Basic types in DWARF repeat frequently and traversing the DIEs using
libdw is relatively slow. Add a simple hashtable based cache for the
processed DIEs.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The compiler may choose not to emit type information in DWARF for all
aliases, but it's possible for each alias to be exported separately.
To ensure we find type information for the aliases as well, read
{section, address} tuples from the symbol table and match symbols also
by address.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Add a basic DWARF parser, which uses libdw to traverse the debugging
information in an object file and looks for functions and variables.
In follow-up patches, this will be expanded to produce symbol versions
for CONFIG_MODVERSIONS from DWARF.
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Currently, 'unsigned long' is used for intermediate variables when
calculating CRCs.
The size of 'long' differs depending on the architecture: it is 32 bits
on 32-bit architectures and 64 bits on 64-bit architectures.
The CRC values generated by genksyms represent the compatibility of
exported symbols. Therefore, reproducibility is important. In other
words, we need to ensure that the output is the same when the kernel
source is identical, regardless of whether genksyms is running on a
32-bit or 64-bit build machine.
Fortunately, the output from genksyms is not affected by the build
machine's architecture because only the lower 32 bits of the
'unsigned long' variables are used.
To make it even clearer that the CRC calculation is independent of
the build machine's architecture, this commit explicitly uses the
fixed-width type, uint32_t.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
free_list() must be called before returning from this for-loop.
Swap 'break' and the combination of free_list() and 'return'.
This reduces the code and minimizes the risk of introducing memory
leaks in future changes.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
To improve readability, reduce the indentation as follows:
- Use 'continue' earlier when the symbol does not match
- flip !sym->is_declared to flatten the if-else chain
No functional changes are intended.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
When a symbol that is already registered is read again from *.symref
file, __add_symbol() removes the previous one from the hash table without
freeing it.
[Test Case]
$ cat foo.c
#include <linux/export.h>
void foo(void);
void foo(void) {}
EXPORT_SYMBOL(foo);
$ cat foo.symref
foo void foo ( void )
foo void foo ( void )
When a symbol is removed from the hash table, it must be freed along
with its ->name and ->defn members. However, sym->name cannot be freed
because it is sometimes shared with node->string, but not always. If
sym->name and node->string share the same memory, free(sym->name) could
lead to a double-free bug.
To resolve this issue, always assign a strdup'ed string to sym->name.
Fixes: 64e6c1e123 ("genksyms: track symbol checksum changes")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
When a symbol that is already registered is added again, __add_symbol()
returns without freeing the symbol definition, making it unreachable.
The following test cases demonstrate different memory leak points.
[Test Case 1]
Forward declaration with exactly the same definition
$ cat foo.c
#include <linux/export.h>
void foo(void);
void foo(void) {}
EXPORT_SYMBOL(foo);
[Test Case 2]
Forward declaration with a different definition (e.g. attribute)
$ cat foo.c
#include <linux/export.h>
void foo(void);
__attribute__((__section__(".ref.text"))) void foo(void) {}
EXPORT_SYMBOL(foo);
[Test Case 3]
Preserving an overridden symbol (compile with KBUILD_PRESERVE=1)
$ cat foo.c
#include <linux/export.h>
void foo(void);
void foo(void) { }
EXPORT_SYMBOL(foo);
$ cat foo.symref
override foo void foo ( int )
The memory leaks in Test Case 1 and 2 have existed since the introduction
of genksyms into the kernel tree. [1]
The memory leak in Test Case 3 was introduced by commit 5dae9a550a
("genksyms: allow to ignore symbol checksum changes").
When multiple init_declarators are reduced to an init_declarator_list,
the decl_spec must be duplicated. Otherwise, the following Test Case 4
would result in a double-free bug.
[Test Case 4]
$ cat foo.c
#include <linux/export.h>
extern int foo, bar;
int foo, bar;
EXPORT_SYMBOL(foo);
In this case, 'foo' and 'bar' share the same decl_spec, 'int'. It must
be unshared before being passed to add_symbol().
[1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=46bd1da672d66ccd8a639d3c1f8a166048cca608
Fixes: 5dae9a550a ("genksyms: allow to ignore symbol checksum changes")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
I do not think the '#' flag is useful here because adding the explicit
'0x' is clearer. Add the '0' flag to zero-pad the CRC values.
This change gives better alignment in the generated *.mod.c files.
There is no impact to the compiled modules.
[Before]
$ grep -A5 modversion_info fs/efivarfs/efivarfs.mod.c
static const struct modversion_info ____versions[]
__used __section("__versions") = {
{ 0x907d14d, "blocking_notifier_chain_register" },
{ 0x53d3b64, "simple_inode_init_ts" },
{ 0x65487097, "__x86_indirect_thunk_rax" },
{ 0x122c3a7e, "_printk" },
[After]
$ grep -A5 modversion_info fs/efivarfs/efivarfs.mod.c
static const struct modversion_info ____versions[]
__used __section("__versions") = {
{ 0x0907d14d, "blocking_notifier_chain_register" },
{ 0x053d3b64, "simple_inode_init_ts" },
{ 0x65487097, "__x86_indirect_thunk_rax" },
{ 0x122c3a7e, "_printk" },
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
A QString constructed from a character literal of length 0, i.e. "", is not
"null" for historical reasons. This does not matter here so use the preferred
method isEmpty() instead.
Also directly construct empty QString objects instead of passing in an empty
character literal that has to be parsed into an empty object first.
Signed-off-by: Rolf Eike Beer <eb@emlix.com>
Link: https://doc.qt.io/qt-6/qstring.html#distinction-between-null-and-empty-strings
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The 'cpio' command is used solely for copying header files to the
temporary directory. However, there is no strong reason to use 'cpio'
for this purpose. For example, scripts/package/install-extmod-build
uses the 'tar' command to copy files.
This commit replaces the use of 'cpio' with 'tar' because 'tar' is
already used in this script to generate kheaders_data.tar.xz anyway.
Performance-wide, there is no significant difference between 'cpio'
and 'tar'.
[Before]
$ rm -fr kheaders; mkdir kheaders
$ time sh -c '
for f in include arch/x86/include
do
find "$f" -name "*.h"
done | cpio --quiet -pd kheaders
'
real 0m0.148s
user 0m0.021s
sys 0m0.140s
[After]
$ rm -fr kheaders; mkdir kheaders
$ time sh -c '
for f in include arch/x86/include
do
find "$f" -name "*.h"
done | tar -c -f - -T - | tar -xf - -C kheaders
'
real 0m0.098s
user 0m0.024s
sys 0m0.131s
Revert commit 69ef0920bd ("Docs: Add cpio requirement to changes.rst")
because 'cpio' is not used anywhere else during the kernel build.
Please note that the built-in initramfs is created by the in-tree tool,
usr/gen_init_cpio, so it does not rely on the external 'cpio' command
at all.
Remove 'cpio' from the package build dependencies as well.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Since commit 13b25489b6 ("kbuild: change working directory to external
module directory with M="), when cross-building host programs for the
linux-headers package, the "Entering directory" and "Leaving directory"
messages appear multiple times, and each object path shown is relative
to the working directory. This makes it difficult to track which objects
are being rebuilt.
In hindsight, using the external module build (M=) was not a good idea.
This commit simplifies the script by leveraging the run-command target,
resulting in a cleaner build log again.
[Before]
$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- bindeb-pkg
[ snip ]
Rebuilding host programs with aarch64-linux-gnu-gcc...
make[5]: Entering directory '/home/masahiro/linux'
make[6]: Entering directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+'
HOSTCC scripts/kallsyms
HOSTCC scripts/sorttable
HOSTCC scripts/asn1_compiler
make[6]: Leaving directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+'
make[5]: Leaving directory '/home/masahiro/linux'
make[5]: Entering directory '/home/masahiro/linux'
make[6]: Entering directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+'
HOSTCC scripts/basic/fixdep
HOSTCC scripts/mod/modpost.o
HOSTCC scripts/mod/file2alias.o
HOSTCC scripts/mod/sumversion.o
HOSTCC scripts/mod/symsearch.o
HOSTLD scripts/mod/modpost
make[6]: Leaving directory '/home/masahiro/linux/debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+'
make[5]: Leaving directory '/home/masahiro/linux'
[After]
$ make ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- bindeb-pkg
[ snip ]
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/basic/fixdep
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/kallsyms
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/sorttable
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/asn1_compiler
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/modpost.o
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/file2alias.o
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/sumversion.o
HOSTCC debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/symsearch.o
HOSTLD debian/linux-headers-6.13.0-rc1+/usr/src/linux-headers-6.13.0-rc1+/scripts/mod/modpost
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
By passing an additional directory to run-parts, allow Debian and its
derivatives to ship maintainer scripts in /usr while at the same time
allowing the local admin to override or disable them by placing hooks of
the same name in /etc. This adds support for the mechanism described in
the UAPI Configuration Files Specification for kernel hooks. The same
idea is also used by udev, systemd or modprobe for their config files.
https://uapi-group.org/specifications/specs/configuration_files_specification/
This functionality relies on run-parts 5.21 or later. It is the
responsibility of packages installing hooks into /usr/share/kernel to
also declare a Depends: debianutils (>= 5.21).
KDEB_HOOKDIR can be used to change the list of directories that is
searched. By default, /etc/kernel and /usr/share/kernel are hook
directories. Since the list of directories in KDEB_HOOKDIR is separated
by spaces, the paths must not contain the space character themselves.
Signed-off-by: Johannes Schauer Marin Rodrigues <josch@mister-muffin.de>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The linux-image package currently includes empty hook directories
(/etc/kernel/{pre,post}{inst,rm}.d/ by default).
These directories were perhaps intended as a fail-safe in case no
hook scripts exist there.
However, they are really unnecessary because the run-parts command is
already guarded by the following check:
test -d ${debhookdir}/${script}.d && run-parts ...
The only difference is that the run-parts command either runs for empty
directories (resulting in a no-op) or is skipped entirely.
The maintainer scripts will succeed without these dummy directories.
The linux-image packages from the Debian kernel do not contain
/etc/kernel/*.d/, either.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Because the `macros` crate exposes procedural macros, it must be
compiled as a dynamic library (so it can be loaded by the compiler at
compile-time).
Before this change the resulting artifact was always named
`libmacros.so`, which works on hosts where this matches the naming
convention for dynamic libraries. However the proper name on macOS would
be `libmacros.dylib`.
This turns out to matter even when the dependency is passed with a path
(`--extern macros=path/to/libmacros.so` rather than `--extern macros`)
because rustc uses the file name to infer the type of the library (see
link). This is because there's no way to specify both the path to and
the type of the external library via CLI flags. The compiler could
speculatively parse the file to determine its type, but it does not do
so today.
This means that libraries that match neither rustc's naming convention
for static libraries nor the platform's naming convention for dynamic
libraries are *rejected*.
The only solution I've found is to follow the host platform's naming
convention. This patch does that by querying the compiler to determine
the appropriate name for the artifact. This allows the kernel to build
with CONFIG_RUST=y on macOS.
Link: https://github.com/rust-lang/rust/blob/d829780/compiler/rustc_metadata/src/locator.rs#L728-L752
Tested-by: Daniel Gomez <da.gomez@samsung.com>
Co-developed-by: Fiona Behrens <me@kloenk.dev>
Signed-off-by: Fiona Behrens <me@kloenk.dev>
Signed-off-by: Tamir Duberstein <tamird@gmail.com>
Tested-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://lore.kernel.org/r/20241216-b4-dylib-host-macos-v7-1-cfc507681447@gmail.com
[ Added `MAKEFLAGS=`s to avoid jobserver warnings. Removed space.
Reworded title. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Commit 4b132aacb0 ("tools: Add xdrgen") adds a tool, which uses Jinja
template files, i.e., files with the j2 file extension, for its lightweight
code generation.
These template files for this tool have proper headers with the SPDX
License information, which are included as Jinja comments by enclosing the
text with '{#' and '#}'. Sofar, the spdxcheck script does not support to
properly parse this license information in Jinja comments and it reports
back with 'Invalid token: #}'.
Parse Jinja comments properly by stripping the known Jinja comment suffix.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@redhat.com>
Link: https://lore.kernel.org/r/20250108125207.57486-1-lukas.bulwahn@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull Kbuild fixes from Masahiro Yamada:
- Fix escaping of '$' in scripts/mksysmap
- Fix a modpost crash observed with the latest binutils
- Fix 'provides' in the linux-api-headers pacman package
* tag 'kbuild-fixes-v6.13-3' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
kbuild: pacman-pkg: provide versioned linux-api-headers package
modpost: work around unaligned data access error
modpost: refactor do_vmbus_entry()
modpost: fix the missed iteration for the max bit in do_input()
scripts/mksysmap: Fix escape chars '$'
The Arch Linux glibc package contains a versioned dependency on
"linux-api-headers". If the linux-api-headers package provided by
pacman-pkg does not specify an explicit version this dependency is not
satisfied.
Fix the dependency by providing an explicit version.
Fixes: c8578539de ("kbuild: add script and target to generate pacman package")
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
The guidelines for git commit ID abbreviation are inconsistent: some
places state to use 12 characters exactly, while other places recommend
12 characters or more. The same issue is present in the checkpatch.pl
script.
E.g. Documentation/dev-tools/checkpatch.rst says:
**GIT_COMMIT_ID**
The proper way to reference a commit id is:
commit <12+ chars of sha1> ("<title line>")
However, scripts/checkpatch.pl has two different checks: one warning
check accepting 12 characters exactly:
# Check Fixes: styles is correct
Please use correct Fixes: style 'Fixes: <12 chars of sha1> (\"<title line>\")'
and a second error check accepting 12-40 characters:
# Check for git id commit length and improperly formed commit descriptions
# A correctly formed commit description is:
# commit <SHA-1 hash length 12+ chars> ("Complete commit subject")
Please use git commit description style 'commit <12+ chars of sha1>
Hence patches containing commit IDs with more than 12 characters are
flagged by checkpatch, and sometimes rejected by maintainers or
reviewers. This is becoming more important with the growth of the
repository, as git may decide to use more characters in case of local
conflicts.
Fix this by settling on at least 12 characters, in both the
documentation and in the checkpatch.pl script.
Fixes: bd17e036b4 ("checkpatch: warn for non-standard fixes tag style")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/1c244040bf6ce304656e31036e5178b4b9dfb719.1733421037.git.geert+renesas@glider.be
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
John wrote:
> kernel-doc gets confused by code like the following:
>
> /**
> * define HOMA_MIN_DEFAULT_PORT - The 16-bit port space is divided into
> * two nonoverlapping regions. Ports 1-32767 are reserved exclusively
> * for well-defined server ports. The remaining ports are used for client
> * ports; these are allocated automatically by Homa. Port 0 is reserved.
> */
> #define HOMA_MIN_DEFAULT_PORT 0x8000
>
> It seems to use the last "-" on the line (the one in "16-bit") rather
> than the first one, so it produces the following false error message:
>
> homa.h:50: warning: expecting prototype for HOMA_MIN_DEFAULT_PORT -
> The 16(). Prototype was for HOMA_MIN_DEFAULT_PORT() instead
>
> There are similar problems if there is a ":" later on the line.
The problem is the regex for the identifier, which is a greedy /.*/ that
matches everything up to the last - or : (i.e. $decl_end). Fix it by
tightening up this regex and not matching those characters as part of the
identifier.
Link: https://lore.kernel.org/all/CAGXJAmzfRzE=A94NT5ETtj3bZc-=2oLg-9E5Kjh4o_-iuw1q8g@mail.gmail.com/
Reported-by: John Ousterhout <ouster@cs.stanford.edu>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Link: https://lore.kernel.org/r/20241221222214.1969823-1-vegard.nossum@oracle.com
With the latest binutils, modpost fails with a bus error on some
architectures such as ARM and sparc64.
Since binutils commit 1f1b5e506bf0 ("bfd/ELF: restrict file alignment
for object files"), the byte offset to each section (sh_offset) in
relocatable ELF is no longer guaranteed to be aligned.
modpost parses MODULE_DEVICE_TABLE() data structures, which are usually
located in the .rodata section. If it is not properly aligned, unaligned
access errors may occur.
To address the issue, this commit imports the get_unaligned() helper
from include/linux/unaligned.h.
The get_unaligned_native() helper caters to the endianness in addition
to handling the unaligned access.
I slightly refactored do_pcmcia_entry() and do_input() to avoid writing
back to an unaligned address. (We would need the put_unaligned() helper
to do that.)
The addend_*_rel() functions need similar adjustments because the .text
sections are not aligned either.
It seems that the .symtab, .rel.* and .rela.* sections are still aligned.
Keep normal pointer access for these sections to avoid unnecessary
performance costs.
Reported-by: Paulo Pisati <paolo.pisati@canonical.com>
Reported-by: Matthias Klose <doko@debian.org>
Closes: https://sourceware.org/bugzilla/show_bug.cgi?id=32435
Reported-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Closes: https://sourceware.org/bugzilla/show_bug.cgi?id=32493
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Optimize the size of guid_name[], as it only requires 1 additional byte
for '\0' instead of 2.
Simplify the loop by incrementing the iterator by 1 instead of 2.
Remove the unnecessary TO_NATIVE() call, as the guid is represented as
a byte stream.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
This loop should iterate over the range from 'min' to 'max' inclusively.
The last interation is missed.
Fixes: 1d8f430c15 ("[PATCH] Input: add modalias support")
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Commit b18b047002 ("kbuild: change scripts/mksysmap into sed script")
changed the invocation of the script, to call sed directly without
shell.
That means, the current extra escape that was added in:
commit ec336aa831 ("scripts/mksysmap: Fix badly escaped '$'")
for the shell is not correct any more, at the moment the stack traces
for nvhe are corrupted:
[ 22.840904] kvm [190]: [<ffff80008116dd54>] __kvm_nvhe_$x.220+0x58/0x9c
[ 22.842913] kvm [190]: [<ffff8000811709bc>] __kvm_nvhe_$x.9+0x44/0x50
[ 22.844112] kvm [190]: [<ffff80008116f8fc>] __kvm_nvhe___skip_pauth_save+0x4/0x4
With this patch:
[ 25.793513] kvm [192]: nVHE call trace:
[ 25.794141] kvm [192]: [<ffff80008116dd54>] __kvm_nvhe_hyp_panic+0xb0/0xf4
[ 25.796590] kvm [192]: [<ffff8000811709bc>] __kvm_nvhe_handle_trap+0xe4/0x188
[ 25.797553] kvm [192]: [<ffff80008116f8fc>] __kvm_nvhe___skip_pauth_save+0x4/0x4
Fixes: b18b047002 ("kbuild: change scripts/mksysmap into sed script")
Signed-off-by: Mostafa Saleh <smostafa@google.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Pull Kbuild fixes from Masahiro Yamada:
- Remove stale code in usr/include/headers_check.pl
- Fix issues in the user-mode-linux Debian package
- Fix false-positive "export twice" errors in modpost
* tag 'kbuild-fixes-v6.13-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
modpost: distinguish same module paths from different dump files
kbuild: deb-pkg: Do not install maint scripts for arch 'um'
kbuild: deb-pkg: add debarch for ARCH=um
kbuild: Drop support for include/asm-<arch> in headers_check.pl
