When a kernel was rebuilt, the previous Module.markers was not cleared.
It caused markers with different format strings to appear as duplicates
when a markers was changed. This problem is present since
scripts/mod/modpost.c started to generate Module.markers, commit
b2e3e658b3
It therefore applies to 2.6.25, 2.6.26 and linux-next.
I merely merged the patches from Roland, Wenji and Takashi here.
Credits to
Roland McGrath <roland@redhat.com>
Wenji Huang <wenji.huang@oracle.com>
and
Takashi Nishiie <t-nishiie@np.css.fujitsu.com>
for providing the individual fixes.
- Changelog :
- Integrated Takashi's Makefile modification to clear Module.markers upon
make clean.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Roland McGrath <roland@redhat.com>
Cc: Wenji Huang <wenji.huang@oracle.com>
Cc: Takashi Nishiie <t-nishiie@np.css.fujitsu.com>
Cc: <stable@kernel.org> [2.6.25.x, 2.6.26.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Kumar Gala <galak@kernel.crashing.org> wrote:
We have a case in powerpc in which we want to link some library
routines with all module objects. The routines are intended for
handling out-of-line function call register save/restore so having
them as EXPORT_SYMBOL() is counter productive (we do also need to
link the same "library" code into the kernel).
Without this patch a powerpc build would error out and fail
to build modules with the added register save/restore module.
There were two obvious solutions:
1) To link the .o file before the modpost stage
2) To ignore the symbols in modpost
Option 1) was ruled out because we do not have any separate
linking stage for single file modules.
This patch implements option 2 - and do so only for powerpc.
The symbols we ignore are all undefined symbols named:
_restgpr_*, _savegpr_*, _rest32gpr_*, _save32gpr_*
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Disable modpost warnings for linkonce sections
My build gives lots of warnings like
WARNING: sound/core/snd.o (.gnu.linkonce.wi.mpspec_def.h.30779716): unexpected section name.
The (.[number]+) following section name are ld generated and not expected.
Did you forget to use "ax"/"aw" in a .S file?
Note that for example <linux/init.h> contains
section definitions for use in .S files.
But for .linkonce. duplicated sections are actually ok and expected.
So just disable the warning for this case.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Only modules that has other MODULE_* content
shall have the MODULE_LICENSE() tag.
This fixes allmodconfig build on my box.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
This patch adds a new command line option -E to modpost, expecting a symbol
file as an argument which is read prior to symbol processing. -E can be
supplied multiple times for as many files as is needed.
When building kernel modules that depend on other modules not in the main
kernel tree, modpost complains about undefined symbols:
# make -C /path/to/linux/kernel M=/path/to/my/module
...
Building modules, stage 2.
....
WARNING: "rt_copy_buf" [/home/rich/osc_etl_rtw/osc_kmod.ko] undefined!
...etc
This situation occurs when modpost processes the new module's symbols. When
it finds symbols not exported by the mainline kernel, it issues this warning.
The patch adds a new command line option -e to modpost which expects a symbol
file as an argument. The symbols listed in this file are added to modpost's
symbol tables during startup. -e can be supplied as often as required.
This patch works together with the second patch. It introduces a new make
variable, KBUILD_EXTRA_SYMBOLS, which is used when calling modpost.
Signed-off-by: Richard Hacker <lerichi@gmx.net>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Adrian Bunk suggested a build time check for
missing MODULE_LICENSE annotation in modules.
The build time check is fatal as we really
want this fixed for all modules.
In-tree modules should all have been fixed up by now.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Adrian Bunk <bunk@kernel.org>
The module alias support in the kernel have a consistency
check where it is checked that the size of a structure
in the kernel and on the build host are the same.
For cross builds this check does not make sense so detect
when we do cross builds and silently skip the check in these
situations.
This fixes a build bug for a wireless driver when cross building
for arm.
Acked-by: Michael Buesch <mb@bu3sch.de>
Tested-by: Gordon Farquharson <gordonfarquharson@gmail.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: stable@kernel.org
XXXINIT_TO_INIT and XXXEXIT_TO_EXIT warnings use the reversed symbol name order
in the suggestion, e.g.:
WARNING: vmlinux.o(.meminit.text+0x36c): Section mismatch in reference from the function free_area_init_core() to the function .init.text:setup_usemap()
The function __meminit free_area_init_core() references
a function __init setup_usemap().
If free_area_init_core is only used by setup_usemap then
annotate free_area_init_core with a matching annotation.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
This adds some new magic in the MODPOST phase for CONFIG_MARKERS. Analogous
to the Module.symvers file, the build will now write a Module.markers file
when CONFIG_MARKERS=y is set. This file lists the name, defining module, and
format string of each marker, separated by \t characters. This simple text
file can be used by offline build procedures for instrumentation code,
analogous to how System.map and Module.symvers can be useful to have for
kernels other than the one you are running right now.
The strings are made easy to extract by having the __trace_mark macro define
the name and format together in a single array called __mstrtab_* in the
__markers_strings section. This is straightforward and reliable as long as
the marker structs are always defined by this macro. It is an unreasonable
amount of hairy work to extract the string pointers from the __markers section
structs, which entails handling a relocation type for every machine under the
sun.
Mathieu :
- Ran through checkpatch.pl
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: David Smith <dsmith@redhat.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
modpost: Use warn() for announcing section mismatches, for easy grepping for
warnings in build logs.
Also change an existing call from fprintf() to warn() while we're at it.
Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
If we cannot determine the symbol then print
(unknown) to hint the reader that we failed to
find the symbol.
This happens with REL relocation records
in arm object files.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
We have several legitimate uses where we export symbols
annotated with one of:
__devinit, __cpuinit, __meminit and their exit counterpart.
So let's stop warning about those being exported in favour
of adding all sorts of workaround to silence the warning.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
We have too many section mismatches detected at the moment.
So silence modpost and prevent the option from being
set in a typical allyesconfig build.
Tell the user how to see all the deteils in the summary
message from modpost.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
If there is a mixture of specifying sections for code in gcc
and assembler then if the assembler code do not add
the "ax" flags the linker will see this as two different sections
and generate unique sections for each. ld does so by adding a dot
and a number.
Teach modpost to warn if a section shows up that match this
pattern - but do this only for non-debug sections.
It will result in warnings like this:
WARNING: vmlinux.o (.sched.text.1): unexpected section name.
The (.[number]+) following section name are ld generated and not expected.
Did you forget to use "ax"/"aw" in a .S file?
Note that for example <linux/init.h> contains
section definitions for use in .S files.
All warnings seen with a defconfig build for:
x86 (32+64bit) and sparc64 has been fixed (via respective maintainers).
arm, powerpc (64 bit), s390 (32 bit), ia64, alpha, sh4 checked - no
warnings seen with a defconfig build.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
If the config option CONFIG_SECTION_MISMATCH is not set and
we see a Section mismatch present the following to the user:
modpost: Found 1 section mismatch(es).
To see additional details select "Enable full Section mismatch analysis"
in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH).
If the option CONFIG_SECTION_MISMATCH is selected
then be verbose in the Section mismatch reporting from mdopost.
Sample outputs:
WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr
The function discover_ebda() references
the variable __initdata ebda_addr.
This is often because discover_ebda lacks a __initdata
annotation or the annotation of ebda_addr is wrong.
WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit()
The variable pci_serial_quirks references
the function __devexit pci_plx9050_exit()
If the reference is valid then annotate the
variable with __exit* (see linux/init.h) or name the variable:
*driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console,
WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu()
The symbol arch_register_cpu is exported and annotated __cpuinit
Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Remove the deprecated __attribute_used__.
[Introduce __section in a few places to silence checkpatch /sam]
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Refactor code so the warning report function
does nothing else than reporting warnings.
As a side effect some other code paths were cleaned
up by this.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
The typical layout is now:
WARNING: vmlinux.o(.text+0x372ec): Section mismatch: reference to .devinit.text:pci_scan_one_pbm in 'psycho_scan_bus'
This is first step towards more readable warnings.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Introducing separate sections for __dev* (HOTPLUG),
__cpu* (HOTPLUG_CPU) and __mem* (MEMORY_HOTPLUG)
allows us to do a much more reliable Section mismatch
check in modpost. We are no longer dependent on the actual
configuration of for example HOTPLUG.
This has the effect that all users see much more
Section mismatch warnings than before because they
were almost all hidden when HOTPLUG was enabled.
The advantage of this is that when building a piece
of code then it is much more likely that the Section
mismatch errors are spotted and the warnings will be
felt less random of nature.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Greg KH <greg@kroah.com>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Adrian Bunk <bunk@kernel.org>
Change the logic in modpost so we identify all the
bad combinations of sections that refer to other
sections.
Compared to the previous approach we are much less
dependent on knowledge of what additional sections
the tool chain uses and thus we can keep the false
positives low.
The implmentation is changed to use a table based
lookup and we now check all combinations in first
pass so we no longer need separate passes for init
and exit sections.
Tested that the same warnings are generated for
an allyesconfig build without CONFIG_HOTPLUG.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Adrian Bunk <bunk@kernel.org>
Split a too long function up in smaller bits to make
prgram logic easier to follow.
A few related changes done due to parameter
changes.
No functional changes.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
The relocation record sometimes contained an address
which was not an exactly match for a symbol.
Implment some simple logic such that if there
is a symbol within 20 bytes of the address contained
in the relocation record then print the name of this
symbol.
With this change modpost could find symbol names
for the remaining .init.text symbols in my
allyesconfig build for x86_64.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
It is very convinient to say:
scripts/mod/modpost mm/built-in.o
to check if any section mismatch errors occured
in mm/ (as an example).
Fix it so this is possible again.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
akpm complained about overly long lines in modpost.c and
when started additional style issues were fixed:
o Updated my copyright
o Removed unneeded {}
o Drop assignments in if ()
o Spaces around operators
o Break long lines
o locate * near variable not type
o Fix a format specifier for sizeof()
o Corrected placement of '{' and '}'
o spaces to tabs (but use tabs only for indention)
modpost.c is not checkpatch clean. Readability were favoured
on top of checkpatch compliance.
But checkpatch were used to find additional stuff to clean up.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
When passing an file name > 1k the stack could be overflowed.
Not really a security issue, but still better plugged.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Fix wrong format strings in modpost exposed by the previous patch.
Including one missing argument -- some random data was printed instead.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
When part of build an external module tree, modpost first reads in the
kernel's and then the external tree's Module.symvers files. From these files
it establishes a symbol => module mapping. When it later reads in each module
built and processes the symbols it finds, it discovers the symbol=>module
mapping from Module.symvers and leaves it as it is.
The problem comes with a module has been re-named or a symbol has moved from
one module to another, since the Module.symvers file was generated. modpost
does not update the symbol=>module mapping when it finds the new location of
the symbol when scanning the newly built modules. This results in the module
containing incorrect dependency information and the new Module.symvers file
written by modpost will also contain the incorrect mappings, perpetuating the
problem to the next build, and so on.
When building the out of kernel development tree for kernel subsystem, like
v4l-dvb or ALSA, deleting the external Module.symvers file before building
(which the kernel build system doesn't do and shouldn't be necessary anyway),
won't fix the problem. modpost still reads the kernel's Module.symvers, and
since we a building a kernel subsystem, it will define the same symbols as the
external modules.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Fix modpost segfault.
Before:
-------
ynezz@ntbk:~/linux-2.6.git$ scripts/mod/modpost vmlinux ath_pci.o
Segmentation fault
After:
------
ynezz@ntbk:~/linux-2.6.git$ scripts/mod/modpost vmlinux ath_pci.o
FATAL: section header offset=815726848 in file 'ath_pci.o' is bigger then filesize=153968
Sam: This seems to warn for a binutils issue. Anyway modpost should not
segfault.
Signed-off-by: Petr Stetiar <ynezz@true.cz>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
With the net namespaces many code leaved the __init section,
thus making the kernel occupy more memory than it did before.
Since we have a config option that prohibits the namespace
creation, the functions that initialize/finalize some netns
stuff are simply not needed and can be freed after the boot.
Currently, this is almost not noticeable, since few calls
are no longer in __init, but when the namespaces will be
merged it will be possible to free more code. I propose to
use the __net_init, __net_exit and __net_initdata "attributes"
for functions/variables that are not used if the CONFIG_NET_NS
is not set to save more space in memory.
The exiting functions cannot just reside in the __exit section,
as noticed by David, since the init section will have
references on it and the compilation will fail due to modpost
checks. These references can exist, since the init namespace
never dies and the exit callbacks are never called. So I
introduce the __exit_refok attribute just like it is already
done with the __init_refok.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is needed on MIPS where the same mechanism as get_user() is used to
intercept bus error exceptions for some hardware probes. Without this
patch modpost will throw spurious warnings:
LD vmlinux
SYSMAP System.map
SYSMAP .tmp_System.map
MODPOST vmlinux
WARNING: arch/mips/sgi-ip22/built-in.o(__dbe_table+0x0): Section mismatch: reference to .init.text:
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The Xtensa architecture places literal pools in sections separate
from the instructions. The corresponsing text sections, therefore,
reference the .literal section, and we have to suppress those
warnings.
The naming convention defines the name for a literal
section as .SECTION.literal, unless .SECTION is .text. In that case
the name is only .literal. Using strncmp() instead of strcmp()
to compare the from-section with .SECTION.init.refok in pattern 0
should not cause any regressions for other architectures.
We also need to suppress warnings for two informational
sections (.xt.lit and .xt.prop) used by the Xtensa architecture.
Signed-off-by: Chris Zankel <chris@zankel.net>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
In the whitelist function of modpost now use the same
check to identify init_section as in other places of modpost.
This has the effect that we now recognize sections named
.init.text.19 as init sections and we no longer warn
when we see these.
At the same time make surrounding code readable by dropping
use of temporary flags.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
We already check and warn about section mismatches from vmlinux
(build as vmlinux.o) during first pass so skip the checks
during the 2nd pass where we process modules.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
This is needed on MIPS where the same mechanism as get_user() is used to
intercept bus error exceptions for some hardware probes. Without this
patch modpost will throw spurious warnings:
LD vmlinux
SYSMAP System.map
SYSMAP .tmp_System.map
MODPOST vmlinux
WARNING: arch/mips/sgi-ip22/built-in.o(__dbe_table+0x0): Section mismatch: reference to .init.text:
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
gcc puts data into .data.rel or .data.rel.* on some architectures (e.g.
ia64) or under certain conditions, so whatever is legal relative to
.data should also be legal for those other sections. Fixes a few
modpost warnings on ia64.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Previously we did do the check on the .o files used to link
vmlinux but that failed to find questionable references across
the .o files.
Create a dedicated vmlinux.o file used only for section mismatch checks
that uses the defualt linker script so section does not get renamed.
The vmlinux.o may later be used as part of the the final link of vmlinux
but for now it is used fo section mismatch only.
For a defconfig build this is instant but for an allyesconfig this
add two minutes to a full build (that anyways takes ~2 hours).
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
arm uses a lot of ops structures named *_timer that has legitimite
references to .init.text.
So let's add this variable to the list of variables that may reference
.init.text without causing any warning.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Replaced this with a __init_refok marker
in front of fb_find_logo().
I think that the __initdata marker for the logo's are
wrong but I have not justified this so I did not remove it.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
The .exit.text section may be discarded either at build or at runtime.
So let modpost warn if this situation is detected.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Move more checks from whitelist to the section check functions.
Remove the redundent pci_fixup check.
Renumber the patterns.
No functional changes.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
There were a great deal of overlap between the two functions
that check which sections may reference .init.text and .exit.text.
Factor out common check to a separate function and
sort entries in the original functions.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
.note* sections are ELF notes, which are typically used by external
tools to examine the kernel image. Since this is removed from any
runtime consideration, it's OK to reference any section from a .note*
section.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
The .paravirtprobe section is obsolete, so modpost doesn't need to handle it.
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
We should do better here by effetively "dereferencing" references to
the .toc (or the .got2) section, but that is much harder.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
