| Age | Commit message (Collapse) | Author |
|---|
|
Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing
CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way
whether the EXPORT_SYMBOL() is placed in *.c or *.S.
For further cleanups, this commit applies a similar approach to the
entire data structure of EXPORT_SYMBOL().
The EXPORT_SYMBOL() compilation is split into two stages.
When a source file is compiled, EXPORT_SYMBOL() will be converted into
a dummy symbol in the .export_symbol section.
For example,
EXPORT_SYMBOL(foo);
EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE);
will be encoded into the following assembly code:
.section ".export_symbol","a"
__export_symbol_foo:
.asciz "" /* license */
.asciz "" /* name space */
.balign 8
.quad foo /* symbol reference */
.previous
.section ".export_symbol","a"
__export_symbol_bar:
.asciz "GPL" /* license */
.asciz "BAR_NAMESPACE" /* name space */
.balign 8
.quad bar /* symbol reference */
.previous
They are mere markers to tell modpost the name, license, and namespace
of the symbols. They will be dropped from the final vmlinux and modules
because the *(.export_symbol) will go into /DISCARD/ in the linker script.
Then, modpost extracts all the information about EXPORT_SYMBOL() from the
.export_symbol section, and generates the final C code:
KSYMTAB_FUNC(foo, "", "");
KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE");
KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct
kernel_symbol that will be linked to the vmlinux or a module.
With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S
files, providing the following benefits.
[1] Deprecate EXPORT_DATA_SYMBOL()
In the old days, EXPORT_SYMBOL() was only available in C files. To export
a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file.
arch/arm/kernel/armksyms.c is one example written in the classic manner.
Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation.
Since then, EXPORT_SYMBOL() can be placed close to the symbol definition
in *.S files. It was a nice improvement.
However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL()
for data objects on some architectures.
In the new approach, modpost checks symbol's type (STT_FUNC or not),
and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly.
There are only two users of EXPORT_DATA_SYMBOL:
EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S)
EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S)
They are transformed as follows and output into .vmlinux.export.c
KSYMTAB_DATA(empty_zero_page, "_gpl", "");
KSYMTAB_DATA(ia64_ivt, "", "");
The other EXPORT_SYMBOL users in ia64 assembly are output as
KSYMTAB_FUNC().
EXPORT_DATA_SYMBOL() is now deprecated.
[2] merge <linux/export.h> and <asm-generic/export.h>
There are two similar header implementations:
include/linux/export.h for .c files
include/asm-generic/export.h for .S files
Ideally, the functionality should be consistent between them, but they
tend to diverge.
Commit 8651ec01daed ("module: add support for symbol namespaces.") did
not support the namespace for *.S files.
This commit shifts the essential implementation part to C, which supports
EXPORT_SYMBOL_NS() for *.S files.
<asm/export.h> and <asm-generic/export.h> will remain as a wrapper of
<linux/export.h> for a while.
They will be removed after #include <asm/export.h> directives are all
replaced with #include <linux/export.h>.
[3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit)
When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses
the directory tree to determine which EXPORT_SYMBOL to trim. If an
EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the
second traverse, where some source files are recompiled with their
EXPORT_SYMBOL() tuned into a no-op.
We can do this better now; modpost can selectively emit KSYMTAB entries
that are really used by modules.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
Commits ffb1b4a41016 ("x86/unwind/orc: Add 'signal' field to ORC
metadata") and fb799447ae29 ("x86,objtool: Split UNWIND_HINT_EMPTY in
two") changed the ORC format. Although ORC is internal to the kernel,
it's the only way for external tools to get reliable kernel stack traces
on x86-64. In particular, the drgn debugger [1] uses ORC for stack
unwinding, and these format changes broke it [2]. As the drgn
maintainer, I don't care how often or how much the kernel changes the
ORC format as long as I have a way to detect the change.
It suffices to store a version identifier in the vmlinux and kernel
module ELF files (to use when parsing ORC sections from ELF), and in
kernel memory (to use when parsing ORC from a core dump+symbol table).
Rather than hard-coding a version number that needs to be manually
bumped, Peterz suggested hashing the definitions from orc_types.h. If
there is a format change that isn't caught by this, the hashing script
can be updated.
This patch adds an .orc_header allocated ELF section containing the
20-byte hash to vmlinux and kernel modules, along with the corresponding
__start_orc_header and __stop_orc_header symbols in vmlinux.
1: https://github.com/osandov/drgn
2: https://github.com/osandov/drgn/issues/303
Fixes: ffb1b4a41016 ("x86/unwind/orc: Add 'signal' field to ORC metadata")
Fixes: fb799447ae29 ("x86,objtool: Split UNWIND_HINT_EMPTY in two")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Link: https://lkml.kernel.org/r/aef9c8dc43915b886a8c48509a12ec1b006ca1ca.1686690801.git.osandov@osandov.com
|
|
The next commit will use it.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
The > comparison should be >= to prevent an out of bounds array
access.
Fixes: 52dc0595d540 ("modpost: handle relocations mismatch in __ex_table.")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
"No build warning" is a strong requirement these days, so you must fix
all issues before enabling a new warning flag.
We often add a new warning to W=1 first so that the kbuild test robot
blocks new breakages.
This commit allows modpost to show extra warnings only when W=1
(or KBUILD_EXTRA_WARN=1) is given.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
For ARM, modpost fails to detect some types of section mismatches.
[test code]
.section .init.data,"aw"
bar:
.long 0
.section .data,"aw"
.globl foo
foo:
.long bar - .
It is apparently a bad reference, but modpost does not report anything.
The test code above produces the following relocations.
Relocation section '.rel.data' at offset 0xe8 contains 1 entry:
Offset Info Type Sym.Value Sym. Name
00000000 00000403 R_ARM_REL32 00000000 .init.data
Currently, R_ARM_REL32 is just skipped.
Handle it like R_ARM_ABS32.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
addend_arm_rel() processes R_ARM_THM_CALL, R_ARM_THM_JUMP24,
R_ARM_THM_JUMP19 in a wrong way.
Here, test code.
[test code for R_ARM_THM_JUMP24]
.section .init.text,"ax"
bar:
bx lr
.section .text,"ax"
.globl foo
foo:
b bar
[test code for R_ARM_THM_CALL]
.section .init.text,"ax"
bar:
bx lr
.section .text,"ax"
.globl foo
foo:
push {lr}
bl bar
pop {pc}
If you compile it with CONFIG_THUMB2_KERNEL=y, modpost will show the
symbol name, (unknown).
WARNING: modpost: vmlinux.o: section mismatch in reference: foo (section: .text) -> (unknown) (section: .init.text)
(You need to use GNU linker instead of LLD to reproduce it.)
Fix the code to make modpost show the correct symbol name. I checked
arch/arm/kernel/module.c to learn the encoding of R_ARM_THM_CALL and
R_ARM_THM_JUMP24. The module does not support R_ARM_THM_JUMP19, but
I checked its encoding in ARM ARM.
The '+4' is the compensation for pc-relative instruction. It is
documented in "ELF for the Arm Architecture" [1].
"If the relocation is pc-relative then compensation for the PC bias
(the PC value is 8 bytes ahead of the executing instruction in Arm
state and 4 bytes in Thumb state) must be encoded in the relocation
by the object producer."
[1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst
Fixes: c9698e5cd6ad ("ARM: 7964/1: Detect section mismatches in thumb relocations")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
When CONFIG_THUMB2_KERNEL is enabled, modpost fails to detect some
types of section mismatches.
[test code]
#include <linux/init.h>
int __initdata foo;
int get_foo(void) { return foo; }
It is apparently a bad reference, but modpost does not report anything.
The test code above produces the following relocations.
Relocation section '.rel.text' at offset 0x1e8 contains 2 entries:
Offset Info Type Sym.Value Sym. Name
00000000 0000052f R_ARM_THM_MOVW_AB 00000000 .LANCHOR0
00000004 00000530 R_ARM_THM_MOVT_AB 00000000 .LANCHOR0
Currently, R_ARM_THM_MOVW_ABS_NC and R_ARM_THM_MOVT_ABS are just skipped.
Add code to handle them. I checked arch/arm/kernel/module.c to learn
how the offset is encoded in the instruction.
One more thing to note for Thumb instructions - the st_value is an odd
value, so you need to mask the bit 0 to get the offset. Otherwise, you
will get an off-by-one error in the nearest symbol look-up.
It is documented in "ELF for the ARM Architecture" [1]:
In addition to the normal rules for symbol values the following rules
shall also apply to symbols of type STT_FUNC:
* If the symbol addresses an Arm instruction, its value is the
address of the instruction (in a relocatable object, the offset
of the instruction from the start of the section containing it).
* If the symbol addresses a Thumb instruction, its value is the
address of the instruction with bit zero set (in a relocatable
object, the section offset with bit zero set).
* For the purposes of relocation the value used shall be the address
of the instruction (st_value & ~1).
[1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
find_fromsym() and find_tosym() are similar - both of them iterate
in the .symtab section and return the nearest symbol.
The difference between them is that find_tosym() allows a negative
distance, but the distance must be less than 20.
Factor out the common part into find_nearest_sym().
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
For ARM defconfig (i.e. multi_v7_defconfig), modpost fails to detect
some types of section mismatches.
[test code]
#include <linux/init.h>
int __initdata foo;
int get_foo(void) { return foo; }
It is apparently a bad reference, but modpost does not report anything.
The test code above produces the following relocations.
Relocation section '.rel.text' at offset 0x200 contains 2 entries:
Offset Info Type Sym.Value Sym. Name
00000000 0000062b R_ARM_MOVW_ABS_NC 00000000 .LANCHOR0
00000004 0000062c R_ARM_MOVT_ABS 00000000 .LANCHOR0
Currently, R_ARM_MOVW_ABS_NC and R_ARM_MOVT_ABS are just skipped.
Add code to handle them. I checked arch/arm/kernel/module.c to learn
how the offset is encoded in the instruction.
The referenced symbol in relocation might be a local anchor.
If is_valid_name() returns false, let's search for a better symbol name.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
addend_arm_rel() processes R_ARM_PC24, R_ARM_CALL, R_ARM_JUMP24 in a
wrong way.
Here, test code.
[test code for R_ARM_JUMP24]
.section .init.text,"ax"
bar:
bx lr
.section .text,"ax"
.globl foo
foo:
b bar
[test code for R_ARM_CALL]
.section .init.text,"ax"
bar:
bx lr
.section .text,"ax"
.globl foo
foo:
push {lr}
bl bar
pop {pc}
If you compile it with ARM multi_v7_defconfig, modpost will show the
symbol name, (unknown).
WARNING: modpost: vmlinux.o: section mismatch in reference: foo (section: .text) -> (unknown) (section: .init.text)
(You need to use GNU linker instead of LLD to reproduce it.)
Fix the code to make modpost show the correct symbol name.
I imported (with adjustment) sign_extend32() from include/linux/bitops.h.
The '+8' is the compensation for pc-relative instruction. It is
documented in "ELF for the Arm Architecture" [1].
"If the relocation is pc-relative then compensation for the PC bias
(the PC value is 8 bytes ahead of the executing instruction in Arm
state and 4 bytes in Thumb state) must be encoded in the relocation
by the object producer."
[1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst
Fixes: 56a974fa2d59 ("kbuild: make better section mismatch reports on arm")
Fixes: 6e2e340b59d2 ("ARM: 7324/1: modpost: Fix section warnings for ARM for many compilers")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
addend_arm_rel() processes R_ARM_ABS32 in a wrong way.
Here, test code.
[test code 1]
#include <linux/init.h>
int __initdata foo;
int get_foo(void) { return foo; }
If you compile it with ARM versatile_defconfig, modpost will show the
symbol name, (unknown).
WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> (unknown) (section: .init.data)
(You need to use GNU linker instead of LLD to reproduce it.)
If you compile it for other architectures, modpost will show the correct
symbol name.
WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> foo (section: .init.data)
For R_ARM_ABS32, addend_arm_rel() sets r->r_addend to a wrong value.
I just mimicked the code in arch/arm/kernel/module.c.
However, there is more difficulty for ARM.
Here, test code.
[test code 2]
#include <linux/init.h>
int __initdata foo;
int get_foo(void) { return foo; }
int __initdata bar;
int get_bar(void) { return bar; }
With this commit applied, modpost will show the following messages
for ARM versatile_defconfig:
WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> foo (section: .init.data)
WARNING: modpost: vmlinux.o: section mismatch in reference: get_bar (section: .text) -> foo (section: .init.data)
The reference from 'get_bar' to 'foo' seems wrong.
I have no solution for this because it is true in assembly level.
In the following output, relocation at 0x1c is no longer associated
with 'bar'. The two relocation entries point to the same symbol, and
the offset to 'bar' is encoded in the instruction 'r0, [r3, #4]'.
Disassembly of section .text:
00000000 <get_foo>:
0: e59f3004 ldr r3, [pc, #4] @ c <get_foo+0xc>
4: e5930000 ldr r0, [r3]
8: e12fff1e bx lr
c: 00000000 .word 0x00000000
00000010 <get_bar>:
10: e59f3004 ldr r3, [pc, #4] @ 1c <get_bar+0xc>
14: e5930004 ldr r0, [r3, #4]
18: e12fff1e bx lr
1c: 00000000 .word 0x00000000
Relocation section '.rel.text' at offset 0x244 contains 2 entries:
Offset Info Type Sym.Value Sym. Name
0000000c 00000c02 R_ARM_ABS32 00000000 .init.data
0000001c 00000c02 R_ARM_ABS32 00000000 .init.data
When find_elf_symbol() gets into a situation where relsym->st_name is
zero, there is no guarantee to get the symbol name as written in C.
I am keeping the current logic because it is useful in many architectures,
but the symbol name is not always correct depending on the optimization.
I left some comments in find_tosym().
Fixes: 56a974fa2d59 ("kbuild: make better section mismatch reports on arm")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
Use PATTERNS() macros to remove unneeded array definitions.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
There is no distinction between TEXT_TO_ANY_EXIT and DATA_TO_ANY_EXIT.
Just merge them.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
You can merge these entries.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
This check is unneeded. Without it, sec_name() will returns the null
string "", then section_mismatch() will return immediately.
Anyway, special section indices rarely appear in these loops.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
r_offset/r_addend holds the offset address from/to which a symbol is
referenced. It is unclear unless you are familiar with ELF.
Rename them to faddr, taddr, respectively. The prefix 'f' means 'from',
't' means 'to'.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
find_tosym() takes 'sym' and stores the return value to another
variable 'to'. You can use the same variable because we want to
replace the original one when appropriate.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
secref_whitelist() does not use the argument 'mismatch'.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
This reverts commit a4d26f1a0958bb1c2b60c6f1e67c6f5d43e2647b.
The variable 'fromsym' never starts with ".L" since commit 87e5b1e8f257
("module: Sync code of is_arm_mapping_symbol()").
In other words, Pattern 6 is now dead code.
Previously, the .LANCHOR1 hid the symbols listed in Pattern 2.
87e5b1e8f257 provided a better solution.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
find_elf_symbol() and find_elf_symbol2() are not good names.
Rename them to find_tosym(), find_fromsym(), respectively.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
find_elf_symbol2() converts the section index to the section name,
then compares the two strings in each iteration. This is slow.
It is faster to compare the section indices (i.e. integers) directly.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
default_mismatch_handler() does not need to compute 'tosec' because
it is calculated by the caller.
Pass it down to default_mismatch_handler() instead of calling
sec_name() twice.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
Merging these two reduces several lines of code. The extable section
mismatch is already distinguished by EXTABLE_TO_NON_TEXT.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
SHF_EXECINSTR is a bit flag (#define SHF_EXECINSTR 0x4).
Compare the masked flag to '!= 0'.
There is no good reason to stop modpost immediately even if a special
section index is given. You will get a section mismatch error anyway.
Also, change the return type to bool.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
report_sec_mismatch() and default_mismatch_handler() are small enough
to be merged together.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
Collect relevant code into one place to clarify all the cases are
covered by 'if () ... else if ... else ...'.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
|
|
This is the last user of get_pretty_name() - it is just used to
distinguish whether the symbol is a function or not. It is not
valuable information.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
report_extable_warnings() prints "from" in a pretty form, but we know
it is always located in the __ex_table section, i.e. a collection of
struct exception_table_entry.
It is very likely to fail to get the symbol name and ends up with
meaningless message:
... in reference from the (unknown reference) (unknown) to ...
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
find_extable_entry_size() is completely broken. It has awesome comments
about how to calculate sizeof(struct exception_table_entry).
It was based on these assumptions:
- struct exception_table_entry has two fields
- both of the fields have the same size
Then, we came up with this equation:
(offset of the second field) * 2 == (size of struct)
It was true for all architectures when commit 52dc0595d540 ("modpost:
handle relocations mismatch in __ex_table.") was applied.
Our mathematics broke when commit 548acf19234d ("x86/mm: Expand the
exception table logic to allow new handling options") introduced the
third field.
Now, the definition of exception_table_entry is highly arch-dependent.
For x86, sizeof(struct exception_table_entry) is apparently 12, but
find_extable_entry_size() sets extable_entry_size to 8.
I could fix it, but I do not see much value in this code.
extable_entry_size is used just for selecting a slightly different
error message.
If the first field ("insn") references to a non-executable section,
The relocation at %s+0x%lx references
section "%s" which is not executable, IOW
it is not possible for the kernel to fault
at that address. Something is seriously wrong
and should be fixed.
If the second field ("fixup") references to a non-executable section,
The relocation at %s+0x%lx references
section "%s" which is not executable, IOW
the kernel will fault if it ever tries to
jump to it. Something is seriously wrong
and should be fixed.
Merge the two error messages rather than adding even more complexity.
Change fatal() to error() to make it continue running and catch more
possible errors.
Fixes: 548acf19234d ("x86/mm: Expand the exception table logic to allow new handling options")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
The section mismatch check relies on the relocation entries.
For REL, the addend value is implicit, so we need some code to compute
it. Currently, EM_386, EM_ARM, and EM_MIPS are supported. This commit
makes sure we covered all the cases.
I believe the other architectures use RELA, where the explicit r_addend
field exists.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux
Pull module updates from Luis Chamberlain:
"The summary of the changes for this pull requests is:
- Song Liu's new struct module_memory replacement
- Nick Alcock's MODULE_LICENSE() removal for non-modules
- My cleanups and enhancements to reduce the areas where we vmalloc
module memory for duplicates, and the respective debug code which
proves the remaining vmalloc pressure comes from userspace.
Most of the changes have been in linux-next for quite some time except
the minor fixes I made to check if a module was already loaded prior
to allocating the final module memory with vmalloc and the respective
debug code it introduces to help clarify the issue. Although the
functional change is small it is rather safe as it can only *help*
reduce vmalloc space for duplicates and is confirmed to fix a bootup
issue with over 400 CPUs with KASAN enabled. I don't expect stable
kernels to pick up that fix as the cleanups would have also had to
have been picked up. Folks on larger CPU systems with modules will
want to just upgrade if vmalloc space has been an issue on bootup.
Given the size of this request, here's some more elaborate details:
The functional change change in this pull request is the very first
patch from Song Liu which replaces the 'struct module_layout' with a
new 'struct module_memory'. The old data structure tried to put
together all types of supported module memory types in one data
structure, the new one abstracts the differences in memory types in a
module to allow each one to provide their own set of details. This
paves the way in the future so we can deal with them in a cleaner way.
If you look at changes they also provide a nice cleanup of how we
handle these different memory areas in a module. This change has been
in linux-next since before the merge window opened for v6.3 so to
provide more than a full kernel cycle of testing. It's a good thing as
quite a bit of fixes have been found for it.
Jason Baron then made dynamic debug a first class citizen module user
by using module notifier callbacks to allocate / remove module
specific dynamic debug information.
Nick Alcock has done quite a bit of work cross-tree to remove module
license tags from things which cannot possibly be module at my request
so to:
a) help him with his longer term tooling goals which require a
deterministic evaluation if a piece a symbol code could ever be
part of a module or not. But quite recently it is has been made
clear that tooling is not the only one that would benefit.
Disambiguating symbols also helps efforts such as live patching,
kprobes and BPF, but for other reasons and R&D on this area is
active with no clear solution in sight.
b) help us inch closer to the now generally accepted long term goal
of automating all the MODULE_LICENSE() tags from SPDX license tags
In so far as a) is concerned, although module license tags are a no-op
for non-modules, tools which would want create a mapping of possible
modules can only rely on the module license tag after the commit
8b41fc4454e ("kbuild: create modules.builtin without
Makefile.modbuiltin or tristate.conf").
Nick has been working on this *for years* and AFAICT I was the only
one to suggest two alternatives to this approach for tooling. The
complexity in one of my suggested approaches lies in that we'd need a
possible-obj-m and a could-be-module which would check if the object
being built is part of any kconfig build which could ever lead to it
being part of a module, and if so define a new define
-DPOSSIBLE_MODULE [0].
A more obvious yet theoretical approach I've suggested would be to
have a tristate in kconfig imply the same new -DPOSSIBLE_MODULE as
well but that means getting kconfig symbol names mapping to modules
always, and I don't think that's the case today. I am not aware of
Nick or anyone exploring either of these options. Quite recently Josh
Poimboeuf has pointed out that live patching, kprobes and BPF would
benefit from resolving some part of the disambiguation as well but for
other reasons. The function granularity KASLR (fgkaslr) patches were
mentioned but Joe Lawrence has clarified this effort has been dropped
with no clear solution in sight [1].
In the meantime removing module license tags from code which could
never be modules is welcomed for both objectives mentioned above. Some
developers have also welcomed these changes as it has helped clarify
when a module was never possible and they forgot to clean this up, and
so you'll see quite a bit of Nick's patches in other pull requests for
this merge window. I just picked up the stragglers after rc3. LWN has
good coverage on the motivation behind this work [2] and the typical
cross-tree issues he ran into along the way. The only concrete blocker
issue he ran into was that we should not remove the MODULE_LICENSE()
tags from files which have no SPDX tags yet, even if they can never be
modules. Nick ended up giving up on his efforts due to having to do
this vetting and backlash he ran into from folks who really did *not
understand* the core of the issue nor were providing any alternative /
guidance. I've gone through his changes and dropped the patches which
dropped the module license tags where an SPDX license tag was missing,
it only consisted of 11 drivers. To see if a pull request deals with a
file which lacks SPDX tags you can just use:
./scripts/spdxcheck.py -f \
$(git diff --name-only commid-id | xargs echo)
You'll see a core module file in this pull request for the above, but
that's not related to his changes. WE just need to add the SPDX
license tag for the kernel/module/kmod.c file in the future but it
demonstrates the effectiveness of the script.
Most of Nick's changes were spread out through different trees, and I
just picked up the slack after rc3 for the last kernel was out. Those
changes have been in linux-next for over two weeks.
The cleanups, debug code I added and final fix I added for modules
were motivated by David Hildenbrand's report of boot failing on a
systems with over 400 CPUs when KASAN was enabled due to running out
of virtual memory space. Although the functional change only consists
of 3 lines in the patch "module: avoid allocation if module is already
present and ready", proving that this was the best we can do on the
modules side took quite a bit of effort and new debug code.
The initial cleanups I did on the modules side of things has been in
linux-next since around rc3 of the last kernel, the actual final fix
for and debug code however have only been in linux-next for about a
week or so but I think it is worth getting that code in for this merge
window as it does help fix / prove / evaluate the issues reported with
larger number of CPUs. Userspace is not yet fixed as it is taking a
bit of time for folks to understand the crux of the issue and find a
proper resolution. Worst come to worst, I have a kludge-of-concept [3]
of how to make kernel_read*() calls for modules unique / converge
them, but I'm currently inclined to just see if userspace can fix this
instead"
Link: https://lore.kernel.org/all/Y/kXDqW+7d71C4wz@bombadil.infradead.org/ [0]
Link: https://lkml.kernel.org/r/025f2151-ce7c-5630-9b90-98742c97ac65@redhat.com [1]
Link: https://lwn.net/Articles/927569/ [2]
Link: https://lkml.kernel.org/r/20230414052840.1994456-3-mcgrof@kernel.org [3]
* tag 'modules-6.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/linux: (121 commits)
module: add debugging auto-load duplicate module support
module: stats: fix invalid_mod_bytes typo
module: remove use of uninitialized variable len
module: fix building stats for 32-bit targets
module: stats: include uapi/linux/module.h
module: avoid allocation if module is already present and ready
module: add debug stats to help identify memory pressure
module: extract patient module check into helper
modules/kmod: replace implementation with a semaphore
Change DEFINE_SEMAPHORE() to take a number argument
module: fix kmemleak annotations for non init ELF sections
module: Ignore L0 and rename is_arm_mapping_symbol()
module: Move is_arm_mapping_symbol() to module_symbol.h
module: Sync code of is_arm_mapping_symbol()
scripts/gdb: use mem instead of core_layout to get the module address
interconnect: remove module-related code
interconnect: remove MODULE_LICENSE in non-modules
zswap: remove MODULE_LICENSE in non-modules
zpool: remove MODULE_LICENSE in non-modules
x86/mm/dump_pagetables: remove MODULE_LICENSE in non-modules
...
|
|
The L0 symbol is generated when build module on LoongArch, ignore it in
modpost and when looking at module symbols, otherwise we can not see the
expected call trace.
Now is_arm_mapping_symbol() is not only for ARM, in order to reflect the
reality, rename is_arm_mapping_symbol() to is_mapping_symbol().
This is related with commit c17a2538704f ("mksysmap: Fix the mismatch of
'L0' symbols in System.map").
(1) Simple test case
[loongson@linux hello]$ cat hello.c
#include <linux/init.h>
#include <linux/module.h>
#include <linux/printk.h>
static void test_func(void)
{
pr_info("This is a test\n");
dump_stack();
}
static int __init hello_init(void)
{
pr_warn("Hello, world\n");
test_func();
return 0;
}
static void __exit hello_exit(void)
{
pr_warn("Goodbye\n");
}
module_init(hello_init);
module_exit(hello_exit);
MODULE_LICENSE("GPL");
[loongson@linux hello]$ cat Makefile
obj-m:=hello.o
ccflags-y += -g -Og
all:
make -C /lib/modules/$(shell uname -r)/build/ M=$(PWD) modules
clean:
make -C /lib/modules/$(shell uname -r)/build/ M=$(PWD) clean
(2) Test environment
system: LoongArch CLFS 5.5
https://github.com/sunhaiyong1978/CLFS-for-LoongArch/releases/tag/5.0
It needs to update grub to avoid booting error "invalid magic number".
kernel: 6.3-rc1 with loongson3_defconfig + CONFIG_DYNAMIC_FTRACE=y
(3) Test result
Without this patch:
[root@linux hello]# insmod hello.ko
[root@linux hello]# dmesg
...
Hello, world
This is a test
...
Call Trace:
[<9000000000223728>] show_stack+0x68/0x18c
[<90000000013374cc>] dump_stack_lvl+0x60/0x88
[<ffff800002050028>] L0\x01+0x20/0x2c [hello]
[<ffff800002058028>] L0\x01+0x20/0x30 [hello]
[<900000000022097c>] do_one_initcall+0x88/0x288
[<90000000002df890>] do_init_module+0x54/0x200
[<90000000002e1e18>] __do_sys_finit_module+0xc4/0x114
[<90000000013382e8>] do_syscall+0x7c/0x94
[<9000000000221e3c>] handle_syscall+0xbc/0x158
With this patch:
[root@linux hello]# insmod hello.ko
[root@linux hello]# dmesg
...
Hello, world
This is a test
...
Call Trace:
[<9000000000223728>] show_stack+0x68/0x18c
[<90000000013374cc>] dump_stack_lvl+0x60/0x88
[<ffff800002050028>] test_func+0x28/0x34 [hello]
[<ffff800002058028>] hello_init+0x28/0x38 [hello]
[<900000000022097c>] do_one_initcall+0x88/0x288
[<90000000002df890>] do_init_module+0x54/0x200
[<90000000002e1e18>] __do_sys_finit_module+0xc4/0x114
[<90000000013382e8>] do_syscall+0x7c/0x94
[<9000000000221e3c>] handle_syscall+0xbc/0x158
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Tested-by: Youling Tang <tangyouling@loongson.cn> # for LoongArch
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
|
|
In order to avoid duplicated code, move is_arm_mapping_symbol() to
include/linux/module_symbol.h, then remove is_arm_mapping_symbol()
in the other places.
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
|
|
After commit 2e3a10a1551d ("ARM: avoid ARM binutils leaking ELF local
symbols") and commit d6b732666a1b ("modpost: fix undefined behavior of
is_arm_mapping_symbol()"), many differences of is_arm_mapping_symbol()
exist in kernel/module/kallsyms.c and scripts/mod/modpost.c, just sync
the code to keep consistent.
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
|
|
modpost now reads CRCs from .*.cmd files, parsing them using strtol().
This is inconsistent with its parsing of Module.symvers and with their
definition as *unsigned* 32-bit values.
strtol() clamps values to [LONG_MIN, LONG_MAX], and when building on a
32-bit system this changes all CRCs >= 0x80000000 to be 0x7fffffff.
Change extract_crcs_for_object() to use strtoul() instead.
Cc: stable@vger.kernel.org
Fixes: f292d875d0dc ("modpost: extract symbol versions from *.cmd files")
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild
Pull Kbuild updates from Masahiro Yamada:
- Support zstd-compressed debug info
- Allow W=1 builds to detect objects shared among multiple modules
- Add srcrpm-pkg target to generate a source RPM package
- Make the -s option detection work for future GNU Make versions
- Add -Werror to KBUILD_CPPFLAGS when CONFIG_WERROR=y
- Allow W=1 builds to detect -Wundef warnings in any preprocessed files
- Raise the minimum supported version of binutils to 2.25
- Use $(intcmp ...) to compare integers if GNU Make >= 4.4 is used
- Use $(file ...) to read a file if GNU Make >= 4.2 is used
- Print error if GNU Make older than 3.82 is used
- Allow modpost to detect section mismatches with Clang LTO
- Include vmlinuz.efi into kernel tarballs for arm64 CONFIG_EFI_ZBOOT=y
* tag 'kbuild-v6.2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (29 commits)
buildtar: fix tarballs with EFI_ZBOOT enabled
modpost: Include '.text.*' in TEXT_SECTIONS
padata: Mark padata_work_init() as __ref
kbuild: ensure Make >= 3.82 is used
kbuild: refactor the prerequisites of the modpost rule
kbuild: change module.order to list *.o instead of *.ko
kbuild: use .NOTINTERMEDIATE for future GNU Make versions
kconfig: refactor Makefile to reduce process forks
kbuild: add read-file macro
kbuild: do not sort after reading modules.order
kbuild: add test-{ge,gt,le,lt} macros
Documentation: raise minimum supported version of binutils to 2.25
kbuild: add -Wundef to KBUILD_CPPFLAGS for W=1 builds
kbuild: move -Werror from KBUILD_CFLAGS to KBUILD_CPPFLAGS
kbuild: Port silent mode detection to future gnu make.
init/version.c: remove #include <generated/utsrelease.h>
firmware_loader: remove #include <generated/utsrelease.h>
modpost: Mark uuid_le type to be suitable only for MEI
kbuild: add ability to make source rpm buildable using koji
kbuild: warn objects shared among multiple modules
...
|
|
Commit 6c730bfc894f ("modpost: handle -ffunction-sections") added
".text.*" to the OTHER_TEXT_SECTIONS macro to fix certain section
mismatch warnings. Unfortunately, this makes it impossible for modpost
to warn about section mismatches with LTO, which implies
'-ffunction-sections', as all functions are put in their own
'.text.<func_name>' sections, which may still reference functions in
sections they are not supposed to, such as __init.
Fix this by moving ".text.*" into TEXT_SECTIONS, so that configurations
with '-ffunction-sections' will see warnings about mismatched sections.
Link: https://lore.kernel.org/Y39kI3MOtVI5BAnV@google.com/
Reported-by: Vincent Donnefort <vdonnefort@google.com>
Reviewed-and-tested-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Tested-by: Vincent Donnefort <vdonnefort@google.com>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
scripts/Makefile.build replaces the suffix .o with .ko, then
scripts/Makefile.modpost calls the sed command to change .ko back
to the original .o suffix.
Instead of converting the suffixes back-and-forth, store the .o paths
in modules.order, and replace it with .ko in 'make modules_install'.
This avoids the unneeded sed command.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
|
|
Similar to other architectures such as arm64, x86, riscv and so on, use
offsets relative to the exception table entry values rather than their
absolute addresses for both the exception location and the fixup.
However, LoongArch label difference because it will actually produce two
relocations, a pair of R_LARCH_ADD32 and R_LARCH_SUB32. Take simple code
below for example:
$ cat test_ex_table.S
.section .text
1:
nop
.section __ex_table,"a"
.balign 4
.long (1b - .)
.previous
$ loongarch64-unknown-linux-gnu-gcc -c test_ex_table.S
$ loongarch64-unknown-linux-gnu-readelf -Wr test_ex_table.o
Relocation section '.rela__ex_table' at offset 0x100 contains 2 entries:
Offset Info Type Symbol's Value Symbol's Name + Addend
0000000000000000 0000000600000032 R_LARCH_ADD32 0000000000000000 .L1^B1 + 0
0000000000000000 0000000500000037 R_LARCH_SUB32 0000000000000000 L0^A + 0
The modpost will complain the R_LARCH_SUB32 relocation, so we need to
patch modpost.c to skip this relocation for .rela__ex_table section.
Signed-off-by: Youling Tang <tangyouling@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
|
|
Breaking long printed messages in multiple lines makes it very hard to
look up where they originated from.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
Since commit 7b4537199a4a ("kbuild: link symbol CRCs at final link,
removing CONFIG_MODULE_REL_CRCS"), module versioning is broken on
some architectures. Loading a module fails with "disagrees about
version of symbol module_layout".
On such architectures (e.g. ARCH=sparc build with sparc64_defconfig),
modpost shows a warning, like follows:
WARNING: modpost: EXPORT symbol "_mcount" [vmlinux] version generation failed, symbol will not be versioned.
Is "_mcount" prototyped in <asm/asm-prototypes.h>?
Previously, it was a harmless warning (CRC check was just skipped),
but now wrong CRCs are used for comparison because invalid CRCs are
just skipped.
$ sparc64-linux-gnu-nm -n vmlinux
[snip]
0000000000c2cea0 r __ksymtab__kstrtol
0000000000c2ceb8 r __ksymtab__kstrtoul
0000000000c2ced0 r __ksymtab__local_bh_enable
0000000000c2cee8 r __ksymtab__mcount
0000000000c2cf00 r __ksymtab__printk
0000000000c2cf18 r __ksymtab__raw_read_lock
0000000000c2cf30 r __ksymtab__raw_read_lock_bh
[snip]
0000000000c53b34 D __crc__kstrtol
0000000000c53b38 D __crc__kstrtoul
0000000000c53b3c D __crc__local_bh_enable
0000000000c53b40 D __crc__printk
0000000000c53b44 D __crc__raw_read_lock
0000000000c53b48 D __crc__raw_read_lock_bh
Please notice __crc__mcount is missing here.
When the module subsystem looks up a CRC that comes after, it results
in reading out a wrong address. For example, when __crc__printk is
needed, the module subsystem reads 0xc53b44 instead of 0xc53b40.
All CRC entries must be output for correct index accessing. Invalid
CRCs will be unused, but are needed to keep the one-to-one mapping
between __ksymtab_* and __crc_*.
The best is to fix all modpost warnings, but several warnings are still
remaining on less popular architectures.
Fixes: 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS")
Reported-by: matoro <matoro_mailinglist_kernel@matoro.tk>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: matoro <matoro_mailinglist_kernel@matoro.tk>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild
Pull Kbuild updates from Masahiro Yamada:
- Remove the support for -O3 (CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE_O3)
- Fix error of rpm-pkg cross-builds
- Support riscv for checkstack tool
- Re-enable -Wformwat warnings for Clang
- Clean up modpost, Makefiles, and misc scripts
* tag 'kbuild-v5.20' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (30 commits)
modpost: remove .symbol_white_list field entirely
modpost: remove unneeded .symbol_white_list initializers
modpost: add PATTERNS() helper macro
modpost: shorten warning messages in report_sec_mismatch()
Revert "Kbuild, lto, workaround: Don't warn for initcall_reference in modpost"
modpost: use more reliable way to get fromsec in section_rel(a)()
modpost: add array range check to sec_name()
modpost: refactor get_secindex()
kbuild: set EXIT trap before creating temporary directory
modpost: remove unused Elf_Sword macro
Makefile.extrawarn: re-enable -Wformat for clang
kbuild: add dtbs_prepare target
kconfig: Qt5: tell the user which packages are required
modpost: use sym_get_data() to get module device_table data
modpost: drop executable ELF support
checkstack: add riscv support for scripts/checkstack.pl
kconfig: shorten the temporary directory name for cc-option
scripts: headers_install.sh: Update config leak ignore entries
kbuild: error out if $(INSTALL_MOD_PATH) contains % or :
kbuild: error out if $(KBUILD_EXTMOD) contains % or :
...
|
|
It is not so useful to have symbol whitelists in arrays. With this
over-engineering, the code is difficult to follow.
Let's do it more directly, and collect the relevant code to one place.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
The ->symbol_white_list field is referenced in secref_whitelist(),
only when 'fromsec' is data_sections.
/* Check for pattern 2 */
if (match(tosec, init_exit_sections) &&
match(fromsec, data_sections) &&
match(fromsym, mismatch->symbol_white_list))
return 0;
If .fromsec is not data sections, the .symbol_white_list member is
not used by anyone.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
This will be useful to define a NULL-terminated array inside a function
call.
Currently, string arrays passed to match() are defined in separate
places:
static const char *const init_sections[] = { ALL_INIT_SECTIONS, NULL };
static const char *const text_sections[] = { ALL_TEXT_SECTIONS, NULL };
static const char *const optim_symbols[] = { "*.constprop.*", NULL };
...
/* Check for pattern 5 */
if (match(fromsec, text_sections) &&
match(tosec, init_sections) &&
match(fromsym, optim_symbols))
return 0;
With the new helper macro, you can list the patterns directly in the
function call, like this:
/* Check for pattern 5 */
if (match(fromsec, PATTERNS(ALL_TEXT_SECTIONS)) &&
match(tosec, PATTERNS(ALL_INIT_SECTIONS)) &&
match(fromsym, PATTERNS("*.contprop.*")))
return 0;
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
Each section mismatch results in long warning messages. Too much.
Make each warning fit in one line, and remove a lot of messy code.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
This reverts commit 77ab21adae509c5540956729e2d03bc1a59bc82a.
Even after 8 years later, GCC LTO has not been upstreamed. Also, it said
"This is a workaround". If this is needed in the future, it should be
added in a proper way.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Jiri Slaby <jirislaby@kernel.org>
|
|
The section name of Rel and Rela starts with ".rel" and ".rela"
respectively (but, I do not know whether this is specification or
convention).
For example, ".rela.text" holds relocation entries applied to the
".text" section.
So, the code chops the ".rel" or ".rela" prefix to get the name of
the section to which the relocation applies.
However, I do not like to skip 4 or 5 bytes blindly because it is
potential memory overrun.
The ELF specification provides a more reliable way to do this.
- The sh_info field holds extra information, whose interpretation
depends on the section type
- If the section type is SHT_REL or SHT_RELA, the sh_info field holds
the section header index of the section to which the relocation
applies.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
|
The section index is always positive, so the argument, secindex, should
be unsigned.
Also, inserted the array range check.
If sym->st_shndx is a special section index (between SHN_LORESERVE and
SHN_HIRESERVE), there is no corresponding section header.
For example, if a symbol specifies an absolute value, sym->st_shndx is
SHN_ABS (=0xfff1).
The current users do not cause the out-of-range access of
info->sechddrs[], but it is better to avoid such a pitfall.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
|
