platform/kernel/linux-rpi.git
3 years agokconfig: omit --oldaskconfig option for 'make config'
Masahiro Yamada [Sun, 21 Feb 2021 13:03:17 +0000 (22:03 +0900)]
kconfig: omit --oldaskconfig option for 'make config'

scripts/kconfig/conf.c line 39 defines the default of input_mode as
oldaskconfig. Hence, 'make config' works in the same way even without
the --oldaskconfig option given. Note this in the help message.

This will be helpful to unify build rules in Makefile in the next
commit.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokconfig: fix 'invalid option' for help option
Masahiro Yamada [Sun, 21 Feb 2021 13:03:16 +0000 (22:03 +0900)]
kconfig: fix 'invalid option' for help option

scripts/kconfig/conf supports -? option to show the help message.
This is not wired up to Makefile, so nobody would notice this, but
it also shows 'invalid option' message.

  $ ./scripts/kconfig/conf -?
  ./scripts/kconfig/conf: invalid option -- '?'
  Usage: ./scripts/kconfig/conf [-s] [option] <kconfig-file>
  [option] is _one_ of the following:
    --listnewconfig         List new options
    --helpnewconfig         List new options and help text
    --oldaskconfig          Start a new configuration using a line-oriented program
    ...

The reason is the '?' is missing in the short option list passed to
getopt_long().

While I fixed this issue, I also changed the option '?' to 'h'.
I prefer -h (or --help, if a long option is also desired).

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokconfig: remove dead code in conf_askvalue()
Masahiro Yamada [Sun, 21 Feb 2021 09:26:23 +0000 (18:26 +0900)]
kconfig: remove dead code in conf_askvalue()

conf_askvalue() is only called for oldconfig, syncconfig, and
oldaskconfig. If it is called for other cases, it is a bug.

So, the code after the switch statement is unreachable.

Remove the dead code, and clean up the switch statement.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokconfig: clean up nested if-conditionals in check_conf()
Masahiro Yamada [Sun, 21 Feb 2021 09:26:22 +0000 (18:26 +0900)]
kconfig: clean up nested if-conditionals in check_conf()

Unify the outer two if-conditionals into one. This decreases the
indent level by one.

Also, change the if-else blocks:

    if (input_mode == listnewconfig) {
            ...
    } else if (input_mode == helpnewconfig) {
            ...
    } else {
            ...
    }

into the switch statement.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokconfig: Remove duplicate call to sym_get_string_value()
Mickaël Salaün [Mon, 15 Feb 2021 18:15:09 +0000 (19:15 +0100)]
kconfig: Remove duplicate call to sym_get_string_value()

Use the saved returned value of sym_get_string_value() instead of
calling it twice.

Signed-off-by: Mickaël Salaün <mic@linux.microsoft.com>
Link: https://lore.kernel.org/r/20210215181511.2840674-2-mic@digikod.net
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoMakefile: Remove # characters from compiler string
Nathan Chancellor [Tue, 16 Feb 2021 21:33:12 +0000 (14:33 -0700)]
Makefile: Remove # characters from compiler string

When using AMD's Optimizing C/C++ Compiler (AOCC), the build fails due
to a # character in the version string, which is interpreted as a
comment:

$ make CC=clang defconfig init/main.o
include/config/auto.conf.cmd:1374: *** invalid syntax in conditional. Stop.

$ sed -n 1374p include/config/auto.conf.cmd
ifneq "$(CC_VERSION_TEXT)" "AMD clang version 11.0.0 (CLANG: AOCC_2.3.0-Build#85 2020_11_10) (based on LLVM Mirror.Version.11.0.0)"

Remove all # characters in the version string so that the build does not
fail unexpectedly.

Link: https://github.com/ClangBuiltLinux/linux/issues/1298
Reported-by: Michael Fuckner <michael@fuckner.net>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoMakefile: reuse CC_VERSION_TEXT
Nick Desaulniers [Fri, 5 Feb 2021 22:01:25 +0000 (14:01 -0800)]
Makefile: reuse CC_VERSION_TEXT

I noticed we're invoking $(CC) via $(shell) more than once to check the
version.  Let's reuse the first string captured in $CC_VERSION_TEXT.

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
[masahiro.yamada:
CC_VERSION_TEXT is assigned by = instead of :=, so this $(shell ) is
evaluated multiple times anyway. The number of $(CC) invocations will
be still the same. Replacing 'grep' with the built-in $(findstring )
will give real performance benefit.]
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: check the minimum linker version in Kconfig
Masahiro Yamada [Tue, 16 Feb 2021 03:10:04 +0000 (12:10 +0900)]
kbuild: check the minimum linker version in Kconfig

Unify the two scripts/ld-version.sh and scripts/lld-version.sh, and
check the minimum linker version like scripts/cc-version.sh did.

I tested this script for some corner cases reported in the past:

 - GNU ld version 2.25-15.fc23
   as reported by commit 8083013fc320 ("ld-version: Fix it on Fedora")

 - GNU ld (GNU Binutils) 2.20.1.20100303
   as reported by commit 0d61ed17dd30 ("ld-version: Drop the 4th and
   5th version components")

This script show an error message if the linker is too old:

  $ make LD=ld.lld-9
    SYNC    include/config/auto.conf
  ***
  *** Linker is too old.
  ***   Your LLD version:    9.0.1
  ***   Minimum LLD version: 10.0.1
  ***
  scripts/Kconfig.include:50: Sorry, this linker is not supported.
  make[2]: *** [scripts/kconfig/Makefile:71: syncconfig] Error 1
  make[1]: *** [Makefile:600: syncconfig] Error 2
  make: *** [Makefile:708: include/config/auto.conf] Error 2

I also moved the check for gold to this script, so gold is still rejected:

  $ make LD=gold
    SYNC    include/config/auto.conf
  gold linker is not supported as it is not capable of linking the kernel proper.
  scripts/Kconfig.include:50: Sorry, this linker is not supported.
  make[2]: *** [scripts/kconfig/Makefile:71: syncconfig] Error 1
  make[1]: *** [Makefile:600: syncconfig] Error 2
  make: *** [Makefile:708: include/config/auto.conf] Error 2

Thanks to David Laight for suggesting shell script improvements.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
3 years agokbuild: remove ld-version macro
Masahiro Yamada [Tue, 16 Feb 2021 03:10:03 +0000 (12:10 +0900)]
kbuild: remove ld-version macro

There is no direct user of ld-version; you can use CONFIG_LD_VERSION
if needed.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
3 years agoscripts: add generic syscallhdr.sh
Masahiro Yamada [Tue, 16 Feb 2021 02:04:12 +0000 (11:04 +0900)]
scripts: add generic syscallhdr.sh

Most of architectures generate syscall headers at the compile time
in a similar way.

As of v5.11-rc1, 12 architectures duplicate similar shell scripts:

  $ find arch -name syscallhdr.sh | sort
  arch/alpha/kernel/syscalls/syscallhdr.sh
  arch/arm/tools/syscallhdr.sh
  arch/ia64/kernel/syscalls/syscallhdr.sh
  arch/m68k/kernel/syscalls/syscallhdr.sh
  arch/microblaze/kernel/syscalls/syscallhdr.sh
  arch/mips/kernel/syscalls/syscallhdr.sh
  arch/parisc/kernel/syscalls/syscallhdr.sh
  arch/powerpc/kernel/syscalls/syscallhdr.sh
  arch/sh/kernel/syscalls/syscallhdr.sh
  arch/sparc/kernel/syscalls/syscallhdr.sh
  arch/x86/entry/syscalls/syscallhdr.sh
  arch/xtensa/kernel/syscalls/syscallhdr.sh

My goal is to unify them into scripts/syscallhdr.sh.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoscripts: add generic syscalltbl.sh
Masahiro Yamada [Tue, 16 Feb 2021 02:04:11 +0000 (11:04 +0900)]
scripts: add generic syscalltbl.sh

Most of architectures generate syscall headers at the compile time
in a similar way.

The syscall table has the same format for all architectures. Each line
has up to 5 fields; syscall number, ABI, syscall name, native entry
point, and compat entry point. The syscall table is processed by
syscalltbl.sh script into header files.

Despite the same pattern, scripts are maintained per architecture,
which results in code duplication and bad maintainability.

As of v5.11-rc1, 12 architectures duplicate similar shell scripts:

  $ find arch -name syscalltbl.sh | sort
  arch/alpha/kernel/syscalls/syscalltbl.sh
  arch/arm/tools/syscalltbl.sh
  arch/ia64/kernel/syscalls/syscalltbl.sh
  arch/m68k/kernel/syscalls/syscalltbl.sh
  arch/microblaze/kernel/syscalls/syscalltbl.sh
  arch/mips/kernel/syscalls/syscalltbl.sh
  arch/parisc/kernel/syscalls/syscalltbl.sh
  arch/powerpc/kernel/syscalls/syscalltbl.sh
  arch/sh/kernel/syscalls/syscalltbl.sh
  arch/sparc/kernel/syscalls/syscalltbl.sh
  arch/x86/entry/syscalls/syscalltbl.sh
  arch/xtensa/kernel/syscalls/syscalltbl.sh

My goal is to unify them into scripts/syscalltbl.sh.

__SYSCALL_WITH_COMPAT should be defined as follows:

32-bit kernel:
  #define __SYSCALL_WITH_COMPAT(nr, native, compat) __SYSCALL(nr, native)

64-bit kernel:

  #define __SYSCALL_WITH_COMPAT(nr, native, compat) __SYSCALL(nr, compat)

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoarch: syscalls: remove $(srctree)/ prefix from syscall tables
Masahiro Yamada [Mon, 15 Feb 2021 00:48:23 +0000 (09:48 +0900)]
arch: syscalls: remove $(srctree)/ prefix from syscall tables

The 'syscall' variables are not directly used in the commands.
Remove the $(srctree)/ prefix because we can rely on VPATH.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoarch: syscalls: add missing FORCE and fix 'targets' to make if_changed work
Masahiro Yamada [Mon, 15 Feb 2021 00:48:22 +0000 (09:48 +0900)]
arch: syscalls: add missing FORCE and fix 'targets' to make if_changed work

The rules in these Makefiles cannot detect the command line change
because the prerequisite 'FORCE' is missing.

Adding 'FORCE' will result in the headers being rebuilt every time
because the 'targets' additions are also wrong; the file paths in
'targets' must be relative to the current Makefile.

Fix all of them so the if_changed rules work correctly.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
3 years agogen_compile_commands: prune some directories
Masahiro Yamada [Thu, 11 Feb 2021 16:11:54 +0000 (01:11 +0900)]
gen_compile_commands: prune some directories

If directories are passed to gen_compile_commands.py, os.walk() traverses
all the subdirectories to search for .cmd files, but we know some of them
are not worth traversing.

Use the 'topdown' parameter of os.walk to prune them.

Documentation about the 'topdown' option of os.walk:
  When topdown is True, the caller can modify the dirnames list
  in-place (perhaps using del or slice assignment), and walk() will
  only recurse into the subdirectories whose names remain in dirnames;
  this can be used to prune the search, impose a specific order of
  visiting, or even to inform walk() about directories the caller
  creates or renames before it resumes walk() again. Modifying
  dirnames when topdown is False has no effect on the behavior of
  the walk, because in bottom-up mode the directories in dirnames
  are generated before dirpath itself is generated.

This commit prunes four directories, .git, Documentation, include, and
tools.

The first three do not contain any C files, so skipping them makes this
script work slightly faster. My main motivation is the last one, tools/
directory.

Commit 6ca4c6d25949 ("gen_compile_commands: do not support .cmd files
under tools/ directory") stopped supporting the tools/ directory.
The current code no longer picks up .cmd files from the tools/
directory.

If you run:

  ./scripts/clang-tools/gen_compile_commands.py --log_level=INFO

then, you will see several "File ... not found" log messages.

This is expected, and I do not want to support the tools/ directory.
However, without an explicit comment "do not support tools/", somebody
might try to get it back. Clarify this.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Nathan Chancellor <nathan@kernel.org>
3 years agokbuild: simplify access to the kernel's version
Sasha Levin [Fri, 12 Feb 2021 16:29:24 +0000 (11:29 -0500)]
kbuild: simplify access to the kernel's version

Instead of storing the version in a single integer and having various
kernel (and userspace) code how it's constructed, export individual
(major, patchlevel, sublevel) components and simplify kernel code that
uses it.

This should also make it easier on userspace.

Signed-off-by: Sasha Levin <sashal@kernel.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: clamp SUBLEVEL to 255
Sasha Levin [Sat, 6 Feb 2021 03:50:32 +0000 (22:50 -0500)]
kbuild: clamp SUBLEVEL to 255

Right now if SUBLEVEL becomes larger than 255 it will overflow into the
territory of PATCHLEVEL, causing havoc in userspace that tests for
specific kernel version.

While userspace code tests for MAJOR and PATCHLEVEL, it doesn't test
SUBLEVEL at any point as ABI changes don't happen in the context of
stable tree.

Thus, to avoid overflows, simply clamp SUBLEVEL to it's maximum value in
the context of LINUX_VERSION_CODE. This does not affect "make
kernelversion" and such.

Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoKconfig: allow explicit opt in to DWARF v5
Nick Desaulniers [Fri, 5 Feb 2021 20:22:20 +0000 (12:22 -0800)]
Kconfig: allow explicit opt in to DWARF v5

DWARF v5 is the latest standard of the DWARF debug info format. GCC 11
will change the implicit default DWARF version, if left unspecified, to
DWARF v5.

Allow users of Clang and older versions of GCC that have not changed the
implicit default DWARF version to DWARF v5 to opt in. This can help
testing consumers of DWARF debug info in preparation of v5 becoming more
widespread, as well as result in significant binary size savings of the
pre-stripped vmlinux image.

DWARF5 wins significantly in terms of size when mixed with compression
(CONFIG_DEBUG_INFO_COMPRESSED).

363M    vmlinux.clang12.dwarf5.compressed
434M    vmlinux.clang12.dwarf4.compressed
439M    vmlinux.clang12.dwarf2.compressed
457M    vmlinux.clang12.dwarf5
536M    vmlinux.clang12.dwarf4
548M    vmlinux.clang12.dwarf2

515M    vmlinux.gcc10.2.dwarf5.compressed
599M    vmlinux.gcc10.2.dwarf4.compressed
624M    vmlinux.gcc10.2.dwarf2.compressed
630M    vmlinux.gcc10.2.dwarf5
765M    vmlinux.gcc10.2.dwarf4
809M    vmlinux.gcc10.2.dwarf2

Though the quality of debug info is harder to quantify; size is not a
proxy for quality.

Jakub notes:
  One thing is GCC DWARF-5 support, that is whether the compiler will
  support -gdwarf-5 flag, and that support should be there from GCC 7
  onwards.

  All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental
  option that enabled some small DWARF subset (initially only a few
  DW_LANG_* codes newly added to DWARF5 drafts).  Only GCC 7 (released
  after DWARF 5 has been finalized) started emitting DWARF5 section
  headers and got most of the DWARF5 changes in...

  Another separate thing is whether the assembler does support
  the -gdwarf-5 option (i.e. if you can compile assembler files
  with -Wa,-gdwarf-5) ... That option is about whether the assembler
  will emit DWARF5 or DWARF2 .debug_line.  It is fine to compile C sources
  with -gdwarf-5 and use DWARF2 .debug_line for assembler files if as
  doesn't support it.

Version check GCC so that we don't need to worry about the difference in
command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to
validate the DWARF Version in the assembler feature detection script.

Most issues with clang produced assembler were fixed in binutils 2.35.1,
but 2.35.2 fixed issues related to requiring the flag -Wa,-gdwarf-5
explicitly. The added shell script test checks for the latter, and is
only required when using clang without its integrated assembler, though
we use for clang regardless as we do not yet have a way to query the
assembler from Kconfig.

Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet
recognize the new additions to the DWARF debug info.

This only modifies the DWARF version emitted by the compiler, not the
assembler.

The DWARF version of a binary can be validated with:
$ llvm-dwarfdump <object file> | head -n 4 | grep version
or
$ readelf --debug-dump=info <object file> 2>/dev/null | grep Version

Parts of the tree don't reuse DEBUG_CFLAGS as they should; such cleanup
is left as a follow up.

Link: http://www.dwarfstd.org/doc/DWARF5.pdf
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1922707
Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Suggested-by: Arvind Sankar <nivedita@alum.mit.edu>
Suggested-by: Caroline Tice <cmtice@google.com>
Suggested-by: Fangrui Song <maskray@google.com>
Suggested-by: Jakub Jelinek <jakub@redhat.com>
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Suggested-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM/Clang v12.0.0-rc1 x86-64
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoKbuild: make DWARF version a choice
Nick Desaulniers [Fri, 5 Feb 2021 20:22:19 +0000 (12:22 -0800)]
Kbuild: make DWARF version a choice

Adds a default CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT which allows
the implicit default version of DWARF emitted by the toolchain to
progress over time.

Modifies CONFIG_DEBUG_INFO_DWARF4 to be a member of a choice, making it
mutually exclusive with CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT. Users
may want to select this if they are using a newer toolchain, but have
consumers of the DWARF debug info that aren't yet ready for newer DWARF
versions' debug info.

Does so in a way that's forward compatible with existing
configs, and makes adding future versions more straightforward. This
patch does not change the current behavior or selection of DWARF
version for users upgrading to kernels with this patch.

GCC since ~4.8 has defaulted to DWARF v4 implicitly, and GCC 11 has
bumped this to v5.

Remove the Kconfig help text  about DWARF v4 being larger.  It's
empirically false for the latest toolchains for x86_64 defconfig, has no
point of reference (I suspect it was DWARF v2 but that's stil
empirically false), and debug info size is not a qualatative measure.

Suggested-by: Arvind Sankar <nivedita@alum.mit.edu>
Suggested-by: Fangrui Song <maskray@google.com>
Suggested-by: Jakub Jelinek <jakub@redhat.com>
Suggested-by: Mark Wielaard <mark@klomp.org>
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Suggested-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agovmlinux.lds.h: add DWARF v5 sections
Nick Desaulniers [Fri, 5 Feb 2021 20:22:18 +0000 (12:22 -0800)]
vmlinux.lds.h: add DWARF v5 sections

We expect toolchains to produce these new debug info sections as part of
DWARF v5. Add explicit placements to prevent the linker warnings from
--orphan-section=warn.

Compilers may produce such sections with explicit -gdwarf-5, or based on
the implicit default version of DWARF when -g is used via DEBUG_INFO.
This implicit default changes over time, and has changed to DWARF v5
with GCC 11.

.debug_sup was mentioned in review, but without compilers producing it
today, let's wait to add it until it becomes necessary.

Cc: stable@vger.kernel.org
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1922707
Reported-by: Chris Murphy <lists@colorremedies.com>
Suggested-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Mark Wielaard <mark@klomp.org>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoKbuild: Make composite object searching more generic
Elliot Berman [Fri, 22 Jan 2021 19:27:17 +0000 (11:27 -0800)]
Kbuild: Make composite object searching more generic

Reduce repeated logic around expanding composite objects.

Signed-off-by: Elliot Berman <eberman@codeaurora.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: use always-y instead of extra-y
Masahiro Yamada [Wed, 20 Jan 2021 06:23:51 +0000 (15:23 +0900)]
kbuild: use always-y instead of extra-y

As commit d0e628cd817f ("kbuild: doc: clarify the difference between
extra-y and always-y") explained, extra-y should be used for listing
the prerequisites of vmlinux.

These targets are not related to vmlinux. always-y is a better fix.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Rob Herring <robh@kernel.org>
3 years agokbuild: stop removing stale <linux/version.h> file
Masahiro Yamada [Wed, 20 Jan 2021 04:04:03 +0000 (13:04 +0900)]
kbuild: stop removing stale <linux/version.h> file

Revert commit 223c24a7dba9 ("kbuild: Automatically remove stale
<linux/version.h> file").

It was more than 6 years ago. I do not expect anybody to start
git-bisect for such a big window.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: doc: remove "Objects which export symbols" section
Masahiro Yamada [Sun, 17 Jan 2021 11:51:56 +0000 (20:51 +0900)]
kbuild: doc: remove "Objects which export symbols" section

EXPORT_SYMBOL is unrelated to makefiles. No need to mention it.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agogenksyms: remove useless case DOTS
Masahiro Yamada [Fri, 15 Jan 2021 23:43:04 +0000 (08:43 +0900)]
genksyms: remove useless case DOTS

This switch statement does not list out all the cases. Since the
'default' covers all the rest, the 'DOTS' case is unneeded.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agogenksyms: remove dead code for ST_TABLE_*
Masahiro Yamada [Fri, 15 Jan 2021 23:43:03 +0000 (08:43 +0900)]
genksyms: remove dead code for ST_TABLE_*

No one sets lexstate to ST_TABLE_*. It is is very old code, and I do
not know what was the plan at that time. Let's remove the dead code.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agogenksyms: make source_file a local variable in lexer
Masahiro Yamada [Fri, 15 Jan 2021 23:43:02 +0000 (08:43 +0900)]
genksyms: make source_file a local variable in lexer

This is only used in yylex() in lex.l

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: check the minimum compiler version in Kconfig
Masahiro Yamada [Fri, 15 Jan 2021 23:35:42 +0000 (08:35 +0900)]
kbuild: check the minimum compiler version in Kconfig

Paul Gortmaker reported a regression in the GCC version check. [1]
If you use GCC 4.8, the build breaks before showing the error message
"error Sorry, your version of GCC is too old - please use 4.9 or newer."

I do not want to apply his fix-up since it implies we would not be able
to remove any cc-option test. Anyway, I admit checking the GCC version
in <linux/compiler-gcc.h> is too late.

Almost at the same time, Linus also suggested to move the compiler
version error to Kconfig time. [2]

I unified the two similar scripts, gcc-version.sh and clang-version.sh
into cc-version.sh. The old scripts invoked the compiler multiple times
(3 times for gcc-version.sh, 4 times for clang-version.sh). I refactored
the code so the new one invokes the compiler just once, and also tried
my best to use shell-builtin commands where possible.

The new script runs faster.

  $ time ./scripts/clang-version.sh clang
  120000

  real    0m0.029s
  user    0m0.012s
  sys     0m0.021s

  $ time ./scripts/cc-version.sh clang
  Clang 120000

  real    0m0.009s
  user    0m0.006s
  sys     0m0.004s

cc-version.sh also shows an error message if the compiler is too old:

  $ make defconfig CC=clang-9
  *** Default configuration is based on 'x86_64_defconfig'
  ***
  *** Compiler is too old.
  ***   Your Clang version:    9.0.1
  ***   Minimum Clang version: 10.0.1
  ***
  scripts/Kconfig.include:46: Sorry, this compiler is not supported.
  make[1]: *** [scripts/kconfig/Makefile:81: defconfig] Error 1
  make: *** [Makefile:602: defconfig] Error 2

The new script takes care of ICC because we have <linux/compiler-intel.h>
although I am not sure if building the kernel with ICC is well-supported.

[1]: https://lore.kernel.org/r/20210110190807.134996-1-paul.gortmaker@windriver.com
[2]: https://lore.kernel.org/r/CAHk-=wh-+TMHPTFo1qs-MYyK7tZh-OQovA=pP3=e06aCVp6_kA@mail.gmail.com

Fixes: 87de84c9140e ("kbuild: remove cc-option test of -Werror=date-time")
Reported-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: LD_VERSION redenomination
Masahiro Yamada [Sat, 12 Dec 2020 16:54:30 +0000 (01:54 +0900)]
kbuild: LD_VERSION redenomination

Commit ccbef1674a15 ("Kbuild, lto: add ld-version and ld-ifversion
macros") introduced scripts/ld-version.sh for GCC LTO.

At that time, this script handled 5 version fields because GCC LTO
needed the downstream binutils. (https://lkml.org/lkml/2014/4/8/272)

The code snippet from the submitted patch was as follows:

    # We need HJ Lu's Linux binutils because mainline binutils does not
    # support mixing assembler and LTO code in the same ld -r object.
    # XXX check if the gcc plugin ld is the expected one too
    # XXX some Fedora binutils should also support it. How to check for that?
    ifeq ($(call ld-ifversion,-ge,22710001,y),y)
        ...

However, GCC LTO was not merged into the mainline after all.
(https://lkml.org/lkml/2014/4/8/272)

So, the 4th and 5th fields were never used, and finally removed by
commit 0d61ed17dd30 ("ld-version: Drop the 4th and 5th version
components").

Since then, the last 4-digits returned by this script is always zeros.

Remove the meaningless last 4-digits. This makes the version format
consistent with GCC_VERSION, CLANG_VERSION, LLD_VERSION.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Acked-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
3 years agokbuild: Remove $(cc-option,-gdwarf-4) dependency from DEBUG_INFO_DWARF4
Masahiro Yamada [Tue, 1 Dec 2020 03:27:48 +0000 (12:27 +0900)]
kbuild: Remove $(cc-option,-gdwarf-4) dependency from DEBUG_INFO_DWARF4

The -gdwarf-4 flag is supported by GCC 4.5+, and also by Clang.

You can see it at https://godbolt.org/z/6ed1oW

  For gcc 4.5.3 pane,    line 37:    .value 0x4
  For clang 10.0.1 pane, line 117:   .short 4

Given Documentation/process/changes.rst stating GCC 4.9 is the minimal
version, this cc-option is unneeded.

Note
----

CONFIG_DEBUG_INFO_DWARF4 controls the DWARF version only for C files.

As you can see in the top Makefile, -gdwarf-4 is only passed to CFLAGS.

  ifdef CONFIG_DEBUG_INFO_DWARF4
  DEBUG_CFLAGS    += -gdwarf-4
  endif

This flag is used when compiling *.c files.

On the other hand, the assembler is always given -gdwarf-2.

  KBUILD_AFLAGS   += -Wa,-gdwarf-2

Hence, the debug info that comes from *.S files is always DWARF v2.
This is simply because GAS supported only -gdwarf-2 for a long time.

Recently, GAS gained the support for --gdwarf-[345] options. [1]
And, also we have Clang integrated assembler. So, the debug info
for *.S files might be improved in the future.

In my understanding, the current code is intentional, not a bug.

[1] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=31bf18645d98b4d3d7357353be840e320649a67d

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
3 years agoMakefile: use smaller dictionary size for xz module compression
Tor Vic [Wed, 16 Dec 2020 12:58:02 +0000 (13:58 +0100)]
Makefile: use smaller dictionary size for xz module compression

By default, xz without parameters uses a dictionary size of 8 MB.
However, most modules are much smaller than that.
The xz manpage states that 'increasing dictionary size usually improves
compression ratio, but a dictionary bigger than the uncompressed file
is waste of memory'.
Use a dictionary size of 2 MB for module compression, resulting in
slightly higher compression speed while still maintaining a good
compression ratio.

Signed-off-by: Tor Vic <torvic9@mailbox.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agoia64: remove generated/nr-irqs.h generation to fix build warning
Masahiro Yamada [Sat, 29 Aug 2020 05:15:24 +0000 (14:15 +0900)]
ia64: remove generated/nr-irqs.h generation to fix build warning

Randy reports the following warning when building ARCH=ia64 with
CONFIG_IA64_PALINFO=m:

../scripts/Makefile.build:68: 'arch/ia64/kernel/palinfo.ko' will not be built even though obj-m is specified.
../scripts/Makefile.build:69: You cannot use subdir-y/m to visit a module Makefile. Use obj-y/m instead.

This message is actually false-positive, and you can get palinfo.ko
correctly built. It is emitted in the archprepare stage, where Kbuild
descends into arch/ia64/kernel to generate include/generated/nr-irqs.h
instead of any kind of kernel objects.

arch/ia64/kernel/nr-irqs.c was introduced by commit 213060a4d699
("[IA64] pvops: paravirtualize NR_IRQS") to pre-calculate:

   NR_IRQS = max(IA64_NATIVE_NR_IRQS, XEN_NR_IRQS, FOO_NR_IRQS...)

Since commit d52eefb47d4e ("ia64/xen: Remove Xen support for ia64"), this
union contains just one field, making NR_IRQS and IA64_NATIVE_NR_IRQS
always match.

So, the following hard-coding now works:

  #define NR_IRQS                IA64_NATIVE_NR_IRQS

If you need to re-introduce NR_IRQS = max(...) gimmick in the future,
please try to implement it in asm-offsets.c instead of a separate file.
It will be possible because the header inclusion has been consolidated
to make asm-offsets.c independent of <asm/irqs.h>.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
3 years agoia64: remove unneeded header includes from <asm/mca.h>
Masahiro Yamada [Sat, 29 Aug 2020 05:15:23 +0000 (14:15 +0900)]
ia64: remove unneeded header includes from <asm/mca.h>

<asm/mca.h> includes too many unneeded headers.

This commit cuts off a lot of header includes.

What we need to include are:

 - <linux/percpu.h> for DECLARE_PER_CPU(u64, ia64_mca_pal_base)
 - <linux/threads.h> for NR_CPUS
 - <linux/types.h> for u8, u64, size_t, etc.
 - <asm/ptrace.h> for KERNEL_STACK_SIZE

The other header includes are actually unneeded.

<asm/mca.h> previously included 436 headers, and now it includes
only 138. I confirmed <asm/mca.h> is still self-contained.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
3 years agoia64: do not typedef struct pal_min_state_area_s
Masahiro Yamada [Sat, 29 Aug 2020 05:15:22 +0000 (14:15 +0900)]
ia64: do not typedef struct pal_min_state_area_s

Documentation/process/coding-style.rst says:

  Please don't use things like ``vps_t``.
  It's a **mistake** to use typedef for structures and pointers.

This commit converts as follows:

  struct pal_min_state_area_s  ->  struct pal_min_state_area
         pal_min_state_area_t  ->  struct pal_min_state_area

My main motivation for this is to slim down the include directives
of <asm/mca.h> in the next commit.

Currently, <asm/mca.h> is required to include <asm/pal.h> directly
or indirectly due to (pal_min_state_area_t *). Otherwise, it would
have no idea what pal_min_state_area_t is.

Replacing it with (struct pal_min_state_area *) will relax the header
dependency since it is enough to tell it is a pointer to a structure,
and to resolve the size of struct pal_min_state_area. It will make
<asm/mca.h> independent of <asm/pal.h>.

<asm/pal.h> typedef's a lot of structures, but it is trivial to
convert the others in the same way.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
3 years agoLinux 5.11-rc7
Linus Torvalds [Sun, 7 Feb 2021 21:57:38 +0000 (13:57 -0800)]
Linux 5.11-rc7

3 years agoMerge tag 'libnvdimm-fixes-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:45:26 +0000 (10:45 -0800)]
Merge tag 'libnvdimm-fixes-5.11-rc7' of git://git./linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A fix for a crash scenario that has been present since the initial
  merge, a minor regression in sysfs attribute visibility, and a fix for
  some flexible array warnings.

  The bulk of this pull is an update to the libnvdimm unit test
  infrastructure to test non-ACPI platforms. Given there is zero
  regression risk for test updates, and the tests enable validation of
  bits headed towards the next merge window, I saw no reason to hold the
  new tests back. Santosh originally submitted this before the v5.11
  window opened.

  Summary:

   - Fix a crash when sysfs accesses race 'dimm' driver probe/remove.

   - Fix a regression in 'resource' attribute visibility necessary for
     mapping badblocks and other physical address interrogations.

   - Fix some flexible array warnings

   - Expand the unit test infrastructure for non-ACPI platforms"

* tag 'libnvdimm-fixes-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  libnvdimm/dimm: Avoid race between probe and available_slots_show()
  ndtest: Add papr health related flags
  ndtest: Add nvdimm control functions
  ndtest: Add regions and mappings to the test buses
  ndtest: Add dimm attributes
  ndtest: Add dimms to the two buses
  ndtest: Add compatability string to treat it as PAPR family
  testing/nvdimm: Add test module for non-nfit platforms
  libnvdimm/namespace: Fix visibility of namespace resource attribute
  libnvdimm/pmem: Remove unused header
  ACPI: NFIT: Fix flexible_array.cocci warnings

3 years agoMerge tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping
Linus Torvalds [Sun, 7 Feb 2021 18:40:48 +0000 (10:40 -0800)]
Merge tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fix from Christoph Hellwig:
 "Fix a 32 vs 64-bit padding issue in the new benchmark code (Barry
  Song)"

* tag 'dma-mapping-5.11-2' of git://git.infradead.org/users/hch/dma-mapping:
  dma-mapping: benchmark: use u8 for reserved field in uAPI structure

3 years agoMerge tag 'irq_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:25:01 +0000 (10:25 -0800)]
Merge tag 'irq_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Borislav Petkov:

 - Prevent device managed IRQ allocation helpers from returning IRQ 0

 - A fix for MSI activation of PCI endpoints with multiple MSIs

* tag 'irq_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Prevent [devm_]irq_alloc_desc from returning irq 0
  genirq/msi: Activate Multi-MSI early when MSI_FLAG_ACTIVATE_EARLY is set

3 years agoMerge tag 'core_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:16:24 +0000 (10:16 -0800)]
Merge tag 'core_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull syscall entry fixes from Borislav Petkov:

 - For syscall user dispatch, separate prctl operation from syscall
   redirection range specification before the API has been made official
   in 5.11.

 - Ensure tasks using the generic syscall code do trap after returning
   from a syscall when single-stepping is requested.

* tag 'core_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  entry: Use different define for selector variable in SUD
  entry: Ensure trap after single-step on system call return

3 years agoMerge tag 'sched_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 18:03:43 +0000 (10:03 -0800)]
Merge tag 'sched_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Borislav Petkov:
 "Revert an attempt to not spread IRQ threads on isolated CPUs which has
  a bunch of problems"

* tag 'sched_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  Revert "lib: Restrict cpumask_local_spread to houskeeping CPUs"

3 years agoMerge tag 'timers_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 17:55:26 +0000 (09:55 -0800)]
Merge tag 'timers_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull timer fixes from Borislav Petkov:
 "Two more timers-related fixes for v5.11:

   - Use a freezable workqueue for RTC sync because the sync can happen
     at any time and trigger suspend assertion checks in the i2c
     subsystem.

   - Correct a previous RTC validation change to check only bit 6 in
     register D because some Intel machines use bits 0-5"

* tag 'timers_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  ntp: Use freezable workqueue for RTC synchronization
  rtc: mc146818: Dont test for bit 0-5 in Register D

3 years agoMerge tag 'x86_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 7 Feb 2021 17:40:47 +0000 (09:40 -0800)]
Merge tag 'x86_urgent_for_v5.11_rc7' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Borislav Petkov:
 "I hope this is the last batch of x86/urgent updates for this round:

   - Remove superfluous EFI PGD range checks which lead to those
     assertions failing with certain kernel configs and LLVM.

   - Disable setting breakpoints on facilities involved in #DB exception
     handling to avoid infinite loops.

   - Add extra serialization to non-serializing MSRs (IA32_TSC_DEADLINE
     and x2 APIC MSRs) to adhere to SDM's recommendation and avoid any
     theoretical issues.

   - Re-add the EPB MSR reading on turbostat so that it works on older
     kernels which don't have the corresponding EPB sysfs file.

   - Add Alder Lake to the list of CPUs which support split lock.

   - Fix %dr6 register handling in order to be able to set watchpoints
     with gdb again.

   - Disable CET instrumentation in the kernel so that gcc doesn't add
     ENDBR64 to kernel code and thus confuse tracing"

* tag 'x86_urgent_for_v5.11_rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/efi: Remove EFI PGD build time checks
  x86/debug: Prevent data breakpoints on cpu_dr7
  x86/debug: Prevent data breakpoints on __per_cpu_offset
  x86/apic: Add extra serialization for non-serializing MSRs
  tools/power/turbostat: Fallback to an MSR read for EPB
  x86/split_lock: Enable the split lock feature on another Alder Lake CPU
  x86/debug: Fix DR6 handling
  x86/build: Disable CET instrumentation in the kernel

3 years agoMerge tag 'kbuild-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Feb 2021 17:37:37 +0000 (09:37 -0800)]
Merge tag 'kbuild-fixes-v5.11-2' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Use the 'python3' command to invoke python scripts because some
   distributions do not provide the 'python' command any more.

 - Clean-up and update documents

 - Use pkg-config to search libcrypto

 - Fix duplicated debug flags

 - Ignore some more stubs in scripts/kallsyms.c

* tag 'kbuild-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kallsyms: fix nonconverging kallsyms table with lld
  kbuild: fix duplicated flags in DEBUG_CFLAGS
  scripts/clang-tools: switch explicitly to Python 3
  kbuild: remove PYTHON variable
  Documentation/llvm: Add a section about supported architectures
  Revert "checkpatch: add check for keyword 'boolean' in Kconfig definitions"
  scripts: use pkg-config to locate libcrypto
  kconfig: mconf: fix HOSTCC call
  doc: gcc-plugins: update gcc-plugins.rst
  kbuild: simplify GCC_PLUGINS enablement in dummy-tools/gcc
  Documentation/Kbuild: Remove references to gcc-plugin.sh
  scripts: switch explicitly to Python 3

3 years agoMerge tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sat, 6 Feb 2021 23:26:28 +0000 (15:26 -0800)]
Merge tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Three small smb3 fixes for stable"

* tag '5.11-rc6-smb3' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: report error instead of invalid when revalidating a dentry fails
  smb3: fix crediting for compounding when only one request in flight
  smb3: Fix out-of-bounds bug in SMB2_negotiate()

3 years agoMerge tag 'riscv-for-linus-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 6 Feb 2021 23:18:10 +0000 (15:18 -0800)]
Merge tag 'riscv-for-linus-5.11-rc7' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:
 "A handful of fixes for this week:

   - A fix to avoid evalating the VA twice in virt_addr_valid, which
     fixes some WARNs under DEBUG_VIRTUAL.

   - Two fixes related to STRICT_KERNEL_RWX: one that fixes some
     permissions when strict is disabled, and one to fix some alignment
     issues when strict is enabled.

   - A fix to disallow the selection of MAXPHYSMEM_2GB on RV32, which
     isn't valid any more but may still show up in some oldconfigs.

  We still have the HiFive Unleashed ethernet phy reset regression, so
  there will likely be something coming next week"

* tag 'riscv-for-linus-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  RISC-V: Define MAXPHYSMEM_1GB only for RV32
  riscv: Align on L1_CACHE_BYTES when STRICT_KERNEL_RWX
  RISC-V: Fix .init section permission update
  riscv: virt_addr_valid must check the address belongs to linear mapping

3 years agoMerge tag 'powerpc-5.11-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 6 Feb 2021 23:14:39 +0000 (15:14 -0800)]
Merge tag 'powerpc-5.11-7' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - A fix for a change we made to __kernel_sigtramp_rt64() which confused
   glibc's backtrace logic, and also changed the semantics of that
   symbol, which was arguably an ABI break.

 - A fix for a stack overwrite in our VSX instruction emulation.

 - A couple of fixes for the Makefile logic in the new C VDSO.

Thanks to Masahiro Yamada, Naveen N.  Rao, Raoni Fassina Firmino, and
Ravi Bangoria.

* tag 'powerpc-5.11-7' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/64/signal: Fix regression in __kernel_sigtramp_rt64() semantics
  powerpc/vdso64: remove meaningless vgettimeofday.o build rule
  powerpc/vdso: fix unnecessary rebuilds of vgettimeofday.o
  powerpc/sstep: Fix array out of bound warning

3 years agoMerge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm
Linus Torvalds [Sat, 6 Feb 2021 23:07:51 +0000 (15:07 -0800)]
Merge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:

 - Fix latent bug with DC21285 (Footbridge PCI bridge) configuration
   accessors that affects GCC >= 4.9.2

 - Fix misplaced tegra_uart_config in decompressor

 - Ensure signal page contents are initialised

 - Fix kexec oops

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: kexec: fix oops after TLB are invalidated
  ARM: ensure the signal page contains defined contents
  ARM: 9043/1: tegra: Fix misplaced tegra_uart_config in decompressor
  ARM: footbridge: fix dc21285 PCI configuration accessors

3 years agoMerge tag 'usb-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 6 Feb 2021 22:59:07 +0000 (14:59 -0800)]
Merge tag 'usb-5.11-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small, last-minute, USB driver fixes for 5.11-rc7

  They all resolve issues reported, or are a few new device ids for some
  drivers. They include:

   - new device ids for some usb-serial drivers

   - xhci fixes for a variety of reported problems

   - dwc3 driver bugfixes

   - dwc2 driver bugfixes

   - usblp driver bugfix

   - thunderbolt bugfix

   - few other tiny fixes

  All have been in linux-next with no reported issues"

* tag 'usb-5.11-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: dwc2: Fix endpoint direction check in ep_from_windex
  usb: dwc3: fix clock issue during resume in OTG mode
  xhci: fix bounce buffer usage for non-sg list case
  usb: host: xhci: mvebu: make USB 3.0 PHY optional for Armada 3720
  usb: xhci-mtk: break loop when find the endpoint to drop
  usb: xhci-mtk: skip dropping bandwidth of unchecked endpoints
  usb: renesas_usbhs: Clear pipe running flag in usbhs_pkt_pop()
  USB: gadget: legacy: fix an error code in eth_bind()
  thunderbolt: Fix possible NULL pointer dereference in tb_acpi_add_link()
  USB: serial: option: Adding support for Cinterion MV31
  usb: xhci-mtk: fix unreleased bandwidth data
  usb: gadget: aspeed: add missing of_node_put
  USB: usblp: don't call usb_set_interface if there's a single alt
  USB: serial: cp210x: add pid/vid for WSDA-200-USB
  USB: serial: cp210x: add new VID/PID for supporting Teraoka AD2000

3 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 6 Feb 2021 22:57:23 +0000 (14:57 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Nothing terribly interesting, just a few fixups"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: xpad - sync supported devices with fork on GitHub
  Input: ariel-pwrbutton - remove unused variable ariel_pwrbutton_id_table
  Input: goodix - add support for Goodix GT9286 chip
  dt-bindings: input: touchscreen: goodix: Add binding for GT9286 IC
  dt-bindings: input: adc-keys: clarify description
  Input: ili210x - implement pressure reporting for ILI251x
  Input: i8042 - unbreak Pegatron C15B
  Input: st1232 - wait until device is ready before reading resolution
  Input: st1232 - do not read more bytes than needed
  Input: st1232 - fix off-by-one error in resolution handling

3 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Sat, 6 Feb 2021 22:42:52 +0000 (14:42 -0800)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "One fix in drivers (lpfc) that stops an oops on resource exhaustion"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: lpfc: Fix EEH encountering oops with NVMe traffic

3 years agoMerge tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 6 Feb 2021 22:40:27 +0000 (14:40 -0800)]
Merge tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A few small regression fixes:

   - NVMe pull request from Christoph:
       - more quirks for buggy devices (Thorsten Leemhuis, Claus Stovgaard)
       - update the email address for Keith (Keith Busch)
       - fix an out of bounds access in nvmet-tcp (Sagi Grimberg)

   - Regression fix for BFQ shallow depth calculations introduced in
     this merge window (Lin)"

* tag 'block-5.11-2021-02-05' of git://git.kernel.dk/linux-block:
  nvmet-tcp: fix out-of-bounds access when receiving multiple h2cdata PDUs
  bfq-iosched: Revert "bfq: Fix computation of shallow depth"
  update the email address for Keith Bush
  nvme-pci: ignore the subsysem NQN on Phison E16
  nvme-pci: avoid the deepest sleep state on Kingston A2000 SSDs

3 years agoMerge tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 6 Feb 2021 22:37:24 +0000 (14:37 -0800)]
Merge tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "Two small fixes that should go into 5.11:

   - task_work resource drop fix (Pavel)

   - identity COW fix (Xiaoguang)"

* tag 'io_uring-5.11-2021-02-05' of git://git.kernel.dk/linux-block:
  io_uring: drop mm/files between task_work_submit
  io_uring: don't modify identity's files uncess identity is cowed

3 years agox86/efi: Remove EFI PGD build time checks
Borislav Petkov [Fri, 5 Feb 2021 10:31:31 +0000 (11:31 +0100)]
x86/efi: Remove EFI PGD build time checks

With CONFIG_X86_5LEVEL, CONFIG_UBSAN and CONFIG_UBSAN_UNSIGNED_OVERFLOW
enabled, clang fails the build with

  x86_64-linux-ld: arch/x86/platform/efi/efi_64.o: in function `efi_sync_low_kernel_mappings':
  efi_64.c:(.text+0x22c): undefined reference to `__compiletime_assert_354'

which happens due to -fsanitize=unsigned-integer-overflow being enabled:

  -fsanitize=unsigned-integer-overflow: Unsigned integer overflow, where
  the result of an unsigned integer computation cannot be represented
  in its type. Unlike signed integer overflow, this is not undefined
  behavior, but it is often unintentional. This sanitizer does not check
  for lossy implicit conversions performed before such a computation
  (see -fsanitize=implicit-conversion).

and that fires when the (intentional) EFI_VA_START/END defines overflow
an unsigned long, leading to the assertion expressions not getting
optimized away (on GCC they do)...

However, those checks are superfluous: the runtime services mapping
code already makes sure the ranges don't overshoot EFI_VA_END as the
EFI mapping range is hardcoded. On each runtime services call, it is
switched to the EFI-specific PGD and even if mappings manage to escape
that last PGD, this won't remain unnoticed for long.

So rip them out.

See https://github.com/ClangBuiltLinux/linux/issues/256 for more info.

Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: http://lkml.kernel.org/r/20210107223424.4135538-1-arnd@kernel.org
3 years agoentry: Use different define for selector variable in SUD
Gabriel Krisman Bertazi [Fri, 5 Feb 2021 18:43:21 +0000 (13:43 -0500)]
entry: Use different define for selector variable in SUD

Michael Kerrisk suggested that, from an API perspective, it is a bad
idea to share the PR_SYS_DISPATCH_ defines between the prctl operation
and the selector variable.

Therefore, define two new constants to be used by SUD's selector variable
and update the corresponding documentation and test cases.

While this changes the API syscall user dispatch has never been part of a
Linux release, it will show up for the first time in 5.11.

Suggested-by: Michael Kerrisk (man-pages) <mtk.manpages@gmail.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210205184321.2062251-1-krisman@collabora.com
3 years agoentry: Ensure trap after single-step on system call return
Gabriel Krisman Bertazi [Wed, 3 Feb 2021 18:00:48 +0000 (13:00 -0500)]
entry: Ensure trap after single-step on system call return

Commit 299155244770 ("entry: Drop usage of TIF flags in the generic syscall
code") introduced a bug on architectures using the generic syscall entry
code, in which processes stopped by PTRACE_SYSCALL do not trap on syscall
return after receiving a TIF_SINGLESTEP.

The reason is that the meaning of TIF_SINGLESTEP flag is overloaded to
cause the trap after a system call is executed, but since the above commit,
the syscall call handler only checks for the SYSCALL_WORK flags on the exit
work.

Split the meaning of TIF_SINGLESTEP such that it only means single-step
mode, and create a new type of SYSCALL_WORK to request a trap immediately
after a syscall in single-step mode.  In the current implementation, the
SYSCALL_WORK flag shadows the TIF_SINGLESTEP flag for simplicity.

Update x86 to flip this bit when a tracer enables single stepping.

Fixes: 299155244770 ("entry: Drop usage of TIF flags in the generic syscall code")
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Kyle Huey <me@kylehuey.com>
Link: https://lore.kernel.org/r/87h7mtc9pr.fsf_-_@collabora.com
3 years agoRevert "lib: Restrict cpumask_local_spread to houskeeping CPUs"
Thomas Gleixner [Fri, 5 Feb 2021 22:28:29 +0000 (23:28 +0100)]
Revert "lib: Restrict cpumask_local_spread to houskeeping CPUs"

This reverts commit 1abdfe706a579a702799fce465bceb9fb01d407c.

This change is broken and not solving any problem it claims to solve.

Robin reported that cpumask_local_spread() now returns any cpu out of
cpu_possible_mask in case that NOHZ_FULL is disabled (runtime or compile
time). It can also return any offline or not-present CPU in the
housekeeping mask. Before that it was returning a CPU out of
online_cpu_mask.

While the function is racy against CPU hotplug if the caller does not
protect against it, the actual use cases are not caring much about it as
they use it mostly as hint for:

 - the user space affinity hint which is unused by the kernel
 - memory node selection which is just suboptimal
 - network queue affinity which might fail but is handled gracefully

But the occasional fail vs. hotplug is very different from returning
anything from possible_cpu_mask which can have a large amount of offline
CPUs obviously.

The changelog of the commit claims:

 "The current implementation of cpumask_local_spread() does not respect
  the isolated CPUs, i.e., even if a CPU has been isolated for Real-Time
  task, it will return it to the caller for pinning of its IRQ
  threads. Having these unwanted IRQ threads on an isolated CPU adds up
  to a latency overhead."

The only correct part of this changelog is:

 "The current implementation of cpumask_local_spread() does not respect
  the isolated CPUs."

Everything else is just disjunct from reality.

Reported-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Nitesh Narayan Lal <nitesh@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: abelits@marvell.com
Cc: davem@davemloft.net
Link: https://lore.kernel.org/r/87y2g26tnt.fsf@nanos.tec.linutronix.de
3 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 5 Feb 2021 21:07:27 +0000 (13:07 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "18 patches.

  Subsystems affected by this patch series: mm (hugetlb, compaction,
  vmalloc, shmem, memblock, pagecache, kasan, and hugetlb), mailmap,
  gcov, ubsan, and MAINTAINERS"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  MAINTAINERS/.mailmap: use my @kernel.org address
  mm: hugetlb: fix missing put_page in gather_surplus_pages()
  ubsan: implement __ubsan_handle_alignment_assumption
  kasan: make addr_has_metadata() return true for valid addresses
  kasan: add explicit preconditions to kasan_report()
  mm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()
  mailmap: add entries for Manivannan Sadhasivam
  mailmap: fix name/email for Viresh Kumar
  memblock: do not start bottom-up allocations with kernel_end
  mm: thp: fix MADV_REMOVE deadlock on shmem THP
  init/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov
  mm/vmalloc: separate put pages and flush VM flags
  mm, compaction: move high_pfn to the for loop scope
  mm: migrate: do not migrate HugeTLB page whose refcount is one
  mm: hugetlb: remove VM_BUG_ON_PAGE from page_huge_active
  mm: hugetlb: fix a race between isolating and freeing page
  mm: hugetlb: fix a race between freeing and dissolving the page
  mm: hugetlbfs: fix cannot migrate the fallocated HugeTLB page

3 years agogenirq: Prevent [devm_]irq_alloc_desc from returning irq 0
Hans de Goede [Mon, 21 Dec 2020 18:56:47 +0000 (19:56 +0100)]
genirq: Prevent [devm_]irq_alloc_desc from returning irq 0

Since commit a85a6c86c25b ("driver core: platform: Clarify that IRQ 0
is invalid"), having a linux-irq with number 0 will trigger a WARN()
when calling platform_get_irq*() to retrieve that linux-irq.

Since [devm_]irq_alloc_desc allocs a single irq and since irq 0 is not used
on some systems, it can return 0, triggering that WARN(). This happens
e.g. on Intel Bay Trail and Cherry Trail devices using the LPE audio engine
for HDMI audio:

 0 is an invalid IRQ number
 WARNING: CPU: 3 PID: 472 at drivers/base/platform.c:238 platform_get_irq_optional+0x108/0x180
 Modules linked in: snd_hdmi_lpe_audio(+) ...

 Call Trace:
  platform_get_irq+0x17/0x30
  hdmi_lpe_audio_probe+0x4a/0x6c0 [snd_hdmi_lpe_audio]

 ---[ end trace ceece38854223a0b ]---

Change the 'from' parameter passed to __[devm_]irq_alloc_descs() by the
[devm_]irq_alloc_desc macros from 0 to 1, so that these macros will no
longer return 0.

Fixes: a85a6c86c25b ("driver core: platform: Clarify that IRQ 0 is invalid")
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20201221185647.226146-1-hdegoede@redhat.com
3 years agocifs: report error instead of invalid when revalidating a dentry fails
Aurelien Aptel [Fri, 5 Feb 2021 14:42:48 +0000 (15:42 +0100)]
cifs: report error instead of invalid when revalidating a dentry fails

Assuming
- //HOST/a is mounted on /mnt
- //HOST/b is mounted on /mnt/b

On a slow connection, running 'df' and killing it while it's
processing /mnt/b can make cifs_get_inode_info() returns -ERESTARTSYS.

This triggers the following chain of events:
=> the dentry revalidation fail
=> dentry is put and released
=> superblock associated with the dentry is put
=> /mnt/b is unmounted

This patch makes cifs_d_revalidate() return the error instead of 0
(invalid) when cifs_revalidate_dentry() fails, except for ENOENT (file
deleted) and ESTALE (file recreated).

Signed-off-by: Aurelien Aptel <aaptel@suse.com>
Suggested-by: Shyam Prasad N <nspmangalore@gmail.com>
Reviewed-by: Shyam Prasad N <nspmangalore@gmail.com>
CC: stable@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
3 years agox86/debug: Prevent data breakpoints on cpu_dr7
Lai Jiangshan [Thu, 4 Feb 2021 15:27:07 +0000 (23:27 +0800)]
x86/debug: Prevent data breakpoints on cpu_dr7

local_db_save() is called at the start of exc_debug_kernel(), reads DR7 and
disables breakpoints to prevent recursion.

When running in a guest (X86_FEATURE_HYPERVISOR), local_db_save() reads the
per-cpu variable cpu_dr7 to check whether a breakpoint is active or not
before it accesses DR7.

A data breakpoint on cpu_dr7 therefore results in infinite #DB recursion.

Disallow data breakpoints on cpu_dr7 to prevent that.

Fixes: 84b6a3491567a("x86/entry: Optimize local_db_save() for virt")
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210204152708.21308-2-jiangshanlai@gmail.com
3 years agox86/debug: Prevent data breakpoints on __per_cpu_offset
Lai Jiangshan [Thu, 4 Feb 2021 15:27:06 +0000 (23:27 +0800)]
x86/debug: Prevent data breakpoints on __per_cpu_offset

When FSGSBASE is enabled, paranoid_entry() fetches the per-CPU GSBASE value
via __per_cpu_offset or pcpu_unit_offsets.

When a data breakpoint is set on __per_cpu_offset[cpu] (read-write
operation), the specific CPU will be stuck in an infinite #DB loop.

RCU will try to send an NMI to the specific CPU, but it is not working
either since NMI also relies on paranoid_entry(). Which means it's
undebuggable.

Fixes: eaad981291ee3("x86/entry/64: Introduce the FIND_PERCPU_BASE macro")
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210204152708.21308-1-jiangshanlai@gmail.com
3 years agoMAINTAINERS/.mailmap: use my @kernel.org address
Nathan Chancellor [Fri, 5 Feb 2021 02:33:03 +0000 (18:33 -0800)]
MAINTAINERS/.mailmap: use my @kernel.org address

Use my @kernel.org for all points of contact so that I am always
accessible.

Link: https://lkml.kernel.org/r/20210126212730.2097108-1-nathan@kernel.org
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Miguel Ojeda <ojeda@kernel.org>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlb: fix missing put_page in gather_surplus_pages()
Muchun Song [Fri, 5 Feb 2021 02:33:00 +0000 (18:33 -0800)]
mm: hugetlb: fix missing put_page in gather_surplus_pages()

The VM_BUG_ON_PAGE avoids the generation of any code, even if that
expression has side-effects when !CONFIG_DEBUG_VM.

Link: https://lkml.kernel.org/r/20210126031009.96266-1-songmuchun@bytedance.com
Fixes: e5dfacebe4a4 ("mm/hugetlb.c: just use put_page_testzero() instead of page_count()")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoubsan: implement __ubsan_handle_alignment_assumption
Nathan Chancellor [Fri, 5 Feb 2021 02:32:57 +0000 (18:32 -0800)]
ubsan: implement __ubsan_handle_alignment_assumption

When building ARCH=mips 32r2el_defconfig with CONFIG_UBSAN_ALIGNMENT:

  ld.lld: error: undefined symbol: __ubsan_handle_alignment_assumption
     referenced by slab.h:557 (include/linux/slab.h:557)
                   main.o:(do_initcalls) in archive init/built-in.a
     referenced by slab.h:448 (include/linux/slab.h:448)
                   do_mounts_rd.o:(rd_load_image) in archive init/built-in.a
     referenced by slab.h:448 (include/linux/slab.h:448)
                   do_mounts_rd.o:(identify_ramdisk_image) in archive init/built-in.a
     referenced 1579 more times

Implement this for the kernel based on LLVM's
handleAlignmentAssumptionImpl because the kernel is not linked against
the compiler runtime.

Link: https://github.com/ClangBuiltLinux/linux/issues/1245
Link: https://github.com/llvm/llvm-project/blob/llvmorg-11.0.1/compiler-rt/lib/ubsan/ubsan_handlers.cpp#L151-L190
Link: https://lkml.kernel.org/r/20210127224451.2587372-1-nathan@kernel.org
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokasan: make addr_has_metadata() return true for valid addresses
Vincenzo Frascino [Fri, 5 Feb 2021 02:32:53 +0000 (18:32 -0800)]
kasan: make addr_has_metadata() return true for valid addresses

Currently, addr_has_metadata() returns true for every address.  An
invalid address (e.g.  NULL) passed to the function when, KASAN_HW_TAGS
is enabled, leads to a kernel panic.

Make addr_has_metadata() return true for valid addresses only.

Note: KASAN_HW_TAGS support for vmalloc will be added with a future
patch.

Link: https://lkml.kernel.org/r/20210126134409.47894-3-vincenzo.frascino@arm.com
Fixes: 2e903b91479782b7 ("kasan, arm64: implement HW_TAGS runtime")
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
Cc: "Paul E . McKenney" <paulmck@kernel.org>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agokasan: add explicit preconditions to kasan_report()
Vincenzo Frascino [Fri, 5 Feb 2021 02:32:49 +0000 (18:32 -0800)]
kasan: add explicit preconditions to kasan_report()

Patch series "kasan: Fix metadata detection for KASAN_HW_TAGS", v5.

With the introduction of KASAN_HW_TAGS, kasan_report() currently assumes
that every location in memory has valid metadata associated.  This is
due to the fact that addr_has_metadata() returns always true.

As a consequence of this, an invalid address (e.g.  NULL pointer
address) passed to kasan_report() when KASAN_HW_TAGS is enabled, leads
to a kernel panic.

Example below, based on arm64:

   BUG: KASAN: invalid-access in 0x0
   Read at addr 0000000000000000 by task swapper/0/1
   Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
   Mem abort info:
     ESR = 0x96000004
     EC = 0x25: DABT (current EL), IL = 32 bits
     SET = 0, FnV = 0
     EA = 0, S1PTW = 0
   Data abort info:
     ISV = 0, ISS = 0x00000004
     CM = 0, WnR = 0

  ...

   Call trace:
    mte_get_mem_tag+0x24/0x40
    kasan_report+0x1a4/0x410
    alsa_sound_last_init+0x8c/0xa4
    do_one_initcall+0x50/0x1b0
    kernel_init_freeable+0x1d4/0x23c
    kernel_init+0x14/0x118
    ret_from_fork+0x10/0x34
   Code: d65f03c0 9000f021 f9428021 b6cfff61 (d9600000)
   ---[ end trace 377c8bb45bdd3a1a ]---
   hrtimer: interrupt took 48694256 ns
   note: swapper/0[1] exited with preempt_count 1
   Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
   SMP: stopping secondary CPUs
   Kernel Offset: 0x35abaf140000 from 0xffff800010000000
   PHYS_OFFSET: 0x40000000
   CPU features: 0x0a7e0152,61c0a030
   Memory Limit: none
   ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---

This series fixes the behavior of addr_has_metadata() that now returns
true only when the address is valid.

This patch (of 2):

With the introduction of KASAN_HW_TAGS, kasan_report() accesses the
metadata only when addr_has_metadata() succeeds.

Add a comment to make sure that the preconditions to the function are
explicitly clarified.

Link: https://lkml.kernel.org/r/20210126134409.47894-1-vincenzo.frascino@arm.com
Link: https://lkml.kernel.org/r/20210126134409.47894-2-vincenzo.frascino@arm.com
Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com>
Reviewed-by: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: "Paul E . McKenney" <paulmck@kernel.org>
Cc: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()
Waiman Long [Fri, 5 Feb 2021 02:32:45 +0000 (18:32 -0800)]
mm/filemap: add missing mem_cgroup_uncharge() to __add_to_page_cache_locked()

Commit 3fea5a499d57 ("mm: memcontrol: convert page cache to a new
mem_cgroup_charge() API") introduced a bug in __add_to_page_cache_locked()
causing the following splat:

  page dumped because: VM_BUG_ON_PAGE(page_memcg(page))
  pages's memcg:ffff8889a4116000
  ------------[ cut here ]------------
  kernel BUG at mm/memcontrol.c:2924!
  invalid opcode: 0000 [#1] SMP KASAN PTI
  CPU: 35 PID: 12345 Comm: cat Tainted: G S      W I       5.11.0-rc4-debug+ #1
  Hardware name: HP HP Z8 G4 Workstation/81C7, BIOS P60 v01.25 12/06/2017
  RIP: commit_charge+0xf4/0x130
  Call Trace:
    mem_cgroup_charge+0x175/0x770
    __add_to_page_cache_locked+0x712/0xad0
    add_to_page_cache_lru+0xc5/0x1f0
    cachefiles_read_or_alloc_pages+0x895/0x2e10 [cachefiles]
    __fscache_read_or_alloc_pages+0x6c0/0xa00 [fscache]
    __nfs_readpages_from_fscache+0x16d/0x630 [nfs]
    nfs_readpages+0x24e/0x540 [nfs]
    read_pages+0x5b1/0xc40
    page_cache_ra_unbounded+0x460/0x750
    generic_file_buffered_read_get_pages+0x290/0x1710
    generic_file_buffered_read+0x2a9/0xc30
    nfs_file_read+0x13f/0x230 [nfs]
    new_sync_read+0x3af/0x610
    vfs_read+0x339/0x4b0
    ksys_read+0xf1/0x1c0
    do_syscall_64+0x33/0x40
    entry_SYSCALL_64_after_hwframe+0x44/0xa9

Before that commit, there was a try_charge() and commit_charge() in
__add_to_page_cache_locked().  These two separated charge functions were
replaced by a single mem_cgroup_charge().  However, it forgot to add a
matching mem_cgroup_uncharge() when the xarray insertion failed with the
page released back to the pool.

Fix this by adding a mem_cgroup_uncharge() call when insertion error
happens.

Link: https://lkml.kernel.org/r/20210125042441.20030-1-longman@redhat.com
Fixes: 3fea5a499d57 ("mm: memcontrol: convert page cache to a new mem_cgroup_charge() API")
Signed-off-by: Waiman Long <longman@redhat.com>
Reviewed-by: Alex Shi <alex.shi@linux.alibaba.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Muchun Song <smuchun@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomailmap: add entries for Manivannan Sadhasivam
Manivannan Sadhasivam [Fri, 5 Feb 2021 02:32:42 +0000 (18:32 -0800)]
mailmap: add entries for Manivannan Sadhasivam

Map my personal and work addresses to korg mail address.

Link: https://lkml.kernel.org/r/20210201104640.108556-1-manivannan.sadhasivam@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomailmap: fix name/email for Viresh Kumar
Viresh Kumar [Fri, 5 Feb 2021 02:32:39 +0000 (18:32 -0800)]
mailmap: fix name/email for Viresh Kumar

For some of the patches the email id was misspelled to linaro.com
instead of linaro.org and for others Viresh Kumar was written as "viresh
kumar" (all small).  Fix both with help of mailmap entries.

Link: https://lkml.kernel.org/r/d6b80b210d7fe0ddc1d4d0b22eff9708c72ef8b3.1612178938.git.viresh.kumar@linaro.org
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomemblock: do not start bottom-up allocations with kernel_end
Roman Gushchin [Fri, 5 Feb 2021 02:32:36 +0000 (18:32 -0800)]
memblock: do not start bottom-up allocations with kernel_end

With kaslr the kernel image is placed at a random place, so starting the
bottom-up allocation with the kernel_end can result in an allocation
failure and a warning like this one:

  hugetlb_cma: reserve 2048 MiB, up to 2048 MiB per node
  ------------[ cut here ]------------
  memblock: bottom-up allocation failed, memory hotremove may be affected
  WARNING: CPU: 0 PID: 0 at mm/memblock.c:332 memblock_find_in_range_node+0x178/0x25a
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper Not tainted 5.10.0+ #1169
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1.fc33 04/01/2014
  RIP: 0010:memblock_find_in_range_node+0x178/0x25a
  Code: e9 6d ff ff ff 48 85 c0 0f 85 da 00 00 00 80 3d 9b 35 df 00 00 75 15 48 c7 c7 c0 75 59 88 c6 05 8b 35 df 00 01 e8 25 8a fa ff <0f> 0b 48 c7 44 24 20 ff ff ff ff 44 89 e6 44 89 ea 48 c7 c1 70 5c
  RSP: 0000:ffffffff88803d18 EFLAGS: 00010086 ORIG_RAX: 0000000000000000
  RAX: 0000000000000000 RBX: 0000000240000000 RCX: 00000000ffffdfff
  RDX: 00000000ffffdfff RSI: 00000000ffffffea RDI: 0000000000000046
  RBP: 0000000100000000 R08: ffffffff88922788 R09: 0000000000009ffb
  R10: 00000000ffffe000 R11: 3fffffffffffffff R12: 0000000000000000
  R13: 0000000000000000 R14: 0000000080000000 R15: 00000001fb42c000
  FS:  0000000000000000(0000) GS:ffffffff88f71000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: ffffa080fb401000 CR3: 00000001fa80a000 CR4: 00000000000406b0
  Call Trace:
    memblock_alloc_range_nid+0x8d/0x11e
    cma_declare_contiguous_nid+0x2c4/0x38c
    hugetlb_cma_reserve+0xdc/0x128
    flush_tlb_one_kernel+0xc/0x20
    native_set_fixmap+0x82/0xd0
    flat_get_apic_id+0x5/0x10
    register_lapic_address+0x8e/0x97
    setup_arch+0x8a5/0xc3f
    start_kernel+0x66/0x547
    load_ucode_bsp+0x4c/0xcd
    secondary_startup_64_no_verify+0xb0/0xbb
  random: get_random_bytes called from __warn+0xab/0x110 with crng_init=0
  ---[ end trace f151227d0b39be70 ]---

At the same time, the kernel image is protected with memblock_reserve(),
so we can just start searching at PAGE_SIZE.  In this case the bottom-up
allocation has the same chances to success as a top-down allocation, so
there is no reason to fallback in the case of a failure.  All together it
simplifies the logic.

Link: https://lkml.kernel.org/r/20201217201214.3414100-2-guro@fb.com
Fixes: 8fabc623238e ("powerpc: Ensure that swiotlb buffer is allocated from low memory")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Wonhyuk Yang <vvghjk1234@gmail.com>
Cc: Thiago Jung Bauermann <bauerman@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: thp: fix MADV_REMOVE deadlock on shmem THP
Hugh Dickins [Fri, 5 Feb 2021 02:32:31 +0000 (18:32 -0800)]
mm: thp: fix MADV_REMOVE deadlock on shmem THP

Sergey reported deadlock between kswapd correctly doing its usual
lock_page(page) followed by down_read(page->mapping->i_mmap_rwsem), and
madvise(MADV_REMOVE) on an madvise(MADV_HUGEPAGE) area doing
down_write(page->mapping->i_mmap_rwsem) followed by lock_page(page).

This happened when shmem_fallocate(punch hole)'s unmap_mapping_range()
reaches zap_pmd_range()'s call to __split_huge_pmd().  The same deadlock
could occur when partially truncating a mapped huge tmpfs file, or using
fallocate(FALLOC_FL_PUNCH_HOLE) on it.

__split_huge_pmd()'s page lock was added in 5.8, to make sure that any
concurrent use of reuse_swap_page() (holding page lock) could not catch
the anon THP's mapcounts and swapcounts while they were being split.

Fortunately, reuse_swap_page() is never applied to a shmem or file THP
(not even by khugepaged, which checks PageSwapCache before calling), and
anonymous THPs are never created in shmem or file areas: so that
__split_huge_pmd()'s page lock can only be necessary for anonymous THPs,
on which there is no risk of deadlock with i_mmap_rwsem.

Link: https://lkml.kernel.org/r/alpine.LSU.2.11.2101161409470.2022@eggly.anvils
Fixes: c444eb564fb1 ("mm: thp: make the THP mapcount atomic against __split_huge_pmd_locked()")
Signed-off-by: Hugh Dickins <hughd@google.com>
Reported-by: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoinit/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov
Johannes Berg [Fri, 5 Feb 2021 02:32:28 +0000 (18:32 -0800)]
init/gcov: allow CONFIG_CONSTRUCTORS on UML to fix module gcov

On ARCH=um, loading a module doesn't result in its constructors getting
called, which breaks module gcov since the debugfs files are never
registered.  On the other hand, in-kernel constructors have already been
called by the dynamic linker, so we can't call them again.

Get out of this conundrum by allowing CONFIG_CONSTRUCTORS to be
selected, but avoiding the in-kernel constructor calls.

Also remove the "if !UML" from GCOV selecting CONSTRUCTORS now, since we
really do want CONSTRUCTORS, just not kernel binary ones.

Link: https://lkml.kernel.org/r/20210120172041.c246a2cac2fb.I1358f584b76f1898373adfed77f4462c8705b736@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Reviewed-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Jessica Yu <jeyu@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/vmalloc: separate put pages and flush VM flags
Rick Edgecombe [Fri, 5 Feb 2021 02:32:24 +0000 (18:32 -0800)]
mm/vmalloc: separate put pages and flush VM flags

When VM_MAP_PUT_PAGES was added, it was defined with the same value as
VM_FLUSH_RESET_PERMS.  This doesn't seem like it will cause any big
functional problems other than some excess flushing for VM_MAP_PUT_PAGES
allocations.

Redefine VM_MAP_PUT_PAGES to have its own value.  Also, rearrange things
so flags are less likely to be missed in the future.

Link: https://lkml.kernel.org/r/20210122233706.9304-1-rick.p.edgecombe@intel.com
Fixes: b944afc9d64d ("mm: add a VM_MAP_PUT_PAGES flag for vmap")
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Axtens <dja@axtens.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm, compaction: move high_pfn to the for loop scope
Rokudo Yan [Fri, 5 Feb 2021 02:32:20 +0000 (18:32 -0800)]
mm, compaction: move high_pfn to the for loop scope

In fast_isolate_freepages, high_pfn will be used if a prefered one (ie
PFN >= low_fn) not found.

But the high_pfn is not reset before searching an free area, so when it
was used as freepage, it may from another free area searched before.  As
a result move_freelist_head(freelist, freepage) will have unexpected
behavior (eg corrupt the MOVABLE freelist)

  Unable to handle kernel paging request at virtual address dead000000000200
  Mem abort info:
    ESR = 0x96000044
    Exception class = DABT (current EL), IL = 32 bits
    SET = 0, FnV = 0
    EA = 0, S1PTW = 0
  Data abort info:
    ISV = 0, ISS = 0x00000044
    CM = 0, WnR = 1
  [dead000000000200] address between user and kernel address ranges

  -000|list_cut_before(inline)
  -000|move_freelist_head(inline)
  -000|fast_isolate_freepages(inline)
  -000|isolate_freepages(inline)
  -000|compaction_alloc(?, ?)
  -001|unmap_and_move(inline)
  -001|migrate_pages([NSD:0xFFFFFF80088CBBD0] from = 0xFFFFFF80088CBD88, [NSD:0xFFFFFF80088CBBC8] get_new_p
  -002|__read_once_size(inline)
  -002|static_key_count(inline)
  -002|static_key_false(inline)
  -002|trace_mm_compaction_migratepages(inline)
  -002|compact_zone(?, [NSD:0xFFFFFF80088CBCB0] capc = 0x0)
  -003|kcompactd_do_work(inline)
  -003|kcompactd([X19] p = 0xFFFFFF93227FBC40)
  -004|kthread([X20] _create = 0xFFFFFFE1AFB26380)
  -005|ret_from_fork(asm)

The issue was reported on an smart phone product with 6GB ram and 3GB
zram as swap device.

This patch fixes the issue by reset high_pfn before searching each free
area, which ensure freepage and freelist match when call
move_freelist_head in fast_isolate_freepages().

Link: http://lkml.kernel.org/r/20190118175136.31341-12-mgorman@techsingularity.net
Link: https://lkml.kernel.org/r/20210112094720.1238444-1-wu-yan@tcl.com
Fixes: 5a811889de10f1eb ("mm, compaction: use free lists to quickly locate a migration target")
Signed-off-by: Rokudo Yan <wu-yan@tcl.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: migrate: do not migrate HugeTLB page whose refcount is one
Muchun Song [Fri, 5 Feb 2021 02:32:17 +0000 (18:32 -0800)]
mm: migrate: do not migrate HugeTLB page whose refcount is one

All pages isolated for the migration have an elevated reference count and
therefore seeing a reference count equal to 1 means that the last user of
the page has dropped the reference and the page has became unused and
there doesn't make much sense to migrate it anymore.

This has been done for regular pages and this patch does the same for
hugetlb pages.  Although the likelihood of the race is rather small for
hugetlb pages it makes sense the two code paths in sync.

Link: https://lkml.kernel.org/r/20210115124942.46403-2-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Yang Shi <shy828301@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlb: remove VM_BUG_ON_PAGE from page_huge_active
Muchun Song [Fri, 5 Feb 2021 02:32:13 +0000 (18:32 -0800)]
mm: hugetlb: remove VM_BUG_ON_PAGE from page_huge_active

The page_huge_active() can be called from scan_movable_pages() which do
not hold a reference count to the HugeTLB page.  So when we call
page_huge_active() from scan_movable_pages(), the HugeTLB page can be
freed parallel.  Then we will trigger a BUG_ON which is in the
page_huge_active() when CONFIG_DEBUG_VM is enabled.  Just remove the
VM_BUG_ON_PAGE.

Link: https://lkml.kernel.org/r/20210115124942.46403-6-songmuchun@bytedance.com
Fixes: 7e1f049efb86 ("mm: hugetlb: cleanup using paeg_huge_active()")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlb: fix a race between isolating and freeing page
Muchun Song [Fri, 5 Feb 2021 02:32:10 +0000 (18:32 -0800)]
mm: hugetlb: fix a race between isolating and freeing page

There is a race between isolate_huge_page() and __free_huge_page().

  CPU0:                                     CPU1:

  if (PageHuge(page))
                                            put_page(page)
                                              __free_huge_page(page)
                                                  spin_lock(&hugetlb_lock)
                                                  update_and_free_page(page)
                                                    set_compound_page_dtor(page,
                                                      NULL_COMPOUND_DTOR)
                                                  spin_unlock(&hugetlb_lock)
    isolate_huge_page(page)
      // trigger BUG_ON
      VM_BUG_ON_PAGE(!PageHead(page), page)
      spin_lock(&hugetlb_lock)
      page_huge_active(page)
        // trigger BUG_ON
        VM_BUG_ON_PAGE(!PageHuge(page), page)
      spin_unlock(&hugetlb_lock)

When we isolate a HugeTLB page on CPU0.  Meanwhile, we free it to the
buddy allocator on CPU1.  Then, we can trigger a BUG_ON on CPU0, because
it is already freed to the buddy allocator.

Link: https://lkml.kernel.org/r/20210115124942.46403-5-songmuchun@bytedance.com
Fixes: c8721bbbdd36 ("mm: memory-hotplug: enable memory hotplug to handle hugepage")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlb: fix a race between freeing and dissolving the page
Muchun Song [Fri, 5 Feb 2021 02:32:06 +0000 (18:32 -0800)]
mm: hugetlb: fix a race between freeing and dissolving the page

There is a race condition between __free_huge_page()
and dissolve_free_huge_page().

  CPU0:                         CPU1:

  // page_count(page) == 1
  put_page(page)
    __free_huge_page(page)
                                dissolve_free_huge_page(page)
                                  spin_lock(&hugetlb_lock)
                                  // PageHuge(page) && !page_count(page)
                                  update_and_free_page(page)
                                  // page is freed to the buddy
                                  spin_unlock(&hugetlb_lock)
      spin_lock(&hugetlb_lock)
      clear_page_huge_active(page)
      enqueue_huge_page(page)
      // It is wrong, the page is already freed
      spin_unlock(&hugetlb_lock)

The race window is between put_page() and dissolve_free_huge_page().

We should make sure that the page is already on the free list when it is
dissolved.

As a result __free_huge_page would corrupt page(s) already in the buddy
allocator.

Link: https://lkml.kernel.org/r/20210115124942.46403-4-songmuchun@bytedance.com
Fixes: c8721bbbdd36 ("mm: memory-hotplug: enable memory hotplug to handle hugepage")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm: hugetlbfs: fix cannot migrate the fallocated HugeTLB page
Muchun Song [Fri, 5 Feb 2021 02:32:03 +0000 (18:32 -0800)]
mm: hugetlbfs: fix cannot migrate the fallocated HugeTLB page

If a new hugetlb page is allocated during fallocate it will not be
marked as active (set_page_huge_active) which will result in a later
isolate_huge_page failure when the page migration code would like to
move that page.  Such a failure would be unexpected and wrong.

Only export set_page_huge_active, just leave clear_page_huge_active as
static.  Because there are no external users.

Link: https://lkml.kernel.org/r/20210115124942.46403-3-songmuchun@bytedance.com
Fixes: 70c3547e36f5 (hugetlbfs: add hugetlbfs_fallocate())
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'nfsd-5.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux
Linus Torvalds [Fri, 5 Feb 2021 18:11:14 +0000 (10:11 -0800)]
Merge tag 'nfsd-5.11-3' of git://git./linux/kernel/git/cel/linux

Pull nfsd fix from Chuck Lever:
 "Fix non-page-aligned NFS READs"

* tag 'nfsd-5.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
  SUNRPC: Fix NFS READs that start at non-page-aligned offsets

3 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 5 Feb 2021 18:03:01 +0000 (10:03 -0800)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "x86 has lots of small bugfixes, mostly one liners. It's quite late in
  5.11-rc but none of them are related to this merge window; it's just
  bugs coming in at the wrong time.

  Of note among the others is "KVM: x86: Allow guests to see
  MSR_IA32_TSX_CTRL even if tsx=off" that fixes a live migration failure
  seen on distros that hadn't switched to tsx=off right away.

  ARM:
  - Avoid clobbering extra registers on initialisation"

[ Sean Christopherson notes that commit 943dea8af21b ("KVM: x86: Update
  emulator context mode if SYSENTER xfers to 64-bit mode") should have
  had authorship credited to Jonny Barker, not to him.  - Linus ]

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: Set so called 'reserved CR3 bits in LM mask' at vCPU reset
  KVM: x86/mmu: Fix TDP MMU zap collapsible SPTEs
  KVM: x86: cleanup CR3 reserved bits checks
  KVM: SVM: Treat SVM as unsupported when running as an SEV guest
  KVM: x86: Update emulator context mode if SYSENTER xfers to 64-bit mode
  KVM: x86: Supplement __cr4_reserved_bits() with X86_FEATURE_PCID check
  KVM/x86: assign hva with the right value to vm_munmap the pages
  KVM: x86: Allow guests to see MSR_IA32_TSX_CTRL even if tsx=off
  Fix unsynchronized access to sev members through svm_register_enc_region
  KVM: Documentation: Fix documentation for nested.
  KVM: x86: fix CPUID entries returned by KVM_GET_CPUID2 ioctl
  KVM: arm64: Don't clobber x4 in __do_hyp_init

3 years agoMerge tag 'iommu-fixes-v5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 5 Feb 2021 17:57:29 +0000 (09:57 -0800)]
Merge tag 'iommu-fixes-v5.11-rc6' of git://git./linux/kernel/git/joro/iommu

Pull IOMMU fix from Joerg Roedel:
 "Fix a possible NULL-ptr dereference in dev_iommu_priv_get() which is
  too easy to accidentially trigger from IOMMU drivers.

  In the current case the AMD IOMMU driver triggered it on some machines
  in the IO-page-fault path, so fix it once and for all"

* tag 'iommu-fixes-v5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu: Check dev->iommu in dev_iommu_priv_get() before dereferencing it

3 years agoMerge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Linus Torvalds [Fri, 5 Feb 2021 17:54:20 +0000 (09:54 -0800)]
Merge tag 'for_linus' of git://git./linux/kernel/git/mst/vhost

Pull vdpa fix from Michael Tsirkin:
 "A bugfix in the mlx driver I got at the last minute"

* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
  vdpa/mlx5: Restore the hardware used index after change map

3 years agoMerge tag 'mmc-v5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
Linus Torvalds [Fri, 5 Feb 2021 17:53:11 +0000 (09:53 -0800)]
Merge tag 'mmc-v5.11-rc6' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "MMC core:
   - Limit retries when analyse of SDIO tuples fails

  MMC host:
   - sdhci: Fix linking err for sdhci-brcmstb"

* tag 'mmc-v5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc:
  mmc: sdhci-pltfm: Fix linking err for sdhci-brcmstb
  mmc: core: Limit retries when analyse of SDIO tuples fails

3 years agoMerge tag 'drm-fixes-2021-02-05-1' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 5 Feb 2021 17:50:21 +0000 (09:50 -0800)]
Merge tag 'drm-fixes-2021-02-05-1' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Fixes for rc7, bit bigger than I'd like at this stage, but most of the
  i915 stuff and some amdgpu is destined for staging and I'd rather not
  hold it up, the i915 changes also pulled in a few precusor code
  movement patches to make things cleaner, but nothing seems that
  horrible, and I've checked over all of it.

  Otherwise there is a nouveau dma-api warning regression, and a ttm
  page allocation warning fix, and some fixes for a bridge chip,

  ttm:
   - fix huge page warning regression

  i915:
   - Skip vswing programming for TBT
   - Power up combo PHY lanes for HDMI
   - Fix double YUV range correction on HDR planes
   - Fix the MST PBN divider calculation
   - Fix LTTPR vswing/pre-emp setting in non-transparent mode
   - Move the breadcrumb to the signaler if completed upon cancel
   - Close race between enable_breadcrumbs and cancel_breadcrumbs
   - Drop lru bumping on display unpinning

  amdgpu:
   - Fix retry in gem create
   - Vangogh fixes
   - Fix for display from shared buffers
   - Various display fixes

  amdkfd:
   - Fix regression in buffer free

  nouveau:
   - fix DMA API warning regression

  drm/bridge/lontium-lt9611uxc:
   - EDID fixes
   - Don't handle hotplug events in IRQ handler"

* tag 'drm-fixes-2021-02-05-1' of git://anongit.freedesktop.org/drm/drm: (29 commits)
  drm/nouveau: fix dma syncing warning with debugging on.
  drm/amd/display: Decrement refcount of dc_sink before reassignment
  drm/amd/display: Free atomic state after drm_atomic_commit
  drm/amd/display: Fix dc_sink kref count in emulated_link_detect
  drm/amd/display: Release DSC before acquiring
  drm/amd/display: Revert "Fix EDID parsing after resume from suspend"
  drm/amd/display: Add more Clock Sources to DCN2.1
  drm/amd/display: reuse current context instead of recreating one
  drm/amd/display: Fix DPCD translation for LTTPR AUX_RD_INTERVAL
  drm/amdgpu: enable freesync for A+A configs
  drm/amd/pm: fill in the data member of v2 gpu metrics table for vangogh
  drm/amdgpu/gfx10: update CGTS_TCC_DISABLE and CGTS_USER_TCC_DISABLE register offsets for VGH
  drm/amdkfd: fix null pointer panic while free buffer in kfd
  drm/amdgpu: fix the issue that retry constantly once the buffer is oversize
  drm/i915/dp: Fix LTTPR vswing/pre-emp setting in non-transparent mode
  drm/i915/dp: Move intel_dp_set_signal_levels() to intel_dp_link_training.c
  drm/i915: Fix the MST PBN divider calculation
  drm/dp/mst: Export drm_dp_get_vc_payload_bw()
  drm/i915/gem: Drop lru bumping on display unpinning
  drm/i915/gt: Close race between enable_breadcrumbs and cancel_breadcrumbs
  ...

3 years agontp: Use freezable workqueue for RTC synchronization
Geert Uytterhoeven [Mon, 25 Jan 2021 14:30:39 +0000 (15:30 +0100)]
ntp: Use freezable workqueue for RTC synchronization

The bug fixed by commit e3fab2f3de081e98 ("ntp: Fix RTC synchronization on
32-bit platforms") revealed an underlying issue: RTC synchronization may
happen anytime, even while the system is partially suspended.

On systems where the RTC is connected to an I2C bus, the I2C bus controller
may already or still be suspended, triggering a WARNING during suspend or
resume from s2ram:

    WARNING: CPU: 0 PID: 124 at drivers/i2c/i2c-core.h:54 __i2c_transfer+0x634/0x680
    i2c i2c-6: Transfer while suspended
    [...]
    Workqueue: events_power_efficient sync_hw_clock
    [...]
      (__i2c_transfer)
      (i2c_transfer)
      (regmap_i2c_read)
      ...
      (da9063_rtc_set_time)
      (rtc_set_time)
      (sync_hw_clock)
      (process_one_work)

Fix this race condition by using the freezable instead of the normal
power-efficient workqueue.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Link: https://lore.kernel.org/r/20210125143039.1051912-1-geert+renesas@glider.be
3 years agovdpa/mlx5: Restore the hardware used index after change map
Eli Cohen [Thu, 4 Feb 2021 07:36:18 +0000 (09:36 +0200)]
vdpa/mlx5: Restore the hardware used index after change map

When a change of memory map occurs, the hardware resources are destroyed
and then re-created again with the new memory map. In such case, we need
to restore the hardware available and used indices. The driver failed to
restore the used index which is added here.

Also, since the driver also fails to reset the available and used
indices upon device reset, fix this here to avoid regression caused by
the fact that used index may not be zero upon device reset.

Fixes: 1a86b377aa21 ("vdpa/mlx5: Add VDPA driver for supported mlx5 devices")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20210204073618.36336-1-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
3 years agosmb3: fix crediting for compounding when only one request in flight
Pavel Shilovsky [Wed, 3 Feb 2021 04:34:32 +0000 (22:34 -0600)]
smb3: fix crediting for compounding when only one request in flight

Currently we try to guess if a compound request is going to
succeed waiting for credits or not based on the number of
requests in flight. This approach doesn't work correctly
all the time because there may be only one request in
flight which is going to bring multiple credits satisfying
the compound request.

Change the behavior to fail a request only if there are no requests
in flight at all and proceed waiting for credits otherwise.

Cc: <stable@vger.kernel.org> # 5.1+
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com>
Reviewed-by: Tom Talpey <tom@talpey.com>
Reviewed-by: Shyam Prasad N <nspmangalore@gmail.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
3 years agodma-mapping: benchmark: use u8 for reserved field in uAPI structure
Barry Song [Fri, 5 Feb 2021 11:33:24 +0000 (00:33 +1300)]
dma-mapping: benchmark: use u8 for reserved field in uAPI structure

The original code put five u32 before a u64 expansion[10] array. Five is
odd, this will cause trouble in the extension of the structure by adding
new features. This patch moves to use u8 for reserved field to avoid
future alignment risk.
Meanwhile, it also clears the memory of struct map_benchmark in tools,
otherwise, if users use old version to run on newer kernel, the random
expansion value will cause side effect on newer kernel.

Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
3 years agoARM: kexec: fix oops after TLB are invalidated
Russell King [Mon, 1 Feb 2021 19:40:01 +0000 (19:40 +0000)]
ARM: kexec: fix oops after TLB are invalidated

Giancarlo Ferrari reports the following oops while trying to use kexec:

 Unable to handle kernel paging request at virtual address 80112f38
 pgd = fd7ef03e
 [80112f38] *pgd=0001141e(bad)
 Internal error: Oops: 80d [#1] PREEMPT SMP ARM
 ...

This is caused by machine_kexec() trying to set the kernel text to be
read/write, so it can poke values into the relocation code before
copying it - and an interrupt occuring which changes the page tables.
The subsequent writes then hit read-only sections that trigger a
data abort resulting in the above oops.

Fix this by copying the relocation code, and then writing the variables
into the destination, thereby avoiding the need to make the kernel text
read/write.

Reported-by: Giancarlo Ferrari <giancarlo.ferrari89@gmail.com>
Tested-by: Giancarlo Ferrari <giancarlo.ferrari89@gmail.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
3 years agoARM: ensure the signal page contains defined contents
Russell King [Fri, 29 Jan 2021 10:19:07 +0000 (10:19 +0000)]
ARM: ensure the signal page contains defined contents

Ensure that the signal page contains our poison instruction to increase
the protection against ROP attacks and also contains well defined
contents.

Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
3 years agousb: dwc2: Fix endpoint direction check in ep_from_windex
Heiko Stuebner [Wed, 27 Jan 2021 10:39:19 +0000 (11:39 +0100)]
usb: dwc2: Fix endpoint direction check in ep_from_windex

dwc2_hsotg_process_req_status uses ep_from_windex() to retrieve
the endpoint for the index provided in the wIndex request param.

In a test-case with a rndis gadget running and sending a malformed
packet to it like:
    dev.ctrl_transfer(
        0x82,      # bmRequestType
        0x00,       # bRequest
        0x0000,     # wValue
        0x0001,     # wIndex
        0x00       # wLength
    )
it is possible to cause a crash:

[  217.533022] dwc2 ff300000.usb: dwc2_hsotg_process_req_status: USB_REQ_GET_STATUS
[  217.559003] Unable to handle kernel read from unreadable memory at virtual address 0000000000000088
...
[  218.313189] Call trace:
[  218.330217]  ep_from_windex+0x3c/0x54
[  218.348565]  usb_gadget_giveback_request+0x10/0x20
[  218.368056]  dwc2_hsotg_complete_request+0x144/0x184

This happens because ep_from_windex wants to compare the endpoint
direction even if index_to_ep() didn't return an endpoint due to
the direction not matching.

The fix is easy insofar that the actual direction check is already
happening when calling index_to_ep() which will return NULL if there
is no endpoint for the targeted direction, so the offending check
can go away completely.

Fixes: c6f5c050e2a7 ("usb: dwc2: gadget: add bi-directional endpoint support")
Cc: stable@vger.kernel.org
Reported-by: Gerhard Klostermeier <gerhard.klostermeier@syss.de>
Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com>
Link: https://lore.kernel.org/r/20210127103919.58215-1-heiko@sntech.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
3 years agousb: dwc3: fix clock issue during resume in OTG mode
Gary Bisson [Mon, 25 Jan 2021 16:19:34 +0000 (17:19 +0100)]
usb: dwc3: fix clock issue during resume in OTG mode

Commit fe8abf332b8f ("usb: dwc3: support clocks and resets for DWC3
core") introduced clock support and a new function named
dwc3_core_init_for_resume() which enables the clock before calling
dwc3_core_init() during resume as clocks get disabled during suspend.

Unfortunately in this commit the DWC3_GCTL_PRTCAP_OTG case was forgotten
and therefore during resume, a platform could call dwc3_core_init()
without re-enabling the clocks first, preventing to resume properly.

So update the resume path to call dwc3_core_init_for_resume() as it
should.

Fixes: fe8abf332b8f ("usb: dwc3: support clocks and resets for DWC3 core")
Cc: stable@vger.kernel.org
Signed-off-by: Gary Bisson <gary.bisson@boundarydevices.com>
Link: https://lore.kernel.org/r/20210125161934.527820-1-gary.bisson@boundarydevices.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
3 years agokallsyms: fix nonconverging kallsyms table with lld
Arnd Bergmann [Thu, 4 Feb 2021 15:29:47 +0000 (16:29 +0100)]
kallsyms: fix nonconverging kallsyms table with lld

ARM randconfig builds with lld sometimes show a build failure
from kallsyms:

  Inconsistent kallsyms data
  Try make KALLSYMS_EXTRA_PASS=1 as a workaround

The problem is the veneers/thunks getting added by the linker extend
the symbol table, which in turn leads to more veneers being needed,
so it may take a few extra iterations to converge.

This bug has been fixed multiple times before, but comes back every time
a new symbol name is used. lld uses a different set of identifiers from
ld.bfd, so the additional ones need to be added as well.

I looked through the sources and found that arm64 and mips define similar
prefixes, so I'm adding those as well, aside from the ones I observed. I'm
not sure about powerpc64, which seems to already be handled through a
section match, but if it comes back, the "__long_branch_" and "__plt_"
prefixes would have to get added as well.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
3 years agokbuild: fix duplicated flags in DEBUG_CFLAGS
Masahiro Yamada [Wed, 3 Feb 2021 07:52:39 +0000 (16:52 +0900)]
kbuild: fix duplicated flags in DEBUG_CFLAGS

Sedat Dilek noticed duplicated flags in DEBUG_CFLAGS when building
deb-pkg with CONFIG_DEBUG_INFO. For example, 'make CC=clang bindeb-pkg'
reproduces the issue.

Kbuild recurses to the top Makefile for some targets such as package
builds.

With commit 121c5d08d53c ("kbuild: Only add -fno-var-tracking-assignments
for old GCC versions") applied, DEBUG_CFLAGS is now reset only when
CONFIG_CC_IS_GCC=y.

Fix it to reset DEBUG_CFLAGS all the time.

Fixes: 121c5d08d53c ("kbuild: Only add -fno-var-tracking-assignments for old GCC versions")
Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Reviewed-by: Mark Wielaard <mark@klomp.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
3 years agoMerge tag 'drm-intel-fixes-2021-02-04' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Fri, 5 Feb 2021 01:03:27 +0000 (11:03 +1000)]
Merge tag 'drm-intel-fixes-2021-02-04' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v5.11-rc7:
- Skip vswing programming for TBT
- Power up combo PHY lanes for HDMI
- Fix double YUV range correction on HDR planes
- Fix the MST PBN divider calculation
- Fix LTTPR vswing/pre-emp setting in non-transparent mode
- Move the breadcrumb to the signaler if completed upon cancel
- Close race between enable_breadcrumbs and cancel_breadcrumbs
- Drop lru bumping on display unpinning

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/87bld0f36b.fsf@intel.com
3 years agoMerge tag 'pci-v5.11-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Fri, 5 Feb 2021 00:05:40 +0000 (16:05 -0800)]
Merge tag 'pci-v5.11-fixes-2' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fix from Bjorn Helgaas:
 "Revert ASPM suspend/resume fix that regressed NVMe devices (Bjorn
  Helgaas)"

* tag 'pci-v5.11-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  Revert "PCI/ASPM: Save/restore L1SS Capability for suspend/resume"

3 years agoMerge tag 'amd-drm-fixes-5.11-2021-02-03' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 4 Feb 2021 23:29:15 +0000 (09:29 +1000)]
Merge tag 'amd-drm-fixes-5.11-2021-02-03' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.11-2021-02-03:

amdgpu:
- Fix retry in gem create
- Vangogh fixes
- Fix for display from shared buffers
- Various display fixes

amdkfd:
- Fix regression in buffer free

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210204041300.4425-1-alexander.deucher@amd.com
3 years agoio_uring: drop mm/files between task_work_submit
Pavel Begunkov [Thu, 4 Feb 2021 19:22:46 +0000 (19:22 +0000)]
io_uring: drop mm/files between task_work_submit

Since SQPOLL task can be shared and so task_work entries can be a mix of
them, we need to drop mm and files before trying to issue next request.

Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
3 years agox86/apic: Add extra serialization for non-serializing MSRs
Dave Hansen [Thu, 5 Mar 2020 17:47:08 +0000 (09:47 -0800)]
x86/apic: Add extra serialization for non-serializing MSRs

Jan Kiszka reported that the x2apic_wrmsr_fence() function uses a plain
MFENCE while the Intel SDM (10.12.3 MSR Access in x2APIC Mode) calls for
MFENCE; LFENCE.

Short summary: we have special MSRs that have weaker ordering than all
the rest. Add fencing consistent with current SDM recommendations.

This is not known to cause any issues in practice, only in theory.

Longer story below:

The reason the kernel uses a different semantic is that the SDM changed
(roughly in late 2017). The SDM changed because folks at Intel were
auditing all of the recommended fences in the SDM and realized that the
x2apic fences were insufficient.

Why was the pain MFENCE judged insufficient?

WRMSR itself is normally a serializing instruction. No fences are needed
because the instruction itself serializes everything.

But, there are explicit exceptions for this serializing behavior written
into the WRMSR instruction documentation for two classes of MSRs:
IA32_TSC_DEADLINE and the X2APIC MSRs.

Back to x2apic: WRMSR is *not* serializing in this specific case.
But why is MFENCE insufficient? MFENCE makes writes visible, but
only affects load/store instructions. WRMSR is unfortunately not a
load/store instruction and is unaffected by MFENCE. This means that a
non-serializing WRMSR could be reordered by the CPU to execute before
the writes made visible by the MFENCE have even occurred in the first
place.

This means that an x2apic IPI could theoretically be triggered before
there is any (visible) data to process.

Does this affect anything in practice? I honestly don't know. It seems
quite possible that by the time an interrupt gets to consume the (not
yet) MFENCE'd data, it has become visible, mostly by accident.

To be safe, add the SDM-recommended fences for all x2apic WRMSRs.

This also leaves open the question of the _other_ weakly-ordered WRMSR:
MSR_IA32_TSC_DEADLINE. While it has the same ordering architecture as
the x2APIC MSRs, it seems substantially less likely to be a problem in
practice. While writes to the in-memory Local Vector Table (LVT) might
theoretically be reordered with respect to a weakly-ordered WRMSR like
TSC_DEADLINE, the SDM has this to say:

  In x2APIC mode, the WRMSR instruction is used to write to the LVT
  entry. The processor ensures the ordering of this write and any
  subsequent WRMSR to the deadline; no fencing is required.

But, that might still leave xAPIC exposed. The safest thing to do for
now is to add the extra, recommended LFENCE.

 [ bp: Massage commit message, fix typos, drop accidentally added
   newline to tools/arch/x86/include/asm/barrier.h. ]

Reported-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20200305174708.F77040DD@viggo.jf.intel.com
3 years agoRevert "x86/setup: don't remove E820_TYPE_RAM for pfn 0"
Mike Rapoport [Thu, 4 Feb 2021 18:12:37 +0000 (20:12 +0200)]
Revert "x86/setup: don't remove E820_TYPE_RAM for pfn 0"

This reverts commit bde9cfa3afe4324ec251e4af80ebf9b7afaf7afe.

Changing the first memory page type from E820_TYPE_RESERVED to
E820_TYPE_RAM makes it a part of "System RAM" resource rather than a
reserved resource and this in turn causes devmem_is_allowed() to treat
is as area that can be accessed but it is filled with zeroes instead of
the actual data as previously.

The change in /dev/mem output causes lilo to fail as was reported at
slakware users forum, and probably other legacy applications will
experience similar problems.

Link: https://www.linuxquestions.org/questions/slackware-14/slackware-current-lilo-vesa-warnings-after-recent-updates-4175689617/#post6214439
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>