUNT_WARN=1 then
the build will warn about mcount callers that are ignored and will just
waste execution time.
Acked-by: NMichal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Link: http://lkml.kernel.org/r/20110421023738.714956282@goodmis.orgSigned-off-by: NSteven Rostedt <rostedt@goodmis.org>

Modifications to recordmcount must be performed on all object
files to stay consistent with what the kernel code may expect.
Add the recordmcount files to the main dependencies to make sure
any change to them causes a full recompile.
Signed-off-by: NMichal Marek <mmarek@suse.cz>
Link: http://lkml.kernel.org/r/20110517133646.GP13293@sepie.suse.czSigned-off-by: NSteven Rostedt <rostedt@goodmis.org>

When mcount is called in a section that ftrace will not modify it into
a nop, we want to warn about this. But not warn about this always. Now
if the user builds the kernel with the option RECORDMCOUNT_WARN=1 then
the build will warn about mcount callers that are ignored and will just
waste execution time.
Acked-by: NMichal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Link: http://lkml.kernel.org/r/20110421023738.714956282@goodmis.orgSigned-off-by: NSteven Rostedt <rostedt@goodmis.org>

Add support for make W=12, make W=123 and so on, to enable warnings from
multiple W= levels. Normally, make W=<level> does not include warnings
from the previous level.
Signed-off-by: NMichal Marek <mmarek@suse.cz>
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Reviewed-By: NValdis Kletnieks <valdis.kletnieks@vt.edu>

Disable the new -Wunused-but-set-variable that was added in gcc 4.6.0
It produces more false positives than useful warnings.

This can still be enabled using W=1
Signed-off-by: NDave Jones <davej@redhat.com>
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Tested-by: NSam Ravnborg <sam@ravnborg.org>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

Building a kernel with "make W=1" produces far too much noise to be
useful.

Divide the warning options in three groups:

    W=1 - warnings that may be relevant and does not occur too often
    W=2 - warnings that occur quite often but may still be relevant
    W=3 - the more obscure warnings, can most likely be ignored

When building the whole kernel, those levels produce:

W=1 - 4859 warnings
W=2 - 1394 warnings
W=3 - 86666 warnings

respectively. Warnings have been counted with Geert's script at

http://www.kernel.org/pub/linux/kernel/people/geert/linux-log/linux-log-summary.pl

Many warnings occur from .h files so fixing one file may have a nice
effect on the total number of warnings.

With these changes I am actually tempted to try W=1 now and then.
Previously there was just too much noise.

Borislav:

- make the W= levels exclusive
- move very noisy and making little sense for the kernel warnings to W=3
- drop -Woverlength-strings due to useless warning message
- copy explanatory text for the different warning levels to 'make help'
- recount warnings per level
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>
Signed-off-by: NBorislav Petkov <bp@alien8.de>
Cc: Dave Jones <davej@redhat.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

The D option of ar is only available in newer versions.
Signed-off-by: NMichal Marek <mmarek@suse.cz>

Signed-off-by: NMichal Marek <mmarek@suse.cz>

Add a 'W=1' Makefile switch which adds additional checking per build
object.

The idea behind this option is targeted at developers who, in the
process of writing their code, want to do the occasional

make W=1 [target.o]

and let gcc do more extensive code checking for them. Then, they
could eyeball the output for valid gcc warnings about various
bugs/discrepancies which are not reported during the normal build
process.

For more background information and a use case, read through this
thread: http://marc.info/?l=kernel-janitors&m=129802065918147&w=2

Cc: Michal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Acked-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NBorislav Petkov <bp@alien8.de>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

cmd_record_mcount is used to locate the _mcount symbols in the object
files, only the files compiled with -pg has the _mcount symbol, so, it
is only needed for such files, but the current cmd_record_mcount is used
for all of the object files, so, we need to fix it and speed it up.

Since -pg may be removed by the method used in kernel/trace/Makefile:

ORIG_CFLAGS := $(KBUILD_CFLAGS)
KBUILD_CFLAGS = $(subst -pg,,$(ORIG_CFLAGS))

Or may be removed by the method used in arch/x86/kernel/Makefile:

CFLAGS_REMOVE_file.o = -pg

So, we must check the last variable stores the compiling flags, that is
c_flags(Please refer to cmd_cc_o_c and rule_cc_o_c defined in
scripts/Makefile.build) and since the CFLAGS_REMOVE_file.o is already
filtered in _c_flags(Please refer to scripts/Makefile.lib) and _c_flags
has less symbols, therefore, we only need to check _c_flags.

---------------
Changes from v1:

  o Don't touch Makefile for CONFIG_FTRACE_MCOUNT_RECORD is enough
  o Use _c_flags intead of KBUILD_CFLAGS to cover CONFIG_REMOVE_file.o = -pg
  (feedback from Steven Rostedt <rostedt@goodmis.org>)
Acked-by: NMichal Marek <mmarek@suse.cz>
Signed-off-by: NWu Zhangjin <wuzhangjin@gmail.com>
LKML-Reference: <3dc8cddf022eb7024f9f2cf857529a15bee8999a.1288196498.git.wuzhangjin@gmail.com>

[ changed if [ .. == .. ] to if [ .. = .. ] to handle dash environments ]
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

When DYNAMIC_FTRACE is enabled and we use the C version of recordmcount,
all objects are run through the recordmcount program to create a
separate section that stores all the callers of mcount.

The build process has a special file: scripts/mod/empty.o. This is
built from empty.c which is literally an empty file (except for a
single comment). This file is used to find information about the target
elf format, like endianness and word size.

The problem comes up when we need to build recordmcount. The
build process requires that empty.o is built first. The build rules
for empty.o will try to execute recordmcount on the empty.o file.
We get an error that recordmcount does not exist.

To avoid this recursion, the build file will skip running recordmcount
if the file that it is building is script/mod/empty.o.

[ extra comment Suggested-by: Sam Ravnborg <sam@ravnborg.org> ]
Reported-by: NIngo Molnar <mingo@elte.hu>
Tested-by: NIngo Molnar <mingo@elte.hu>
Cc: Michal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

The C version of recordmcount is compiled to a binary, which will
end up located in the objtree. If the kernel is built with O=path,
the srctree will not include the binary recordmcount caller.

Cc: Michal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

This patch adds the support for the C version of recordmcount and
compile times show ~ 12% improvement.

After verifying this works, other archs can add:

 HAVE_C_MCOUNT_RECORD

in its Kconfig and it will use the C version of recordmcount
instead of the perl version.

Cc: <linux-arch@vger.kernel.org>
Cc: Michal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Cc: John Reiser <jreiser@bitwagon.com>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

On ARM, we have two ABIs, and the ABI used is controlled via a config
option.  Object files built with one ABI can't be merged with object
files built with the other ABI.  So, record_mcount.pl needs to use the
same compiler flags as the kernel when generating the object file with
the mcount locations.  Ensure this by passing CFLAGS to the script.
Acked-by: NSteven Rostedt <rostedt@goodmis.org>
Signed-off-by: NRabin Vincent <rabin@rab.in>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

It is now possible to assign options to AS and CC
on the command line - which is only used for built-in code.

{A,C}FLAGS_KERNEL was used both in the top-level Makefile
in the arch makefiles, thus users had no way to specify
additional options to AS, CC without overriding
the original value.

Introduce a new set of variables KBUILD_{A,C}FLAGS_KERNEL
that is used by arch specific files and free up
{A,C}FLAGS_KERNEL so they can be assigned on
the command line.

All arch Makefiles that used the old variables has been updated.
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

It is now possible to assign options to AS, CC and LD
on the command line - which is only used when building modules.

{A,C,LD}FLAGS_MODULE was all used both in the top-level Makefile
in the arch makefiles, thus users had no way to specify
additional options to AS, CC, LD when building modules
without overriding the original value.

Introduce a new set of variables KBUILD_{A,C,LD}FLAGS_MODULE
that is used by arch specific files and free up
{A,C,LD}FLAGS_MODULE so they can be assigned on
the command line.

All arch Makefiles that used the old variables has been updated.

Note: Previously we had a MODFLAGS variable for both
AS and CC. But in favour of consistency this was dropped.
So in some cases arch Makefile has one assignmnet replaced by
two assignmnets.

Note2: MODFLAGS was not documented and is dropped
without any notice. I do not expect much/any breakage
from this.
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Chen Liqin <liqin.chen@sunplusct.com>
Acked-by: Mike Frysinger <vapier@gentoo.org> [blackfin]
Acked-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com> [avr32]
Signed-off-by: NMichal Marek <mmarek@suse.cz>

commit 37a8d9f6 tried to combine some
duplicate code and accidentally broke how KBUILD_SYMTYPES worked

This fixes the code to match the original intention by the author who
originally added the code I believe.

The fixes include:
- removing extra whitespaces in the if-statements
- moving the if-statement from around the -r to the -T
- adding a second arg to cmd_gensymtypes to simplify the options passed
  to genksyms.

Tested by instrumenting genksyms and seeing what options were passed in
during a make, KBUILD_SYMTYPES make, and when a foo.symref was created.

Everything compiled and looked ok.
Signed-off-by: NDon Zickus <dzickus@redhat.com>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

This patch fixes the link error "built-in.o: no such file or directory".
The problem happens if "dirx/Makefile" contains only "obj-m += diry/
dirz/" and the empty "dirx/built-in.o" is missing. Adding $(subdir-m)
into check for builtin-target fixes this error.
Signed-off-by: NJiafu He <jay@goldhive.com>
Signed-off-by: NMichal Marek <mmarek@suse.cz>

MIPS and some other architectures need this argument to handle
big/little endian respectively.
Signed-off-by: NWu Zhangjin <wuzj@lemote.com>
Cc: Nicholas Mc Guire <der.herr@hofr.at>
Cc: zhangfx@lemote.com
Cc: Wu Zhangjin <wuzhangjin@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: linux-kernel@vger.kernel.org
Cc: linux-mips@linux-mips.org
Patchwork: http://patchwork.linux-mips.org/patch/674/Acked-by: NSteven Rostedt <rostedt@goodmis.org>
Signed-off-by: NRalf Baechle <ralf@linux-mips.org>

Adding a reference to <linux/linkage.h> to x86's <asm/cache.h> causes
the x86 linker script to have syntax errors, because the ALIGN and
ENTRY keywords get redefined to the assembly implementations of those.
One could fix this by adjusting the include structure, but I think any
solution based on that approach would be fragile.

Currently, it is impossible when writing a header to do something
different for assembly files and linker scripts, even though there are
clearly cases where one wants them to define macros differently for
the two (ENTRY being an excellent example).
So I think the right solution here is to introduce a new preprocessor
definition, called LINKER_SCRIPT that is set along with __ASSEMBLY__
for linker scripts, and to use that to not define ALIGN and ENTRY in
linker scripts.
I suspect we'll find other uses for this mechanism in
the future.
Signed-off-by: NTim Abbott <tabbott@ksplice.com>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Albin Tonnerre <albin.tonnerre@free-electrons.com> reported:

    Bash 4 filters out variables which contain a dot in them.
    This happends to be the case of CPPFLAGS_vmlinux.lds.
    This is rather unfortunate, as it now causes
    build failures when using SHELL=/bin/bash to compile,
    or when bash happens to be used by make (eg when it's /bin/sh)

Remove the common definition of CPPFLAGS_vmlinux.lds by
pushing relevant stuff to either Makefile.build or the
arch specific kernel/Makefile where we build the linker script.

This is also nice cleanup as we move the information out where
it is used.

Notes for the different architectures touched:

arm - we use an already exported symbol
cris - we use a config symbol aleady available
       [Not build tested]
mips - the jiffies complexity has moved to vmlinux.lds.S where we need it.
       Added a few variables to CPPFLAGS - they are only used by
       the linker script.
       [Not build tested]
powerpc - removed assignment that is not needed
          [not build tested]
sparc - simplified it using $(BITS)
um - introduced a few new exported variables to deal with this
xtensa - added options to CPP invocation
         [not build tested]

Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

When this script fails the build should fail too. Otherwise there
are mysterious build failures later.
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

I had some problems with record_mcount in the Makefile and it was hard
to track down. Echo it by default to make it easier to diagnose.
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Following patch introduce support for setting options
to gcc that has effect for current directory and all
subdirectories.

The typical use case are an architecture or a subsystem that
decide to cover all files with -Werror.
Today alpha, mips and sparc uses -Werror in almost all their
Makefile- with subdir-ccflag-y it is now simpler to do so
as only the top-level directories needs to be covered.

Likewise if we decide to cover a full subsystem such
as net/ with -Werror this is done by adding a single
line to net/Makefile.
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>

This reverts commit ad7a953c.

And commit: ("allow stripping of generated symbols under CONFIG_KALLSYMS_ALL")
            9bb48247

These stripping patches has caused a set of issues:

1) People have reported compatibility issues with binutils due to
   lack of support for `--strip-unneeded-symbols' with objcopy 2.15.92.0.2
   Reported by: Wenji
2) ccache and distcc no longer works as expeced
   Reported by: Ted, Roland, + others
3) The installed modules increased a lot in size
   Reported by: Ted, Davej + others
Reported-by: NWenji Huang <wenji.huang@oracle.com>
Reported-by: N"Theodore Ts'o" <tytso@mit.edu>
Reported-by: NDave Jones <davej@redhat.com>
Reported-by: NRoland McGrath <roland@redhat.com>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

In IA64, module build and kernel build use different option.
Make recordmcount.pl differentiate the two cases.
Signed-off-by: NShaohua Li <shaohua.li@intel.com>
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This patch changes the way __crc_ symbols are being resolved from
using ld to do so to using the assembler, thus allowing these symbols
to be marked local (the linker creates then as global ones) and hence
allow stripping (for modules) or ignoring (for vmlinux) them. While at
this, also strip other generated symbols during module installation.

One potentially debatable point is the handling of the flags passeed
to gcc when translating the intermediate assembly file into an object:
passing $(c_flags) unchanged doesn't work as gcc passes --gdwarf2 to
gas whenever is sees any -g* option, even for -g0, and despite the
fact that the compiler would have already produced all necessary debug
info in the C->assembly translation phase. I took the approach of just
filtering out all -g* options, but an alternative to such negative
filtering might be to have a positive filter which might, in the ideal
case allow just all the -Wa,* options to pass through.
Signed-off-by: NJan Beulich <jbeulich@novell.com>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Avoid duplicating long list of options in two places
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Sometimes it is preferable to avoid changes of exported symbol checksums
(to avoid breaking externally provided modules).  When a checksum change
occurs, it can be hard to figure out what caused this change: underlying
types may have changed, or additional type information may simply have
become available at the point where a symbol is exported.

Add a new --reference option to genksyms which allows it to report why
checksums change, based on the type information dumps it creates with the
--dump-types flag.  Genksyms will read in such a dump from a previous run,
and report which symbols have changed (and why).

The behavior can be controlled for an entire build as follows: If
KBUILD_SYMTYPES is set, genksyms uses --dump-types to produce *.symtypes
dump files.  If any *.symref files exist, those will be used as the
reference to check against.  If KBUILD_PRESERVE is set, checksum changes
will fail the build.
Signed-off-by: NAndreas Gruenbacher <agruen@suse.de>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Impact: cleanup

Sam Ravnborg pointed out that I could condense the code for the parameters of
recordmcount.pl by using an $(if ...) condition.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Acked-by: NSam Ravnborg <sam@ravnborg.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

The recordmcount script requires that the actual arch is passed in.
This works well when ARCH=i386 or ARCH=x86_64 but does not handle the
case of ARCH=x86.

This patch adds a parameter to the function to pass in the number of
bits of the architecture. So that it can determine if x86 should be
run for x86_64 or i386 archs.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

CHK     include/linux/version.h
  CHK     include/linux/utsrelease.h
  CC      scripts/mod/empty.o
/bin/sh: /usr/src/25/scripts/recordmcount.pl: Permission denied

We shouldn't assume that files have their `x' bits set.  There are various
ways in which file permissions get lost, including use of patch(1).

It might not be correct to assume that perl lives in $PATH?
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

I'm seeing when I use separate src/build dirs:

make[3]: *** [arch/x86/kernel/time_32.o] Error 1
/bin/sh: scripts/recordmcount.pl: No such file or directory
make[3]: *** [arch/x86/kernel/irq_32.o] Error 1
/bin/sh: scripts/recordmcount.pl: No such file or directory
make[3]: *** [arch/x86/kernel/ldt.o] Error 1
/bin/sh: scripts/recordmcount.pl: No such file or directory
make[3]: *** [arch/x86/kernel/i8259.o] Error 1
/bin/sh: scripts/recordmcount.pl: No such file or directory

This fixes it.
Signed-off-by: NIngo Molnar <mingo@elte.hu>

This patch creates a section in the kernel called "__mcount_loc".
This will hold a list of pointers to the mcount relocation for
each call site of mcount.

For example:

objdump -dr init/main.o
[...]
Disassembly of section .text:

0000000000000000 <do_one_initcall>:
   0:   55                      push   %rbp
[...]
000000000000017b <init_post>:
 17b:   55                      push   %rbp
 17c:   48 89 e5                mov    %rsp,%rbp
 17f:   53                      push   %rbx
 180:   48 83 ec 08             sub    $0x8,%rsp
 184:   e8 00 00 00 00          callq  189 <init_post+0xe>
                        185: R_X86_64_PC32      mcount+0xfffffffffffffffc
[...]

We will add a section to point to each function call.

   .section __mcount_loc,"a",@progbits
[...]
   .quad .text + 0x185
[...]

The offset to of the mcount call site in init_post is an offset from
the start of the section, and not the start of the function init_post.
The mcount relocation is at the call site 0x185 from the start of the
.text section.

  .text + 0x185  == init_post + 0xa

We need a way to add this __mcount_loc section in a way that we do not
lose the relocations after final link.  The .text section here will
be attached to all other .text sections after final link and the
offsets will be meaningless.  We need to keep track of where these
.text sections are.

To do this, we use the start of the first function in the section.
do_one_initcall.  We can make a tmp.s file with this function as a reference
to the start of the .text section.

   .section __mcount_loc,"a",@progbits
[...]
   .quad do_one_initcall + 0x185
[...]

Then we can compile the tmp.s into a tmp.o

  gcc -c tmp.s -o tmp.o

And link it into back into main.o.

  ld -r main.o tmp.o -o tmp_main.o
  mv tmp_main.o main.o

But we have a problem.  What happens if the first function in a section
is not exported, and is a static function. The linker will not let
the tmp.o use it.  This case exists in main.o as well.

Disassembly of section .init.text:

0000000000000000 <set_reset_devices>:
   0:   55                      push   %rbp
   1:   48 89 e5                mov    %rsp,%rbp
   4:   e8 00 00 00 00          callq  9 <set_reset_devices+0x9>
                        5: R_X86_64_PC32        mcount+0xfffffffffffffffc

The first function in .init.text is a static function.

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices

The lowercase 't' means that set_reset_devices is local and is not exported.
If we simply try to link the tmp.o with the set_reset_devices we end
up with two symbols: one local and one global.

 .section __mcount_loc,"a",@progbits
 .quad set_reset_devices + 0x10

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices
                 U set_reset_devices

We still have an undefined reference to set_reset_devices, and if we try
to compile the kernel, we will end up with an undefined reference to
set_reset_devices, or even worst, it could be exported someplace else,
and then we will have a reference to the wrong location.

To handle this case, we make an intermediate step using objcopy.
We convert set_reset_devices into a global exported symbol before linking
it with tmp.o and set it back afterwards.

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 T set_reset_devices

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 T set_reset_devices

00000000000000a8 t __setup_set_reset_devices
000000000000105f t __setup_str_set_reset_devices
0000000000000000 t set_reset_devices

Now we have a section in main.o called __mcount_loc that we can place
somewhere in the kernel using vmlinux.ld.S and access it to convert
all these locations that call mcount into nops before starting SMP
and thus, eliminating the need to do this with kstop_machine.

Note, A well documented perl script (scripts/recordmcount.pl) is used
to do all this in one location.
Signed-off-by: NSteven Rostedt <srostedt@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Signed-off-by: NRobert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

Setting the option DEBUG_SECTION_MISMATCH will
report additional section mismatch'es but this
should in the end makes it possible to get rid of
all of them.

See help text in lib/Kconfig.debug for details.
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>

When multiple built-in modules (especially drivers) provide the same
capability, they're prioritized by link order specified by the order
listed in Makefile.  This implicit ordering is lost for loadable
modules.

When driver modules are loaded by udev, what comes first in
modules.alias file is selected.  However, the order in this file is
indeterministic (depends on filesystem listing order of installed
modules).  This causes confusion.

The solution is two-parted.  This patch updates kbuild such that it
generates and installs modules.order which contains the name of
modules ordered according to Makefile.  The second part is update to
depmod such that it generates output files according to this file.

Note that both obj-y and obj-m subdirs can contain modules and
ordering information between those two are lost from beginning.
Currently obj-y subdirs are put before obj-m subdirs.

Sam Ravnborg cleaned up Makefile modifications and suggested using awk
to remove duplicate lines from modules.order instead of using separate
C program.
Signed-off-by: NTejun Heo <htejun@gmail.com>
Acked-by: NGreg Kroah-Hartman <gregkh@suse.de>
Cc: Bill Nottingham <notting@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Jon Masters <jonathan@jonmasters.org>
Signed-off-by: NSam Ravnborg <sam@ravnborg.org>