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.
Linus stated negative opinions about this slowness in commits:
- 5cf0fd591f ("Kbuild: disable TRIM_UNUSED_KSYMS option")
- a555bdd0c5 ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding")
We can do this better now. The final data structures of EXPORT_SYMBOL
are generated by the modpost stage, so modpost can selectively emit
KSYMTAB entries that are really used by modules.
Commit f73edc8951 ("kbuild: unify two modpost invocations") is another
ground-work to do this in a one-pass algorithm. With the list of modules,
modpost sets sym->used if it is used by a module. modpost emits KSYMTAB
only for symbols with sym->used==true.
BTW, Nicolas explained why the trimming was implemented with recursion:
https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/
Actually, we never achieved that level of optimization where the chain
reaction of trimming comes into play because:
- CONFIG_LTO_CLANG cannot remove any unused symbols
- CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux,
but not modules
If deeper trimming is required, we need to revisit this, but I guess
that is unlikely to happen.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
fixdep is designed only for parsing text files. read_file() appends
a terminating null byte ('\0') and parse_config_file() calls strstr()
to search for CONFIG options.
rustc outputs *.rlib, *.rmeta, *.so to dep-info. fixdep needs them in
the dependency, but there is no point in parsing such binary files.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
The dep files (*.d files) emitted by C compilers usually contain the
deduplicated list of included files.
One exceptional case is when a header is included by the -include
command line option, and also by #include directive.
For example, the top Makefile adds the command line option,
"-include $(srctree)/include/linux/kconfig.h". You do not need to
include <linux/kconfig.h> in every source file.
In fact, include/linux/kconfig.h is listed twice in many .*.cmd files
due to include/linux/xarray.h having "#include <linux/kconfig.h>".
I did not fix that since it is a small redundancy.
However, this is more annoying for rustc. rustc emits the dependency
for each emission type.
For example, cmd_rustc_library emits dep-info, obj, and metadata.
So, the emitted *.d file contains the dependency for those 3 targets,
which makes fixdep parse the same file 3 times.
$ grep rust/alloc/raw_vec.rs rust/.alloc.o.cmd
rust/alloc/raw_vec.rs \
rust/alloc/raw_vec.rs \
rust/alloc/raw_vec.rs \
To skip the second parsing, this commit adds a hash table for parsed
files, just like we did for CONFIG options.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
Reviewed-by: Vincenzo Palazzo <vincenzopalazzodev@gmail.com>
Change the hash table code so it will be easier to add the second table.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
fixdep parses dependency files (*.d) emitted by the compiler.
*.d files are Makefiles describing the dependencies of the main source
file.
fixdep understands minimal Makefile syntax. It works well enough for
GCC and Clang, but not for rustc.
This commit improves the parser a little more for better processing
comments, escape sequences, etc.
My main motivation is to drop comments. rustc may output comments
(e.g. env-dep). Currentyly, rustc build rules invoke sed to remove
comments, but it is more efficient to do it in fixdep.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
The cmd-check macro compares $(cmd_$@) and $(cmd_$1), but a pitfall is
that you cannot use cmd_<target> as the variable name for the command.
For example, the following code will not work in the top Makefile
or ./Kbuild.
quiet_cmd_foo = GEN $@
cmd_foo = touch $@
targets += foo
foo: FORCE
$(call if_changed,foo)
In this case, both $@ and $1 are expanded to 'foo', so $(cmd_check)
is always empty.
We do not need to use the same prefix for cmd_$@ and cmd_$1.
Rename the former to savedcmd_$@.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
This is unneeded since commit 69304379ff ("fixdep: use fflush() and
ferror() to ensure successful write to files").
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
To enable Clang randstruct support, move the structure layout
randomization seed generation out of scripts/gcc-plugins/ into
scripts/basic/ so it happens early enough that it can be used by either
compiler implementation. The gcc-plugin still builds its own header file,
but now does so from the common "randstruct.seed" file.
Cc: linux-hardening@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20220503205503.3054173-6-keescook@chromium.org
Currently, fixdep checks the return value from (v)printf(), but it does
not ensure the complete write to the .cmd file.
printf() just writes data to the internal buffer, which usually succeeds.
(Of course, it may fail for another reason, for example when the file
descriptor is closed, but that is another story.)
When the buffer (4k?) is full, an actual write occurs, and printf() may
really fail. One of typical cases is "No space left on device" when the
disk is full.
The data remaining in the buffer will be pushed out to the file when
the program exits, but we never know if it is successful.
One straight-forward fix would be to add the following code at the end
of the program.
ret = fflush(stdout);
if (ret < 0) {
/* error handling */
}
However, it is tedious to check the return code in all the call sites
of printf(), fflush(), fclose(), and whatever can cause actual writes
to the end device. Doing that lets the program bail out at the first
failure but is usually not worth the effort.
Instead, let's check the error status from ferror(). This is 'sticky',
so you need to check it just once. You still need to call fflush().
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: David Laight <david.laight@aculab.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
The pattern prefixed with '/' matches files in the same directory,
but not ones in sub-directories.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Miguel Ojeda <ojeda@kernel.org>
Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Andra Paraschiv <andraprs@amazon.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Gabriel Krisman Bertazi <krisman@collabora.com>
To build host programs, you need to add the program names to 'hostprogs'
to use the necessary build rule, but it is not enough to build them
because there is no dependency.
There are two types of host programs: built as the prerequisite of
another (e.g. gen_crc32table in lib/Makefile), or always built when
Kbuild visits the Makefile (e.g. genksyms in scripts/genksyms/Makefile).
The latter is typical in Makefiles under scripts/, which contains host
programs globally used during the kernel build. To build them, you need
to add them to both 'hostprogs' and 'always-y'.
This commit adds hostprogs-always-y as a shorthand.
The same applies to user programs. net/bpfilter/Makefile builds
bpfilter_umh on demand, hence always-y is unneeded. In contrast,
programs under samples/ are added to both 'userprogs' and 'always-y'
so they are always built when Kbuild visits the Makefiles.
userprogs-always-y works as a shorthand.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:
struct foo {
int stuff;
struct boo array[];
};
By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.
Also, notice that, dynamic memory allocations won't be affected by
this change:
"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]
sizeof(flexible-array-member) triggers a warning because flexible array
members have incomplete type[1]. There are some instances of code in
which the sizeof operator is being incorrectly/erroneously applied to
zero-length arrays and the result is zero. Such instances may be hiding
some bugs. So, this work (flexible-array member conversions) will also
help to get completely rid of those sorts of issues.
This issue was found with the help of Coccinelle.
[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 7649773293 ("cxgb3/l2t: Fix undefined behaviour")
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Here are 3 SPDX patches for 5.7-rc1.
One fixes up the SPDX tag for a single driver, while the other two go
through the tree and add SPDX tags for all of the .gitignore files as
needed.
Nothing too complex, but you will get a merge conflict with your current
tree, that should be trivial to handle (one file modified by two things,
one file deleted.)
All 3 of these have been in linux-next for a while, with no reported
issues other than the merge conflict.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-----BEGIN PGP SIGNATURE-----
iG0EABECAC0WIQT0tgzFv3jCIUoxPcsxR9QN2y37KQUCXodg5A8cZ3JlZ0Brcm9h
aC5jb20ACgkQMUfUDdst+ykySQCgy9YDrkz7nWq6v3Gohl6+lW/L+rMAnRM4uTZm
m5AuCzO3Azt9KBi7NL+L
=2Lm5
-----END PGP SIGNATURE-----
Merge tag 'spdx-5.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/spdx
Pull SPDX updates from Greg KH:
"Here are three SPDX patches for 5.7-rc1.
One fixes up the SPDX tag for a single driver, while the other two go
through the tree and add SPDX tags for all of the .gitignore files as
needed.
Nothing too complex, but you will get a merge conflict with your
current tree, that should be trivial to handle (one file modified by
two things, one file deleted.)
All three of these have been in linux-next for a while, with no
reported issues other than the merge conflict"
* tag 'spdx-5.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/spdx:
ASoC: MT6660: make spdxcheck.py happy
.gitignore: add SPDX License Identifier
.gitignore: remove too obvious comments
If *q is '\0', the condition (isalnum(*q) || *q == '_') is false anyway.
It is redundant to ensure non-zero *q.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
This is probably stale code. In old days (~ Linux 2.5.59), Kbuild made
genksyms generate include/linux/modules/*.ver files.
The currenct Kbuild does not generate *.ver files at all.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
In old days, the "host-progs" syntax was used for specifying host
programs. It was renamed to the current "hostprogs-y" in 2004.
It is typically useful in scripts/Makefile because it allows Kbuild to
selectively compile host programs based on the kernel configuration.
This commit renames like follows:
always -> always-y
hostprogs-y -> hostprogs
So, scripts/Makefile will look like this:
always-$(CONFIG_BUILD_BIN2C) += ...
always-$(CONFIG_KALLSYMS) += ...
...
hostprogs := $(always-y) $(always-m)
I think this makes more sense because a host program is always a host
program, irrespective of the kernel configuration. We want to specify
which ones to compile by CONFIG options, so always-y will be handier.
The "always", "hostprogs-y", "hostprogs-m" will be kept for backward
compatibility for a while.
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Kbuild descends into scripts/basic/ even before the Kconfig.
I do not expect any other host programs added to this Makefile.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
When there is not enough space on your storage device, the build will
fail with 'No space left on device' error message.
The reason is obvious from the message, so you will free up some disk
space, then you will resume the build.
However, sometimes you may still see a mysterious error message:
unterminated call to function 'wildcard': missing ')'.
If you run out of the disk space, fixdep may end up with generating
incomplete .*.cmd files.
For example, if the disk-full error occurs while fixdep is running
print_dep(), the .*.cmd might be truncated like this:
$(wildcard include/config/
When you run 'make' next time, this broken .*.cmd will be included,
then Make will terminate parsing since it is a wrong syntax.
Once this happens, you need to run 'make clean' or delete the broken
.*.cmd file manually.
Even if you do not see any error message, the .*.cmd files after any
error could be potentially incomplete, and unreliable. You may miss
the re-compilation due to missing header dependency.
If printf() cannot output the string for disk shortage or whatever
reason, it returns a negative value, but currently fixdep does not
check it at all. Consequently, fixdep *successfully* generates a
broken .*.cmd file. Make never notices that since fixdep exits with 0,
which means success.
Given the intended usage of fixdep, it must respect the return value
of not only malloc(), but also printf() and putchar().
This seems a long-standing issue since the introduction of fixdep.
In old days, Kbuild tried to provide an extra safety by letting fixdep
output to a temporary file and renaming it after everything is done:
scripts/basic/fixdep $(depfile) $@ '$(make-cmd)' > $(dot-target).tmp;\
rm -f $(depfile); \
mv -f $(dot-target).tmp $(dot-target).cmd)
It was no help to avoid the current issue; fixdep successfully created
a truncated tmp file, which would be renamed to a .*.cmd file.
This problem should be fixed by propagating the error status to the
build system because:
[1] Since commit 9c2af1c737 ("kbuild: add .DELETE_ON_ERROR special
target"), Make will delete the target automatically on any failure
in the recipe.
[2] Since commit 392885ee82 ("kbuild: let fixdep directly write to
.*.cmd files"), .*.cmd file is included only when the corresponding
target already exists.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Add SPDX license identifiers to all Make/Kconfig files which:
- Have no license information of any form
These files fall under the project license, GPL v2 only. The resulting SPDX
license identifier is:
GPL-2.0-only
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
My main motivation of this commit is to clean up scripts/Kbuild.include
and scripts/Makefile.build.
Currently, CONFIG_TRIM_UNUSED_KSYMS works with a tricky gimmick;
possibly exported symbols are detected by letting $(CPP) replace
EXPORT_SYMBOL* with a special string '=== __KSYM_*===', which is
post-processed by sed, and passed to fixdep. The extra preprocessing
is costly, and hacking cmd_and_fixdep is ugly.
I came up with a new way to find exported symbols; insert a dummy
symbol __ksym_marker_* to each potentially exported symbol. Those
dummy symbols are picked up by $(NM), post-processed by sed, then
appended to .*.cmd files. I collected the post-process part to a
new shell script scripts/gen_ksymdeps.sh for readability. The dummy
symbols are put into the .discard.* section so that the linker
script rips them off the final vmlinux or modules.
A nice side-effect is building with CONFIG_TRIM_UNUSED_KSYMS will
be much faster.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Nicolas Pitre <nico@linaro.org>
Commit 8370edea81 ("bin2c: move bin2c in scripts/basic") moved bin2c
to the scripts/basic/ directory, incorrectly stating "Kexec wants to
use bin2c and it wants to use it really early in the build process.
See arch/x86/purgatory/ code in later patches."
Commit bdab125c93 ("Revert "kexec/purgatory: Add clean-up for
purgatory directory"") and commit d6605b6bbe ("x86/build: Remove
unnecessary preparation for purgatory") removed the redundant
purgatory build magic entirely.
That means that the move of bin2c was unnecessary in the first place.
fixdep is the only host program that deserves to sit in the
scripts/basic/ directory.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Underscores in symbol names are translated into slashes for path names.
Filesystems treat consecutive slashes as if there was only one, so
let's do the same in the dependency list for easier grepping, etc.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
The idea of using fixdep was inspired by Kconfig, but autoksyms
belongs to a different group. So, I want to move those touched
files under include/config/ksym/ to include/ksym/.
The directory include/ksym/ can be removed by 'make clean' because
it is meaningless for the external module building.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
kconfig.h was excluded from consideration by fixdep by
6a5be57f0f (fixdep: fix extraneous dependencies) to avoid some false
positive hits
(1) include/config/.h
(2) include/config/h.h
(3) include/config/foo.h
(1) occurred because kconfig.h contains the string CONFIG_ in a
comment. However, since dee81e9886 (fixdep: faster CONFIG_ search), we
have a check that the part after CONFIG_ is non-empty, so this does not
happen anymore (and CONFIG_ appears by itself elsewhere, so that check
is worthwhile).
(2) comes from the include guard, __LINUX_KCONFIG_H. But with the
previous patch, we no longer match that either.
That leaves (3), which amounts to one [1] false dependency (aka stat() call
done by make), which I think we can live with:
We've already had one case [2] where the lack of include/linux/kconfig.h in
the .o.cmd file caused a missing rebuild, and while I originally thought
we should just put kconfig.h in the dependency list without parsing it
for the CONFIG_ pattern, we actually do have some real CONFIG_ symbols
mentioned in it, and one can imagine some translation unit that just
does '#ifdef __BIG_ENDIAN' but doesn't through some other header
actually depend on CONFIG_CPU_BIG_ENDIAN - so changing the target
endianness could end up rebuilding the world, minus that small
TU. Quoting Linus,
... when missing dependencies cause a missed re-compile, the resulting
bugs can be _really_ subtle.
[1] well, two, we now also have CONFIG_BOOGER/booger.h - we could change
that to FOO if we care
[2] https://lkml.org/lkml/2018/2/22/838
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
The string CONFIG_ quite often appears after other alphanumerics,
meaning that that instance cannot be referencing a Kconfig
symbol. Omitting these means make has fewer files to stat() when
deciding what needs to be rebuilt - for a defconfig build, this seems to
remove about 2% of the (wildcard ...) lines from the .o.cmd files.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
uml-config.h hasn't existed in this decade (87e299e5c7 - x86, um: get
rid of uml-config.h). The few remaining UML_CONFIG instances are defined
directly in terms of their real CONFIG symbol in common-offsets.h, so
unlike when the symbols got defined via a sed script, anything that uses
UML_CONFIG_FOO now should also automatically pick up a dependency on
CONFIG_FOO via the normal fixdep mechanism (since common-offsets.h
should at least recursively be a dependency). Hence I believe we should
actually be able to ignore the HELLO_CONFIG_BOOM cases.
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Richard Weinberger <richard@nod.at>
Cc: user-mode-linux-devel@lists.sourceforge.net
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
str_ends_with() tests if the given token ends with a particular string.
Currently, it is used to check file paths without $(srctree).
Actually, we have one more place where this helper is useful. Use it
to check if CONFIG option ends with _MODULE.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
parse_dep_file() has too much indentation, and puts the code far to
the right. This commit refactors the code and reduces the one level
of indentation.
strrcmp() computes 'slen' by itself, but the caller already knows the
length of the token, so 'slen' can be passed via function argument.
With this, we can swap the order of strrcmp() and "*p = \0;"
Also, strrcmp() is an ambiguous function name. Flip the logic and
rename it to str_ends_with().
I added a new helper is_ignored_file() - this returns 1 if the token
represents a file that should be ignored.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
I do not mind global variables where they are useful enough. In this
case, I do not see a good reason to use global variables since they
are just referenced in shallow places. It is easy to pass them via
function arguments.
I squashed print_cmdline() into main() since it is just one line code.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Each token in the depfile is copied to the temporary buffer 's' to
terminate the token with zero. We do not need to do this any more
because the parsed buffer is now writable. Insert '\0' directly in
the buffer without calling memcpy().
<limits.h> is no longer necessary. (It was needed for PATH_MAX).
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Now, do_config_files() and print_deps() are almost the same. Only
the difference is the parser function called (parse_config_file vs
parse_dep_file).
We can reduce the code duplication by factoring out the common code
into read_file() - this function allocates a buffer and loads a file
to it. It returns the pointer to the allocated buffer. (As before,
it bails out by exit(2) for any error.) The caller must free the
buffer when done.
Having empty source files is possible; fixdep should simply skip them.
I deleted the "st.st_size == 0" check, so read_file() allocates 1-byte
buffer for an empty file. strstr() will immediately return NULL, and
this is what we expect.
On the other hand, an empty dep_file should be treated as an error.
In this case, parse_dep_file() will error out with "no targets found"
and it is a correct error message.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Commit dee81e9886 ("fixdep: faster CONFIG_ search") changed how to
read files in which CONFIG options are searched. It used malloc()
and read() instead of mmap() because it needed to zero-terminate the
buffer in order to use strstr(). print_deps() was left untouched
since there was no reason to change it.
Now, I have two motivations to change it in the same way.
- do_config_file() and print_deps() do quite similar things; they
open a file, load it onto memory, and pass it to a parser function.
If we use malloc() and read() for print_deps() too, we can factor
out the common code. (I will do this in the next commit.)
- parse_dep_file() copies each token to a temporary buffer because
it needs to zero-terminate it to be passed to printf(). It is not
possible to modify the buffer directly because it is mmap'ed with
O_RDONLY. If we load the file content into a malloc'ed buffer, we
can insert '\0' after each token, and save memcpy(). (I will do
this in the commit after next.)
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
<arpa/inet.h> was included for ntohl(), but it was removed by
commit dee81e9886 ("fixdep: faster CONFIG_ search").
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
do_config_file() should exit with an error code on internal run-time
errors, and not return if it fails as then the error in do_config_file()
would go unnoticed in the current code and allow the build to continue.
The exit with error code will make the build fail in those very
exceptional cases. If this occurs, this actually indicates a deeper
problem in the execution of the kernel build process.
Now, in these error cases, we do not explicitly free memory and close
the file handlers in do_config_file(), as this is covered by exit().
This issue in the fixdep script was introduced with its initial
implementation back in 2002 by the original author Kai Germaschewski with
this commit 04bd72170653 ("kbuild: Make dependencies at compile time")
in the linux history git tree, i.e.,
git://git.kernel.org/pub/scm/linux/kernel/git/history/history.git.
This issue was identified during the review of a previous patch that
intended to address a memory leak detected by a static analysis tool.
Link: https://lkml.org/lkml/2017/12/14/736
Suggested-by: Nicholas Mc Guire <der.herr@hofr.at>
Suggested-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
This is a bunch of trivial fixes and cleanups.
Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Do you think kernel build is 100% dominated by gcc? You are wrong!
One small utility called "fixdep" consistently manages to sneak into
profile's first page (unless you have small monitor of course).
The choke point is this clever code:
for (; m < end; m++) {
if (*m == INT_CONF) { p = (char *) m ; goto conf; }
if (*m == INT_ONFI) { p = (char *) m-1; goto conf; }
if (*m == INT_NFIG) { p = (char *) m-2; goto conf; }
if (*m == INT_FIG_) { p = (char *) m-3; goto conf; }
4 branches per 4 characters is not fast.
Use strstr(3), so that SSE2 etc can be used.
With this patch, fixdep is so deep at the bottom, it is hard to find it.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
bin2c is used to create a valid C file out of a binary file where two
symbols will be globally defined: <name> and <name>_size. <name> is
passed as the first parameter of the host binary.
Building using goto-cc reported that the purgatory binary code (the only
current user of this utility) declares kexec_purgatory_size as 'size_t'
where bin2c generate <name>_size to be 'int' so in a 64-bit host where
sizeof(size_t) > sizeof(int) this type mismatch will always yield the
wrong value for big-endian architectures while for little-endian it will
be wrong if the object laid in memory directly after
kexec_purgatory_size contains non-zero value at the time of reading.
This commit changes <name>_size to be size_t instead.
Note:
Another way to fix the problem is to change the type of
kexec_purgatory_size to be 'int' as there's this check in code:
(kexec_purgatory_size <= 0)
Signed-off-by: Michael Tautschnig <tautschn@amazon.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
Like with kconfig options, we now have the ability to compile in and
out individual EXPORT_SYMBOL() declarations based on the content of
include/generated/autoksyms.h. However we don't want the entire
world to be rebuilt whenever that file is touched.
Let's apply the same build dependency trick used for CONFIG_* symbols
where the time stamp of empty files whose paths matching those symbols
is used to trigger fine grained rebuilds. In our case the key is the
symbol name passed to EXPORT_SYMBOL().
However, unlike config options, we cannot just use fixdep to parse
the source code for EXPORT_SYMBOL(ksym) because several variants exist
and parsing them all in a separate tool, and keeping it in synch, is
not trivially maintainable. Furthermore, there are variants such as
EXPORT_SYMBOL_GPL(pci_user_read_config_##size);
that are instanciated via a macro for which we can't easily determine
the actual exported symbol name(s) short of actually running the
preprocessor on them.
Storing the symbol name string in a special ELF section doesn't work
for targets that output assembly or preprocessed source.
So the best way is really to leverage the preprocessor by having it
output actual symbol names anchored by a special sequence that can be
easily filtered out. Then the list of symbols is simply fed to fixdep
to be merged with the other dependencies.
That implies the preprocessor is executed twice for each source file.
A previous attempt relied on a warning pragma for each EXPORT_SYMBOL()
instance that was filtered apart from stderr by the build system with
a sed script during the actual compilation pass. Unfortunately the
preprocessor/compiler diagnostic output isn't stable between versions
and this solution, although more efficient, was deemed too fragile.
Because of the lowercasing performed by fixdep, there might be name
collisions triggering spurious rebuilds for similar symbols. But this
shouldn't be a big issue in practice. (This is the case for CONFIG_*
symbols and I didn't want to be different here, whatever the original
reason for doing so.)
To avoid needless build overhead, the exported symbol name gathering is
performed only when CONFIG_TRIM_UNUSED_KSYMS is selected.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Coverity has recently added a check that will find when we don't check
the return code from fstat(2). Copy/paste the checking logic that
print_deps() has with an appropriate re-wording of the perror() message.
Signed-off-by: Tom Rini <trini@konsulko.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
strrcmp only performs read access to the memory addressed by its
arguments so make them const pointers.
Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Signed-off-by: Michal Marek <mmarek@suse.com>
The clear_config() is called just once at the beginning of this
program, but the global variable hashtab[] is already zero-filled
at the start-up.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
If the target string matches "CONFIG_", move the pointer p
forward. This saves several 7-chars adjustments.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Michal Marek <mmarek@suse.com>
currently bin2c builds only if CONFIG_IKCONFIG=y. But bin2c will now be
used by kexec too. So make it compilation dependent on CONFIG_BUILD_BIN2C
and this config option can be selected by CONFIG_KEXEC and CONFIG_IKCONFIG.
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Greg Kroah-Hartman <greg@kroah.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: WANG Chao <chaowang@redhat.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch series does not do kernel signature verification yet. I plan
to post another patch series for that. Now distributions are already
signing PE/COFF bzImage with PKCS7 signature I plan to parse and verify
those signatures.
Primary goal of this patchset is to prepare groundwork so that kernel
image can be signed and signatures be verified during kexec load. This
should help with two things.
- It should allow kexec/kdump on secureboot enabled machines.
- In general it can help even without secureboot. By being able to verify
kernel image signature in kexec, it should help with avoiding module
signing restrictions. Matthew Garret showed how to boot into a custom
kernel, modify first kernel's memory and then jump back to old kernel and
bypass any policy one wants to.
This patch (of 15):
Kexec wants to use bin2c and it wants to use it really early in the build
process. See arch/x86/purgatory/ code in later patches.
So move bin2c in scripts/basic so that it can be built very early and
be usable by arch/x86/purgatory/
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Greg Kroah-Hartman <greg@kroah.com>
Cc: Dave Young <dyoung@redhat.com>
Cc: WANG Chao <chaowang@redhat.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current use-case for fixdep is: a source file is run through a single
processing step, which creates a single dependency file as a side-effect,
which fixdep transforms into the file used by the kernel build process.
In order to transparently run the C pre-processor on device-tree files,
we wish to run both gcc -E and dtc on a source file in a single rule.
This generates two dependency files, which must be transformed together
into the file used by the kernel build process. This change modifies
fixdep so it can process the concatenation of multiple separate input
dependency files, and produce a correct unified output.
The code changes have the slight benefit of transforming the loop in
parse_dep_file() into more of a lexer/tokenizer, with the loop body being
more of a parser. Previously, some of this logic was mixed together
before the loop. I also added some comments, which I hope are useful.
Benchmarking shows that on a cross-compiled ARM tegra_defconfig build,
there is less than 0.5 seconds speed decrease with this change, on top
of a build time of ~2m24s. This is probably within the noise.
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Rob Herring <rob.herring@calxeda.com>
Masahiro Yamada